*Edited by Ilze Strumfa and Guntis Bahs*

Pathology is a diagnostic medical specialty dealing with the evaluation of tissues and body fluids to diagnose disease and predict prognosis or response to treatment. In particular, a biopsy is the "gold standard" in the diagnostics of certain diseases, especially tumours. *Pathology - From Classics to Innovations* is a collection of original peer-reviewed studies and review articles by a truly global scientific team on the recent advances in pathology. Chapters discuss classic surgical pathology and the application of microscopic tissue studies in anatomic research, immunohistochemistry, molecular pathology, liquid biopsy, and digital pathology.

ISBN 978-1-83881-858-6

Pathology - From Classics to Innovations

Published in London, UK © 2021 IntechOpen © Stepan Khadzhi / iStock

## Pathology From Classics to Innovations

*Edited by Ilze Strumfa and Guntis Bahs*

## Pathology - From Classics to Innovations

*Edited by Ilze Strumfa and Guntis Bahs*

Published in London, United Kingdom

### *Supporting open minds since 2005*

Pathology - From Classics to Innovations http://dx.doi.org/10.5772/intechopen.87426 Edited by Ilze Strumfa and Guntis Bahs

#### Contributors

Joe Yeong, Siting Goh, Yue Da Chua, Yiyu Cai, Justina Lee, Meir Warman, Rona Bourla, Elchanan Zloczower, Monica Huszar, Stefania Scarpino, Umberto Malapelle, Rodrigo Vismari de Oliveira, Valeria Cecilia Denninghoff, Ruoyu Li, Junzhou Huang, Mehveş Ece Genç, Emine Nur Özdamar, Ruth L. Katz, Xin Ye, Xiao Zheng Yang, Roberta Carbone, Iris Barschack, Shawn Baldacchino, Rajani Singh

#### © The Editor(s) and the Author(s) 2021

The rights of the editor(s) and the author(s) have been asserted in accordance with the Copyright, Designs and Patents Act 1988. All rights to the book as a whole are reserved by INTECHOPEN LIMITED. The book as a whole (compilation) cannot be reproduced, distributed or used for commercial or non-commercial purposes without INTECHOPEN LIMITED's written permission. Enquiries concerning the use of the book should be directed to INTECHOPEN LIMITED rights and permissions department (permissions@intechopen.com).

Violations are liable to prosecution under the governing Copyright Law.

Individual chapters of this publication are distributed under the terms of the Creative Commons Attribution 3.0 Unported License which permits commercial use, distribution and reproduction of the individual chapters, provided the original author(s) and source publication are appropriately acknowledged. If so indicated, certain images may not be included under the Creative Commons license. In such cases users will need to obtain permission from the license holder to reproduce the material. More details and guidelines concerning content reuse and adaptation can be found at http://www.intechopen.com/copyright-policy.html.

#### Notice

Statements and opinions expressed in the chapters are these of the individual contributors and not necessarily those of the editors or publisher. No responsibility is accepted for the accuracy of information contained in the published chapters. The publisher assumes no responsibility for any damage or injury to persons or property arising out of the use of any materials, instructions, methods or ideas contained in the book.

First published in London, United Kingdom, 2021 by IntechOpen IntechOpen is the global imprint of INTECHOPEN LIMITED, registered in England and Wales, registration number: 11086078, 5 Princes Gate Court, London, SW7 2QJ, United Kingdom Printed in Croatia

British Library Cataloguing-in-Publication Data A catalogue record for this book is available from the British Library

Additional hard and PDF copies can be obtained from orders@intechopen.com

Pathology - From Classics to Innovations Edited by Ilze Strumfa and Guntis Bahs p. cm. Print ISBN 978-1-83881-858-6 Online ISBN 978-1-83881-859-3 eBook (PDF) ISBN 978-1-83881-860-9

## We are IntechOpen, the world's leading publisher of Open Access books Built by scientists, for scientists

5,600+

Open access books available

137,000+

International authors and editors

170M+ Downloads

156 Countries delivered to Our authors are among the

Top 1% most cited scientists

12.2%

Contributors from top 500 universities

Selection of our books indexed in the Book Citation Index (BKCI) in Web of Science Core Collection™

### Interested in publishing with us? Contact book.department@intechopen.com

Numbers displayed above are based on latest data collected. For more information visit www.intechopen.com

## Meet the editors

Professor Ilze Strumfa, MD, Ph.D., is an outstanding medical lecturer, actively involved in the research in pathology. She graduated from the Medical Academy of Latvia with distinction in 1998, underwent board certification in pathology in 2001, and received a Ph.D. in 2005. Currently, she is a professor and head of the Department of Pathology, Riga Stradins University (RSU), Latvia. Her twelve years of teaching experience have culminat-

ed with the RSU "Lecturer of the Year" Annual Award (2018), given to the most distinguished teachers. As the head of the Department of Pathology, she is leading a skilled, motivated team of teachers and scientists that have won awards such as Best Academic Unit (2011), Best Ph.D. Student (2012), and Best Digital Junior Teacher (2016). Prof. Strumfa is an author/co-author of more than 100 peer-reviewed journal articles and 16 chapters in scientific monographs and medical textbooks. She has been acting as the leading expert in several European and national research projects devoted to the development of diagnostic technologies, neuroendocrine and endocrine tumors, breast cancer, laboratory training in research, and tumor microenvironment. Her main research interests include morphological and molecular diagnostics and prognostic assessment of tumors as well as digital pathology and other innovations in pathology and cytology.

Professor Guntis Bahs, MD, Ph.D., is a dedicated physician who actively translates his excellent clinical experience into research, medical education, and administrative work. He has published numerous peer-reviewed scientific articles and book chapters on biomarkers in the prognosis and pathogenesis of cardiovascular and pulmonary diseases, also addressing adiposity and metabolic and biochemical events in human disease. Prof. Bahs is the head

of study courses in pulmonology and cardiovascular diseases, as well as the supervisor of the Medicine and Paediatrics study programs in Riga Stradins University (RSU), Latvia. He has served as the dean of the Medical Faculty and is currently the vice-rector for Health Studies at RSU.

Contents

**Section 1**

**Preface XI**

Classic Surgical Pathology and Anatomic Studies **1**

**Chapter 1 3**

**Chapter 2 15**

Molecular Pathology **29**

**Chapter 3 31**

**Chapter 4 49**

Digital Pathology **65**

**Chapter 5 67**

**Chapter 6 83**

Antrochoanal Polyp: Updated Clinical Approach, Histology

Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic Markers in Breast Cancer

Molecular Pathology in the New Age of Personalized Medicine

Fast Regions-of-Interest Detection in Whole Slide Histopathology

Approaches for Handling Immunopathological and Clinical Data Using Deep Learning Methodology: Multiplex IHC/IF Data

*by Siting Goh, Yueda Chua, Justina Lee, Joe Yeong and Yiyu Cai*

Tracking of Fascicles of Cutaneous Nerves of Thigh:

Characteristics, Diagnosis and Treatment *by Warman Meir, Rona Bourla, Monica Huszar* 

*and Elchanan Zloczower*

A Histological Study *by Rajani Singh*

*by Rodrigo Vismari de Oliveira*

*by Valeria Cecilia Denninghoff*

*by Junzhou Huang and Ruoyu Li*

**Section 2**

**Section 3**

Images

as a Paradigm

### Contents



Preface

Pathology is a diagnostic medical specialty dealing with the evaluation of tissues and body fluids to reach a diagnosis and predict prognosis or response to treatment. A biopsy is a well-known "gold standard" in the diagnostics of most cancers, and its value extends beyond oncology to other subspecialties of medicine. Currently, significant innovations in pathology have pushed the field towards less invasive and more exact diagnostic approaches, going deeper – from the tissues to the molecules. In this journey, microscopes and human minds are being supported by digital pathology and artificial

This book is a collection of original peer-reviewed studies and review articles on recent advances in pathology: from classic surgical pathology and application of microscopic tissue evaluation in anatomic studies to immunohistochemistry, molecular pathology,

The book is divided into five sections: "Classic Surgical Pathology and Anatomic Studies," "Molecular Pathology," "Digital Pathology," "Liquid Biopsy," and "Pathophysiology."

The "Classic Surgical Pathology and Anatomic Studies" section includes a brilliant review in Chapter 1 on antrochoanal polyp, a benign unilateral polyp originating from the maxillary sinus and expanding through the accessory or natural ostia into the nasal cavity and choanae. The authors link pathogenesis and histology to clinical features and treatment in a comprehensive and interesting narrative. Regarding anatomic studies, Chapter 2 on tracking cutaneous nerves of the thigh is remarkable not only by its direct findings and meticulous performance but even more so by the ground-breaking

Diagnostics of breast cancer represents the brightest, paradigm-setting example of the clinical application of molecular pathology. Therefore, Chapter 3 in the "Molecular Pathology" section reviews advances in the molecular and immunohistochemical assessment of breast cancer. Along with the classic molecular subtypes of breast cancer, the author summarizes current views on the role of germline BRCA gene mutations, cancer stem cells, immunological features of breast cancer and its microenvironment, as well as the molecular events in advanced breast cancer. To further expand the discussion on molecular pathology, Chapter 4 describes the technological prerequisites of molecular testing, including methodological constraints in the work with formalin-fixed, paraffinembedded tissues. Pre-analytical conditions, such as cold ischemia and tissue fixation, are discussed along with downstream technological steps, for example, tumour microdissection for tumour cell enrichment, PCR assays, and next-generation sequencing.

The "Digital Pathology" section includes two original studies. Chapter 5 presents a computer-based assessment of whole-slide digital pathological images. The authors have targeted the hot topic: how to achieve an objective assessment of very complex images in a reasonable time. They propose a two-stage, superpixel-based approach for the identification of regions of interest, explain the necessity for the innovations in the given point, describe the method, and provide its mathematical basis. Chapter 6 presents an approach to handling multiplex immunohistochemical data by using deep

intelligence.

liquid biopsy, and digital pathology.

technology that can be used in future studies.

learning-based artificial intelligence methodology.

## Preface

Pathology is a diagnostic medical specialty dealing with the evaluation of tissues and body fluids to reach a diagnosis and predict prognosis or response to treatment. A biopsy is a well-known "gold standard" in the diagnostics of most cancers, and its value extends beyond oncology to other subspecialties of medicine. Currently, significant innovations in pathology have pushed the field towards less invasive and more exact diagnostic approaches, going deeper – from the tissues to the molecules. In this journey, microscopes and human minds are being supported by digital pathology and artificial intelligence.

This book is a collection of original peer-reviewed studies and review articles on recent advances in pathology: from classic surgical pathology and application of microscopic tissue evaluation in anatomic studies to immunohistochemistry, molecular pathology, liquid biopsy, and digital pathology.

The book is divided into five sections: "Classic Surgical Pathology and Anatomic Studies," "Molecular Pathology," "Digital Pathology," "Liquid Biopsy," and "Pathophysiology."

The "Classic Surgical Pathology and Anatomic Studies" section includes a brilliant review in Chapter 1 on antrochoanal polyp, a benign unilateral polyp originating from the maxillary sinus and expanding through the accessory or natural ostia into the nasal cavity and choanae. The authors link pathogenesis and histology to clinical features and treatment in a comprehensive and interesting narrative. Regarding anatomic studies, Chapter 2 on tracking cutaneous nerves of the thigh is remarkable not only by its direct findings and meticulous performance but even more so by the ground-breaking technology that can be used in future studies.

Diagnostics of breast cancer represents the brightest, paradigm-setting example of the clinical application of molecular pathology. Therefore, Chapter 3 in the "Molecular Pathology" section reviews advances in the molecular and immunohistochemical assessment of breast cancer. Along with the classic molecular subtypes of breast cancer, the author summarizes current views on the role of germline BRCA gene mutations, cancer stem cells, immunological features of breast cancer and its microenvironment, as well as the molecular events in advanced breast cancer. To further expand the discussion on molecular pathology, Chapter 4 describes the technological prerequisites of molecular testing, including methodological constraints in the work with formalin-fixed, paraffinembedded tissues. Pre-analytical conditions, such as cold ischemia and tissue fixation, are discussed along with downstream technological steps, for example, tumour microdissection for tumour cell enrichment, PCR assays, and next-generation sequencing.

The "Digital Pathology" section includes two original studies. Chapter 5 presents a computer-based assessment of whole-slide digital pathological images. The authors have targeted the hot topic: how to achieve an objective assessment of very complex images in a reasonable time. They propose a two-stage, superpixel-based approach for the identification of regions of interest, explain the necessity for the innovations in the given point, describe the method, and provide its mathematical basis. Chapter 6 presents an approach to handling multiplex immunohistochemical data by using deep learning-based artificial intelligence methodology.

Liquid biopsy represents a spectrum of technologies for the molecular analysis of blood and other biological liquids (e.g., urine, saliva, tears, or others) to detect cell-free/tumour nucleic acids, exosomes, microRNAs, tumour-educated platelets, and circulating or free-floating tumour or foetal cells, depending on the clinical context. In contrast to classic tumour markers (e.g., prostate-specific antigen (PSA) or carcinoembryonic antigen (CEA)), liquid biopsy is distinguished by high specificity, as it provides genomic, proteomic, and cellular characteristics of the disease. The anticipated outstanding reliability of these tests has been reflected in the term itself: "liquid biopsy" is expected to be at least as informative as tissue biopsy, used as the gold standard in certain diagnostic fields, especially in oncology.

The advantages of liquid biopsy include a non-invasive approach that is patient-friendly, associated with low risk of complications, and technically feasible even in patients who are in serious general condition or affected by tumour or metastases that are not easily accessible by conventional tissue biopsy. The clinically simple application allows the repeated use of liquid biopsy resulting in real-time follow-up for the disease course. Liquid biopsy is also a much-awaited tool to overcome the limitations set by tumour heterogeneity upon conventional tissue biopsy representing only a small part of the whole tumour.

In this book, the "Liquid Biopsy" section (Chapters 7–9) includes extensive in-depth reviews, mainly focusing on circulating tumour DNA and circulating tumour cells; extracellular vesicles are also discussed. The clinical use of liquid biopsies is assessed, highlighting the manifold practical consequences in oncology, from screening and early diagnostics to selection of personalized treatment. The approval status of certain diagnostic tests and indications for their application are described, providing a practically useful general evaluation of the current situation in the field of liquid biopsy. Technological background, the mainstay of any laboratory analysis, is appropriately widely considered. In particular, a fluorescence in situ hybridization approach to detect circulating tumour cells represents an interesting and promising innovation.

In the "Pathophysiology" section, Chapter 10 on dopamine turnover concludes the book, emphasizing that pathology comprises not only morphology but also dynamic understanding of pathophysiological processes in disease.

The book was created by a truly global team of scientists including researchers from Israel, the United Kingdom, Italy, Singapore, Argentina, Brazil, China, the United States, India, and Turkey. We would like to thank sincerely all the authors for their excellent contributions.

We also express our sincere gratitude to the IntechOpen editorial team, especially Mrs. Mia Vulovic and Ms. Ana Simcic for their continuous support and professional help during the editorial process.

*"Random discoveries only happen to prepared minds." Blaise Pascal*

> **Ilze Strumfa, MD, Ph.D. and Guntis Bahs, MD, Ph.D.** Professor, Riga Stradiņš University, Latvia

Section 1

## Classic Surgical Pathology and Anatomic Studies

#### **Chapter 1**

## Antrochoanal Polyp: Updated Clinical Approach, Histology Characteristics, Diagnosis and Treatment

*Warman Meir, Rona Bourla, Monica Huszar and Elchanan Zloczower*

#### **Abstract**

Antrochoanal polyp (ACP) is a benign unilateral polyp, originating from the maxillary sinus and expanding through the accessory or natural ostia into the nasal cavity and choanae. It has a 2: 1 male predominance and is more common in children and young adults. The exact pathophysiology is unclear, and it is thought to have less of the inflammatory reactions as opposed to typical bilateral nasal polyps which are commonly seen in diffused chronic rhinosinusitis. The presenting symptoms of ACP are unilateral nasal obstruction and rhinitis. Epistaxis, pain, and foul-smelling secretions are not typically seen and point towards a different etiology. Diagnosis is mainly clinical via endoscopic examination and supported by Computed tomography (CT) imaging. In CT images the three components of the polyp can be identified; an intramaxillary portion, intranasal and choanal components. Treatment is surgical, where Endoscopic sinus surgery (ESS) is the main technique used with other assisting approaches to reach the more challenging anterior and inferior areas of the maxillary sinus. Successful resection depends on complete removal of the intramaxillary component of the polyp to avoid polyp regrowth. The typical histologic characteristics are cyst formation, fibrosis and squamous metaplasia that are significantly more common in ACP than diffused nasal polyps.

**Keywords:** histoloy, Immunohistochemistry, antrochoanal polyp, nasal polyps

#### **1. Introduction**

Antrochoanal polyp (ACP) is a benign, unilateral polyp originating from the maxillary sinus, extending through the natural or accessory ostia into the nasal cavity. This finding is more common in children and young adults [1] with 2:1 male to female ratio. Its etiology is vague and varies from neoplasia to inflammatory polyp or cystic degeneration of intramaxillary retention cyst. The exact anatomic origin of ACP inside the maxillary sinus is not agreed upon in the literature. The medial and posterior walls are the most common origin sites [2, 3], but the polyp may grow from virtually any site inside the maxillary sinus. ACP exits the maxillary sinus through the accessory ostium in at least 70% of cases [4], which may explain why the polyp grows inferiorly and posteriorly into the nasopharynx. Recent

publications show evidence that nearly all ACPs extend through the accessory ostium [2, 5]. The most common symptoms of ACP are nasal obstruction and anterior nasal discharge, while epistaxis and pain point towards a different etiology necessitating further workup. The treatment of choice for ACP is surgical resection [1]. While different surgical techniques were described in the past, endoscopic removal of both the intranasal and intramaxillary parts of the polyp is the common practice today. ACP is common in the pediatric population. While it represents only 4–6% of all nasal polyps in adults, up to 35% of nasal polyps in children will eventually be diagnosed as ACP [6]. The common symptoms are the same as with adults, however additional sinus pathologies are rarely seen in children. Oropharyngeal descent is more prevalent in children compared with adults [7]. In addition, children generally present with more advanced disease, probably as a result of delayed diagnosis. The recurrence rate of ACP after endoscopic surgical treatment is not significantly different between children and adults [8]. A meta-analysis conducted by Galluzzi demonstrated a 15% recurrence rate in children with significantly higher rates in patients who underwent endoscopic surgical treatment alone compared with combined approach (i.e. endoscopic and trans-canine sinusoscopy or mini-Caldwell-Luc) [9].

#### **2. Pathophysiology**

There are different theories regarding the pathogenesis of ACP; Early studies suggested that ACP grows from an antral mucous retention cyst, a quite common finding in the general population (8–10%) [10]. In their attempt to explain why ACP occurs in only a minority of patients with retention cysts, Frosini et al. hypothesized that increased intra-sinus pressure caused by partial occlusion of the natural ostium due to inflammatory changes and edema is leading an antral cyst to herniate through the accessory ostium [5]. Histologic features of ACP, which include a high rate of inflammatory cells, may support this theory.

The association between ACP and allergy is controversial. While the exact pathogenesis of ACP is unknown, a relationship between ACP and allergic rhinitis or ipsilateral maxillary sinusitis has been shown in pediatric patients [7]. Moreover, increased recurrence rates of ACP after endoscopic surgery were noted in children who were exposed to cigarette smoke (aka 'passive smokers'); Mantilla described a series of 27 cases of recurrent ACP in children in which nearly half of the subjects were considered as passive smokers [11]. While this data may point to a causal correlation between smoking and the development of ACP, such a relationship is not documented elsewhere and more research is needed in this area.

#### **3. Differential diagnosis**

The diagnosis of ACP may be challenging, mainly in young children (5–8 years). In this age group, adenoid hypertrophy is a very common finding and the symptoms may resemble those of ACP, like nasal obstruction, chronic rhinorrhea and snoring. Even though the pre-operative management in these cases include nasal endoscopy and/or lateral plain films of the neck, sometimes the diagnosis of ACP may be overlooked. Another unilateral nasal pathology to be ruled out in children is foreign body but it usually manifests with unilateral foul-smelling rhinorrhea. Epistaxis is not a usual clinical feature of ACP. In these cases, vascular lesions (such as juvenile nasopharyngeal angiofibroma, hemangioma or hemangiopericytoma) and neoplasia (inverted papilloma or malignant tumors) should be excluded [12]. The key to


*Antrochoanal Polyp: Updated Clinical Approach, Histology Characteristics, Diagnosis… DOI: http://dx.doi.org/10.5772/intechopen.96329*

#### **Table 1.**

*Differential diagnosis of pediatric nasal obstruction [14].*

differentiate between ACP and other pathologies is a thorough and detailed history along with meticulous physical examination. In cases of limited physical examination, imaging may contribute to the diagnosis. One should keep in mind that adenoid to nasopharynx ratio decreases with age (especially in children >8 years) due to a change in nasopharynx width [13]. Therefore, children older than 8 years must undergo complete nasal flexible endoscopy to rule out nasal polyp (**Table 1**).

#### **4. Clinical manifestations**

#### **4.1 History**

The most common presenting symptoms of ACP are nasal obstruction and anterior rhinorrhea. Nasal Obstruction may be unilateral or bilateral, depends on the evolution of growth of the polyp. When it emerges from the maxillary sinus ostium to the nasal cavity the patient will complain on unilateral nasal obstruction. However, as the polyp further descends into the choana it may cause bilateral obstruction, as commonly seen in hypertrophic obstructive adenoid tissue. Rhinorrhea is usually unilateral and watery; purulence is rarely seen. Other symptoms may include mouth breathing, snoring and sleep disorders, although ACP does no lead to truly obstructive sleep apnea (OSA). The cystic component is very typical to ACP. Some patients report of a sudden watery or yellow drainage followed by a relief of the nasal obstruction implying to a spontaneous rapture of the cystic part in the ACP.

Very large polyps may descend into the oropharynx and cause a foreign body sensation. As previously noted, the presentation of bilateral nasal obstruction is possible due to expansion of the polyp from one choanae to the other, however true bilateral ACP is extremely rare [15].

#### **5. Imaging**

Computed tomography (CT) imaging with nasal endoscopy represent the gold standard in the diagnosis of ACP [5]. All patients must have preoperative sinonasal CT scan, as it is a crucial part of the diagnosis and provides critical information of nasal and sinus bony landmarks prior to surgical intervention.

The classic appearance of ACP in CT is a hypo-attenuating unilateral soft tissue mass that completely occupies the maxillary sinus. It extends through the accessory maxillary ostium into the nasal cavity, medially to the inferior turbinate with progression towards the nasopharynx (**Figure 1**). Less commonly, the polyp

#### **Figure 1.**

*Computed tomography (CT) imaging of right-sided antrochoanal polyp (ACP). (A) coronal image showing total opacification of the right maxillary sinus and nasal cavity. The antrochoanal polyp has both an intramaxillary component (black asterisk) and an intranasal component (black arrow) this view also demonstrates the enlarged accessory maxillary ostium (white arrow) through which the intramaxillary and intranasal portions are connected via a thin stalk. (B) Coronal view of choanal component of the polyp (white asterisk) obstructing the nasopharynx on the ipsilateral side. (C) and (D) axial and sagittal views demonstrating the different components of the antrochoanal polyp intra-maxillary (black asterisk), intranasal (black arrow) and choanal / nasopharyngeal (white asterisk) potions.*

*Antrochoanal Polyp: Updated Clinical Approach, Histology Characteristics, Diagnosis… DOI: http://dx.doi.org/10.5772/intechopen.96329*

extends anteriorly to the middle turbinate and the anterior inferior turbinate region [16]. Bony changes (bone erosion, destruction or sclerosis) are not typically seen with ACP, although widening of the accessory maxillary ostium may occur, usually due to enlarging cystic portion of the polyp leading to the appearance of expansile maxillary mass (**Figure 1**) [8]. In cases of suspected bone destruction in CT, other pathologies such as malignancy should be considered. However, studies have shown that thinning of alveolar bone in the maxillary sinus may occur secondary to the progressive growing of ACP [2]. Lee classified 3 stages of ACP based on the radiological appearance of the lesion on CT [3, 17]: Stage I (antronasal polyp without extension to the nasopharynx), Stage II (full occlusion of the maxillary sinus ostium with extension to the nasopharynx) and Stage III (partially occlusion of the maxillary sinus ostium with polyp extension to the nasopharynx). In children, advanced CT stages (stage II, III) are more commonly seen due to delayed diagnosis in this population, as previously noted [7]. Magnetic resonance imaging (MRI) shows a hypointense T1 and enhanced T2 signals. With gadolinium administration, the cystic part of the polyp is peripherally enhanced. Although CT is the preferred imaging modality in the diagnosis of any nasal or sinus pathology including ACP, MRI may be considered in children (due to the lack of radiation exposure) and in cases of total unilateral nasal and sinus opacification in CT scans (in order to distinguish between sinus secretions and the mass itself). In nasal endoscopy, ACP appears as a gray-white colored mass with a smooth round surface. Unlike other allergic or inflammatory nasal polyps, ACP has a unique course from the maxillary sinus to the choana and has a bulging expansile behavior due to its cystic component.

