**4. Conclusions**

In this book chapter, we provide a survey about the most common processing frameworks for information extraction in medical HSI. First, we show the main motivations on the usage of HS technology for biomedical data: the interaction between the light and tissue provides useful information for diagnostic applications. Second, we survey the most common approaches for HSI processing in the medical field: inverse optical modeling and machine learning approaches.

Within the ML approaches, we show there is a big variety in the methods which are used, mainly in two different types: traditional machine learning approaches based handcrafted features and recent DL techniques. Even within each subfield, the variety of options to extract information in medical HSI is still wide.

In **Table 1** we provide a summary of the applications of the different methods which have been described in this book chapter. Such table relates the main information extraction methods and the biomedical applications of HSI. Further literature revision about the different biomedical HSI applications are out of the scope of this chapter. However, we recommend readers who are interested in further information about the usage of HSI for different biomedical applications to refer the different literature reviews in this context mentioned in the introduction section.

The main challenge in HS medical image processing is to determine which processing framework is the most appropriate for clinical applications. Nowadays, the current trend for researchers working with medical HS data is to collect their own data, and then propose a processing framework to address a certain problem. Normally such processing frameworks are customized for their particular applications. In order to reach an agreement by the research community on the most successful information extraction methods for HSI, there is the need of further investigations with comparisons among the most promising processing approaches. To this end, the availability of large public datasets would help. However, although there is no general

**51**

**Table 1.**

*Information Extraction Techniques in Hyperspectral Imaging Biomedical Applications*

Monte Carlo Simulations

PCA and supervised classification

t-SNE and supervised classification

classification

Band selection with Mahalanobis distance

Band selection with optimization techniques

Wavelet transformation and supervised classification

Fourier series and supervised

Spatial and spectral features in supervised classification

Superpixel segmentation and supervised classification

Supervised classification and K-NN spatial filtering

*.*

Deep learning 2D-CNN and 1D-DNN In-vivo brain tumor detection‡ [69]

**Algorithm Application Ref.**

SVM Intestinal ischemia identification [36]

RF In-vivo oral cancer [41] MLR Ulcerative colitis in histological slides [42] SVM, RF Brain cancer in histological slides [43] SVM, RF, LDA Head and neck tumor [44]

enhancement

PCA and false color Melanocytic lesions visualization [53]

delineation

detection

Band selection with mRMR Ex-vivo breast cancer detection [61]

slides

2D-CNN (Inception v4) Head and neck cancer [71]

2D-CNN and 3D-CNN Head and neck cancer [73] 2D-CNN (U-Net) Tongue cancer detection [74]

GAN HS image generation from RGB [76] RNNs, 2D-CNN and 3D-CNN Head and neck cancer detection [77]

Orthogonal projections Retina analysis for Alzheimer's

PCA Biliary trees visualization

colon cancer samples

Cholesterol identification in skin [22] Arthritis identification in skin [23] Blood volume fraction estimation in

Gastric cancer detection [37] Prostate cancer [38] Tongue cancer [39] Skin cancer [40]

Detection of in-vivo oral cancer [54] Prostate cancer in histological slides [55]

In-vivo brain tumor detection [59]

Prostate cancer in mice models [60]

Breast cancer detection [61]

Gastric cancer identification [62]

In-vivo head and neck cancer [63]

In-vivo brain tumor detection ‡ [64]

Leukemia detection in blood smear [66] Red blood cell counting [67]

In-vivo brain tumor detection‡ [27]

Salivary gland cancer [72]

Breast cancer [75]

Brain tumor detection in histological

The identification of white blood cells in blood smear slides

Intraoperative brain tumor

[24]

[52]

[56]

[57]

[58]

[68]

*DOI: http://dx.doi.org/10.5772/intechopen.93960*

Optical inverse modeling Light transport models and

**Information extraction** 

Pixel-wise classification

Feature extraction and feature selection

Spatialspectral classification

*Publicly available datasets are marked with ‡*

*Summary of information extraction methods for medical HSI.*

**method**

Feature learning


*Information Extraction Techniques in Hyperspectral Imaging Biomedical Applications DOI: http://dx.doi.org/10.5772/intechopen.93960*

*Publicly available datasets are marked with ‡*

#### **Table 1.**

*Summary of information extraction methods for medical HSI.*

*Multimedia Information Retrieval*

**4. Conclusions**

HSI by Trajanovski *et al.* for tongue cancer detection with a 2D input data using all HS channels for semantic segmentation of ex-vivo specimens [74]. Additionally, Kho et al. used ex-vivo specimens from patients with breast cancer and applied a standard U-net with 2D input HS data using all spectral channels for semantic segmentation [75]. More recently, several modern DL approaches with origins in computer-vision have been applied to medical HSI experimentally. In [76], a generative adversarial network (GAN) was applied to use DL to learn the association of RGB images and HS images to learn the ability to generate HS digital histology images from standard RGB digital histology images of breast cancer. Another modern approach is long-short-term-memory (LSTM) and recurrent neural networks (RNN) which can utilize spatial-spectral and time-based inputs to operate in real-time video approaches. In [77], RNNs are compared to and outperform 2D- and 3D-CNN methods for in-vivo cancer detection with the goal of real-time video endoscopy. The use of DL for HS processing is currently a hot topic in the research community in different fields. The main advantage of DL approaches in HSI is their capability to exploit jointly the spatial and the spectral information for image processing tasks. Currently, researchers are experimenting with different DL architectures in order to find the most appropriate DL model for HSI [78]. In the context of medical HSI, the use of DL in medical HS have shown good performance in different applications, but its usage is still limited compared to other ML approaches. The main reason is the limited number of data due to the novelty of the technology. More publicly available datasets with a large number of patients are required in order to definitively establish an adequate comparative of DL and traditional ML techniques.

In this book chapter, we provide a survey about the most common processing frameworks for information extraction in medical HSI. First, we show the main motivations on the usage of HS technology for biomedical data: the interaction between the light and tissue provides useful information for diagnostic applications. Second, we survey the most common approaches for HSI processing in the medical field: inverse optical modeling and machine learning approaches.

Within the ML approaches, we show there is a big variety in the methods which are used, mainly in two different types: traditional machine learning approaches based handcrafted features and recent DL techniques. Even within each subfield,

In **Table 1** we provide a summary of the applications of the different methods which have been described in this book chapter. Such table relates the main information extraction methods and the biomedical applications of HSI. Further literature revision about the different biomedical HSI applications are out of the scope of this chapter. However, we recommend readers who are interested in further information about the usage of HSI for different biomedical applications to refer the different literature reviews in this context mentioned in the introduction section. The main challenge in HS medical image processing is to determine which processing framework is the most appropriate for clinical applications. Nowadays, the current trend for researchers working with medical HS data is to collect their own data, and then propose a processing framework to address a certain problem. Normally such processing frameworks are customized for their particular applications. In order to reach an agreement by the research community on the most successful information extraction methods for HSI, there is the need of further investigations with comparisons among the most promising processing approaches. To this end, the availability of large public datasets would help. However, although there is no general

the variety of options to extract information in medical HSI is still wide.

**50**

processing framework, the different information extraction techniques together with HS medical data have demonstrated several advantages for biomedical applications.