#### **6. Histology/histopathology**

Macroscopically, ACP is composed of a cystic part filling the maxillary sinus and a solid part emerging through the maxillary ostia and filling the nasal cavity. It has a gross appearance of a "dumbbell" shape with a narrow stalk connecting between the cystic and solid components (**Figure 2**). Microscopically, the antral (or intramaxillary portion) part of ACP demonstrates a central cystic cavity surrounded by a homogeneous edematous stroma with few cells [5]. The intranasal portion of the polyp is covered with a respiratory epithelium similar to the normal mucosa of the sino-nasal tract and the choanal portion occasionally shows squamous metaplasia and reactive fibrosis (**Figure 3**). In comparison to allergic polyp, ACP is characterized by higher inflammatory cell infiltration and edema, lower eosinophilic infiltration and less submucosal glands [18]. These findings indicate that inflammatory changes are the main pathophysiological processes in the pathogenesis of ACP while allergy plays only a minor role. In addition, the paucity of submucosal glands suggests that ACP results from edematous hypertrophy of the respiratory epithelium rather than from distention of the glandular structure, which is the event responsible for the development of ordinary nasal polyps [18]. Angiogenesis is significantly less evident in ACP compared to nasal polyps resulting from chronic rhinosinusitis, with lower expression of angiogenic markers vasculo-endothelial growth factor (VEGF) and CD-34 [12]. These findings further support the idea that ACP is a result of a local inflammatory process and could also explain why ACP has less tendency to bleed compared with other types of polyps, both as a presenting symptom or during endoscopic surgery. ACP is characterized with a significantly high prevalence of intramural cysts [19, 20]. It is speculated that these cysts may have a role in the pathogenesis of ACP, and they contribute to the gross cystic appearance of both its intramaxillary and intranasal components. Moreover, the presence of intramural cysts supports

#### **Figure 2.**

*Combined radiologic and intraoperative views of a left-sided Antrochoanal polyp. (A) & (B). Coronal and axial images showing total opacification of the left maxillary sinus and nasal cavity. The intra-maxillary portion (black asterisk) and the intranasal portion (black arrow) are connected through the enlarged accessory maxillary ostium (white arrow). (C). Endoscopic view of the same patient: The intranasal component of the polyp (black arrow) is medialized with a sinus-seeker (white cross) exposing the stalk (white arrow) that connects it to the intramaxillary component (black asterisk). (D). Gross appearance of the antrochoanal polyp after resection. The intranasal (black arrow) and the choanal (white asterisk) portions are seen clearly, the stalk preserved (white arrow) is seen after separating it from the intra-maxillary portion. MT = middle turbinate. S = nasal septum.*

#### **Figure 3.**

*Typical histologic characteristics of ACP. Image (A) shows a cystic portion of ACP with cuboidal epithelium (H&E original magnification X200). Image (B) demonstrate the intranasal portion of the ACP, edema is seen (H&E X100). Images (C) & (D) demonstrate squamous metaplasia of choanal portion of the ACP (C- H&E X200, (D)- monoclonal P63 antibody stain x200).*

*Antrochoanal Polyp: Updated Clinical Approach, Histology Characteristics, Diagnosis… DOI: http://dx.doi.org/10.5772/intechopen.96329*

Berg's theory [10, 20] that the cystic part of the polyp develops from obstruction in the acinar glands or lymphatic ducts secondary to persistent inflammation. The pressure generated in the process of the polyp's growth through the accessory sinus ostium may be the cause for the substantial edema that is seen.

An explanation of why ACP presents with more cystic changes than diffuse chronic rhinosinusitis with nasal polyps (d-CRS) may be related to their different origins. ACPs develop from the maxillary sinus, characterized by typical respiratory epithelium with thin lamina propria, cyst formation and fewer submucosal glands. On the contrary, nasal polyps in d-CRS typically originate from the ethmoid sinus, which has a thick submucosal layer [21].

When comparing ACP with d-CRS preparations, Warman et al. found that ACP exhibits typical histologic features like cyst formation and edema. ACP demonstrated significantly increased edema when compared to the d-CRS (82.5% vs. 44.4% respectively, p < 0.001), and higher cyst formation (40% vs. 6.2% P = 0.02). More over ACP preparations demonstrate lower degrees of inflammatory markers than d-CRS [22]. The lack of an inflammatory drive in the pathogenesis of ACP may explain why anti-inflammatory treatment is futile in this population, leading to the common notion that ACP is a rather surgical issue than a medical one.

#### **7. Treatment**

Surgery is the standard of care in the treatment for ACP. Since its first description by Killian in 1906, many surgical techniques have been proposed for exposing the maxillary region [4]. Successful ACP resection depends on complete removal of the intramaxillary component of the polyp. The ideal procedure should facilitate excellent approach to all maxillary sinus walls and yet be minimally invasive as possible, especially in children. Currently, various surgical approaches are available: endoscopic sinus surgery (ESS) with polyp removal via either inferior meatus or middle meatus, or a combined inferior and middle meatal naso-antral window. Other options such as ESS with adjuvant canine fossa puncture, or ESS with "mini Caldwel-Luc" procedure aim to facilitate visualization of the anterior and inferior walls of the maxillary sinus [4, 23, 24].

#### **8. Endoscopic inferior meatal antrostomy (EIMA)**

Described by Mikulicz in 1887, inferior meatal antrostomy (known as intranasal antrostomy) was a common surgical procedure in the management of maxillary sinus disease. However, the popularity of this technique has declined with the increased use of middle meatal antrostomy due to the growing recognition that an opening in the inferior meatus does not improve sinus drainage, and might even harm the maxillary sinus mucociliary clearance mechanism. Nevertheless, endoscopic approach via inferior meatal antrostomy has the advantage of inferior meatal naso-antral window that avoids violation of the ostiomeatal complex (OMC) and provides better access to anterior-inferior maxillary sinus lesions. Arguments against inferior meatal antrostomy include: persistent sinus disease following surgery, low patency rates, possible injury to the nasolacrimal duct or to developing canine teeth, and technical difficulties associated with the procedure [24, 25]. While these arguments were substantial using anterior rhinoscopy approach, they are not valid with endoscopic approach in EIMA. As the inferior turbinate is carefully medialized, the opening of the nasolacrimal duct (Hasner's valve) is clearly seen and preserved. Then, the maxillary wall is penetrated posterior to that point, and an antrostomy of 8–10 mm is created. Once a

satisfactory exposure is achieved, view of the posterior, lateral and anterior portions of the sinus walls is possible with 0- and 45-degree endoscope in respect. The lesion is then removed with straight and curved instruments. At the end of the procedure, the inferior turbinate is lateralized back to its original position [24, 25].

Landsberg and Warman reported 56 patients with multiple maxillary pathologies (45% of them with ACP) in which EIMA was the primary approach for revision surgery. In a follow-up period for at least a year, 93% of patients had no evident sinus disease recurrence. There were no cases of ACP recurrence, and recirculation was not observed during the follow-up period. In addition, no major complications such as nasolacrimal duct injury or bleeding were observed [24].

#### **9. Endoscopic middle meatal antrostomy (EMMA)**

Endoscopic sinus antrostomy via the middle meatus (EMMA) is currently considered the gold standard treatment for ACP resection. It is generally recommended that the antral portion should be completely removed together with its stalk to minimize polyp regrowth. As a result, the intranasal and choanal components of the polyp should be resected first (**Figure 2**). Occasionally when the choanal portion is too large, it is easier to push it back to the oropharynx and remove it trans-orally.

Next, the cystic part of the polyp is resected through maxillary antrostomy. The maxillary sinus natural ostium is identified and usually connected with the already enlarged accessory ostium. Resecting the intramaxillary portion includes −45°- 70°- endoscopes to better visualize and identify the origin of the polyp. Removal of this intramaxillary portion is extremely important as to minimize post-operative recurrence [4, 26, 27].

Recurrence rate after EMMA is low. Cook et al. observed no recurrence in 33 patients with ACP [28]. Sometimes the intramaxillary portion is tightly adherent to the anterior or antero-inferior walls of the sinus, which makes the dissection a challenging task. In these cases, usage of angled instrumentation is strongly recommended. Nevertheless, the recurrence rate in these cases may increase up to 20% [17, 24, 26, 27].

Ozer et al. reviewed 42 patients who underwent ESS for ACP removal. Transcanine sinoscopy and Caldwell Luc approach were used in addition in 14 and 13 patients respectively. They found recurrence in 3/15 patients after ESS alone (20%), yet there was no recurrence after combined ESS and transcanine sinoscopy or the Caldwell Luc approach [29]. They postulated that the relative high recurrence rate may be due to improper identification of the attachment site of the polyp inside the maxillary sinus (50% of all cases). As a result, they advised considering combined approaches in cases when the attachment site is not clearly recognized. Hong et al. recommended powered instrumentation (Hummer, Stryker Instruments, Kalamazoo, MI) during ESS as an effective technique for removing ACP, especially the antral portion. They found an improvement rate of 96.4% with no significant complications when powered instrumentation was used [29, 30]. Complications following ACP resection are rare.

#### **10. Combining endoscopic middle meatal antrostomy and transcanine sinusocopy**

Lee and Huang used the transnasal endoscopic approach for ACPs originated from the inferior and posterior walls of the maxillary sinus, saving the more invasive combined endoscopic and transcanine approach for polyps originated from the lateral wall or in revision surgery. They reported success rate of the transnasal

*Antrochoanal Polyp: Updated Clinical Approach, Histology Characteristics, Diagnosis… DOI: http://dx.doi.org/10.5772/intechopen.96329*

endoscopic approach and the combined endoscopic middle meatal and transcanine approach as 76.9% and 100%, respectively [31].

As mentioned earlier, Ozer et al. found no recurrence after combined ESS and transcanine sinoscopy approach [29].

Transcanine exposure has some complications such as facial swelling pain and rarely injury to the infraorbital nerve. Although rare these complications yet are against using transcanine procedure in ACP resection, especially if the polyp is approachable via EIMA.

#### **11. Combination of ESS and "mini Caldwell-Luc" approach**

Kelles et al. retrospectively reviewed 46 patients treated for ACP during a 7-year period. 20 patients underwent endoscopic endonasal surgery (ESS) with mini-Caldwell operation (performing a canine fossa window of 0.5–0.6 cm), while 26 patients underwent ESS alone. The only statistically significant difference between the groups was the recurrence rate, which was higher in the ESS group compared with ESS plus mini-Caldwell group (P < 0.05).

In the ESS group, bleeding, synechia, and ostium stenosis were more evident than in the ESS plus mini-Caldwell group, but these differences were not statistically significant. Therefore, Kelles theorized that adding the mini Caldewell-Luc approach allowed better visualization of the maxillary sinus walls and subsequently easier resection of the remnant polyp [23].

Atighechi et al. used a mini-Caldwell approach with ESS in their patients. They reported minimal recurrence and low complication rates, deciding that the technique is useful for the completely removal of ACP [32].

The traditional Caldwell-Luc approach offers good exposure and ensures complete removal of the polyp with the associated antral mucosa. Nevertheless, this approach has been largely abandoned in the treatment of maxillary sinus pathologies, because it does not address the natural ostium of the maxillary hence considered non-functional. Complications include: cheek anesthesia, sensory deficits, cheek swelling and risks for normal teeth development in children [4, 23, 29, 33, 34].

#### **12. Special consideration in ACP resection; ESS in children**

As previously noted, the incidence of ACP is higher in children and young adults. Although no difference in the pathophysiology or histology were seen between children and adults, children are at higher risk for recurrence. It is reasonable to believe that the anatomically narrow sinuses, the not-yet erupted teeth, and concern of maxillary growth may affect the surgeon's decision regarding the surgical approach, leading to higher failure rate [17, 31, 35].

In his review of 200 patients with ACP, Forsini described recurrence in 4 patients (2%) all of which were children <7 years of age, in whom only polypectomy was performed. Eventually, in all cases of recurrence ESS was performed without evidence of recurrence [4].

#### **13. Recurrence and follow up**

As evident by various published series, recurrence rates range from 0% reported by Tsukidate to 64% reported by Saito and collaborators. Recurrence rates vary

between different surgical approaches, patient's age and other factors such as accompanying sinus pathologies [36, 37]. This raises the question – how long should we follow patients ACP resection?

Lee and Huang determined that 65% of their pediatric patients with ACPs had associated chronic sinusitis. Similarly, some authors have also identified association of ACPs with allergic disease. The main hypothesis is the challenge of removing the entire sick mucosa with the origin of the polyp once there is chronic inflammation [31]. Natasha Choudhury reported 29 patients after EMMA surgery for ACP. They described no polyp recurrence, with a mean follow-up period of 14.7 months [8]. Galluzzi reviewed 13 studies and found that recurrence in children is higher than in adults, mostly because of reasons described earlier. The review showed that combined approach had the lowest recurrence rate, with a range of follow-up between 6 to 120 months. Most recurrences were noted between 5 months to 3 years after initial surgery [17]. Some authors claim that different anatomic variations in the nasal cavity such as septal deviation, conchal hypertrophy, and concha bullosa may increase the intramaxillary pressure, hence predisposing for the development of ACP. While these variations were documented in up to 80% of patients with ACP, none of them were linked to increased rates of recurrence [4, 17, 23, 24, 30]. In most relevant studies, the time of recurrence was 1.2 ± 0.6 years. Therefore, it is advised to monitor ACP patients for at least 2 years after surgery in order to detect 95% of recurrent cases [35].

#### **14. Conclusion**

ACP originates in the maxillary sinus of children and young adults. Its etiology is speculative, currently considered a benign cystic polyp with limited inflammatory characteristics. It has a consistent three component structure intramaxillary, intranasal and choanal portions. ACP has a typical imaging characteristic and the gold standard of treatment is complete surgical resection. Special attention should be given to identify and resect the intramaxillary portion to prevent recurrence. Long term follow-up is needed to rule out polyp regrowth.

#### **Author details**

Warman Meir1,3\*, Rona Bourla1 , Monica Huszar2 and Elchanan Zloczower1

1 Department of Otorhinolaryngology, Head and Neck Surgery, Kaplan Medical Center, Rehovot, Israel

2 Department of Pathology, Kaplan Medical Center, Rehovot, Israel

3 Hebrew University, Hadassah Medical School, Jerusalem, Israel

\*Address all correspondence to: meirwarma@gmail.com

© 2021 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

*Antrochoanal Polyp: Updated Clinical Approach, Histology Characteristics, Diagnosis… DOI: http://dx.doi.org/10.5772/intechopen.96329*

#### **References**

[1] Lee TJ, Huang SF, et al. Endoscopic sinus surgery for antrochoanal polyps in children. *Otolaryngol Head Neck Surg.* 2006; 135(5):688-692.

[2] Bidkar VG, Sajjanar AB, et al. Role of computed tomography findings in the quest of understanding origin of antrochoanal polyp. *Indian J Otolaryngol. Head Neck Surg*. 2019; 71(3):1800-1804.

[3] Lee DH, Yoon TM, et al. Difference of antrochoanal polyp between children and adults. *Int J Pediatr Otolaryngolol.* 2016; 84:143-146.

[4] Frosini P, Picarella G, et al. Antrochoanal polyp: analysis of 200 cases. *Acta Otolaryngology Ital*. 2009; 29 (1):21-26.

[5] Yaman H, Yilmaz S, et al. Evaluation and management of antrochoanal polyps. *Clin Exp Otorhinolaryngol* 2010; 3: 110-114.

[6] Larsen K, Tos M, et al. The estimated incidence of symptomatic nasal polyps. *Acta Otolaryngol*. 2002; 122: 179-182.

[7] Balikci HH, Ozkul MH, et al. Antrochoanal polyposis: analysis of 34 cases. *Eur Arch Otolaryngology.* 2013; 270 (5):1651-1654.

[8] Choudhury N. Hariri A, et al. Endoscopic management of antrochoanal polyp: a single UK centre's experience. *Eur Arch Otorhinolaryngol.* 2015; 272(9) 2305-2311.

[9] Mantilla E, Villamor P, et al. Combined approach for paediatric recurrent antrochoanal polyp: a singlecentre case series of 27 children. *J Laryngo Otol.* 2019; 133(7): 627-631.

[10] Min YG, Chung J, et al. Histologic Structure of Antrochoanal Polyps, *Acta Oto-Laryngologica.* 1995; 115:4, 543-547.

[11] Gendeh BS, Long YT, et al. Antrochoanal Polyps: clinical presentation and the role of powered endoscopic polypectomy. *Asian J Surg*. 2004; 27(1):22-25.

[12] Hirshoren N, Neuman T, et al. Angiogenesis in chronic rhinosinusitis with nasal polyps and antrochoanal polyps. *Inflamm Res.* 2011; 60 (4):321-327.

[13] Cohen O, Betito HR, et al. Development of the nasopharynx: A radiological study of children. *Clin Anat.* 2020 Oct;33(7):1019-1024.

[14] Flint P.W (2020) Cummings Otolaryngology - Head and Neck Surgery (7th Ed.) chapter 200; Pediatric Chronic Rhinosinusitis 2970-2978. Elsevier

[15] Oner F, Sakat M, et al. Bilateral Antrochoanal Polyp. *J Craniofacial Surg*. 2015; 26 (7): 661-662.

[16] Peric A, Vucadinovic T, et al. Choanal polyps in children and adults: 10-year experience from a tertiary hospital. *Eur Arch Otorhinolaryngol.* 2019; 276(1) 107-113.

[17] Galluzzi F, Pignataro L, et al. Recurrences of surgery for antrochoanal polyps in children: A systematic review. *Int J Pediatr Otolaryngol.* 2018; 106:26-30

[18] Stammberger H, Hawke M. Essentials of functional endoscopic sinus surgery. St Louis: Mosby 1993. p. 103-105.

[19] Maldonado M, Martínez A, et al. The antrochoanal polyp. *Rhinology*. 2004;42(4):178—182.

[20] Berg O, Carenfelt C, et al. Origin of the choanal polyp. *Arch Otolaryngol Head Neck Surg* 1988; 114:1270-1271.

[21] Latta JS, Schall RF. LXXVIII The Histology of the Epithelium of the Paranasal sinuses under Various Conditions. *Ann Otol Rhinol Laryngol*. 1934;43(4):945-971. Doi:10.1177/000348943434300401

[22] Warman M, Kamar Matias A, et al. Inflammatory Profile of Antrochoanal polyps in the Caucasian Population – A Histologic Study. *Am J Rhinol Allergy* 2020 Accepted for publication*.*

[23] Kelles M, Toplu Y, et al. Antrochoanal polyp: clinical presentation and retrospective comparison of endoscopic sinus surgery and endoscopic sinus surgery plus mini-Caldwell surgical procedures. *J Craniofac Surg*. 2014;25(5):1779-1781

[24] Landsberg R, Warman M, et al. The Rationale for Endoscopic Inferior Meatal Antrostomy. *ORL* 2019; 81:41-47.

[25] Yanagisawa E, Christmas DA Jr, et al. Endoscopic view of a long-term inferior meatal antrostomy. *Ear Nose Throat J*. 2007;86(6):318-319.

[26] Pagella F, Emanuelli E, et al. Clinical features and management of antrochoanal polyps in children: Cues from a clinical series of 58 patients. *Int J Pediatr Otorhinolaryngol.* 2018; 114:87-91.

[27] Stammberger H, Posawetz W, et al. Functional endoscopic sinus surgery. Concept, indications and results of the Messerklinger technique. *Eur Arch Otorhinolaryngol*. 1990;247(2):63-76.

[28] Cook PR, Davis WE, et al. Antrochoanal polyposis: a review of 33 cases. *Ear Nose Throat J*. 72:401- 410, 1993.

[29] Ozer F, Ozer C, et al. Surgical approaches for antrochoanal polyp: a comparative analysis. *B-ENT*. 2008;4(2):93-99.

[30] Hong SK, Min YG, et al. Endoscopic removal of the antral portion of

antrochoanal polyp by powered instrumentation. *Laryngoscope*. 2001;111(10):1774-1778.

[31] Ta-Jen Lee, Shiang-Fu Huang, et al. Endoscopic sinus surgery for antrochoanal polyps in children. Otolaryngol Head Neck Surg 2006; 135:688-692.

[32] Atighechi S, Baradaranfar MH, et al. Antrochoanal polyp: a comparative study of endoscopic endonasal surgery alone and endoscopic endonasal plus mini-Caldwell technique. Eur Arch Otorhinolaryngol 2009; 266:1245-1248

[33] Paul W. Flint*. Cummings Otolaryngology--Head & Neck Surgery*. 5 th ed. Elsevier Saunders, 2014. ISBN: 9781455746965

[34] Datta RK, Viswanatha B, et al. Caldwell Luc Surgery: Revisited. *Indian J Otolaryngol Head Neck Surg*. 2016;68(1):90-93.

[35] Chaiyasate S, Roongrotwattanasiri. K, et al. Antrochoanal Polyps: How Long Should Follow-Up Be after Surgery? *Int J Otolaryngol*. 2015; 297417.

[36] Tsukidate T, Haruna S, et al. Long term evaluation after endoscopic sinus surgery for chronic pediatric sinusitis with polyps. *Auris Nasus Larynx* 2012;39(6)583-7.

[37] Saito H, Honda N, et al. Intractable pediatric chronic sinusitis with antrochoanal polyp, *Int. J Pediatr Otorhinolaryngol.* 2000 31;54(2-3):11-6.

#### **Chapter 2**

## Tracking of Fascicles of Cutaneous Nerves of Thigh: A Histological Study

*Rajani Singh*

#### **Abstract**

Present study uncovers the secrets of internal morphology of femoral nerve branches namely, cutaneous trunk, subcutaneous trunks, saphenous, medial cutaneous and intermedius cutaneous nerves innervating the skin of anteromedial thigh at fascicular level. Therefore, the aim of the study is to track, correlate, interpret and identify the pathways of fascicles through histological slides. The femoral nerve and its branching points were calibrated in distances from inguinal ligament. These trunks and nerves of a cadaver were processed for histological slides staining with haematoxylin and eosin. The fascicles in the histological slides were identified, tracked, correlated and interpreted from cranial most slide to the last terminal slides of these nerves and trunks. The correlation of the pathways of fascicles revealed that these fascicles are continuous, consistent and traceable interrupted by split, fusion and multiplexing. Femoral nerve branches/fascicles/nerve fibres if damaged, impair the sensation of corresponding area of skin of anteromedial thigh creating helm of neurological complications. Hence the injured fascicles can be repaired with the help of identification and correlation of fascicular pathways carried out in this study with least invasion. The findings of present study will be of paramount importance for intraoperative stimulation to diagnose and identify the fascicle for microneurosurgical repair/graft/regenerate/neurotisation in the cutaneous branches of femoral nerve at fascicular level.

**Keywords:** cutaneous nerves, microanatomy, transformational process, pathways of fascicles

#### **1. Introduction**

There are many variations of femoral nerve and these have been classified by Singh et al. [1]. A femoral nerve cropped from a cadaver was type II of classification of Singh et al. [1]. This femoral nerve bifurcated into muscular and cutaneous trunks at one centimetre below the inguinal ligament. The cutaneous trunk further splits into sub-cutaneous trunk of thigh and the saphenous (S) nerve. The subcutaneous trunk, then, bifurcates into intermedius cutaneous (IC) and medial cutaneous (MC) nerves of the thigh. The group of afferent fascicles of S, IC and MC supply skin of anteromedial thigh.

Though the injuries to the IC and MC nerves have hardly been reported yet few cases of pain and paraesthesia over the anterior and medial aspects of thigh, as a

result of engagement of IC and MC nerves of the thigh, are described. However, sensory loss on the medial side of the thigh, leg and foot up to the ball of the great toe because of engagement of the saphenous nerve through iatrogenic lapses or otherwise are well reported. The outcome of injuries may not be fatal or produce unbearably serious signs and symptoms so the patients may not be opting for costly neurosurgical diagnosis (MRI) for detection of location and degree of injury and procedures.

The neuro-therapy of neuropathological morbidity requires accurate diagnosis and treatment. There is very limited scope of investigating location and identification of injured fascicle or nerve fibres under current knowledge of internal morphology of nerve. Though Chhabra et al. claims that location and degree of injury can be identified through MR advanced neurography [2] yet it has its own limitations regarding resolution and image defects. Therefore, a micro-anatomic study has been planned to track fascicles in histological slides of subcutaneous trunk, S, IC and MC for improving identification of injured fascicle, its location and degree of injury for diagnosis and imagery interpretation together with non-invasive neurosurgical repair, grafting and regeneration of injured nerve fibres.

#### **2. Tracking and correlation of fascicles**

A24 slide was prepared from the femoral nerve just above the inguinal ligament. This femoral nerve just below the inguinal ligament bifurcated into muscular and cutaneous trunks. Cutaneous trunk then bifurcated into saphenous nerve and subcutaneous trunk which further divided into intermedius and medial cutaneous nerves. Saphenous, intermedius and medial cutaneous nerves innervate skin of anteromedial thigh. Histological slides of cutaneous, saphenous nerve, subcutaneous trunk, medial and intermediate cutaneous nerves were prepared and stained with haematoxylin and eosin. The fascicles in these nerves/trunks were identified, tracked, correlated and interpreted.

#### **3. Naming scheme of fascicles**

For deciphering the fascicles of individual nerves and to avoid confusing duplicate numbers, the name of the fascicles at the point of transformational processes was changed in sequential order with prefix from CF of composite fascicles in femoral nerve to CCF in cutaneous trunk and SCCFs in subcutaneous trunk extending it to SCF in S nerve, MCCFs and ICCFs in MC and IC respectively.

The composite fascicles (CFs) in slide A24 1 have been identified as CFs {(303, 304); (280, 257, 270, 312, 313): (316, 317, 318)}\* of both cutaneous and muscular trunks. CFs 316, 317, 318 corresponds muscular trunk and CFs {(303, 304); (280, 257, 270, 312, 313)} to cutaneous trunk (**Figure 1**).

#### **4. Tracking and correlation of fascicles in cutaneous trunk**

The cutaneous trunk was cut into 6 pieces and six blocks C1, C2, C3, C4, C5 and C6 were prepared. From C1, 13 slides were processed. C1 13 was the cranial most slide of C1. Similarly variable number of slides were prepared from C2, C3, C4, C5 and C6.

In C24 1, composite fascicles, {(303, 304); (280, 257, 270, 312, 313)} correspond to cutaneous trunk, but CF313 of A24 1 splits and formed CCF 318 and 319

#### **Figure 1.** *Correlation between A24 1-C1 13. MT- muscular trunk, CT- cutaneous trunk.*

in C1 13. So C1 13 consists of {(303, 304); (280, 257, 270, 312, 318, 319)} (**Figure 1**). These fascicles continuously consistent up to slide C2 16, the cranial most slide of C2 block.

The fascicles in C2 16 were continuous, consistent and correctable up to slide C2 12. In slide C2 12, CF312 and CCF318 fused forming CCF320 and CF280 in slide C2 12 split into CCF321, 322 and 323 in C2 11 (**Figure 2**). C2 11 is traceable to C2 1 meaning that the fascicles which were present in slide C2 11 were continuing up to slide C2 1.

In C2 1 slide CCF321, 322 and 257 fused forming CCF324 in C3 3 (**Figure 3**). So the slide C3 3 consists of CFs 303, 304, CCFs 319, 320, 323, 324 and 270 fascicles. CCF223 changed the shape forming CCF 323. Fascicles of C3 3 were traceable, continuous, and consistent up to slide C3 1.

CCF270 and CCF324 in C3 1 fused forming CCF325 in C4 4 (**Figure 4**). So slide C4 4 consists of CFs 303, 304, CCFs 320, 323, 325 and 319 fascicles. Fascicles of C4 4 were traceable, continuous, and consistent up to slide C4 1.

CCF325 in C4 1 split into CCF326 and 327 in C5 5 (**Figure 5**). So slide C5 5 consists of CFs 303, 304, CCFs 319, 320, 323, 326 and 327 fascicles. C5 5 is traceable, continuous, and consistent up to C6 1.

In C6 1 fascicles CFs 304 and 303 and CCFs 319, 320, 323, 326 and 327 were found to be surrounded by internal epineurium (**Figure 6**) indicating subcutaneous trunk and saphenous nerve are formed and ready to emerge.

**Figure 2.** *Correlation between C2 12-C2 11.*

After C6 1, cutaneous trunk bifurcated into a) S nerve containing CCFs 319, 320, 323, 326 and 327 as observed in S1 1 of S1 block and b) sub-cutaneous trunk consisting of CFs 304, 305 and 306 in SCT1 1, slide of SCT1 block (**Figure 7**).

#### **5. Tracking and correlation of fascicles in saphenous nerve**

S nerve was cut into 6 pieces and six blocks S1, S2, S3, S4, S5 and S6 were prepared. Slides prepared from these six blocks were stained with haematoxylin and eosin and fascicles of S nerve were correlated starting from S1 block to S6 block.

The S nerve having CCFs 319, 320, 323, 326 and 327 in S1 1 emerged out after C6 1 (**Figure 7**). These CCFs were traceable from the slides of block S1 through S2 1 the caudal most slide of S2 block.

The CCFs 319 and 320 in S2 1 fused together forming SCF 328 in S3 13 the cranial most slide of S3 block (**Figure 8**). So, S3 13 slide consists of 323, 326, 327 and 328 fascicles. These fascicles were continuous, consistent and traceable up to slide S3 1.

**Figure 3.** *Correlation of C2 1 with C3 3.*

CCF328 in S3 1 slide split into SCFs 329 and 330 in S4 12, the cranial most slide of S4 block (**Figure 9**). So, the slide S4 12 consists of 323, 326, 327, 329 and 330 fascicles. The fascicles of S4 12 slide were continuous, consistent and traceable up to slide S4 1.

CCF330 emerged out as branch of S nerve after S4 1 as this fascicle was not found in next slide. CCFs323 in S4 1 split into SCFs331 and 332 S5 12, CCFs326 in S4 1 into SCF333 and 334 in S5 12, CCF329 in S4 1 into SCF335 and 336 in S5 12. So slide S5 12 consists of 331, 332, 333, 334, 335, 336 and 327 (**Figure 10**). These fascicles in S5 12 were continuous, consistent and traceable up to slide S5 1.

The fascicles from S5 1 reorganise their position through migration and rearrange in S6 11. The fascicles got separated laterally into infrapatellar branch having SCFs (331, 332, 335, 336) and main S nerve having SCFs (327,333, 334) in S6 11 (**Figure 11**) continued caudally.

The fascicles are traceable between S6 11 and S6 7. Again reorganisation of fascicles is taking place from S6 7 to S6 6 (**Figure 12**). These fascicles in S6 6 were continuous, consistent and traceable up to slide S6 1. Infrapatellar branch emanated laterally from S nerve after S6 1.

**Figure 4.** *Correlation of C3 1 and C4 4.*

**Figure 5.** *Correlation of C4 1 and C5 5.*

*Correlation of C5 1 and C6 1. SCT- fascicles of subcutaneous trunk, S- fascicles of saphenous nerve.*

**Figure 7.** *Correlation of C6 1 and S1 1 and SCT1 1.*

**Figure 8.** *Correlation of S2 1 and S3 13.*

**Figure 9.** *Correlation of S3 1 and S4 12.*

**Figure 10.** *Correlation of S4 1 and S5 12.*

**Figure 11.** *Correlation of S5 1 and S6 11.*

**Figure 12.** *Correlation of S6 7 and S6 6. IP- infrapatellar branch, S- main saphenous nerve.*

### **6. Tracking and correlation of fascicles in subcutaneous trunk**

Subcutaneous trunk was cut into 3 pieces and 3 blocks SCT1, SCT2 and SCT3 were prepared. Subcutaneous trunk bifurcated into MC and IC. One block of

**Figure 13.** *Correlation of C6 1 with SCT1. SCT- subcutaneous trunk, S- saphenous nerve.*

#### **Figure 14.** *Correlation of SCT3 1 with IC1 1and MC1 3.*

**Figure 15.** *Correlation between IC1 1 and IC2 6.*

MC,MC1 and two blocks of IC, IC1 and IC2 were prepared and slides from aforementioned blocks were stained with haematoxylin and eosin and fascicles were correlated as elaborated below:

The CFs 303 and 304 constituted subcutaneous trunk which was separated laterally from cutaneous trunk after C6 1 (**Figure 6**). Then CF303 split into SCCF 305 and 306 in SCT1 1 slide (**Figure 13**). So the subcutaneous trunk now consists of fascicle CFs 304, SCCFs 305 and 306. After the slide SCT1 1, the CFs 304, SCCFs 305 and 306 were traceable from SCT1 1 up to SCT3 1.

The subcutaneous trunk bifurcated into IC and MC nerves after SCT3 1 slide. The CF304 constitute MC and 305 and 306 IC (**Figure 14**). Then CF304 split into MCCF307, 308, 309, 310 in MC1 3 slide. The SCCF305 split into ICCF307, 308, 309 and SCCF306 into ICCF310, 311 in IC1 1 respectively (**Figure 14**).

ICCF308 IC1 1 split into ICCF314 and 315, ICCF309 IC1 1 split into ICCF312 and 313. ICCF310 split into ICCF316 and 317, ICCF311 split into ICCF318 and 319 (**Figure 15**). These fascicles form fine nervelets innervating skin of thigh.

#### **7. Clinical significance of this study**

No such study has been carried out involving sensory nerves of femoral nerve. If the fascicles of S, MC and IC are damaged, the communication of sensory information from the innervated area will be interrupted leading to aggravation of clinical problems. So it is not merely nerve where injury should be investigated rather injured fascicles should be targeted for diagnosis for complications which may occur anywhere in entire fascicular path from origin to point of innervation. The diagnosis of neural insults requires not only the location and degree of injury but also identification, isolation, orientation, directivity, and matching of shape and size of injured nerve CFs for planning surgical repair, grafting and regeneration [3].

The location and degree of injury is investigated by the high resolution MRI advanced neurography [2, 4]. But this has its own limitations of recording and interpretation. This generates uncertainty in diagnosis and thereby in treatment. Thus the radiologists and neurosurgeons face the impediments of pinpointing the probable position of injury and identification of fascicles. Therefore, the imagery coupled with our internal morphological study together can refine the interpretation for identification of injured CF and location of injury. Methodically, it can be done by one to one correlation between images of transverse histological and high resolution MRI advanced neurographic sections at the same position from inguinal ligament. The distance of location of injured fascicle from inguinal ligament may be computed in MRI neurography and then the calibrated histological sections of cutaneous, subcutaneous trunks, S, MC and IC nerves at the same level may be compared and examined for confirmation of identified injured fascicle. After identification of injured fascicle, the idea of shape, size, location and orientation can also be derived from histological slides for matching, alignment and directivity of nerve fibres for repair and grafting.

#### **7.1 Personal communication**

The neurosurgery at fascicular level is currently uncommon however, with upcoming science and technology in future, present study will be highly useful for neurosurgeons. The study will help in carrying out less invasive surgery as stimulation of identified injured fascicles will not involve other fascicles which in case of nerve stimulation may be stimulated causing discomfort to the patient.

### **8. Conclusion**

The histological slides of cutaneous, subcutaneous trunks, S, MC and IC nerves brought out correlation of fascicles and grouped fascicles together with their configuration data present the straight, continuous and identified pathways of CFs interrupted by transformational processes calibrated in distance from inguinal ligament. This data will be of utmost importance to imagery guided microneurosurgical interventions more precisely at fascicular level together with the imagery interpretation to radiologists and neurosurgeons to assess injury and its location in an identified fascicular pathway to plan for its repair and surgical access. Fascicular electrode may be designed and developed like nerve cuff electrode [5] to improve neural microsurgery tremendously at fascicular level.

### **9. Limitation**

This study involves variations in the branching pattern of sensory nerves in one individual. Further studies are recommended to encampass variations in other individuals**.**

### **Author details**

Rajani Singh Department of Anatomy, UP University of Medical Sciences, Saifai Etawah, UP, India

\*Address all correspondence to: nani\_sahayal@rediffmail.com

© 2020 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

### **References**

[1] Singh R, Tubbs RS, Singla, M. Classification and fascicular analysis of variant branching pattern of femoral nerve for microsurgical intervention. A series of thirteen cadavers. Int J Morphol 2016; 34(2):561-569.

[2] Chhabra Avneesh, Lianxin Zhao, John A Carrino, Eo Trueblood, Saso Koceski, Filip Shteriev et al. 2013. MR Neurography: Advances. Radiology Research and Practice. Volume 2013, 14 pages Article ID 809568.

[3] Payne S Houston. Nerve Repair and Grafting in the Upper Extremity. J South Orthop Assoc 2001; 10(2). https://www. medscape.com/viewarticle/423216\_6

[4] Bäumer Philipp, Sabine Heiland, Martin Bendszus, Mirko Pham. MR Neurography - Diagnostic Criteria to Determine Lesions of Peripheral Nerves. Clinical Neurology 2012; page 10-14. Available at Magnetom Flash · 2/2012 www. siemens. com/magnetom-world.

[5] Chandra Naresh, Singh Rajani. Tracking of Fascicles of Sartorius and Pectineus Nerves-A key to Neurosurgery. Journal of Clinical and Diagnostic Research 2019; 13(1):AC01-AC08.

## Section 2 Molecular Pathology

#### **Chapter 3**

## Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic Markers in Breast Cancer

*Rodrigo Vismari de Oliveira*

### **Abstract**

In the last two decades, new discoveries concerning on breast cancer have contributed to important changes on its classification, from purely morphologic to molecular embased, to establish better correlation with clinicopathologic features. The classification in molecular subtypes, based on hormonal receptor and *HER-2* status, have been remarkable not only for its more accurated clinical correlations, but also for its easy applicability in diagnostic routine, better replication of tumor microenvironment through the selection of paraffinized tumor amounts and cost-effectiveness of the detection method, the immunohistochemistry. Hence, this classification may predict the breast cancer prognosis and became an important target for therapy with hormonal and *HER-2* antagonist drugs. Other study models, like cancer-stem cell hypothesis and immunological aspects of human cancer, have brought new emerging ideas regarding on molecular pathways and accurated prognostic preditions. Putative stem-cell markers and *PD-1/PDL-1*, have highlighted among several emerging molecular markers because of the bad cancer prognosis determinated by stem-cell markers expression and for emerging new drugs with selective action to PD-1/PDL-1, with promising results. The therapy of breast cancer have became diverse, target directed and personalized, in order to take in consideration the clinicopathologic cancer aspects, molecular tumor profile and clinical status of the patient.

**Keywords:** breast cancer, target therapy, molecular markers, prognostic markers, immunomarkers, cancer stem-cell

#### **1. Introduction**

Breast cancer is the one of the three most frequent human neoplastic disease worldwide and is the most common female cancer, remaining with considerable impact on general mortality. Worldwide, in the last 10 years, the incidence is growing up, with approximately 2.1 million of new cases per year and estimated mortality of 15%, at about 300.000 per year [1, 2].

Breast cancer remains as an heterogeneous group of disease from the point of view of biological behavior, therapeutic issues and prognostic features, determining different tracks of overall and free of disease survivals [3, 4]. Thus, the

clinicopathologic classification of breast cancer has been challenging over the last years, since the isolated simple morphologic classification of the tumor on histology examination is not necessarily related to the precise biologic behavior of the disease [5, 6].

In this way, especially over the last two decades, important researches revealing novel molecular markers expressed by cancer cells has been published in the literature. The new discoveries have improved the breast cancer classification, which has been progressed from a purely morphologic classification, based on histologic patterns, to a molecular classification, based on expression of oncoproteins and hormonal receptors, detected mainly by immunohistochemical techniques, in paraffin-embedded tumoral specimens [6, 7].

The novel molecular classification of breast cancer seems to exhibit more accurated correlation to the clinicopathological aspects of the tumor, as proliferative index, invasiveness and potential to metastatic spread. Furthermore, some of these molecular markers allowed the development of new drugs with specific actions on populations of cancer cells with specific genes alterations, improving considerably the therapeutic, prognostic and survival issues [7].

Instead of the recent advances on new therapeutic protocols under a new molecular perspective, early breast cancer on clinicopathological classification still remains the single one potencially curative [8]. The management of advanced clinicopathologic stage tumors and some established molecular groups of cancer, especially the 'triple negative' disease, remains with lacks of consensus. Anyway, the molecular markers have just improved the pathophysiology pathways knowledge, with potential future development of promising new drugs for target therapy of breast cancer [8–10].

#### **2. The molecular subtypes of breast cancer of clinicopathologic importance**

In the beginning of the 21st century, breast cancer was classified mainly on histologic basis. The WHO current histologic classification of breast cancer is demonstrated in **Table 1**. Photomicrographies of the most frequent histologic subtypes of invasive breast cancer are represented on **Figure 1**. The hormonal status receptors (estrogen and progesterone) expression by the neoplastic cells was just evaluated by immunohistochemistry on paraffin-embedded specimens of tumor (core needle biopsy or the surgical excision specimen) [6, 7].

Breast cancer is known for its heterogeneous behavior [3, 4]. The histologic classification has been satisfactory for malignancy determination [6]. Though, the clinical division based on hormonal status was not enough for accurate prediction of the prognosis and of clinical response to the therapy [5]. Thus, until the last decade of 20th century, the clinical treatment of breast cancer was based on unespecific chemotherapy and hormonal therapy with drugs like tamoxifen, a known hormonal receptor antagonist [12].

The hormone positive breast cancer is more "differentiated" than the negative one, as the cancer cells maintain the epithelial original cell feature of hormonal receptor expression and, therefore, the hormonal antagonist drugs are effective against these tumors [8]. On the other hand, the approaching of hormonal negative cancers were variable, since it was forming a kindly heterogeneous group, with different aggressiveness potentials, imprecise therapeutic response and doubtful prognosis [6, 8, 10].

In the first decade of the current century, it was emerged a promising classification of breast cancer, proposing a division of the disease in 3 molecular subtypes:


*Current histologic (morphologic) classification of epithelial breast tumors (WHO, 2019, 5th edition). This classification considers the tumors histologic patterns of tumors. The most common histologic breast cancer subtype is the infiltrating duct carcinoma NOS (or invasive ductal carcinoma non-special type), accounting for 65–80% of all breast cancers. Invasive lobular carcinoma corresponds to around 5% of all breast malignancies.*

#### **Figure 1.**

*Photomicrographies of hematoxylin & eosin (H&E) slides illustrating the most frequent histologic subtypes of infiltrating (invasive) breast cancer. (A) Infiltrating duct carcinoma (Invasive ductal carcinoma NOS) is the most frequent histologic subtype of breast cancer (nearly 75–80% of all invasive breast cancer), constituted of cohesive cancer cells forming infiltrative ductal and ribbons structures (4×); (B) Lobular invasive carcinoma is the second most frequent invasive breast cancer (5–15% of all invasive breast cancer), composed of infiltrating cancer cells with diffuse single-file pattern (10×). In this subtype, the cancer cells lose the cohesion (e-cadherin, an immunomarker important for cell adhesion evaluation, is negative on immunohistochemistry); (C) Mucinous carcinoma represents approximately 2% of breast invasive cancer, composed of groups of cancer cells outling ductal structures, immersed in mucin pools, with delicate fibrous strands containing capilaries (10×); (D) Tubular carcinoma represents around 2% of invasive breast cancer, composed of haphazard arrangement of small well-differentiated duct structures, forming tubules (4×). The other listed invasive breast cancers are uncommon, with each one histologic subtype representing 1% or less (figures extracted from [11]).*

luminal, *HER-2* overexpressed and "triple negative" (**Table 2**). This new classification has demonstrated better correlation with the breast cancer behavior. Thus, it was adopted on diagnostic routine of breast cancer. Since this study was published, besides of evaluate the histologic patterns and report the pathologic tumor stage, the pathologist has been required to determine the molecular cancer profile, which has became indispensable to therapy planning [12, 13].

The luminal subtype cancer is the hormonal positive tumors. This kind of cancer is frequently well or moderately differentiated on histology, formed by lower grades of cells, with lower proliferative index, which is evaluated by antibody *Ki-67/MIB-1* on immunohistochemistry. The majority of breast cancers are classified as this subtype (**Figure 2**). Eventually, luminal cancer can overexpress or amplify at the same time the protein called human epithelial growth factor receptor 2 (*HER-2*), codified by the oncogen *ERBB2* [14, 15].

*ERBB2* is a oncogen localized in chromosome 17, which codifies the *HER-2* protein, a type I transmembrane protein with an extracellular and an intracellular domains, activating signaling pathways from extracellular signals. In last instance, the overexpression/amplification of *HER-2* overactivates the intracellular protein kinases, dysregulating the cell cycle, disrupting the cell adhesion and cell polarity and promoting the invasive phenotype [16].

*Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic… DOI: http://dx.doi.org/10.5772/intechopen.94462*


#### **Table 2.**

*Molecular subtypes of breast carcinoma. The reported absolute incidences of each molecular subtype of breast carcinoma are variable among several studies.*

#### **Figure 2.**

*Photomicrographies of immunohistochemical assessment of invasive breast cancer hormonal expression, in an example of infiltrating duct carcinoma (Invasive ductal carcinoma NOS, WHO 2019), which is the most frequent histologic subtype of breast cancer, constituted by ductal and ribbons structures of cancer cells infiltrating the breast stroma (A). Any kind of nuclear positivity of estrogen receptor (B) and progesterone receptor (C) allows to consider the tumor as positive to hormonal receptor on immunohistochemistry, even when rare cells are positive (C). The hormonal receptors positivities on immunohistochemistry are evaluated for intensity (mild, moderate or strong) and percentages of positive cells (0–100%). Examples of mild positivity (black arrow, C), moderate positivity (red arrow, B) and strong positivity (green arrow, C). Ki-67/ MIB-1 assesses the tumor proliferative index (D), its positivity is nuclear and is expressed in percentages of positive cells (0–100%).*

The breast cancer classified as *HER-2* subtype is necessarily negative for hormonal receptors and is featured by overexpression or amplification of *HER-2.* This subtype is frequently less differentiated than the luminal ones on histology, constituted by high grades of cancer cells, with high proliferative index. The presence of elevated concentration of intratumoral lymphocytes (TIL) is not an uncommon finding in these tumors [17].

#### **Figure 3.**

*(a) Photomicrographies of immunohistochemical assessment of HER-2 expression status by tumoral cells in histologic paraffinized specimen of breast cancer. Score 0 (negative): none tumoral cell is labeled. Score 1+ (negative): incomplete positivity with low intensity in part of tumoral cells. Score 2+ (equivocal): complete positivity with low intensity in majority of tumoral cells. Score 3+ (positive): complete positivity with strong intensity in majority of tumoral cells. (b) Photomicrographies of amplification of HER-2 gene performed through fluorescence in situ hibridization (FISH) in a HER-2-overexpressed breast carcinoma on immunohistochemistry (Score 3+, E). HER-2 gene copies are the orange signals (B) and chromosome 17 centromeres (CEP17) are the green signals (C). The signals of HER-2 gene and CEP17 are present in tumoral cell nuclei (blue, A and D). CEP17 is an internal control on the same chromosome to compare with HER-2 signals in tumoral cell nucleus. According to American Society of Clinical Oncology/College of American Pathologists (ASCO-CAP) guidelines, a HER-2/CEP17 ratio > 2.0 defines a positive result for amplification of HER-2 gene. If HER-2/CEP17 ratio is < 2.0, an average HER-2 copy number > 6.0 signals/cell defines a positive result for amplification of HER-2 gene, an average HER-2 copy number < 4.0 signals/cell defines a negative result for amplification of HER-2 gene and an average HER-2 copy number > 4.0 and < 6.0 signals/cell defines an equivocal result for amplification of HER-2 gene (extracted from [20]).*

*Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic… DOI: http://dx.doi.org/10.5772/intechopen.94462*

This new receptor was one of the pioneers for target therapy in molecular era of breast cancer approaching, as it was developed a new class of drug, called trastuzumab, with selective action against the cancer cells overexpressing/amplifying *HER-2*. Besides the *HER-2* subtype tumors, this drug is also recommended for the luminal ones with positive status for *HER-2* [18, 19].

The status of *HER-2* expression is analyzed through immunohistochemistry of paraffin-embedded specimens of the breast cancer (**Figure 3a**). The tumor is considered negative for *HER-2* if it is not labeled (score zero) or the cell membrane is partially labelled for the *HER-2* antibody (score 1+). The tumor is positive for *HER-2* if all the cell membranes outlines are strongly labeled for this antibody (score 3+). Finally, in part of the cases, the *HER-2* antibody can label totally the cancer cell membrane, but with low intensity or can label partially the cell membrane with high intensity. In these situations, the *HER-2* status is considered equivocal (score 2+). The confirmation of overexpression/amplification must be evaluated through fluorescence "in situ" hybridization (FISH) (**Figure 3b**) [21, 22].

The "triple negative" breast cancer is negative for hormonal receptors and *HER-2*. It is the less differentiated tumor subtype on histology, formed by highest grades cancer cells, with highest proliferative index, presenting the worst prognosis among the 3 molecular subtypes. This tumor still does not present an specific therapy, which is chosen depending on the clinicopathological stage. In metastatic disease, the treatment focuses on quality of life and palliation. In "triple negative" tumors, the evaluation of *BRCA* status is mandatory [8, 21].

#### **3. Germline mutations of** *BRCA-1/BRCA-2* **genes: increased risk of breast cancer development during the life**

Identified in 1994, *BRCA-1*/*BRCA-2* are tumoral suppressor genes, respectively located in chromosome 17 and 13. Mutations of these genes are related to hereditary breast cancer, estimated in 5–10% of all breast malignancies. *BRCA-1/BRCA-2* play a central role in DNA repair [23, 24]. Mutations of these genes increase the susceptibility for DNA damages. "Triple negative" subtypes carry more frequently mutations of *BRCA-1* and mutations of *BRCA-2* increase the risk for luminal subtypes of breast cancer. *HER-2* overexpression is inversely correlated to *BRCA* mutations [24, 25].

It was observed in some studies that "triple negative" breast cancers with *BRCA* mutations present more chemosensitivity than the ones without *BRCA* mutations. Chemotherapy with DNA-damaging drugs, like the alkylating agents and anthracycline, can prolong the free of disease survival for tumors of triple negative phenotypes. This found is expected, since *BRCA* mutation prejudices the DNA repair and, consequently, increase the sensibility to DNA damages of cancer cells by these drugs. Neither therapeutic response nor free of disease survival of luminal subtypes of breast cancer seems to be influenced by *BRCA* mutations [8, 24, 26].

Regarding on prognosis, multiples studies present conflicting results. The prognosis depends on tumor features, especially the molecular subtypes and the clinicopathologic stage. The predictive value depends on the administrated therapy. Thus, *BRCA-1* mutated breast cancer probably present worse prognosis than the *BRCA-2* mutated ones, since *BRCA-1* mutated tumors are mainly of "triple negative" phenotype, therefore intrinsically more aggressive than the luminal subtypes harboring *BRCA-2* mutations [24, 27].

The tumoral suppressor proteins codified by *BRCA-1/BRCA-2* act on homologous recombination repair of double stranded DNA breaks. Homologous recombination mechanism protect the integrity of genome in proliferating cells. *BRCA-1* recognize DNA damage and recruit DNA repair proteins. *BRCA-2* mediates the

recruitment of another protein, called *RAD51*, to double stranded DNA breaks, allowing for homologous recombination repair [24, 28].

In *BRCA*-mutant breast tumors, the base excision repair pathway is important for cancer cell survival, in response to single stranded DNA breaks. Polyadenosine diphosphate-ribose (*PARP*) is a family of DNA repair enzymes, playing a key role in base excision repair mechanism. These enzymes are recruited to the site of DNA damage and add ADP-ribose to target nuclear proteins, causing post-translational modifications and restarting stalled DNA replication. *BRCA*-mutant breast cancer presents deficiency of homologous recombination repair, with overactivated *PARP*, leading the cancer cell to avoid apoptosis [24, 26, 28].

The inhibition of *PARP* cause persistance of single stranded break, resulting in stalled replication and double strand breaks. This mechanism leads to accumulation of DNA damage, causing cell cycle arrest and apoptosis. The *PARP* inhibitors form an emerging class of drugs, which have been recommended to chemotherapy for *BRCA*-mutant breast cancer and empirically for metastatic breast cancer, with promising results [24, 25, 28].

#### **4. Cancer stem-cell hypothesis: impact in breast cancer prognosis**

In the last two decades, experimental evidences in several studies of neoplastic tissues have revealed a population of cancer cell with properties of self-renewal, differentiation to multiple lineages ability and low proliferative index. These properties have been considered cancer stem-cell like features and attributed to a possible cancer stem-cell lineage present in the tumor bulk [29, 30].

Cancer stem-cell has awaked interest in the context of breast cancer because of its characteristic heterogeneity of biological behavior and therapeutic response. It has been hypothesized that cancer stem-cell might be one of the causes of the high variability of biological and prognostic spectrum of breast cancer. Cancer-stem cells might play an important role on therapeutic resistance and progression of disease, affecting the overall and free of disease survival [31, 32].

Thus, an important feature which allows possible cancer stem-cell resistance to chemotherapy is its low expression of surface proteins. Because of its self-renewal properties, cancer stem-cell does not depends on signaling from other cells to proceed its functions in tumoral tissues. Furthermore, for its low antigenicity and low proliferation index, there are few alternatives for drug interactions. DNA damage agents are poor effective against these cells possibly for a lack of proliferation, as well new classes of drugs, like *PARP* inhibitors, which better act on cells in proliferative phase [31, 33].

One of possible pathways for breast cancer therapeutic resistance acquired along the time might be explained by populations of cancer stem-cells not eliminated, selected by multiple chemotherapy cycles. Tumoral cells in active proliferation phase are more hitten, increasing the proportion of indolent cells with stem-like features in cancer cell population. Through the capacity of multilineage differentiation, cancer stem cells might generate new daughter cells with more aggressiveness and chemoresistance [32, 34].

The identification of cancer stem-cells is challenging. First, because of its irregular distribution in selected tumor amounts. Second, for definition, these cells are frequently scarces in tumor bulk. In this way, these cells are better identified through "in vitro" methods, like cellular cultures. However, the mainly disadvantage of this technique is the fact of stem cells behave in a different fashion in artificial environment, since the cell phenotype expression depends on their interactions [32, 35].

#### *Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic… DOI: http://dx.doi.org/10.5772/intechopen.94462*

Thus, several studies with cancer stem-cells in different neoplastic tissues have been accomplished with conflicting results. An interesting method to identify these cells in their original environment is the immunohistochemistry performed on amounts of paraffin-embedded neoplastic tissues, with the advantages to allow the evaluation of phenotype expression next to the reality and to be easily performed and cost-effectiveness in diagnostic routine [35].

In the last years, some putative stem-cell markers detected by immunohistochemistry have been tested in paraffinized tissues of breast cancer. Multiple studies have demonstrated that expression of putative stem-cell markers by tumoral cells seems to worse the prognosis and survival in breast cancer. The most frequent studied stem-cell markers are *CD24*, *CD44*, *CD133* and *EPCAM*, with two identified putative stem-cell phenotypes: *CD24* low/*CD44* enriched and co-expression of *CD133* and *EPCAM* (**Figure 4**). Besides of the scarcity of stem-cells in neoplastic tissues, the conflictous results of these studies might be explained by a necessity to qualitative analysis of these markers expression, exactly for the rarity of stem-cells [32, 36].

In some studies, identification of a stem-cell like phenotype *CD24* low/*CD44* enriched have prejudiced the free of disease survival, especially in cases of early stages of breast cancer, with more occurrence of distant metastasis and cancer recurrence after surgical and adjuvant treatments. The presence of cancer cells with positivity for cancer stem-cell phenotype *CD133/EPCAM* is has been related to poor overall survival in breast cancer, with more adjuvant therapeutic fail [32].

For the moment, these putative stem-cell phenotypes seems to be independent prognostic factors in breast cancer. "Triple negative" breast cancer and *BRCA-1* mutant breast cancer have been associated to stem-cell like phenotype *CD24* low/*CD44* enriched. These putative stem-cell markers may become possible future targets for new drugs in the future [30, 32].

#### **Figure 4.**

*Photomicrographies of double-labeled simple stained putative CSC antibodies (400×, original magnification, immunoperoxidase and DAB). (A) CD133: cytoplasm positivity (immunoperoxidase); (B) EPCAM: membrane positivity (DAB); (B) CD133+/EPCAM+: CSC profile (black arrow: membrane positivity to DAB and cytoplasm positivity to immunoperoxidase at the same cell); (C) CD24: cytoplasm positivity (immunoperoxidase); (D) CD24*−*/CD44+: CSC profile (black arrow: membrane positivity only to DAB)*

#### **5. Immunologic aspects related to breast cancer**

In the context of cancer, the immune system can suppress the tumor growth by the destruction of cancer cells or inhibition of their outgrowth. On the other hand, immune system can play a role on tumor progression by the selection of tumor cells which are adapted to survive in an immunocompetent host or modifying the tumor environment to facilite the tumor outgrowth [37].

Elevated levels of *CD4+* regulatory T lymphocytes (*Tregs*) found in many cancers are associated to poor prognosis. *Tregs* create a favorable immunosuppressive microenvironment to the outgrowth and progression of the tumor. On this way, *FOXP3* is expressed by the *Tregs* and can be detected by immunohistochemistry. *FOXP3* is responsible for induction and maintenance of tolerance to self antigens in normal cells, as well this immunotolerance can be performed by the *Tregs* with cancer cell antigens [37, 38].

Another example of cancer cell escape mechanism from the immune system is *caspase-8* mutations present in "triple negative" breast cancers and other solid malignant tumors. These mutations abolish the death induced by cytotoxic lymphocytes *CD8+* in tumoral cells [37, 39].

The activation of T lymphocytes by foreign antigens occurs by concomitant major histocompatibility complex (*MHC*) antigen presentation and co-expression of T-cell receptor (*TCR*). At the same time, a family of T-cell transmembrane proteins *CD28/B7*, called "immune checkpoints", produces co-inhibitory or co-stimulatory signals. The immune checkpoints regulates the T-cell immunotolerance to protect the tissues from undesirable damages. Cancer cells may produce signals to inhibit T-cell action, through cytotoxic T-lymphocyte associated antigen-4 (*CTLA-4*), programmed cell death-1 (*PD-1*) and its ligands (*PDL-1*) [37, 40].

*PD-1* is an inhibitory "immune checkpoint" expressed on the surface of T-cells, B-cells and NK-cells. When T-cells have been activated by their *TCR*, the cells express at the same time *PD-1*, which is a possibility to the attacked cell to escape from the immune reaction (**Figure 5**). Cancer cells express the ligand *PDL-1* on their surfaces, activating *PD-1* of T-cells, escaping from the attack [37, 40].

*PD-L1* expression has been associated with large tumor size, high grade, high proliferation, estrogen receptor (*ER*)-negative status, and human epidermal growth factor receptor-2 (*HER2*)-positive status in breast cancer. Survival in breast cancer is inversely related to *PD-1/PDL-1* levels. *PDL-1* expression increases tumor

#### **Figure 5.**

*Simplified schematic illustration of PD-1/PDL-1 interactions in immune responses against cancer cell. Tumoral antigens (Ag) are presented via T-cell by major histocompatibility complex (MHC) of dendritic cells. T-cell recognize tumoral Ag via TCR (T-cell receptor). Interaction Ag-TCR induces an positive immune response against tumoral Ag. Though, there is a scape mechanism of cancer cell from the T-cell attack: interaction of programmed death cell ligands (PDL-1/2) expressed by cancer cell with PD-1 expressed by T-cell inhibit the T-cell action. This scape mechanism of cancer cell mimics the regulation action to avoid immune responses of T-cell against self antigens. The principle of immune therapy is the inhibition of PD-1/PDL-1 (extracted from [40]).*

*Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic… DOI: http://dx.doi.org/10.5772/intechopen.94462*

aggressiveness, stimulating tumorigenesis, invasiveness and ability to escape from cytotoxic T *CD8+* lymphocytes attacks [39, 41]. The immunohistochemical evaluation of *PDL-1* is shown in **Figure 6**.

Immune therapies with anti-*CTLA-4* and anti-*PD1*/anti-*PDL-1* agents have been promising for treating several cancers. In breast cancer, some researches reported positive results around 20% of breast tumors on treatment with these agents, mainly the "triple negative" and *HER-2* subtypes, for their higher antigenicity. In general, breast cancer present lower immunogenicity than other cancers and breast cancer cells frequently create an immunosuppressor tumor microenvironment by signaling [37, 43].

The presence of tumor infiltrating lymphocytes (TIL) in some breast cancers has been related to a favorable prognosis, especially in "triple negative" and *HER-2* subtypes. TIL are formed mainly by T-cells *CD3+/CD56* negative, which are either *CD4+* or *CD8+*. A minority component of B-cells *CD20+* and *NK-cells* may be present. The attraction of TIL by cancer cells have been related to their expression of some chemokines, like *CXCL9* and *CXCL13* [37, 44].

In "triple negative" and *HER-2* subtypes of breast cancer, the presence of TIL is related to a better response to neoadjuvant therapy, as well neoadjuvant treatment may modify the tumor microenvironment to attract TIL to tumor site. Furthermore, when the TIL are not attracted instead of neoadjuvance, it is indicative for bad prognosis [44].

#### **Figure 6.**

*Examples of PDL-1 expression in breast cancer using 3 different antibodies: Dako 22C3 (D,E and F), Ventana SP263 (G,H and I) and BioCare RbM CAL10 (A, B and C). PDL-1 scoring is divided into 3 groups: zero staining is negative, 1–49% of positive cells are considered "low PDL-1 expression" and 50% of more positive cells are considered "high PDL-1 expression". Examples of negative, low and high PDL-1 expression are represented on A, B and C for BioCare antibody (extracted from [42]).*

#### **6. Advanced stage breast cancer: considerations under current approach and futures perspectives**

Metastatic breast cancer is considered incurable nowadays with currently therapies. Therapy of metastatic disease aims to guarantee quality of life, palliation of symptoms and prolongation of the patient survival. Advanced stage disease is becoming increasingly chronic, controlled by sequencial therapies, with more personalized approach than the early stage breast cancer [8].

Systemic therapy is frequently the first choice of metastatic disease. Before the new therapeutic decision, it is necessary to consider the previous treatments. If possible, it is recommended to re-evaluate the histologic features and molecular subtype status of the metastatic lesion through a new biopsy, with new immunohistochemical study for hormonal receptor and *HER-2* status. Some studies reported until 40% of discrepances of metastatic lesion histologic features and molecular subtype status *versus* primary tumor histologic and immunohistochemical aspects [45].

The metastatic disease therapeutic choices search for positive targets to hit more effectively the neoplastic cells. Thereby, expression of hormonal receptors by the metastatic lesion is elective for endocrine therapy. Endocrine drugs include tamoxifen, aromatase inhibitors, fulvestrant and progestins. The use of these drugs in metastasis with hormone receptor positive status have demonstrated increase of free of disease survival in several studies [8, 45].

Furthermore, new generation of drugs which inhibit the cyclin dependant kinase (*CDK*) have been successful in prolongation of free of disease survival in luminal subtype *HER-2* negative metastatic disease. *CDK4/6* is a holoenzyme responsible for several extracellular signaling pathways to cell cycle transitions. *CDK4/6* fosforilates and inactivates retinoblastoma tumor supressor protein (*Rb*). Extracellular signals regulate the expression of cyclins and *CDK* inhibitors, like *p16*INK4a [46].

In human cancer, this circuit is dysregulated by either overexpression of cyclin D1, loss of *p16*Ink4a, the mutation of *CDK4* to an *Ink4*-refractory state, or the loss of *Rb* itself. The primary target of *CDK4* is the *Rb* protein, though this holoenzyme either can phosphorylate factors involved in cell differentiation affecting their transcriptional activity, apoptotic factors affecting their activity and other factors that can directly affect mitochondrial function [8, 46, 47].

Therefore, *CDK* inhibitors act in tumor microenvironment, blocking *Rb* phosphorilation and leading to cell cycle exit. Moreover, *CDK* have kinase activity towards *SPOP*, an ubiquitin protein that interacts with *PDL-1*. *CDK* inhibitors lead to inhibition of *SPOP* phosphorilation with blockade of *PDL-1* and stimulus to *PD-1* expression by T-cells, attracting T-cell infiltration to the tumor. In this way, the combined use of *CDK* inhibitors and *PDL-1/PD-1* inhibitors may be promising, requiring more future studies [46–48].

For the moment, hormonal receptors and *HER-2* status are the few validated molecular targets of clinical importance on metastatic breast cancer approaching through chemotherapy and endocrine therapy. For *HER-2* positive metastatic disease, anti-HER-2 treatment with trastuzumab is well established and is recommended as soon as possible. Immune therapy is not standardized for metastatic breast cancer, since metastatic breast disease is highly heterogeneous. Though, it is a promising therapy for the future, as well the target molecular therapies, which become more effective with discovery of novel pathways and mutations by new studies to be developed [8].

A resume of main biomarkers of clinicopathologic importance for breast cancer management is shown in **Table 3** and a proposal of a algorithm for clinicopathologic evaluation of breast cancer is presented in **Table 4**.

*Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic… DOI: http://dx.doi.org/10.5772/intechopen.94462*


#### **Table 3.**

*Resume of main biomarkers of clinicopathologic importance for breast cancer management.*

#### **Table 4.** *Proposal of an algorithm for clinicopathologic evaluation of breast cancer.*

#### **7. Conclusion and final considerations**

In the 21st century, breast cancer classification and diagnosis advanced considerably from a purely morphologic/histologic approaching to a immune and molecular basis, with remarkable improvement of the correlation between classification and prediction of biological behavior and prognosis.

The adoption of a clinicopathologic classification based on molecular subtypes of breast cancer in the last decade has modified decisively the management of the disease in the way of molecular era, opening new ways to discovering of multiple targets for novel therapies.

Innovative concepts related to immune reactions related to human cancers, which have been unveiled in the recent years, particularly the immune checkpoints, have offered new treatment tools for several human cancers with promising results, although not still established for breast cancer.

In the molecular era of cancer, the integration of novel knowledges in a direction of more accurated diagnosis and prediction of prognosis to allow personalized therapies is the key to future human cancer management, including the breast cancer.

#### **Author details**

Rodrigo Vismari de Oliveira Rede D'Or São Luiz, São Paulo, Brazil

\*Address all correspondence to: rodrigovismari@gmail.com

© 2020 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

*Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic… DOI: http://dx.doi.org/10.5772/intechopen.94462*

#### **References**

[1] Ahmad, A. (Ed.). (2019). *Breast Cancer Metastasis and Drug Resistance. Advances in Experimental Medicine and Biology.* doi:10.1007/978-3-030-20301-6

[2] Coughlin SS. Epidemiology of Breast Cancer in Women. *Adv Exp Med Biol*. 2019;**1152**:9-29. DOI: 10.1007/978-3-030-20301-6\_2

[3] Zardavas D, Irrthum A, Swanton C, Piccart M. Clinical management of breast cancer heterogeneity. *Nat Rev Clin Oncol*. 2015;**12**(7):381-394. DOI: 10.1038/nrclinonc.2015.73

[4] Joseph C, Papadaki A, Althobiti M, Alsaleem M, Aleskandarany MA, Rakha EA. Breast cancer intratumour heterogeneity: current status and clinical implications. *Histopathology*. 2018;**73**(5):717-731. DOI: 10.1111/ his.13642

[5] Wesolowski R, Ramaswamy B. Gene expression profiling: changing face of breast cancer classification and management. *Gene Expr*. 2011;**15**(3):105-115. DOI: 10.3727/105221 611x13176664479241

[6] Tsang JYS, Tse GM. Molecular Classification of Breast Cancer. *Adv Anat Pathol*. 2020;**27**(1):27-35. DOI: 10.1097/PAP.0000000000000232

[7] Marchio C, Balmativola D, Castiglione R, Annaratone L, Sapino A. Predictive Diagnostic Pathology in the Target Therapy Era in Breast Cancer. *Curr Drug Targets*. 2017;**18**(1):4-12. DOI: 10.2174/1389450116666150203121218

[8] Harbeck N, Penault-Llorca F, Cortes J, et al. Breast cancer. *Nat Rev Dis Primers*. 2019;5(1):66. Published 2019 Sep 23. doi:10.1038/s41572-019-0111-2

[9] Tryfonidis K, Senkus E, Cardoso MJ, Cardoso F. Management of locally advanced breast cancer-perspectives

and future directions [published correction appears in Nat Rev Clin Oncol. 2015 Jun;12(6):312]. *Nat Rev Clin Oncol*. 2015;**12**(3):147-162. DOI: 10.1038/nrclinonc.2015.13

[10] Jézéquel P, Kerdraon O, Hondermarck H, et al. Identification of three subtypes of triple-negative breast cancer with potential therapeutic implications. *Breast Cancer Res*. 2019;21(1):65. Published 2019 May 17. doi:10.1186/s13058-019-1148-6

[11] Simpson, Sanders. Differential Diagnosis in Surgical Pathology. Breast: Wolters Kluwer; 2017

[12] Du XL, Key CR, Osborne C. Community-based assessment of adjuvant hormone therapy in women with breast cancer, 1991-1997. *Breast J.* 2004;**10**(5):433-439. DOI: 10.1111/j.1075-122X.2004.21357.x

[13] Prat A, Pineda E, Adamo B, et al. Clinical implications of the intrinsic molecular subtypes of breast cancer. *Breast*. 2015;**24**(Suppl 2):S26-S35. DOI: 10.1016/j.breast.2015.07.008

[14] Bueno GAM. Características clínicas y pronósticas de los subtipos moleculares de cáncer de mama determinados por inmunohistoquímica. Arequipa, Perú [Clinical and prognostic characteristics of the molecular subtypes of breast cancer determined by immunohistochemistry. Arequipa, Peru]. *Rev Peru Med Exp Salud Publica*. 2017;**34**(3):472-477. DOI: 10.17843/ rpmesp.2017.343.2530

[15] Irigoyen MA, García FV, Iturriagagoitia AC, Beroiz BI, Martínez MS, Guillén Grima F. Subtipos moleculares del cáncer de mama: implicaciones pronósticas y características clínicas e inmunohistoquímicas [Molecular subtypes of breast cancer: prognostic implications and clinical and immunohistochemical characteristics]. *An Sist Sanit Navar*. 2011;**34**(2):219-233. DOI: 10.4321/s1137-66272011000200008

[16] Moasser MM. The oncogene HER2: its signaling and transforming functions and its role in human cancer pathogenesis. *Oncogene*. 2007;**26**(45):6469-6487. DOI: 10.1038/ sj.onc.1210477

[17] Hida AI, Watanabe T, Sagara Y, et al. Diffuse distribution of tumorinfiltrating lymphocytes is a marker for better prognosis and chemotherapeutic effect in triple-negative breast cancer. *Breast Cancer Res Treat*. 2019;**178**(2):283- 294. DOI: 10.1007/s10549-019-05390-x

[18] Ahmed S, Sami A, Xiang J. HER2 directed therapy: current treatment options for HER2-positive breast cancer. *Breast Cancer*. 2015;**22**(2):101-116. DOI: 10.1007/s12282-015-0587-x

[19] Kolarova I, Vanasek J, Odrazka K, et al. Therapeutic significance of hormone receptor positivity in patients with HER-2 positive breast cancer. *Biomed Pap Med Fac Univ Palacky Olomouc Czech Repub*. 2019;**163**(4):285- 292. DOI: 10.5507/bp.2019.060

[20] Press MF, Villalobos I, Santiago A, Guzman R, Cervantes M, Gasparyan A, et al. Assessing the New American Society of Clinical Oncology/College of American Pathologists Guidelines for *HER2* Testing by Fluorescence *In Situ* Hybridization: Experience of an Academic Consultation Practice. Arch Pathol Lab Med. 2016 Nov;**140**(11):1250-1258. DOI: 10.5858/ arpa.2016-0009-OA Epub 2016 Apr 15. PMID: 27081877

[21] Azizun-Nisa BY, Raza F, Kayani N. Comparison of ER, PR and HER-2/ neu (C-erb B 2) reactivity pattern with histologic grade, tumor size and lymph node status in breast cancer. *Asian Pac J Cancer Prev*. 2008;**9**(4):553-556

[22] Ciampa A, Xu B, Ayata G, et al. HER-2 status in breast cancer: correlation of gene amplification by FISH with immunohistochemistry expression using advanced cellular imaging system. *Appl Immunohistochem Mol Morphol*. 2006;**14**(2):132-137. DOI: 10.1097/01.pai.0000150516.75567.13

[23] Talhouet S, Peron J, Vuilleumier A, et al. Clinical outcome of breast cancer in carriers of BRCA1 and BRCA2 mutations according to molecular subtypes. *Sci Rep*. 2020;10(1):7073. Published 2020 Apr 27. doi:10.1038/ s41598-020-63759-1

[24] Turk AA, Wisinski KB. PARP inhibitors in breast cancer: Bringing synthetic lethality to the bedside. *Cancer*. 2018;**124**(12):2498-2506. DOI: 10.1002/cncr.31307

[25] Sonnenblick A, de Azambuja E, Azim HA Jr, Piccart M. An update on PARP inhibitors--moving to the adjuvant setting. *Nat Rev Clin Oncol*. 2015;**12**(1):27-41. DOI: 10.1038/ nrclinonc.2014.163

[26] Guney Eskiler G, Cecener G, Egeli U, Tunca B. Triple negative breast cancer: new therapeutic approaches and BRCA status. *APMIS*. 2018;**126**(5):371- 379. DOI: 10.1111/apm.12836

[27] Baretta Z, Mocellin S, Goldin E, Olopade OI, Huo D. Effect of BRCA germline mutations on breast cancer prognosis: A systematic review and meta-analysis. *Medicine (Baltimore)*. 2016;**95**(40):e4975. DOI: 10.1097/ MD.0000000000004975

[28] Tangutoori S, Baldwin P, Sridhar S. PARP inhibitors: A new era of targeted therapy. *Maturitas*. 2015;**81**(1):5-9. DOI: 10.1016/j.maturitas.2015.01.015

[29] Dontu G. Breast cancer stem cell markers - the rocky road to clinical applications. *Breast Cancer Res*. 2008;**10**(5):110. DOI: 10.1186/bcr2130 *Advances in Molecular and Immunohistochemical Detection of Prognostic and Therapeutic… DOI: http://dx.doi.org/10.5772/intechopen.94462*

[30] Honeth G, Bendahl PO, Ringnér M, et al. The CD44+/CD24- phenotype is enriched in basal-like breast tumors. *Breast Cancer Res*. 2008;**10**(3):R53. DOI: 10.1186/bcr2108

[31] Dawood S, Austin L, Cristofanilli M. Cancer stem cells: implications for cancer therapy. *Oncology (Williston Park)*. 2014;**28**(12):1101-1110

[32] Oliveira RV, Souza VB, Souza PC, et al. Detection of Putative Stem-cell Markers in Invasive Ductal Carcinoma of the Breast by Immunohistochemistry: Does It Improve Prognostic/Predictive Assessments? *Appl Immunohistochem Mol Morphol*. 2018;**26**(10):760-768. DOI: 10.1097/PAI.0000000000000513

[33] Czerwinska P, Kaminska B. Regulation of breast cancer stem cell features. *Contemp Oncol (Pozn)*. 2015;**19**(1A):A7-A15. DOI: 10.5114/ wo.2014.47126

[34] Das PK, Rakib MA, Khanam JA, Pillai S, Islam F. Novel Therapeutics Against Breast Cancer Stem Cells by Targeting Surface Markers and Signaling Pathways. *Curr Stem Cell Res Ther*. 2019;**14**(8):669-682. DOI: 10.2174/ 1574888X14666190628104721

[35] Neumeister V, Agarwal S, Bordeaux J, Camp RL, Rimm DL. In situ identification of putative cancer stem cells by multiplexing ALDH1, CD44, and cytokeratin identifies breast cancer patients with poor prognosis. *Am J Pathol*. 2010;**176**(5):2131-2138. DOI: 10.2353/ajpath.2010.090712

[36] Joseph C, Arshad M, Kurozomi S, et al. Overexpression of the cancer stem cell marker CD133 confers a poor prognosis in invasive breast cancer. *Breast Cancer Res Treat*. 2019;**174**(2):387- 399. DOI: 10.1007/s10549-018-05085-9

[37] Bayraktar S, Batoo S, Okuno S, Glück S. Immunotherapy in breast cancer. *J Carcinog*. 2019;18:2. Published 2019 May 23. doi:10.4103/jcar.JCar\_2\_19 [38] Shou J, Zhang Z, Lai Y, Chen Z, Huang J. Worse outcome in breast cancer with higher tumor-infiltrating FOXP3+ Tregs : a systematic review and meta-analysis. *BMC Cancer*. 2016;16(1):687. Published 2016 Aug 26. doi:10.1186/s12885-016-2732-0

[39] Cano-González A, Mauro-Lizcano M, Iglesias-Serret D, Gil J, López-Rivas A. Involvement of both caspase-8 and Noxa-activated pathways in endoplasmic reticulum stressinduced apoptosis in triple-negative breast tumor cells. *Cell Death Dis*. 2018;9(2):134. Published 2018 Jan 26. doi:10.1038/s41419-017-0164-7

[40] Hartkopf AD, Taran FA, Wallwiener M, et al. PD-1 and PD-L1 Immune Checkpoint Blockade to Treat Breast Cancer. *Breast Care (Basel)*. 2016;**11**(6):385-390. DOI: 10.1159/000453569

[41] Sabatier R, Finetti P, Mamessier E, et al. Prognostic and predictive value of PDL1 expression in breast cancer. *Oncotarget*. 2015;**6**(7):5449-5464. DOI: 10.18632/oncotarget.3216

[42] Karnik T, Kimler BF, Fan F, Tawfik O. PD-L1 in breast cancer: comparative analysis of 3 different antibodies. Hum Pathol. 2018 Feb;**72**:28- 34. DOI: 10.1016/j.humpath.2017.08.010 Epub 2017 Aug 31. PMID: 28843709

[43] Chen X, Shao Q, Hao S, et al. CTLA-4 positive breast cancer cells suppress dendritic cells maturation and function. *Oncotarget*. 2017;**8**(8):13703- 13715. DOI: 10.18632/oncotarget.14626

[44] Wimberly H, Brown JR, Schalper K, et al. PD-L1 Expression Correlates with Tumor-Infiltrating Lymphocytes and Response to Neoadjuvant Chemotherapy in Breast Cancer. *Cancer Immunol Res*. 2015;**3**(4):326-332. DOI: 10.1158/2326- 6066.CIR-14-0133

[45] Li Z, Kang Y. Emerging therapeutic targets in metastatic progression:

A focus on breast cancer. *Pharmacol Ther*. 2016;**161**:79-96. DOI: 10.1016/j. pharmthera.2016.03.003

[46] Klein ME, Kovatcheva M, Davis LE, Tap WD, Koff A. CDK4/6 Inhibitors: The Mechanism of Action May Not Be as Simple as Once Thought. *Cancer Cell*. 2018;**34**(1):9-20. DOI: 10.1016/j. ccell.2018.03.023

[47] Deng J, Wang ES, Jenkins RW, et al. CDK4/6 Inhibition Augments Antitumor Immunity by Enhancing T-cell Activation. *Cancer Discov*. 2018;**8**(2):216-233. DOI: 10.1158/2159- 8290.CD-17-0915

[48] Lim S, Kaldis P. Cdks, cyclins and CKIs: roles beyond cell cycle regulation. *Development*. 2013;**140**(15):3079-3093. DOI: 10.1242/dev.091744

#### **Chapter 4**

## Molecular Pathology in the New Age of Personalized Medicine

*Valeria Cecilia Denninghoff*

#### **Abstract**

Personalized medicine is a new approach that allows the identification of patients that can benefit from targeted therapies because of the molecular characteristics of the tumors they present. The molecular profile of the tumor can be studied at the genomic (DNA), transcriptomic (RNA) or protein (protein) level. The next generation sequencing is a useful tool for the study of molecular profile from DNA/RNA. This tool requires molecular pathologists highly trained in preanalytic processes, tumor area microdissection for tumor cell enrichment, methodology analysis and results. The in-depth study of molecular alterations in patients allows optimizing molecular diagnosis and selecting candidates for receive novel treatments against specific molecular targets. These patients are expected to benefit from multidisciplinary approach and learning. The aim of this chapter is to show the implications of molecular pathology in personalized medicine with an actual approach from the methodological limitations of formalin-fixed paraffin embedded (FFPE) tissues and their pre-analytical conditions.

**Keywords:** molecular pathology, personalized medicine, next generation sequencing, NGS, clinical benefit, multidisciplinary approach

#### **1. Introduction**

Personalized medicine is a new approach that allows the identification of patients that can benefit from targeted therapies, since the molecular characteristics of their tumors could be identified. Over the last decade, new drugs have been incorporated into the treatment, including the development of immunotherapy and treatment against specific molecular targets [1]. Thus, patients can receive specific treatments according to the biology of their tumor, turning oncology a tool for personalized medicine. In order to do so, the development of new DNA/RNA sequencing technologies was required, as well as the development of specific antibodies identifying mutated or altered proteins, and the design of new in situ hybridization techniques. The latter has enabled the selection via genetic biomarkers of patients, who can benefit from therapies targeted against specific molecular alterations [2]. Based on the detection of these point molecular alterations, with a clear oncogenic role, treatments have been developed to block the activation of mutated, amplified proteins or product of translocations by specific drugs. The identification of patients with therapeutic molecular targets in their tumors is currently a standard of care. Notwithstanding that, the initial morphological diagnosis and the eventual tumor classification by immunohistochemistry (IHC), as well as the acquisition, handling and processing of tumor tissue play a pivotal role.

In advanced-stage patients, a relatively small amount of tissue obtained at a single procedure must be used most efficiently for all studies [3]. In this sense, consensus exists about making histopathological diagnosis using as little material as possible, which should be kept for molecular studies [4, 5]. The combination of less invasive techniques that provide very small samples to carry out an increasing number of determinations is controversial, since it does not allow to increase the amount of tumor cells. Consequently, more sensitive and specific molecular determinations are required [6].

Although several methods are being developed, such as free tumor DNA detection in peripheral blood, most of these determinations are currently experimental and few are validated for clinical use [7, 8]. Therefore, until more sophisticated techniques for these and other molecular markers are validated, the amount/size of the samples should be considered.

The aim of this chapter is to show the implications of molecular pathology in personalized medicine with an actual approach from the methodological limitations of formalin-fixed paraffin embedded (FFPE) tissues and their pre-analytical conditions.

#### **2. Pre-analytical processes**

In molecular pathology, several variables should be considered for optimal results, and pre-analytical conditions are evaluated.

#### **2.1 Cold ischemia**

One of the crucial phases in tissue management is the period of time immediately after the sample is extracted from the patient until it is placed in a fixation solution (cold ischemia). In an experimental animal model, significant differences in pH values were found between organs at the same cold-ischemia time, and in the same organ at different times. However, no differences were seen in the RNA quality assessed by its integrity number or absorbance ratios [9]. These results reveal a high pH in tissues undergoing ischemia. Firstly, although RNA integrity number (RIN) is a powerful tool to analyze the ribosomal profile and to further infer RNA quality from fresh and frozen tissues (and to compare samples RIN values given the same organism/tissue/extraction method), it is not enough to predict the integrity of mRNA transcripts or to describe the real biological conditions. Secondly, acidic duodenal pH has been reported to alter gene expression in the pancreas of a cystic fibrosis mouse. Upon correction of duodenal pH, either genetically (breeding CFTR-null with gastrin-null mice) or pharmacologically (proton pump inhibitor omeprazole), expression levels of genes measured by quantitative RT-PCR were significantly normalized [10]. Whether alkalosis is secondary to ischemic cell damage, or it may contribute to ischemic cell damage, is yet unknown. Thus, tissue alkalosis in cold-ischemia time may be an underlying mechanism of gene expression changes. Therefore, tissue-pH regulation after organ removal may minimize biological stress in human tissue samples. To date, no consensus exists about the optimal preservation solution. Further optimization of the composition of preservation solutions is required to prolong organ preservation time, and to maximize the yield of successful transplantations by improving the quality and function of organs [11]. Most laboratories have neither control nor record of how long it takes between tissue removal and immersion in the fixer, and its arrival in the laboratory. In addition, most automatic tissue processor machines include a fixation step that further increases the fixation time, which is not often considered.

#### **2.2 Tissue fixation**

Once the tissue has been obtained it should be fixed and 10% Neutral buffered formalin (NBF) fixation is recommended. Pre-fixation in alcoholbased fixative, decalcifying acidic solutions, acidic fixatives (such as Bouin) or those containing metallic salts may alter DNA antigenicity or integrity. Setting a period of more than 6 hours and less than 48 hours is recommended [12]. Short or excessive fixation time may have deleterious effects on DNA and protein antigenic epitopes [13, 14]. The most frequently described effect of formalin in DNA is its fragmentation into small pieces. The use of polymerase chain reaction (PCR) techniques in formalin fixed paraffin embedded (FFPE) tissues is associated with a higher incidence of sequence artifacts and risk of misinterpretation in PCR results, compared with the use of fresh samples [15, 16]. After the inclusion of the tissue in paraffin, the sample remains stable and is preserved against oxidative damage or other degenerative effects. However, in addition to fixation, the type of storage is another documented source that can damage DNA and cause artifacts in the PCR. For a better preservation of DNA, FFEP blocks should be stored below 27°C in humidity-free conditions. Although humidity can affect DNA stability, the acceptable humidity control range is not described. In our experience, up to ten-year-old FFEP blocks have been used. Provided that storage is accurately done and the pre-analytical parameters indicated in this chapter are met, blocks can be preserved up to this time [17]. Since FFPE tissue is currently used for genetic analysis, results should always be carefully interpreted. Mutations detected from FFPE samples by sequencing must be confirmed by independent PCR reaction. Determining the nature and duration of fixation is a great challenge to pathology laboratory, which receives samples from other centers. Therefore, it was suggested that the cold ischemia time, the type and time of tissue fixation should be registered in the pathology report [18].

#### **3. Tumor area microdissection for tumor cell enrichment**

For a molecular analysis, the following data are required: type of biopsy (primary tumor or metastasis), type of block, and percentage of tumor cells needed for each method.

#### **3.1 PCR amplicon size**

As above mentioned, fixation breaks the genetic material into small fragments, and then PCR of FFPE tissue needs a design of specific-sequences primers that flank targets with molecular weight less than 300 bp. Should the designed primers flank a fragmented-amplicon, they fail to perform the enzyme amplification because they need the continuity of the DNA/RNA mold to generate a strand, thus leading to lower sensitivity or false-negative results. Thus, the input for a PCR reaction performed from FFPE tissue requires mandatory quantification with DNA/ RNA calculator spectrophotometer. Thus, each methodology uses a different sample input to obtain the analytical sensitivity (LOD). Every PCR requires a balance between its reaction components, and then the sample input has a direct relationship with the concentration of the primers.

Therefore, somatic mutations, which are generated in tumors and are not present in normal cells, require a minimum percentage for each method.

#### **3.2 Tumor cell enrichment**

Based on the premise that somatic mutations occur, for the most part, in one of the alleles present in human genome, knowing that in humans there are two equal alleles on somatic chromosomes, one of maternal and one of paternal origin, we must understand that if we seek a tumor marker, we must enrich our input in this allele (**Figure 1**).

Sequencing of tumors is now routine and guides personalized cancer therapy. Mutant allele fractions (MAFs, or the 'mutation dose') of a driver gene may reveal the genomic structure of tumors and influence response to targeted therapies [19]. Mutation fraction can be defined as the ratio between mutant and wild-type (wt) alleles in a tumor sample. Allelic fraction is generally applied to a single mutation in a tumor, and is therefore distinct from allelic frequency, which examines the frequency of an allele in a population. To date, however, these terminuses are unfortunately exchanged. Dideoxynucleotide sequencing is a routine method for identifying genetic changes. Since both alleles are amplified in this method, enough input of mutant allele (as compared to the input of normal allele) must be detected. However, this detection requires at least 10–20% of allelic presence. Mutations below this threshold due to normal cells high contamination or tumor heterogeneity could not be detected by this method [20]. Low percentages of neoplastic cells are sometimes associated with unreliable results. Therefore, the percentage of tumor cells must be estimated either through microdissection technique or selection of block interest region [5, 21]. The normal tissue and the lymphocyte infiltration areas must be removed from the tissue for analysis since both are nucleated elements that provide normal DNA. Areas of necrosis should be also removed, since the cell causing necrosis cannot be identified and may be normal or neoplastic. As we know, cell/tumor free DNA drained by biological mechanisms such as secretion, apoptosis and necrosis can be amplified by new generation methods that require smaller chain fragments, this allows us to infer that necrotic cell DNA can be amplified too, considering that an amount of intact nucleic acid chains still present in necrotic masses, unknowing the normal/tumor cell origin. In case microdissection is performed, higher sensitivity is obtained and more chance to detect a tumor specific mutation.

Depending on the method of extraction, hematic areas might be removed. However, they fail to provide normal DNA, because they are anucled cells, but hemoglobin is one of the main polymerase inhibitors in PCR [22]. Regarding the use of clots, a DNA purification method is required to extract hemoglobin. In this sense, specific columns for FFPE tissues are of value. In several cases, Fine Needle Aspiration (FNA) is the first (and often the sole) diagnostic technique, given its low invasiveness, with

**Figure 1.** *Mutant allele fractions (MAFs, or the 'mutation dose').*

#### *Molecular Pathology in the New Age of Personalized Medicine DOI: http://dx.doi.org/10.5772/intechopen.94927*

the clot being all the material available for molecular studies. Here formalin fixation is recommended, and although some reports propose 70% ethanol as an alternative, as above mentioned, DNA antigenicity or integrity may be altered by alcohol-based fixatives [5, 6, 23]. To increase the sensitivity of Sanger sequencing, and to discriminate from technical background, at least 70% of tumor cells are required [5]. The chromatogram obtained failed to discriminate specific signal from background. Such chromatogram type may be determined by pre-analytical conditions (pre-fixing, fixative type or fixation time).

As expected, there was a statistically significant difference between large and small samples DNA concentration. However, no significant differences were observed in concentration, fragments number or tumor initial percentage among different small sample types [18]. We can infer that all these types of tissue samples are similarly useful and depend on interdisciplinary medical team (surgeons, radiologists, clinicians, pathologists and oncologists) [6]. Large samples are blocks from surgical specimens, while small samples could be a core biopsy (yielding tissue samples approximately 1 mm in diameter), biopsies from bronchoscopy, nodal biopsies obtained by mediastinoscopy, and fine needle aspiration resulting in cytological specimens and clots. However, no significant differences were observed in concentration, fragments number or tumor initial percentage among different small sample types. **Figure 2** shows that the amount of tissue obtained from small biopsies is often inadequate for a complete evaluation [18].

Over the last decade, genomic research of various solid tumors has suddenly progressed through the discovery of several molecular biomarkers that eventually impact on the prognosis and treatment of most common cancers. Recent technical innovations, such as "next or second generation" sequencing or "massively parallel" sequencing, have the potential to detect many abnormalities in a single assay, and are probably the solution to tissue shortage [24, 25].

This definitely results into multiple activities for surgeons and pathologists, who must obtain and process samples, write a pathology report, choose the material for molecular biology. In furtherance, those molecular biosciences technicians performing studies must draw up guidelines to standardize these practices, and algorithms to cover cyto- and histopathological diagnoses, IHC and molecular studies [4, 5, 23, 24, 26].

**Figure 2.** *Performance of IHC and molecular study of large and small biopsies.*

#### **4. Methodological analysis and results**

Detection of tumor-derived mutations in FFPE is challenging because the tumor DNA is often scarce, fragmented, at a very low concentration and diluted by the presence of a background of non-mutant DNA (both tumor and non-tumor origin). Once the area of tumor cells is selected to be processed, the method of purification of the macromolecules must be chosen. Although manual non-expensive forms (phenol-chloroform-PK) exist, they fail to provide the necessary amount and quality of DNA. There are affinity columns for DNA, RNA or DNA/RNA together, which can be used on a low scale; and finally automated nucleic acid extraction equipment. Some years ago manual extraction was used for FFPE tissue because the columns were developed only for fresh samples. In the last decade the advent of personalized medicine boosted the development of new methodologies for this purpose. Heydt *et al.* used FFPE tissue samples for the comparison of five automated DNA extraction systems, the BioRobot M48, the QIAcube and the QIAsymphony SP all from Qiagen (Hilden, Germany), the Maxwell 16 from Promega (Mannheim, Germany) and the InnuPure C16 from Analytik Jena (Jena, Germany). The results revealed that the Maxwell 16 from Promega seems to be the superior system for DNA extraction from FFPE material. This study also evaluated DNA quantification systems using the three most common techniques, UV spectrophotometry, fluorescent dye-based quantification, and quantitative PCR. The comparison of quantification methods showed inter-method variations, but all methods could be used to estimate the right amount for PCR amplification and for massively parallel sequencing. DNA extracts were quantified as follows: NanoDrop 2000c spectrophotometer (Thermo Fisher Scientific), Quant-iT dsDNA HS Assay on the Qubit 2.0 fluorometer (Life Technologies), QuantiFluor dsDNA Sample Kit on the QuantiFluor-ST fluorometer (Promega) and Quant-iT Pico-Green dsDNA reagent (Life Technologies) on the LightCycler 480 Instrument (Roche). No difference was observed in mutation analysis based on the results of the quantification methods. These findings emphasize that it is particularly important to choose the most reliable and constant DNA extraction system, especially when using small biopsies and low elution volumes [27]. Once DNA/RNA has been obtained and quantified, analysis requires highly sensitive and specific assays. Different techniques with their own advantages and disadvantages can be used to identify and monitor mutations.

#### **4.1 Real-time qPCR assays**

A real-time PCR or quantitative PCR (qPCR) amplifies, both quantitatively and semi-quantitatively, a targeted DNA molecule during the PCR process. There exist at least two methods for the detection of PCR products: non-specific fluorescent dyes that bind double-stranded DNA molecules by intercalating between the DNA bases. This method is used in qPCR because the fluorescence can be measured at the end of each amplification cycle to determine, either relatively or absolutely, how much DNA has been amplified. The other method is sequence-specific DNA probes consisting of oligonucleotides that are labeled with a fluorescent reporter, which permits detection only after hybridization of the probe with its complementary sequence (TaqMan).

There is also a revolutionary method that uses PlexZyme™ technology. The revolution in this technology is given by a structure called partzyme (A and B). Each partzyme has 3 different regions: (I) the region that joins the target sequence of DNA, (II) the catalytic constituent region, and (III) the region that joins the probe. Once the primers generate the amplicons, both partzymes join their complementary sequences through the region (I), acquiring a characteristic structure thanks to the region (II) that allows the region (III) to be exposed. The fluorescently *Molecular Pathology in the New Age of Personalized Medicine DOI: http://dx.doi.org/10.5772/intechopen.94927*

labeled reporter probe also binds to the partzymes in the region (III) exposed, and once the active catalytic core is formed, the probe is cleaved, producing a signal that is indicative of successful amplification of the target gene. This technology can produce a robust quintuplex with five target assays into a single reaction tube that contained 10 partzymes (5 A and 5 B), 10 primers (5 forwards and 5 revers), and 5 probes, with a 5 different fluorophores. All consumables required for sample preparation and RT-PCR amplification and detection are provided in a single cartridge loaded into the Idylla™ system. Handling time is less than two minutes per sample, with the liquid-tight, disposable cartridges greatly reducing the risk of contamination (Biocartis NV, Belgium).

#### **4.2 ddPCR assay**

In the non-sequencing space, digital PCR (ddPCR), is a highly sensitive and specific technique for the detection of mutations. DNA molecules are split into droplets that form a water oil emulsion. Droplets are like individual test tubes or wells on a plate where a PCR reaction occurs from a DNA template. Each drop is analyzed or read to determine the fraction of positive droplets in the total sample and can accurately and sensitively quantify a mutation. The creation of thousands of drops means that a single sample can generate thousands of data, which are statistically analyzed. For digital PCR the assays are limited to specific single mutations or sets of highly related mutations at the same locus. The analysis of broader genomic regions using ddPCR is not feasible. However, discriminatory multiplex ddPCR assays can be developed, which enable very rapid and cost-effective monitoring for a limited number of mutations in serial plasma samples [28].

#### **4.3 Sanger capillary sequencing**

Sanger sequencing is a DNA sequencing method based on the selective incorporation of chain-terminating dideoxynucleotides by DNA polymerase during in vitro DNA replication [29, 30]. This method was first developed by Frederick Sanger and colleagues in 1977, and became the most widely used sequencing method for over 40 years. However, the Sanger method remains widely used for smaller-scale projects and for validation of NGS results.

#### **4.4 Next-generation sequencing (NGS)**

In this decade, the treatment of cancer patients has evolved with the addition of new massive sequencing technologies. This contributed to the study of tumor biology with an accurate and highly covered diagnostic method that allows the selection of those patients likely to benefit most from target-specific targeted therapies. NGS, massively parallel or deep sequencing, refers to a DNA sequencing technology that has revolutionized genomic research. NGS can be used to sequence the whole human genome within a single day. In contrast, the previous Sanger sequencing technology used to decipher the human genome took over a decade to deliver the final draft [31]. Over the last years, massively parallel sequencing has rapidly evolved and has now transitioned into molecular pathology routine laboratories. This is an interesting platform for the simultaneous analysis of multiple genes with low input material. Therefore, laboratories working with FFPE material and high sample throughput largely require high-quality DNA obtained from automated DNA extraction systems. The spectrum of DNA variation in a human genome comprises small base changes (substitutions), insertions and deletions of DNA, large genomic deletions of exons or whole genes and rearrangements, such as inversions and translocations. Traditional Sanger sequencing focuses on the discovery of substitutions and small insertions and deletions.

There are a number of different NGS platforms using different sequencing technologies, but all these platforms sequence millions of small fragments of DNA in parallel. The aim of bioinformatics analyses is to piece together these fragments by mapping the individual reads to the human reference genome (pipelines). Each of the three billion bases in the human genome is sequenced several times, in order to provide accurate data and an insight into unexpected DNA variation. NGS can be used to sequence either whole genomes or specific genomic areas of interest, including all 22,000 coding genes, the whole-genome sequencing (WGS), the whole exome sequencing (WES). This is a genomic technique for sequencing all of the protein-coding regions of genes in a genome, known as the exome; or small numbers of individual genes (NGS panels).

Parallel sequencing requires target enrichment, which is a pre-sequencing step that only allows part of a whole-genome be sequenced, or regions of interest, without sequencing the entire genome of a sample. The two most commonly used techniques for NGS target enrichment are capture hybridization and amplicon-based (multiplex PCR). In capture hybridization, genomic DNA is cut to produce small fragments that join sequencer-specific adaptors and indexes to prepare the library, and then the sample is hybridized with biotinylated RNA library primers. Target regions are extracted with magnetic streptavidin beads, amplified and sequenced. Capture hybridization is a screening method for large genetic panels and a large DNA input (more than 1 ug DNA), with a laborious and complex workflow, but a better performance. In amplicon sequencing, custom oligo capture probes are designed to flank DNA specific regions without fragmenting. Extension/ligation takes place between hybridized probes. Finally, the uniquely labeled amplicon library is ready for cluster generation and sequencing. The extension/ligation occurs between hybrid probes which determines a uniquely tagged amplicon library ready for cluster generation and sequencing. Amplification sequencing is used for small gene panels or somatic mutation hotspots (target from kb to Mb), with lower DNA input (100 ng). It has a simple and fast protocol (combining sample preparation and enrichment in one assay), but it is more liable to false positive and negative calls. Considering the WGS method in the same fresh and FFPE samples, hybrid capture sequencing showed higher sensitivity compared to amplicon sequencing, while maintaining 100% specificity using Sanger sequencing as a validation method. Amplicon method has higher target rates. Hybridization capture-based approaches demonstrated that many of them could be false positives or negatives [32]. These results reveal advantages and disadvantages of both methods. Therefore, a greater number of trials must be undertaken to demonstrate both clinical usefulness and socioeconomic benefits. On occasions, an extremely sensitive method is not worth using given its clinical implications.

The basic premise of cancer genomics is that cancer is caused by somatically acquired mutations, and is therefore a disease of the genome. Capillary-based cancer sequencing has been ongoing for over a decade. However, these investigations were restricted to relatively few samples and small numbers of candidate genes. Tumor heterogeneity and the addition of new molecular targets have become a challenge that needs a multidisciplinary approach and learning, with the study of the molecular profile of the tumor at the genomic (DNA), transcriptomic (RNA) or protein (protein) level. NGS technique is a useful and novel tool for the study of molecular profile from DNA/RNA. To do the library using amplicon methods it is only necessary to obtain 10 ng of DNA just from the tumor, and 10 ng of RNA, which is feasible, even from small samples, fixed in formalin and included in paraffin [31].

Thus, three of the major technical drawbacks of the massive analysis required for the approach of multiple specific biomarkers for the treatment are resolved. These

drawbacks include the small size of biopsy sample and material scarcity, paraffin fixation of tissues and its effect on DNA/RNA and the impossibility to collect and store fresh material in standard clinical practice. Therefore, this type of studies is necessary to optimize the quality of patient care, avoiding errors and false positives or negatives. Thus, the use of NGS panels with small and overlapping amplicons would solve all these drawbacks, always associated with a bioinformatics algorithm (pipeline) that allows the overlap of the fragments obtained with a reference sequence.

#### **5. In-depth study of molecular alterations**

The prevalence of molecular alterations with targeted treatment may vary according to different variables, such as the region of the world, race and gender [33, 34]. About 86% of tumors have molecular alterations that can potentially be treatable with approved or developing drugs, of which approximately 30% have clinically available drugs. The distribution of these alterations in patients with metastatic disease varies compared to those observed in resected tumors at earlier stages [35].

Different analysis options may be combined according to the molecular target to be identified, the type of molecular alteration and the type of sample required. Regarding the KRAS gene, a GTPase which functions as an upstream regulator of the MAPK and PI3K pathways, it is frequently mutated in various cancer types including pancreatic, colorectal and lung cancers [36].

KRAS was one of the first markers to be used as a therapeutic target in colorectal cancer (CRC) in clinical practice since the approval of cetuximad in the second line in 2008. Both the European Medical Agency (EMEA) and the Food and Drug Administration (FDA) in 2008 approved the use of anti-EGFR monoclonal antibodies in patients with tumors with non-mutated KRAS (KRAS-wt). The selection of patients for anti-EGFR treatment based on the mutational status of codons 12 and 13 of the KRAS gene is highly specific to non-responder patients. At that time, the tissue was not macro-dissected, biopsies containing more than 70% of tumor cells were processed by sequencing for the reasons mentioned above, and approximately 30% of cases could not be evaluated since they failed to meet these criteria. Codon 12 and 13 of exon 2 of the KRAS gene were studied and the type of mutation found was irrelevant. For exon 2, 40% of the CRC patients were mutated and 60% were wt (codon 12 and 13). Results showed that 95% of patients with mutated CRC for KRAS did not benefit from anti-EGFR treatment. However, it was not sensitive enough because only half of patients with KRAS-wt tumors responded to treatment [37]. Then, the 59 and 61 codons of exon 3 and the 117 and 146 codons of exon 4 were eventually added. Automated qPCR methods were developed, which covered these hot-spots and dually reported wt or mutated. Nowadays, these binomial methods (wt/mutated) would not serve to identify the G12C amino acid change (c.34G > T p.Gly12Cys). Target therapies like KRAS G12C covalent inhibitors, such as AMG-510, are currently in early phase clinical trials and show promising results for the treatment of KRAS G12 mutant lung cancer patients. However, KRAS G12C colorectal cancer patients have not shown the same response. KRAS mutation testing was carried out using 13 technologies and assays. Limits of detection (LD) of the 13 methods were showed in the following table. Of 13 assays evaluated in this work, 9 showed relatively similar levels of accuracy and reliability in detecting KRAS mutations at low levels with varying sensitivities (50 copies mutant allele frequency by each technology). The best performances were obtained by three assays: Oncomine Focus Assay, Idylla KRAS Mutation Test and UltraSEEK, with high sensitivity and specificity across the entire cell line panel. The worst performances in detection were Illumina Nextera Rapid Capture Custom Lung Panel and Sanger capillary sequencing [38].


#### *Molecular Pathology in the New Age of Personalized Medicine DOI: http://dx.doi.org/10.5772/intechopen.94927*

The NGS study may infer biological mechanisms that may explain primary resistance (absence of response to tyrosine kinase inhibitors and disease progression as a better response). This information is required for decision-making of the allelic frequency data for DNA sequence variants, amplified reads for fusions, or the number of copies of amplified genes, since in order to determine that a sequence variant has a clear oncogenic role in the tumor, its representative presence is required. One of the most common false positives with NGS, partly due to its high sensitivity, is the amplification and sequencing of variants from clonal hematopoiesis. Obtaining DNA from FFPE is a methodology used for more than decades, with satisfactory results, since the DNA obtained was degraded by fixation-paraffinization process, as well as its opposite effect which is the deparafinization of tissue. Obtaining RNA from this type of sample is most controversial given its increased lability, and was recently accepted due to the incorporation of new purification strategies. Therefore, obtaining RNA from FFPE was the greatest difficulty of this DNA/ RNA NGS method, and required this minimum learning curve to achieve optimal 80% performance (**Figure 3**). The effectiveness of RNA isolation was calculated, taking into account criterion >5000 reads as evaluable sample, for each run/chip. Increased performance was achieved as the long runs occurred. The initial yield was less than 50%, reaching 80% maximum, because the fixation of the tissue as well as the deparanization process are counterproductive effects for obtaining RNA. Pre-analytical pathological processes for NGS take a crucial role.

This has been especially relevant in RNA sequencing from paraffin block. A learning curve is required before using this methodology in the clinical field. The acquisition of macromolecules management is critical. On the other hand, multidisciplinary work is crucial for the correct interpretation of the information provided by these new technologies. Crude data alone, without associated bioinformatics information, should not be used for the treatment of patients. The main pitfall of NGS in the clinical setting is the infrastructure, such as computer capacity and

**Figure 3.** *RNA performance (1= > 5000 reads, 0 < 5000 reads).*

storage, and personnel trained in comprehensive analyses and interpretation of the subsequent data. In addition, and in order to obtain clinically relevant information in a clear and robust interface, the volume of data needs to be proficiently managed. However, to make NGS cost effective one would have to run large batches of samples which may require supra-regional centralization. The objective of implementing new technologies is to develop personalized treatment strategies that result in prolongation of survival of patients with a better quality of life.

#### **6. Conclusion**

The analysis of the biology of tumors, using NGS, allows to expand the number of molecular alterations to be studied, and allows to detect more patients who can benefit from targeted treatments, modifying the survival in patients with detected and treated molecular alterations. A continuous and inexorable shift in surgical pathology can be observed, with histological diagnosis being just one of its components. The molecular profile is nowadays an essential tool for anatomic pathology practice, which invariably requires highly trained specialists. The in-depth study of molecular alterations in patients allows optimizing molecular diagnosis and selecting patients to receive novel treatments, targeted against specific molecular targets for the clinical benefit of patients, through a multidisciplinary approach and learning.

#### **Acknowledgements**

The author thanks Dr. Boris Elsner & Dr. Alejandra Avagnina for their mentorship; Dr. Alejo Garcia**†** for being an excellent scientific partner; and Valeria Melia for proofreading of the manuscript.

#### **Conflict of interest**

The author declares no conflict of interest.

#### **Author details**

Valeria Cecilia Denninghoff CEMIC-CONICET Interacting Units, (CEMIC: Center for Medical Education and Clinical Research "Norberto Quirno" – CONICET: National Scientific and Technical Research Council) and Buenos Aires University (UBA), Ciudad Autónoma de Buenos Aires, Argentina

\*Address all correspondence to: vcdennin@gmail.com

© 2021 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

*Molecular Pathology in the New Age of Personalized Medicine DOI: http://dx.doi.org/10.5772/intechopen.94927*

#### **References**

[1] Carlisle JW, Ramalingam SS: A banner year for immunotherapy and targeted therapy. Nat Rev Clin Oncol 2019;16:79-80. DOI: 10.1038/ s41571-018-0138-4

[2] Camidge DR, Doebele RC, Kerr KM: Comparing and contrasting predictive biomarkers for immunotherapy and targeted therapy of NSCLC. Nat Rev Clin Oncol 2019;2016:341-335. DOI: 10.1038/s41571-019-0173-9

[3] Tsao MS, Sakurada A, Cutz JC, Zhu CQ, Kamel-Reid S, Squire J, Lorimer I, Tong Zhang, Ni Liu, Daneshmand M, Marrano P, da Cunha Santos G, Lagarde A, Richardson F, Seymour L, Whitehead M, Ding K, Pater J, Shepherd FA: Erlotinib in Lung Cancer - Molecular and Clinical Predictors of Outcome. N Engl J Med 2005;353:133- 144. DOI: 10.1056/NEJMoa050736

[4] Pirker R, Herth F, Kerr K, Filipits M, Taron M, Gandara D, Hirsch FR, Grunenwald D, Popper H, Smit E, Dietel M, Marchetti A, Manegold C, Schimacher P, Thomas M, Rosell R, Capuzzo F, Stahel R, European EGFR Workshop Group: Consensus for EGFR Mutation Testing in Nonsmall Cell Lung Cancer. Results from a European Workshop. J Thorac Oncol 2010;5:1706-1713. DOI: 10.1097/ JTO.0b013e3181f1c8de

[5] Lindeman N, Cagle P, Beasley M, Chitale DA, Dacic S, Giaccone G, Jenkins RB, Kwiatowski DJ, Saldivas JS, Squire J, Thunnissen E, Ladanyi M: Molecular testing guideline for selection of lung cancer patients for EGFR and ALK tyrosine kinase inhibitors: guideline from the College of American Pathologists, International Association for the Study of Lung Cancer, and Association for Molecular Pathology. J Thorac Oncol 2013; 8: 823-859. DOI: 10.1097/JTO.0b013e318290868f

[6] Ellis P, Blais N, Soulieres D,Ionescu DN, Kashyap M, Liu G, Melosky B, Reiman T, Romeo P, Shepherd FA, Tsao MS, Leigh NB: Systematic Review and Canadian Consensus. Recommendations on the Use of Biomarkers in the Treatment of Nonsmall Cell Lung Cancer. J Thorac Oncol 2011;6:1379-1391. DOI: 10.1097/ JTO.0b013e318220cb8e

[7] Aparicio S, Caldas C: The Implications of Clonal Genome Evolution for Cancer Medicine. N Engl J Med 2013; 368:842-851. DOI: 10.1056/ NEJMra1204892

[8] Diaz L, Bardelli A: Liquid Biopsies: Genotyping Circulating Tumor DNA. J Clin Oncol 2014;32:579-586. DOI: 10.1200/JCO.2012.45.2011

[9] Denninghoff V, Olivieri EHR, Fresno C, Uceda A, Mota L, Suenega AP, Carraro DM, Martins VR, Avagnina A, Soarres FA, Fróes Marques Campos AHJ: Tissue alkalosis in cold-ischemia time. Sci Rep 2017;7:10867. DOI: 10.1038/ s41598-017-11284-z

[10] Kaur S, Norkina O, Ziemer D, Samuelson LC, De Lisle RC: Acidic duodenal pH alters gene expression in the cystic fibrosis mouse pancreas. AJP Gastrointestinal and Liver Physiology;287:G480-490. DOI: 10.1152/ajpgi.00035.2004

[11] Petrenko A, Carnevale M, Somova A, Osorio J, Rodríguez J, Guibert E, Fuller B, Froghi F: Organ Preservation into the 2020s: The Era of Dynamic Intervention. Transfus Med Hemother 2019;46:151-172. DOI: 10.1159/000499610

[12] Wolff AC, Hammond MEH, Schwartz JN, Hagerty KL, Allred DC, Cote RJ, Dowsett M, Ftzgibbons PL, Hanna WM, Lamger A, McShane LM, Paik S, Pegram MD, Perez EA, Press MF, Rhodes A, Sturgeon C, Taube SE, Tubbs R, Vance GH, van de Vijver M, Wheeler TM, Hayes DF, American Society of Clinical Oncology/College of American Pathologists: American Society of Clinical Oncology/College of American Pathologists guideline recommendations for human epidermal growth factor receptor 2 testing in breast cancer. Arch Pathol Lab Med 2007; 131:18-43. DOI: 10.1043/1543-2165(2007)131[18:ASOCC O]2.0.CO;2

[13] Eberhard D, Giaccone G, Johnson B, Non-Small-Cell Lung Cancer Working Group: Non-Small-Cell Lung Cancer Working Group. Biomarkers of response to epidermal growth factor receptor inhibitors in Non-Small-Cell Lung Cancer Working Group: standardization for use in the clinical trial setting. J Clin Oncol 2008;26:983-994. DOI: 10.1200/ JCO.2007.12.9858

[14] Greer C, Lund J, Manos M: PCR amplification from paraffin embedded tissues: recommendations on fixatives for longterm storage and prospective studies. PCR Methods Appl 1991,1:46- 50. DOI: 10.1101/gr.1.1.46

[15] Williams C, Ponten F, Moberg C, Söderkvist P, Uhlén M, Pontén J, Sitbon G, J Lundeberg J: A high frequency of sequence alterations is due to formalin fixation of archival specimens. Am J Pathol 1999,155:1467-1471. DOI: 10.1016/ S0002-9440(10)65461-2

[16] Akbari M, Hansen M, Halgunset J, Skorpen F, Krokan HE: Low copy number DNA template can render polymerase chain reaction error prone in a sequence-dependent manner. J Mol Diagn 2005;7:36-39. DOI: 10.1016/ s1525-1578(10)60006-2

[17] Kapila SN, Boaz K, Natarajan S: The post-analytical phase of histopathology practice: Storage, retention and use of human tissue specimens. Int J Appl Basic Med Res 2016; 6:3-7. doi: 10.4103/2229-516X.173982

[18] García A, Cuello MT, dos Santos M, Elsner B, Avagnina A, Denninghoff V: Performance Assessment of Epidermal Growth Factor Receptor Gene Sequencing According to Sample Size in Daily Practice Conditions. Appl Immunohistochem Mol Morphol 2018;26:495-500. DOI: 10.1097/ PAI.0000000000000455

[19] Dienstmann R, Elez E, Argiles G, Matos I, Sanz-Garcia E, Ortiz C, Macarulla T, Capdevila J, Alsina M, Sauri T, Verdaguer H, Vilaro M, Ruiz-Pace F, Viaplana C, Garcia A, Landolfi S, Palmer HG, Nuciforo P, Rodon J, Vivancos A, Tabernero J: Analysis of mutant allele fractions in driver genes in colorectal cancer – biological and clinical insights. Mol Oncol 2017; 11: 1263-1272. DOI: 10.1002/1878-0261.12099

[20] Do H, Dobrovic A: Limited copy number-high resolution melting (LCN-HRM) enables the detection and identification by sequencing of low level mutations in cancer biopsies. Mol Cancer 2009;8:82. DOI: 10.1186/1476-4598-8-82

[21] García-Foncillas J, Garrido P, Gómez J, Palacios J, Tarón M: Recomendaciones para la determinación de las mutaciones del gen EGFR en el carcinoma de pulmón no microcítico. Rev Esp Patol 2011;44:17-31. DOI: 10.1016/j.patol.2011.02.003

[22] Al-Soud WA, Radström P: Purification and characterization of PCR-inhibitory components in blood cells. J Clin Microbiol 2001;39:485-493. DOI: 10.1128/JCM.39.2.485-493.2001

[23] Kerr KM, Wistuba I, Yatabe Y. Sample Acquisition, Processing and General Diagnostic Procedures. In: Tsao MS, Hirsch FR, Yatabe Y, editors. IASLC Atlas of ALK Testing in Lung Cancer. International Association for the Study of Lung Cancer. IASLC Press: Colorado USA; 2013. p. 13-16.

*Molecular Pathology in the New Age of Personalized Medicine DOI: http://dx.doi.org/10.5772/intechopen.94927*

[24] Jung CY: Biopsy and Mutation Detection Strategies in Non-Small Cell Lung Cancer. Tuberc Respir Dis 2013;75:181-187. DOI: 10.4046/ trd.2013.75.5.181

[25] de Biase D, Visani M, Malapelle U, Simonato F, Cesari V, Bellevicine C, Pession A, Troncone G, Fassina A, Tallini G: Next-Generation Sequencing of Lung Cancer EGFR Exons 18-21 Allows Effective Molecular Diagnosis of Small Routine Samples (Cytology and Biopsy). Plos One 2013;8:e83607. DOI: 10.1371/journal. pone.0083607

[26] Kerr KM, Bubendorf L, Edelman MJ, Marchetti A, Mok T, Novello S, O'Byrne K, Stahel R, Peters S, Felip E, Panel Members: Second ESMO consensus conference on lung cancer: pathology and molecular biomarkers for non-small-cell lung cancer. Ann Oncol. 2014;25:1681-1690. DOI: 10.1093/ annonc/mdu145

[27] Heydt C, Fassunke J, Künstlinger H, Ihle MA, König K, Heukamp LC, Schildhaus HU, Odenthal M, Büttner R, Merkelbach-Bruse S: Comparison of pre-analytical FFPE sample preparation methods and their impact on massively parallel sequencing in routine diagnostics. PLoS One 2014;9:e104566. DOI: 10.1371/journal.pone.0104566

[28] Rowlands V, Rutkowski AJ, Meuser E, Carr TH, Harrington EA, Barrett JC: Optimisation of robust singleplex and multiplex droplet digital PCR assays for high confidence mutation detection in circulating tumour DNA. Scientific Reports 2019;9:12620. DOI: 10.1038/ s41598-019-49043-x

[29] Sanger F; Coulson AR (May 1975). A rapid method for determining sequences in DNA by primed synthesis with DNA polymerase. J Mol Biol 1975;94:441-448. DOI:10.1016/0022-2836(75)90213-2

[30] Sanger F; Nicklen S; Coulson AR: DNA sequencing with chainterminating inhibitors. PNAS 1977;74:5463-5467. DOI:10.1073/ pnas.74.12.5463

[31] Behjati S, Tarpey PS: What is next generation sequencing? Arch Dis Child Educ Pract Ed 2013;98:236-238. DOI: 10.1136/archdischild-2013-304340

[32] Hung SS, Maissner B, Chavez EA, Ben-Neriah S, Ennishi D, Jones MR, Shulha HP, Chun-Chan F, Boyle M, Kridel R, Gascoyne RD, Mungdall AL, Marra MA, Scott DW, Connors JM, Steidl C: Assessment of Capture and Amplicon-Based Approaches for the Development of a Targeted Next-Generation Sequencing Pipeline to Personalize Lymphoma Management. J Mol Diagn 2018;20:203-214. DOI: 10.1016/j.jmoldx.2017.11.010

[33] Sholl LM, Aisner DL, Varella-Garcia M, Berry LD, Dias-Santagata D, Wistuba II, Chen H, Fujimoto J, Kugler K, Franklin WA, Iafrate J, Ladanyi M, Kris MG, Johnson BE, Bunn PA, Minna JD, Kwiatkowski DJ, and on behalf of the LCMC Investigators. Multi-institutional Oncogenic Driver Mutation Analysis in Lung Adenocarcinoma: The Lung Cancer Mutation Consortium experience. J Thorac Oncol 2015;10:768-777. DOI: 10.1097/JTO.0000000000000516

[34] Kohno T, Nakaoku T, Tsuta K, Tsuchihara K, Matsumoto S, Yoh K, Goto K: Beyond ALK-RET, ROS1 and other oncogene fusions in lung cancer. Transl Lung Cancer Res 2015;4:156-164. DOI: 10.3978/j.issn.2218-6751.2014.11.11

[35] Devarakonda S, Morgensztern D, Govindan R. Genomic alterations in lung adenocarcinoma. Lancet Oncol 2015;16:e342-351. DOI: 10.1016/ S1470-2045(15)00077-7

[36] The AACR Project GENIE Consortium: AACR Project GENIE: powering precision medicine through an international consortium.Cancer Discov 2017;7:818-831. DOI: 10.1158/2159-8290. CD-17-0151

[37] Perazzo F, Piaggio F, Krupitzki H, García A, Avagnina A, Elsner B, Denninghoff V: Clinical-pathological features and gene profile in colorectal cancer Medicina (B Aires) 2013;73:417-422.

[38] Sherwood JL, Brown H, Rettino A, Schreieck A, Clark G, Claes B, Agrawal B, Chaston R, Kong BSG, Choppa P, Nygren AOH, Deras IL, Kohlmann A: Key differences between 13 KRAS mutation detection technologies and their relevance for clinical practice. ESMO Open 2017; 2:e000235. DOI: 10.1136/ esmoopen-2017-000235

## Section 3 Digital Pathology

#### **Chapter 5**

## Fast Regions-of-Interest Detection in Whole Slide Histopathology Images

*Junzhou Huang and Ruoyu Li*

#### **Abstract**

Detecting and localizing pathological region of interest (ROI) over whole slide pathological image (WSI) is a challenging problem. To reduce computational complexity, we introduced a two-stage superpixel-based ROI detection approach. To efficiently construct superpixels with fine details preserved, we utilized a novel superpixel clustering algorithm which cluster blocks of pixel in a hierarchical fashion. The major reduction of complexity is attributed to the combination of boundary update and coarse-to-fine refinement in superpixel clustering. The former maintains the accuracy of segmentation, meanwhile, avoids most of unnecessary revisit to the 'non-boundary' pixels. The latter reduces the complexity by faster localizing those boundary blocks. Detector of RoI was trained using handcrafted features extracted from super-pixels of labeled WSIs. Extensive experiments indicates that the introduced superpixel clustering algorithm showed lifted accuracy on lung cancer WSI detection at much less cost, compared to other classic superpixel clustering approaches. Moreover, the clustered superpixels do not only facilitate a fast detection, also deliver a boundary-preserving segmentation of ROI in whole slide images.

**Keywords:** region of interest, whole slide histopathology images, superpixel, segmentation, detection, unsupervised learning

#### **1. Introduction**

At our age, many hazardous infectious diseases, e.g. bird flu, and many different kind of cancers, e.g. lung cancer, are still the top threats to our personal health and the public sanitation as well. Automatic searching and localizing Regions of Interest (ROIs) on histopathological images is a crucial intermediate step between largescale images acquisition and the computer-aided automated diagnosis that we pursue. As the fast development of deep learning techniques and the introduction of neural network models, e.g. convolutional neural networks (CNNs), to medical image understanding area, we are finally able to extend the boundary of modern medical image saliency detection, classification and segmentation [1–3]. Whole Slide Images (WSIs) are the digitized histopathology images taken over an entire slide of tissue, which retrains as much intact pathological information as possible. Therefore, a typical WSI, that usually has resolution at scale of 10<sup>6</sup> <sup>10</sup>6, is 1.5 2.0 Gigabyte large on disk, which is thousands times larger than those images from deep learning benchmark datasets, like MNIST [4] and CIFAR [5].

Therefore, traditional fully convolutional networks, used to work perfectly for medical image segmentation [6], are no longer applicable, because of the parameter scale that may explode and the rising risk of under-fitting along with lack of labeled WSIs for training. We need a brand-new cost-efficient solution designed especially for WSIs to handle such magnificent scale of data without losing too much performance. As far as we know, there are no existing convolutional neural networks who claim themselves to directly work on raw images at WSI scale without any downsampling or patching. The most popular walk-around for extracting features from WSIs is to first sample a bag of patches over WSIs and then train and execute inference on patches respectively. Then, aggregating the prediction from patch level to WSI level is to give final model output. Patch-based network [2] successfully handled classification task on WSIs, [7] enabled survival time inference purely based on tumor tissue WSIs. Although, these models applied to WSIs successfully saved most of computational cost by patching, they also dumped lots of taskrelevant information hidden in those patches not being sampled. Besides, losing topological spatial information of patches after being sampled from WSI makes predictor treat patches equally, which is obviously not the optimal strategy.

Considering the practical clinic scenarios for image detection and segmentation techniques applied to CT [6] and MRI [8] and the associated pathophysiological procedures, we summarized some challenging but necessary technical requirements for any ROI detection and segmentation solutions for WSIs:


Regions of interest (ROI) could have different definition according to particular scenarios. In this article, we name ROIs as the local regions filled with tumor cell cluster or other cancer-related cells such as lymphocyte. In past related works, ROI detection and segmentation are usually treated separately as two different tasks. The former is to quickly search and localize any suspicious regions on image according to predefined patterns. The output of this task may not have to be finedetailed at pixel level, due to computational efficiency concern, and sometimes a bounding box that surrounds, at least partly, the ground-truth ROI is enough satisfactory. While, the latter task is to give a pixel-accurate contour of each detected ROIs, which is significantly more expensive. In fact, detection and segmentation are not strictly isolated, and on the opposite, the two tasks could be combined as one under some circumstances. Many CNNs based image segmentation models are indeed end-to-end solutions directly extract and learn hierarchical feature pyramid from raw channels of images to execute pixel-clustering at different level of granularity. Semantic segmentation network [9] is to obtain object detection and segmentation in single forward-pass of network. The advantages of applying deep neural network is from treating the feature design work as an optimization problem, and therefore CNNs are able to discover hidden representations that better serve prediction tasks than handcrafted descriptors, who are either overlocalized or not robust. The requirement on high recall rules out patch-based WSI solutions. And patch-based methods obviously cannot handle segmentation of entire WSIs. However, in order to directly work on WSI input, the networks either

#### *Fast Regions-of-Interest Detection in Whole Slide Histopathology Images DOI: http://dx.doi.org/10.5772/intechopen.94238*

make the receptive field of convolutional operators large enough to cover any potential region of interest, or stack more layers with relatively small kernel to aggregate local features from entire ROI to form its high-level representations for classification. No matter what architecture is chosen, the total count of parameter in WSI segmentation network is going to be magnificent. Due to the expense of having quality annotation of all ROIs (i.e. tumor cells) on WSIs, the annotated groundtruths of segmentation for training models is quite constrained and very likely not enough to train a wide and deep network as described above that directly works on raw WSIs.

To work around this difficulty, in the method to be introduced, we first chose to still rely on handcrafted features as descriptors of patches to save massive feature aggregation calculations in CNNs, and in the meanwhile we also utilized the hierarchical pyramid structures appear between feature maps of consecutive convolutional layers of CNNs. While, different from what happened in CNNs, in the pyramid of introduced multi-level iterative method, feature vectors of descriptor are not changed along with level, because we did not have gradients back-propagated from loss to update feature formations, while the spatial segmentation did get updated at different granularity of patching. Without having ground-truth of segmentation of ROIs, we introduced superpixel clustering as an unsupervised way to learn spatial segmentation of image, since we do not have gradient to update the assignment of segmentations as well. At different level granularity, we divide the entire WSI into patches of different scale, then the introduced superpixel clustering method [10] is going to cluster patches based on several handcrafted local textual descriptors, preserving both topological consistency and appearance similarity. After superpixel constructed, we run a pre-trained classifier, e.g. SVM or CRF, to classify superpixels represented by the averaged descriptors of patches. Averaging of patch descriptors is to avoid additional difficulty of training a classifier for superpixels of different size and varying shape. This is also the biggest challenge for building an end-to-end fully convolutional network fed with clustered superpixels, since the shape of input tensor to any neural network cannot be undefined.

The main contributions of article is to decouple and reformulate ROI detection and semantic segmentation, that requires dense annotation, into an iterative execution of unsupervised superpixel clustering and classification at coarse-to-fine level of patching granularity. This semi-supervised approach largely replies on quality of superpixel clustering. To obtain better fine-detailed superpixels, we introduced a novel topology-preserved superpixel clustering algorithm to this problem. Besides, the approach introduced is also dependent on accurate classification of superpixels, especially at coarser levels, because any mistaken classification of coarse superpixel cannot be compensated in fine-grained superpixel refinement at next level of granularity. The recall of ROIs will benefit from the increased classification accuracy. Therefore, we trained compact but robust classifier, e.g. SVM, with minimal data requirement. On the other hand, without fine-tune, an improved segmentation of superpixels will automatically boost accuracy of a pre-trained classifier.

#### **2. Related work**

Superpixel is a common replacement of pixel with purposes more than saving computational cost. It clusters nearby pixels of similar attributes together as fundamental operational unit in downstream tasks, e.g. object detection, segmentation and even real-time tracking. In this session we introduced the state-of-the-art superpixel clustering algorithms and the combination of superpixel with deep neural networks (DNNs) in medical image understanding.

#### **2.1 Superpixel clustering**

One important feature of superpixel construction is that this is a pure unsupervised approach in which there are no annotated ground-truths in any format for guiding the label assignment on pixels. The pixels are clustered purely based on the attributes, such as appearance and physical location, etc. SLIC [11] is an iterative K-mean superpixel clustering that walk through all pixels. It is able to generate almost equally-sized superpixels with outstanding boundary adherence. And the time complexity could be further reduced by limiting search space to a small nearby area. While, iterating over entire pixels is still too expensive, stopping SLIC from being applied on large images like WSIs. If compromise part of accuracy, SEEDS [12], that started from randomly initialized superpixel partitions, focused on updating boundary pixel allocation only and proposed a fast energy function to evaluate each adaption of pixel label assignment by enforcing color homogeneity. Linear spectral clustering, a.k.a. LSC, combined normalized cut and K-mean clustering after discovering optimizing these two objective functions are in fact equivalent on the condition that defines similar function as inner product of feature vectors [13]. LSC also achieved satisfactory boundary adherence and color consistency within segmented superpixels with Oð Þ *N* complexity, where *N* is the pixel number. Compared to SLIC, LSC saved computations from pre-allocation of pixel to large regions by eigenvector-based normalized cuts. And different from the twostage Ncuts [14], LSC accomplished Ncuts and K-mean in one-stage. Similar to LSC, the computational complexity of SEEDS and SLIC is also approximated as Oð Þ *N* . Therefore, within visual comparison in **Figure 1**, we did not include expensive solutions such as ERS [15] with <sup>O</sup> *<sup>N</sup>*<sup>2</sup> log *<sup>N</sup>* and EneOpt0 [16] with <sup>O</sup> *<sup>N</sup>*<sup>3</sup> complexity. Because we only consider those approaches who are potentially feasible for segmenting whole slide images.

#### **Figure 1.**

*Example of superpixel clustering on image with three classic solutions: LSC [13] (left), SEEDS [12] (middle) and SLIC [11] (right). The upper row is the edges of superpixels displayed on image. The middle row is the contours of superpixels. The bottom row is the segmentation mask filled with different color on different superpixel.*

#### **2.2 ROI and superpixel**

Regions of interest (ROI) in histopathology whole slide images (WSIs) are usually those disease-related cells or the tissues of specific patterns, but they do not have descriptive definitions to form a category of objects. Due to the magnificent scale of WSI, the major challenge would be the scalability and the memory efficiency of algorithms. Bejnordi et al. [17] relied on cheap segmentation of superpixels on downsampled WSIs to filter out those regions irrelevant to ROIs. However, it did not correctly notice the inevitable influence of wrong classification of coarse superpixels, because the algorithm completely ruled out those regions from later more accurate segmentation and classification. Besides, the classifiers had to be trained multiple times with patches extracted from the superpixels of different magnification to work on different levels of granularity. Litjens et al. [18] reduced the workload of labeling and grading by two ways: by excluding the areas of definitely normal tissues within a single specimen or by excluding entire specimens which do not contain any tumor cells. Litjens et al. [18] presented a multiresolution cancer detection algorithm to boost the latter. While it also suffered from the loss of recall as [17]. Another superpixel automated segmentation method is [19], which trained a classifier to predict where mitochondrial boundaries occurs using diverse cues from superpixel graph. While, because the selected superpixel clustering approach [11] did not offer satisfactory boundary adherence, the classifier encumber the overall detection performance. As summary, in order to accomplish a quick detection and segmentation of ROIs in WSI, a combination of superpixel clustering and pre-trained classifier seems a popular choice, while the performance bottleneck was the tradeoff between the efficiency and the quality of superpixel clustering, which directly determined classifier accuracy.

To reduce the intense computational cost in superpixel clustering, the algorithm to be introduced creatively combined the coarse-to-fine scheme [20] and the boundary-only update strategy proposed in SPSS [21]. In our method, clustering manipulated the rectangular blocks of pixel as basic unit and a coarse segmentation of superpixel would be constructed before a more fine-detailed refinement got executed. on each level of construction, only boundary blocks or their nearby neighbors got chance of label update. **Figure 2** illustrated the procedures of introduced superpixel clustering. Furthermore, the introduced boundary-only update strategy on next level would emphasize on differentiating foreground and background blocks, considering the boundaries between superpixels within ROIs are less important. The improvement brought by our algorithm on ROI detection accuracy has been proved and verified in [10, 22], where the method had quantitatively verify the improvement of the accuracy of ROI detection in histopathology images,

#### **Figure 2.**

*An example of the coarse-to-fine/boundary-only update based superpixel segmentation algorithm first presented in [10]. The basic manipulation unit is the rectangular block instead of pixels during each stage. We start from a coarse segmentation and end with pixel-level refinement on superpixel boundary. The block size is respectively* 10 10*,* 2 2*,* 1 1 *(single pixel) from left to right.*

e.g. lung cancer H&E-stained WSIs. **Figure 3** shows comparison of classic superpixel methods [11–13] on cancer patients WSI.

#### **2.3 DNNs on superpixel**

As success of deep neural networks in computer vision, many works have extended application of DNNs onto superpixel. Gadde et al. [23] introduced a bilateral inceptions module to accelerate convergence of CNNs with superpixel as network input for semantic segmentation. Kwak et al. [24] treated superpixels as "pooling" layer in neural network, but preserving low-level structures. Therefore, their framework trained semantic segmentation network without pixel-level ground-truth. To construct superpixels for small objects of complicated boundaries, [25] introduced a superpixel segmentation based on pixel features trained with affinity loss and segmentation error. In medical images domain, superpixels are also utilized as a topology-preserving simplification of data for deep network. The organ segmentation network in [26] worked on the descriptors extracted from superpixels clustered in CT images. And then CNN simply did a pixel-wise refinement based on the coarse segmentation given by superpixel. Different from previous works who simply utilized superpixels as reduction of image primitives, [27] proposed an end-to-end" Superpixel Sampling Network" (SSN) which contains differentiable superpixel construction together with learning a task-specific prediction.

The rest of article is organized as following: we first introduce the multiresolution fast superpixel clustering with coarse-to-fine and boundary-only strategy to increase efficiency. Both mathematical explanation and illustrative examples will be given in Section 3. Then we elaborate the numerical results on classification accuracy and visual comparison of superpixels with classic methods on TCGA WSI dataset in Session 4. Lastly, conclusion and future work will be given in Session 5.

#### **Figure 3.**

*Example of pathological whole slide image with ROI annotations and the superpixels generated by three classic solutions of linear complexity: (1) LSC [13], (2) SEEDS [12] and (3) SLIC [11].*

#### **3. Methodology**

The detection framework introduced is not only going to propose bounding box to surround ROIs, but also is going to offer fine-detailed, boundary-adherent superpixel segmentation of them. On the other hand, an improved superpixel construction contributes the differentiation of ROI from background as well. Therefore, the proposed approach comprised two components: fine-detailed superpixel segmentation and superpixel classification. For reduction of computational expense, we chose not to accomplish superpixel segmentation at finest level in one shot. For instead, we first obtain a coarse superpixel segmentation from clustering big pixel blocks (e.g. 500 � 500). A pre-trained binary classifier then predicts label (ROI v.s. background) of superpixels. Afterwards, those superpixels labeled as ROI along with their neighbors will move to next round of segmentation at finer resolution. The process will be repeated until quality becomes satisfactory. Different from previous superpixel clustering methods [11, 21], the introduced algorithm gave topology-preserving superpixels. A better detection recall is expected as well, since our method did not completely rule out negatively labeled superpixel at coarse stage as [17, 18], and for instead we include negative neighbor superpixels to next level of segmentation.

#### **3.1 Superpixel clustering and detection**

#### *3.1.1 Energy function*

Think of superpixels of flexible number of blocks **S** ¼ f g *s*0, ⋯, *sK*�<sup>1</sup> , and the blocks belong to superpixel **S***<sup>k</sup>* as f g *b*0, ⋯, *bM*�<sup>1</sup> , we devised two representations of block: appearance and position. Appearance representation of block is the averaged RGB color over pixels in block as **C**. Position representation of block is the relative position coordinates at center point of block as **P**. At superpixel level, Θ ¼ ð Þ *θ*0, ⋯, *θ<sup>K</sup>*�<sup>1</sup> and Ξ ¼ *ξ*0, ⋯, *ξ<sup>K</sup>*�<sup>1</sup> ð Þ are the center positions and the mean color vectors of superpixels. The objective function to be minimized consists of a series of energy functions and penalty terms. For appearance, total variance of three color channels are color energy function of superpixel **S***<sup>k</sup>* defined as:

$$E\_{col}(\mathbf{S}\_k) = \sum\_{q=0}^{2} \frac{\mathbf{1}}{||\mathbf{s}||} \sum\_{b \in \mathbf{S}\_k} \left( c\_b^q - \xi\_k^q \right)^2,\tag{1}$$

also known as appearance coherence. For position, the averaged *l*2 distance from block position **P***<sup>b</sup>* to the center position of its superpixel is the position energy function, *Epos*ð Þ¼ **S***<sup>k</sup>* 1 ∥*Sk*∥ P *b* ∈**S***<sup>k</sup>* <sup>∥</sup>**P***<sup>b</sup>* � *<sup>θ</sup>k*∥<sup>2</sup> . This is to ensure clustered blocks are geophysically close. Besides, to avoid seeing any superpixels with sophisticated boundary, we use the total boundary length as boundary penalty function. Furthermore, we constrain the minimal size of finalized superpixel to be at least 25% of initial size. If any update of block's belonging violates this constrain, we give infinity penalty to this update, therefore, the algorithm will reject such label assignment update.

$$P\_{size}(\mathbf{S}\_k) = \begin{cases} +i\boldsymbol{\mu}^\*, & \text{size}(\mathbf{S}\_k) < \mathbf{0}.25 \times initialsize \\ \mathbf{0}, & \text{otherwise}. \end{cases} \tag{2}$$

Similar penalty would be applied, if the update causes any isolated blocks who are surrounded by blocks from other superpixels. This is to enforce all generated superpixels to be topologically connected.

#### *3.1.2 Boundary-only update*

To define boundary energy function, we need to define boundary block and length. If a block has any neighbor block from other superpixel, then it is a boundary block. The boundary length of block is the number of neighbor blocks that belong to other superpixel.

$$P\_b(\mathbf{s}) = \sum\_{b \in \mathbf{S\_k}} \sum\_{b\_n \in Nights(b)} \mathbf{S(S\_k, b\_n)},\tag{3}$$

where *S*ð Þ **S***k*, *bn* is the indicator function of superpixel belonging for block, which return 0 if *bn* ∈*Sk*, otherwise 1. In our algorithm, we first stack entire initial boundary blocks into a queue, then the iterative superpixel clustering algorithm will work on boundary blocks only for consideration of updating label (i.e. superpixel assignment) of block. This is so-called 'boundary-only update'. In other words, the non-boundary blocks will not be considered for label change until they become boundary blocks. When the algorithm decides to update the label of a block, its neighbor will be considered to become new boundary blocks. When using the boundary-only update, there are two things to notice: 1) when update the label of block, it definitely change the list of boundary blocks; 2) we need to append the new boundary block to the end of the list because and follow the FIFO principle when deciding the order of blocks for consideration of changing label, in order to avoid the risk of divergence given by correlated dimensions in coordinate descent optimization. The candidate superpixel labels for a boundary block to swap are limited to its neighbor superpixels, otherwise it will trigger the topology connectivity penalty by having an isolated block. Given a trial of label update, the algorithm compares the objective function values before and after the change to see whether and how much the change is able to drive energy down.

We elaborate objective function each step of updating block-wise superpixel label assignment as below:

$$E(\mathbf{S}) = \sum\_{\mathfrak{s}} \left( E\_{col}(\mathfrak{s}) + \lambda\_{pos} E\_{pos}(\mathfrak{s}) + \lambda\_{b} P\_{b}(\mathfrak{s}) + P\_{topo}(\mathfrak{s}) + P\_{size}(\mathfrak{s}) \right), \tag{4}$$

where *λpos*, *λ<sup>b</sup>* are respectively the tradeoff coefficients for position energy function and boundary length penalty term. In practice, the regularization on superpixel size and topological connectivity will give infinite penalty on those superpixels of over-small size as *Psize*ð Þ *Sk* ≈*inf* and those of isolated blocks, i.e. *Ptopo*ð Þ *Sk* ≈*inf* . Therefore, the algorithm will always reject such label proposal that violates topology connectivity and size regularization. When superpixel assignment of a boundary block is updated, the algorithm will add its neighbor blocks to queue, because those non-boundary blocks are now next to other superpixels. The convergence will arrive when the queue is empty.

**Algorithm 1** Multi-resolution ROI Detection (MROID).

```
superpixel number - K
for l = 1 to levelMax do
  if l = 1 then
     1. Initialize blocks B on level l size on entire image;
     2. Initialize K superpixels S; initialize Θ, Ξ
  else
```
1. Initialize blocks **B** on level *l* size within positive superpixels and their neighbor superpixels **S**^. Initialize Θ, Ξ for **S**^. end if Compute the mean color and position in each block; Initialize *L*, the queue of boundary blocks on level *l*; **while** length(*L*) 6¼ 0 **do** Pop out block *bl <sup>i</sup>* from the queue; *Ebefore* ¼ *E*ð Þ **S** ; **for** *bn* ∈ Neighbor(*b<sup>l</sup> i* ) **do** change label of *b<sup>l</sup> <sup>i</sup>* to neighbor *bn*'s label; *Eafter*ð Þ¼ *bn E*ð Þ **S** ; end for find the ^ *bn* <sup>¼</sup> *arg* min *bn* <sup>∈</sup> *Neighbor b<sup>l</sup>* ð Þ*<sup>i</sup> Eafter*ð Þ *bn* ; if *Eafter*ð Þ *bn* <sup>&</sup>lt;*Ebefore* then update label of *<sup>b</sup><sup>l</sup> <sup>i</sup>* to that of ^ *bn*. append *Neighbor b<sup>l</sup> i* � � to *<sup>L</sup>*. end while run binary classifier on superpixels to predict ROI. end for

#### *3.1.3 Coarse-to-fine detection*

Instead of processing WSI at different resolutions [17], we cluster superpixels at coarse-to-fine level of resolution. Yao et al. [10] adopted boundary-only update as well to save unnecessary revisit to non-boundary blocks, while the boundary blocks on WSI may still be too much for extensive iterations. To further reduce the amount of data brought to finer update with more intense computation, we utilized a pre-trained classifier, e.g. SVM, to predict whether the superpixel belongs part of ROI. For any superpixel moved to finer update, smaller blocks will be initialized within its region. For example, a 10 � 10 block will be divided into 25 block of size 2 � 2 arranged at 5�5 grid. Boundary block queue will be refilled with 2 � 2 blocks who sit on superpixel boundaries. The classifier was trained using features extracted from patches sampled from ROI and non-ROI regions over annotated WSIs. To deal with different cardinality of patch per superpixel, we use pooling patch features at inference time. Given that we did not downsample images, therefore, the classifier trained on raw WSIs is able to be reused with different level of superpixel. See **Figure 4** as illustration.

#### **3.2 Complexity analysis**

Pixel-wise superpixel constructions [11, 12] have Oð Þ *N* complexity, where *N* is number of pixel, while it made them infeasible on WSIs of trillions of pixels. The introduced algorithm is able to reduce the complexity to scale of number of block i.e. <sup>O</sup> <sup>P</sup>*<sup>K</sup>*�<sup>1</sup> *<sup>k</sup>*¼<sup>0</sup> ∥S*k*∥ � � <sup>≪</sup> Oð Þ *<sup>N</sup>* . The boundary-only update, first presented in [10], further constrains involved blocks to those boundary blocks. Considering the purpose of clustered superpixel, our algorithm combined detection and superpixel clustering together, and it only executes finer segmentation within those coarse superpixels who were classified as ROI. It saved the calculations wasted on updating the superpixels that do not contribute to ROI detection. Due to the reduced dimensionality, the convergence comes faster than pixel-wise clustering methods.

#### **Figure 4.**

*An illustration of multi-resolution process of ROI detection on WSI. The example has 3 level of granularity in term of block size. Note that we did not downsample the WSI directly, which dump falsely many local details, and we still include neighbor superpixels close to positive ones at coarse classification to next level. If the bounding box is the ROI (a rough identifier), as resolution goes high, superpixels cover and surround the bounding box will get fine-detailed update.*

#### **4. Experiments**

#### **4.1 ROIs in lung cancer histopathology WSI**

In histopathology images like lung cancerWSIs, the regions of interest are those areas consist of cancer cells or other tissues that may be related to tumor diagnosis. A fast detection approach of ROIs is to search and localize those regions on image at WSI scale, that usually have trillions of pixels. Traditional pixel-wise methods and neural network cannot directly work on WSI, due to the extraordinary data scale and image dimensionality. Downsampling of WSI reduces complexity but also loses local fine-detailed features. Superpixels first cluster those pixels of similar spacial, color and topological properties as whole, and then in downstream tasks e.g. detection and segmentation, the superpixels will act as minimal manipulatable unit, reducing image primitives and complexity. If superpixels were well constructed, the downstream will not be affected by the sparse representation of image. The tumor cells of lung cancer patients (not only for lung cancer, but also generally appear in other subtypes of cancer) infest as cell mass. If treat the regions where tumor cell mass appears as ROIs, we can easily see that the H&E stained histopathology images that those tumor cells are more deeply colored due to the massive reproduction of genetic materials inside tumor nuclei (See **Figure 5**).

#### **4.2 Experimental setup**

In the experimental stage, a random forest and a support-vector-machine (a.k.a. SVM) classifier were trained with local features extracted from regions defined by

*Fast Regions-of-Interest Detection in Whole Slide Histopathology Images DOI: http://dx.doi.org/10.5772/intechopen.94238*

#### **Figure 5.**

*The comparison of several superpixel clustering on lung cancer H*&*E stained WSI: 1) the origin (with ROI annotated), 2) SLIC [11], 3) SPSS [21], 4) our method. The ROI is contoured by green line.*

the superpixels given by Algorithm 1. The total 384 dimensional features include local binary patterns and statistics derived from the histogram of the three-channel HSD color model as well as common texture features, e.g. color SIFT. The introduced method was compared against the superpixels generated by SLIC [11] and tetragonum (i.e. rectangular patches). The experiments used the adenocarcinoma and squamous cell carcinoma lung cancer WSIs from the NLST (National Lung Screening Trial) Data Portal1<sup>1</sup> . In superpixel classification, we executed feature extraction on the sampled patches (100 100) with 10% overlap with each other within each superpixel, we rule out patches sit across boundary avoiding noise. Lastly, we averaged the feature vectors of patches as representation of superpixel. When deciding ROI belonging for superpixel, if any part of ground-truth ROI fall into a superpixel, it will count as positive. The setup is rooted at the extremely high recall requirement for medical diagnosis. Given this setup, for better detection precision, superpixels should be better boundary adherent and clearly separated from background.

#### **4.3 Experimental results**

Due to the overwhelming fidelity of superpixels given by our algorithm, the classifier operated over the regions segmented by superpixels is able to deliver better classification results (See **Table 1**). Since the feature descriptors were built

<sup>1</sup> https://biometry. nci.nih.gov/cdas/studies/nlst/


#### **Table 1.**

*The table presents the comparison results of the proposed solution, MROID (numbers in bold), SLIC and tetragonum (non-superpixel) in term of classification statistics including: the rate of error classification, precision and recall. Tetragonum: Sliding rectangular windows.*

on the patches segmented by contours of superpixels, the better the superpixel adhere to the boundaries, the better differentiability the features have for superpixel classification.

**Figure 6** demonstrated the introduced multi-resolution coarse-to-fine superpixel segmentation in a lung cancer histopathology images. The algorithm first manipulated large block (180 180) to cluster superpixels, then move to finer segmentation with 10 10 blocks on the superpixels selected by the classifier. The recursive refinement continues until the block queue run out, which means energy loss converges. In **Table 1**, we compared the classification recall and precision using superpixels given by SLIC and our method as well as simply patches without any preprocessing like superpixel clustering. Our results showed that, compared to simple patching, utilizing superpixel may not always increase ROI recall but definitely lift precision. Compared to superpixel given by SLIC with sophisticated boundary, out method outperformed on both recall and precision. We also observed that, if superpixels do not adhere to boundary, a detection based on classification of superpixels of low segmentation accuracy leads to worse accuracy than a trivial patch based method. While, our method delivered best results at both recall and precision.

#### **Figure 6.**

*A coarse-to-fine superpixel clustering on a lung cancer WSI from NLST. 1) coarse segmentation of superpixels using large blocks (180180); 2) refined segmentation with small blocks within selected superpixels.*

### **5. Conclusion**

In the chapter, we presented a novel local feature based solution to fast search and detection of regions of interest (ROI) in whole slide lung cancer histopathology image. For superpixel clustering, we introduced coarse-to-fine multi-resolution segmentation of superpixel by manipulating blocks of different size. Besides, boundary-only update strategy also reduced the computational complexity to the scale of superpixel boundary length, irrelevant of image size.

We creatively embedded the ROI classification into superpixel clustering algorithm. Iteratively executing superpixel construction and ROI detection. A better superpixel will accelerate detection and lift accuracy, while on the other hand, a better classification of ROI on coarse superpixel guides superpixel segmentation at finer level. Our algorithm performed a faster and finer ROI detection and segmentation. The effectiveness and efficiency of our algorithm has been verified on large histopathology WSI database, e.g. NLST.

In future, as the development of neural network capable of flexible input size [28, 29], it is likely to merge superpixel construction and downstream tasks, e.g. semantic segmentation, classification together in neural network architecture, in which superpixels are clustered using hidden features, while superpixels boost feature learning as well.

### **Author details**

Junzhou Huang\* and Ruoyu Li 1 500 UTA Boulevard, The University of Texas at Arlington, Arlington, TX, United States

\*Address all correspondence to: jzhuang@uta.edu

© 2020 The Author(s). Licensee IntechOpen. This chapter is distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/ by/3.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.

### **References**

[1] Takács P, Manno-Kovacs A. MRI Brain Tumor Segmentation Combining Saliency and Convolutional Network Features. In2018 International Conference on Content-Based Multimedia Indexing (CBMI) 2018 Sep 4 (pp. 1–6). IEEE.

[2] Hou L, Samaras D, Kurc TM, Gao Y, Davis JE, Saltz JH. Patch-based convolutional neural network for whole slide tissue image classification. InProceedings of the ieee conference on computer vision and pattern recognition 2016 (pp. 2424–2433).

[3] Bándi P, van de Loo R, Intezar M, Geijs D, Ciompi F, van Ginneken B, van der Laak J, Litjens G. Comparison of different methods for tissue segmentation in histopathological whole-slide images. In2017 IEEE 14th International Symposium on Biomedical Imaging (ISBI 2017) 2017 Apr 18 (pp. 591–595). IEEE.

[4] LeCun Y, Cortes C, Burges CJ. MNIST handwritten digit database.

[5] Krizhevsky A, Hinton G. Convolutional deep belief networks on cifar-10. Unpublished manuscript. 2010 Aug;40(7):1–9.

[6] Zhou Z, Siddiquee MM, Tajbakhsh N, Liang J. Unet++: A nested u-net architecture for medical image segmentation. InDeep Learning in Medical Image Analysis and Multimodal Learning for Clinical Decision Support 2018 Sep 20 (pp. 3–11). Springer, Cham.

[7] Yao J, Zhu X, Jonnagaddala J, Hawkins N, Huang J. Whole slide images based cancer survival prediction using attention guided deep multiple instance learning networks. Medical Image Analysis. 2020 Jul 19:101789.

[8] Milletari F, Navab N, Ahmadi SA. Vnet: Fully convolutional neural networks for volumetric medical image segmentation. In2016 fourth international conference on 3D vision (3DV) 2016 Oct 25 (pp. 565–571). IEEE.

[9] Chan L, Hosseini MS, Rowsell C, Plataniotis KN, Damaskinos S. Histosegnet: Semantic segmentation of histological tissue type in whole slide images. InProceedings of the IEEE International Conference on Computer Vision 2019 (pp. 10662–10671).

[10] Yao J, Boben M, Fidler S, Urtasun R. Real-time coarse-to-fine topologically preserving segmentation. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2015 (pp. 2947–2955).

[11] Achanta R, Shaji A, Smith K, Lucchi A, Fua P, Süsstrunk S. SLIC superpixels compared to state-of-the-art superpixel methods. IEEE transactions on pattern analysis and machine intelligence. 2012 May 29;34(11):2274–82.

[12] Van den Bergh M, Boix X, Roig G, de Capitani B, Van Gool L. Seeds: Superpixels extracted via energy-driven sampling. InEuropean conference on computer vision 2012 Oct 7 (pp. 13–26). Springer, Berlin, Heidelberg.

[13] Li Z, Chen J. Superpixel segmentation using linear spectral clustering. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2015 (pp. 1356–1363).

[14] Ren X, Malik J. Learning a classification model for segmentation. Innull 2003 Oct 13 (p. 10). IEEE.

[15] Liu MY, Tuzel O, Ramalingam S, Chellappa R. Entropy rate superpixel segmentation. InCVPR 2011 2011 Jun 20 (pp. 2097–2104). IEEE.

[16] Veksler O, Boykov Y, Mehrani P. Superpixels and supervoxels in an

*Fast Regions-of-Interest Detection in Whole Slide Histopathology Images DOI: http://dx.doi.org/10.5772/intechopen.94238*

energy optimization framework. InEuropean conference on Computer vision 2010 Sep 5 (pp. 211–224). Springer, Berlin, Heidelberg.

[17] Bejnordi BE, Litjens G, Hermsen M, Karssemeijer N, van der Laak JA. A multi-scale superpixel classification approach to the detection of regions of interest in whole slide histopathology images. InMedical Imaging 2015: Digital Pathology 2015 Mar 19 (Vol. 9420, p. 94200H). International Society for Optics and Photonics.

[18] Litjens G, Bejnordi BE, Timofeeva N, Swadi G, Kovacs I, Hulsbergen-van de Kaa C, van der Laak J. Automated detection of prostate cancer in digitized whole-slide images of H and E-stained biopsy specimens. InMedical Imaging 2015: Digital Pathology 2015 Mar 19 (Vol. 9420, p. 94200B). International Society for Optics and Photonics.

[19] Lucchi A, Smith K, Achanta R, Lepetit V, Fua P. A fully automated approach to segmentation of irregularly shaped cellular structures in EM images. InInternational Conference on Medical Image Computing and Computer-Assisted Intervention 2010 Sep 20 (pp. 463–471). Springer, Berlin, Heidelberg.

[20] Van den Bergh M, Roig G, Boix X, Manen S, Van Gool L. Online video seeds for temporal window objectness. InProceedings of the IEEE international conference on computer vision 2013 (pp. 377–384).

[21] Yamaguchi K, McAllester D, Urtasun R. Efficient joint segmentation, occlusion labeling, stereo and flow estimation. InEuropean Conference on Computer Vision 2014 Sep 6 (pp. 756– 771). Springer, Cham.

[22] Li R, Huang J. Fast regions-ofinterest detection in whole slide histopathology images. In: International Workshop on Patch-based Techniques

in Medical Imaging 2015 Oct 9 (pp. 120-127). Springer, Cham.

[23] Gadde R, Jampani V, Kiefel M, Kappler D, Gehler PV. Superpixel convolutional networks using bilateral inceptions. InEuropean Conference on Computer Vision 2016 Oct 8 (pp. 597– 613). Springer, Cham.

[24] Kwak S, Hong S, Han B. Weakly supervised semantic segmentation using superpixel pooling network. InAAAI 2017 Feb 4 (Vol. 1, p. 2).

[25] Tu WC, Liu MY, Jampani V, Sun D, Chien SY, Yang MH, Kautz J. Learning superpixels with segmentation-aware affinity loss. InProceedings of the IEEE Conference on Computer Vision and Pattern Recognition 2018 (pp. 568–576).

[26] Liu X, Guo S, Yang B, Ma S, Zhang H, Li J, Sun C, Jin L, Li X, Yang Q, Fu Y. Automatic organ segmentation for CT scans based on super-pixel and convolutional neural networks. Journal of digital imaging. 2018 Oct 1;31(5):748–60.

[27] Jampani V, Sun D, Liu MY, Yang MH, Kautz J. Superpixel sampling networks. InProceedings of the European Conference on Computer Vision (ECCV) 2018 (pp. 352–368).

[28] Li R, Wang S, Zhu F, Huang J. Adaptive graph convolutional neural networks. arXiv preprint arXiv: 1801.03226. 2018 Jan 10.

[29] Yang F, Sun Q, Jin H, Zhou Z. Superpixel Segmentation with Fully Convolutional Networks. InProceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition 2020 (pp. 13964–13973).

#### **Chapter 6**

## Approaches for Handling Immunopathological and Clinical Data Using Deep Learning Methodology: Multiplex IHC/IF Data as a Paradigm

*Siting Goh, Yueda Chua, Justina Lee, Joe Yeong and Yiyu Cai*

### **Abstract**

Recent advancements in deep learning based artificial intelligence have enabled us to analyse complex data in order to provide patients with improved cancer prognosis, which is an important goal in precision health medicine. In this chapter, we would be discussing how deep learning could be applied to clinical data and immunopathological images to accurately determine survival rate prediction for patients. Multiplex immunohistochemistry/immunofluorescence (mIHC/IF) is a relatively new technology for simultaneous detection of multiple specific proteins from a single tissue section. To adopt deep learning, we collected and pre-processed the clinical and mIHC/IF data from a group of patients into three branches of data. These data were subsequently used to train and validate a neural network. The specific process and our recommendations will be further discussed in this chapter. We believe that our work will help the community to better handle their data for AI implementation while improving its performance and accuracy.

**Keywords:** immunopathology, deep learning, multiplex IHC/IF

#### **1. Introduction**

Improved cancer prognosis is a vital goal of precision health medicine. Advancements in Deep Learning (DL) based Artificial Intelligence (AI) technologies enable modelling of complex data providing deeper insights and patients with more reliable results. Machine Learning (ML) is the process of enabling machines to make predictions from data that is fed into it. This includes DL, a type of approach created from the development of artificial neural networks [1]. The DL network consists of multiple layers of artificial neural networks including an input, an output and multiple hidden layers [2, 3]. Predictions are made after datasets are generated from and trained against these hidden layers. Recent advancements in computational processing power has sparked interest in tapping into the vast research on DL and applying it to digital pathology. Digital pathology is the process of digitizing Whole-Slide Images (WSI) using advanced slide-scanning techniques and AI-based methods for detecting, segmenting, diagnosing and analysing digitized images [4].

DL is the engine of advancement in artificial learning in both computer and clinical sciences. It is a collection of computer learning algorithm layers that uses the raw data input to first generate generalise features that are subsequently used to progressively extract higher level features such as tumour stroma count, and assign them class labels. Eventually, the system will distinguish different interest categories via the identified ideal data features. The DL approaches are widely accepted due to the ability of discovering patterns and signals from data too large for human comprehension. Furthermore, the multiple layers allow modelling of highly complex non-linear problems. On top of having higher accuracy, the DL approaches are also easily applied.

#### **1.1 The importance of deep learning in digital pathology and mIHC**

In current clinical practice, pathologists base their medical diagnosis on the quantification and visual recognition of the analysed sample details, which could lead to diagnostic discrepancies and potential suboptimal patient care [5]. The increased adoption of non-invasive clinical procedures to acquire diagnostic samples has also severely reduced the quantity and quality of samples obtained, which compounds the workload of pathologists. In view of the inter-variabilities in analysing samples manually and the limitations of available samples, the use of DL analysis has thus been researched on and progressively applied in the clinical practice.

DL in digital pathology aims to improve the workload of pathologists by automating time-consuming tasks, hence allowing additional time to be spent on disease presentations with complex features. AI applications in digital pathology can also be applied to develop prognostic assays that evaluate the severity of diseases and make an accurate prognosis in response to therapy. This could be applied to various image processing and classification tasks, such as low-level jobs revolving around image recognition issues including detection and segmentation, as well as high-level tasks such as prognosis of response to therapy based on patterns of images [6, 7]. Such AI approaches are designed to extract relevant image renditions to train machines to be used as specific segmentation, diagnostic or prognostic tools.

One of the most extensively used DL models in pathology image analysis is the Convolutional Neural Networks (CNN). The CNN is a class of deep, feedforward networks, comprising several layers which extrapolate an output from an input and contains multiple convolutional sheets. These convolution sheets are the foundation of a CNN in which the network learns and extrapolates feature maps from images using filters between the input and output layers [4]. These layers in CNN are not connected as the neurons in one layer only interact with a specific region of the previous layer instead of all its neurons. The CNN also contains pooling layers which primarily function to scale down or reduce the dimensionalities of the features. CNN DL-based approaches are used for image-based detection and segmentation tasks to distinguish and quantify cells, histological features or highlight regions of interest [4]. CNN DL-based approaches have also been developed to automatically distinguish and segment blurry areas in digitised WSIs with high accuracy.

Another type of DL approach is the Fully Convolutional Network (FCN) which learns representations from every pixel and makes a potential feature detection that may occur sparsely in the entire pathology image [4]. FCNs uses co-registered Haemotoxylin and Eosin (H&E) images with multimodal microscopy techniques to classify WSIs into 4 classes: cancer, non-malignant epithelium, background and other tissues. When FCN was used to detect invasive breast cancer regions on WSIs, it had a diagnostic accuracy of 71% (Sørensen-Dice coefficient) when compared to an expert breast pathologist's assessment [8]. With better technologies and further research, FCNs can potentially automate these tasks with a higher accuracy, reducing the workload of pathologists.

#### *Approaches for Handling Immunopathological and Clinical Data Using Deep Learning… DOI: http://dx.doi.org/10.5772/intechopen.96342*

AI-based approaches such as Generative Adversarial Network (GAN)-based approaches can be used for training to automatically score tumoral programmed cell death 1 ligand 1 (PD-L1) expression in biopsy sample images [4]. They reduce the number of inputs required from pathologists and make up for lack of tissue samples available in biopsy specimens. Novel GAN-based approaches propose converting H&E staining of WSIs to virtual immunohistochemistry staining, thus eliminating the need for destructive IHC tissue testing.

Many have also trialled DL in the field of immunohistochemistry (IHC). Traditional IHC is a common diagnostic tool used in pathology, but its application is significantly limited by its ability of only allowing single marker labelling per tissue section [9]. Alternatively, multiplex immunohistochemistry/immunofluorescence (mIHC/IF) technologies permit simultaneous detection of several markers on a single tissue section [9]. However, analysing large samples with multiple markers in conventional and manual ways by pathologists are highly time-consuming, laborious and susceptible to human error. By combining mIHC/IF with DL to analyse digitized WSIs, this will overcome the limitations.

In conclusion, the research and diagnostic fields have come a long way since the introduction of IHC. With the introduction of Al-based approaches in the application of IHC, higher accuracy and productivity could be achieved not just in the diagnostic level but also providing us with a platform to further venture into areas of medical knowledge yet to be fully understood.

#### **1.2 mIHC/IF Technologies**

To-date, our understanding of cancer immunotherapy has evolved and led to multiple studies investigating and refining strategies targeting negative regulators. Many have studied the use of checkpoint blockade immunotherapy such as programmed cell death receptor 1 (PD-1), PD-L1 and cytotoxic T-lymphocyte– associated protein 4 (CTLA-4) in a variety of cancers. The subsequent success of checkpoint blockade inhibition in clinical trials has led to the Food and Drug Administration's approval of various drugs such as Ipillimumab and Prembrolizumab for melanoma treatment of non-small cell lung cancer (NSCLC) respectively [10]. Furthermore, trial of combination immunotherapy has shown clinical efficacy in various cancers [11, 12]. However, other studies have also suggested that efficacy of these immunotherapy in various cancers may depend on the expression of biomarkers. For example, PD-L1 is suggested as a useful predictive marker in patients with NSCLC receiving Prembrolizumab [13]. However, this is not the case in patients with stage III melanoma [14]. To further discover potential biomarkers that could determine the efficacy of immunotherapy in various cancers, IHC has been introduced as a platform for these clinical studies.

Since its introduction in the 1940s [15], conventional IHC has been widely used in field of pathology and research. It involves the process of staining tissues samples using antibodies specific to antigens present within the samples. This specificity allows microscopic visualisation for diagnosis of neoplasm and obtaining valuable prognostic information. Despite this, it does have several limitations. The inability of labelling more than one marker per tissue sample has resulted in loss of potential information for analysis. For instance, the prediction of prognosis to an immunotherapy such as PD-L1/PD-1 checkpoint blockade may depend on the expression of an individual biomarker or in combination with other biomarkers [16–18]. Furthermore, the immune system can potentially be better understood, if the analysis of various biomarkers' expression patterns are done simultaneously, or cellular interactions within the tumour microenvironment can be visualised [19].

Moreover, IHC involves many critical steps which have high inter-user variability. For instance, antigens such as Ki-67 are more vulnerable to ischemia. As such, over fixation could result in irreversible damage to these antigens [20]. The concern of IHC's reproducibility such as for Ki-67 and its implications was also mentioned in the 2017 St. Gallen International Expert Consensus Conference [21]. However, multiple studies have since demonstrated that analytical variability can be negated with the use of digital analysis to calculate biomarkers index [22, 23].

Although conventional IHC is a cost-effective diagnostic and prognostic tool, it has been replaced with the introduction of mIHC. mIHC has been used to overcome the shortcoming of single biomarker labelling in conventional IHC. The use of mIHC has proven to provide an even more accurate analysis as seen in the study by Yeong *et al.,* where the simultaneous quantification of three different PD-L1 antibodies (22C3, SP142 and SP263) by mIHC scoring had moderate-to-strong correlation (with 67%–100% individual concordance rates and Spearman's rank correlation coefficient values up to 0.88 [24]) when compared with manual scoring from four different pathologists.. This demonstrated the use of mIHC as a promising tool for an even more accurate analysis.

The use of mIHC has played a significant role in both research and clinical studies of cancer immunotherapy. mIHC is a relatively new tool to study the spatial tumour microenvironment especially those of limited tissue specimens. It has great potential in clinical and translational application. This was demonstrated by Halse *et al.*, who used mIHC to reveal a close relationship between the presence of CD8+ T cells within the tumour and the expression of PD-L1 in melanoma [25]. A systemic review and metaanalysis of studies also reported that mIHC improved results in predicting responses to PD-1/PD-L1 checkpoint blockade immunotherapy in various solid tumour types when compared to using conventional IHC analysis [26]. Several studies have also used various types of mIHC to obtain data for analysis. For instance, TSA-based mIHC was used to profile PD-1 to PD-L1 proximity in 166 metastatic melanoma samples and 42 Merkel cell carcinoma samples in two respective studies [27, 28]. As aforementioned, understanding the tumour microenvironment could potentially provide a foundation upon which interpretation of immunotherapy response could be made.

#### **1.3 Use of mIHC in combination with digital pathology**

mIHC can be powered by digital pathology analysis software, such as inForm (Akoya Biosciences, California, USA) [29–31] and HALO TM (Indica Labs) [28, 32].


**Table 1.**

*Digital pathology softwares, InForm, HALO, and Oncotopix and their software features for multiplex IHC/IF.*

#### *Approaches for Handling Immunopathological and Clinical Data Using Deep Learning… DOI: http://dx.doi.org/10.5772/intechopen.96342*

These software resolve the restrictions of labeling a single marker per tissue section by precisely evaluating the unique localization of multiple simultaneously detected biomarkers and their co-expressions or interactions between cells [33].

For example, although Ki-67/PD-L1 labeling is useful by itself, a multiplex approach enables several markers to be interrogated simultaneously [34–36]. However, only analytical digital pathology solutions for Ki67 and PD-L1 scoring are currently commercially available as listed in **Table 1** [33, 37]. The involvement of digital pathology has also decreased intra- and inter-observer variability seen in manual scoring as previously highlighted. Consequently, using mIHC in conjunction with digital analysis software will resolve the restrictions of conventional IHC, thus providing us with an accurate and powerful tool in the interpretation of immune response in various fields.

#### **2. Proposed deep learning framework for analysing immunopathological and clinical data**

This section presents a holistic guiding framework to select and develop a DL architecture for multi-dimensional analysis. The entire pipeline can be broken down into 3 parts: [1] data pre-processing, [2] feature engineering and [3] model selection, validation, and evaluation (**Figure 1**). This includes treating the data input, selecting the appropriate model for the type of data and using the preferable method to validate the selected model.

To demonstrate the clinical application of the framework, a total of 107 clinical as well as mIHC/IF data from patients with breast cancer (BC) previously published [37]. The clinical data consists of parameters such as age and tumour grade as stated in **Table 2** Row 1, while the mIHC data comprised of antibody-based spectral unmixing result obtained from stained mIHC image of tumour section labeling markers such as cytokeratin, CD68, CD8, CD20, FOXP3, PD-L1, and CK/EpCAM (**Figure 2**).

#### **2.1 Data pre-processing**

The first step in data pre-processing involves analysis of the dataset. This process consists of four main components: [1] one-hot encoding, [2] data normalization, [3] data enhancement and [4] data shape conformity.

#### *2.1.1 One-hot encoding*

One-hot encoding is the process of converting any non-numerical data existing in the clinical dataset to a categorical numerical representation that is readable by

**Figure 1.** *General overview of the DL framework.*


**Table 2.**

*Data of Clinical Dataset.*

#### **Figure 2.**

*Representative images of breast tissue stained using multiplex immunohistochemistry/immunofluorescence (mIHC/IF) [DAPI (blue), CD8 (red), CD20 (white), CD68 (green), FOXP3 (cyan), PD-L1 (yellow), CK/ EpCAM (magenta)]. (Magnification, 200X).*

the computer. Any non-numerical data within each category is split into the number of categories it has and encoded with a binary 0/1. For example, in the case of our clinical data, the columns, "Lymphovascular Invasion*"* contains 3 possible values: positive, possible, negative (**Table 3**). This represents 3 categories and is the prime candidate to be one-hot encoded. The input column is subsequently expended to 3 columns, one for each category in this input as shown in **Tables 3** and **4**.

#### *2.1.2 Data normalisation*

Most CNN research and models are developed with the intention for application in Computer Vision, where an entire input image data points are all pixels with ranges from zero to 255. Non-imaging datasets are more complicated as each input parameters have different units of measurements that might range from ones


#### **Table 3.** *Lymphovascular invasion data before One-hot encoding.*

*Approaches for Handling Immunopathological and Clinical Data Using Deep Learning… DOI: http://dx.doi.org/10.5772/intechopen.96342*


**Table 4.**

*Lymphovascular invasion data after One-hot encoding.*

to hundreds of thousands. Using such models with disparate values meant that a model with a large input parameter could easily outweigh another with a smaller value range. Therefore, data normalisation is needed to ensure that the dataset has comparable values across the data inputs while still maintaining their distribution within each data input. Data normalisation was done by scaling each input column to carry a mean of 0 and a standard deviation, by applying the following formula:

$$X\_{N\text{low}} = \frac{X - \mu}{\sigma}$$

An example of a segment of clinical dataset following one-hot encoding is as shown in **Table 5**. A notable feature of the clinical dataset is the disparate values across the columns which arose due to the different units of measurements used across the columns, such as categorical numbers, months, and millimetres. As such, these numbers could not be directly compared. To obtain a more comparable data, normalisation of these values was done, while maintaining the distribution within each column (**Table 6**).

#### *2.1.3 Data enhancements*

When working with a medical dataset, it will be advantageous to have medical insights augment the data, as it can improve the result. The use of medical insights is however dependent on the context of the problem and is subjective to the augmentation or removal of features and/or any dataset. In this study, clinically relevant data was augmented to the cell dataset to count the number of stroma and cancer cells of each patient. Subsequently this was evaluated with a simple 12-layer dense neural network and the obtained results were compared with and without data enhancements on 10000 epochs. It was discovered that there was a marked improvement in the reduction of mean absolute error by 14.8% when the clinical dataset was enhanced with more relevant information. However, the reduction in mean absolute error was highly dependent on clinical dataset used and thus varies with its application.

#### *2.1.4 Ensuring data conformity*

The CNN requires the dataset to be homogeneous in its shape, which is achievable in the classical Computer Vision problems where images could be resized to a uniform rectangular shape. However, in the case of medical dataset, the dimensions are mostly dependent on the source of the data, which is usually 3-dimensional or more. There are two ways to homogenise medical dataset, either by appending nonmeaningful data to the clinical dataset, or selectively removed data until the shape is uniform. This process requires a higher dimensional visualisation which is best explained using a tangible example as follows:

