**Diagnostic Decision Support System in Dysmorphology**

Kaya Kuru and Yusuf Tunca

Additional information is available at the end of the chapter

http://dx.doi.org/10.5772/51118

## **1. Introduction**

[6] Groot, M. J., Schilt, R., Ossenkoppele, J. S., Berende, P. L. M., & Haasnoot, W. (1998). Combinations of growth promoters in veal calves: consequences for screening and

[7] Elfaki, A. O., Muthaiyah, S., Ho, C. K., & Phon-Amnuaisuk, S. (2011). Knowledge representation and validation in a Decision Support System: Introducing a variability modelling technique. *In: C.S. Jao (ed.), Efficient decision support systems: practice and*

[8] Worachartcheewana, A., Nantasenamata, C., Isarankura-Na-Ayudhyaa, C., Pidetch‐ ab, P., & Prachayasittikula, V. (2010). Identification of metabolic syndrome using de‐

[9] Das, A., & Bhattacharya, M. (2011). Computerized Decision Support System for Mass Identification in Breast Using Digital Mammogram: A Study on GA-Based Neuro-Fuzzy Approaches. *In: H.R. Arabnia and Q.-N. Tran, Software tools and algorithms for bi‐ ological systems, Advances in Experimental Medicine and BiologySpringer Science and*

[10] Sidiropoulos, K., Glotsos, D., Kostopoulos, S., Ravazoula, P., Kalatzis, I., Cavouros, D., & Stonham, J. (2012). Real time decision support system for diagnosis of rare can‐ cers, trained in parallel, on a graphics processing unit. *Computers in Biology and Medi‐*

[11] Kacprzyk, J., & Zadrożny, S. (2009). Supporting decision making via verbalization of data analysis using linguistic data summaries. *In: E.Rakus-Andersson et al., Recent Ad‐*

[12] Rass, S., Machot, F. A., & Kyamakya, K. (2011). Fine-grained diagnosticsof ontologies with assurance. *In: C.S. Jao (ed.), Efficient decision support systems: practice and challenges*

[13] Chen, G., Yan, P., & Wei, Q. (2009). Discovering associations with uncertainty from large databases. *In: E. Rakus-Andersson et al., Recent Advances in Decision Making*,

[14] Barthelemy, J. P., Bisdorff, R., & Coppin, G. (2002). Human centered processes and decision support systems. *European Journal of Operational Research*, 136, 233-252.

*vances in Decision Making*, 45-66, Springer Verlag, Berlin, Heidelberg.

*from current to future*, 79-98, Intech, Rijeka, Croatia.

45-66, Springer Verlag, Berlin, Heidelberg.

confirmation methods. *J. Vet. Med. A*, 45, 425-440.

*Business Media*.

66 Decision Support Systems

*cine*, 42, 376-386.

*challenges from current to future*, 29-48, Intech, Rijeka, Croatia.

cision tree analysis. *Diabetes Research and Clinical Practice*, 90, e15-e18.

Dysmorphology is the aspect of clinical genetics concerned with syndrome diagnosis in pa‐ tients who have a combination of congenital malformations and unusual facial features, of‐ ten with delayed motor and cognitive development [1]. Making a diagnosis for a dysmorphic patient requires a high degree of experience and expertise [2] since many dys‐ morphic diseases are very rare. The human brain possesses a remarkable ability to recognize what is familiar or unfamiliar in a face. The general public can recognize the face shape of individuals with Down's syndrome and experienced clinical geneticists develop this innate skill to such a degree that they recognize subtle facial features associated with several hun‐ dred dysmorphic syndromes [3]. In some areas of the world, however, genetic diagnosis is generally performed by general practitioners, dermatologists or pediatricians, not particu‐ larly trained in dysmorphology, rather than trained geneticists due to the lack in the num‐ bers of geneticists. Dysmorphologic diagnosis is usually performed by referring the images or terms standardized and specified in some limited number of catalogs and databases. However, not being very familiar with the terminology, especially for practitioners in rural areas, is an important handicap to dig into the right diagnosis. This can lead to diagnostic inaccuracy which in turn curtails both the right cure of patients that will best suit their par‐ ticular needs and the right guiding of their parents who may be at risk of having a new dys‐ morphic child. The whole process of reaching a genetic diagnosis can be very lengthy and entail struggling with referrals to numerous medical professionals and waiting for appoint‐ ments. Throughout this process, parents find themselves dragged from one medical profes‐ sional to another until discovering the accurate diagnosis of their child. These parents often experience high levels of burdensome anxiety and frustration. Moreover, delay in diagnosis may also delay access to critical services such as clinical trials and a patient's referral to sup‐ portive services including early intervention, physical or occupational therapy. Syndrome recognition and diagnosis is of clinical importance for several reasons according to Smithson

© 2012 Kuru and Tunca; licensee InTech. This is an open access article 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. © 2012 Kuru and Tunca; licensee InTech. This is a paper 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.

[1]: first, it influences patient management because awareness of the pattern of anomalies as‐ sociated with a particular syndrome highlights the investigations that need to be undertak‐ en; second, diagnosis provides information about the long-term prognosis and may help to identify options for treatment: for example, bone marrow transplantation or enzyme re‐ placement therapy can now be offered for some inborn errors of metabolism (e.g. in Fabry disease); third, the diagnosis determines what genetic advice can be given, including an esti‐ mation of genetic risks and possible means of prenatal diagnosis. Thus, delays in early treat‐ ment can have a miserable impact on the patient's health and can dramatically influence the chances of the child catching up to his/her peers and leading to a normal life. Therefore, reaching a thorough genetic diagnosis at an early stage is crucial. Not only does this affect the choice of cure for the patient, but it also enables the proper guidance, counseling and support for the parents to ensure an overall improvement in the patient's quality of life and care. Most physicians, neurologists or pediatricians are capable of noticing these early signs in a child, but are not equipped to perform a precise genetic diagnosis on their own especial‐ ly for very rare diseases. Since there are thousands of possible genetic conditions to be taken into account and each condition is in itself very rare, a specialist's evaluation generally seems the best path to ensure a proper genetic diagnosis is reached though it is sometimes very difficult even for geneticists to diagnose these orphan diseases.

population of comparable age and sex [7]. A most recent study of Hammond [7] has re‐ viewed the surface-based image capture in 3D in syndrome delineation and discrimination, in the categorization of individual facial dysmorphology and in phenotype–genotype stud‐ ies which is a very complex implementation of dysmorphological diagnosis process in a clinical environment. For dysmorphic syndromes with known genetic causes, molecular and/or cytogenetic analysis is the appropriate route of investigation in order to confirm a di‐ agnosis. However, applying right analysis method throughout many probable analyses is very much dependent to the accurate diagnosis considered before genotyping is undertak‐ en. The aim of this chapter is to propose a new methodology by observing the characteristic key components of facial dysmorphology associated with genetic disorder to indicate the

Diagnostic Decision Support System in Dysmorphology

http://dx.doi.org/10.5772/51118

69

Douglas [8] has stated that classification of faces based on facial patterns in isolation is un‐ likely to be accepted by dysmorphologists unless the mathematical features extracted and identified by feature selection algorithms to be discriminatory can be related to facial ap‐ pearance. Principal component analysis (PCA), independent component analysis (ICA), ker‐ nel principal component analysis (KPCA), local feature analysis (LFA), probability density estimation (PDA), multi-linear analysis (MLA), elastic graph matching (EGM), kernel dis‐ criminant analysis (KDA), support vector machine (SVM), Gabor wavelet and Fischer's line‐ ar discriminant analysis (LDA) exist to analyze the features of a face. Among them, the methods of PCA using eigenface, elastic graph matching, Gabor wavelet and Fischer's LDA are popular for face recognition. PCA can be computed as an optimal compression scheme that minimizes the mean squared error between an image and its reconstruction [9] as well as it may achieve good results of up to 96% recognition under ideal conditions [10] which can be provided with either capturing images in ideal environments by having a good illu‐ mination and using a good capturing device or preprocessing of images with several image processing techniques before extracting features. PCA using eigenfaces have been identified as computational efficient on even reduced hardware [11]. Additionally it is used for dimen‐ sion reduction as in our study from 2D to 1D to ease and speed up the calculations. That is to say, the dimensionality of the required space for all images can be reduced to the number

of input images instead of the pixel count that all images have in total.

City block distance, Euclidean distance, sub/space method, multiple similarity method, Bayes decision method and Mahalanobis distance are known typical distance functions [9]. Kapoor [9] highlights that the Mahalanobis distance is the most effective of the typical eval‐ uation distance-based approaches that calculate the distance from a point to a particular point in the data set. We have tested the Mahalanobis distance and Euclidean distance in our study to find the genotype-phenotype correlations. The well/known Mahalanobis dis‐

Gaussian [9]. Kapoor [9] indicates the difference such that Mahalanobis distance is a dis‐ tance measure based on correlations between variables by which different patterns can be identified and analyzed. It is a useful way of determining similarity of an unknown sample

2 Interested reader may found detailed information about Mahalanobis distance from Gul's thesis [14].

is based on the assumption that the underlying probability distributions are

right diagnostic criteria.

tance classifier2

The face is widely recognized as an attribute which best distinguishes a person, even at first glance. Facial features give lots of clues about the identity, gender, age and ethnicity. The face develops under the influence of many genes and in many cases a face provides impor‐ tant information to diagnose a syndrome. Thus, facial appearance can be a significant hint in the initial identification of genetic anomalies generally associated with cognitive impair‐ ments. There are specific properties, especially for facial dysmorphology caused by genetic syndromes and these properties are used by geneticists to pre-diagnose even before a clini‐ cal examination and genotyping are undertaken. Analyzing of properties in faces is some‐ times sufficient to diagnose for some cases, however, it is necessary to analyze other specific properties of the body such as the structure of the skeleton and the characteristics of speech produced for some other cases. Diagnosing of genotype-phenotype correlations correctly among many syndromes seems beyond the capability of human especially for very rare dis‐ eases. Most of the genetic diagnostic decision support systems (DDSS) have been applied in terms of the anthropometry and more specifically craniofacial anthropometry including stereophotogrammetry1 so far. It might be possible to diagnose a good number of syn‐ dromes correctly by using computer-assisted face analysis DSS as asserted by some scien‐ tists such as Farkas [4], Loos [5], Boehringer [2], and Hammond [6, 7]. Farkas [4] pioneered techniques for studying facial morphology using direct anthropometry for nearly 40 years ago [4]. His approach, using a ruler, calipers, tape measure and protractor, has been applied widely in the analysis of facial dysmorphology [7]. Many clinicians undertake such a man‐ ual craniofacial assessment and compare a patient's phenotype to the norms of a control

<sup>1</sup> Anthropometry is defined as the biological science of measuring the size, weight, and proportions of the human body [4]. Craniofacial anthropometry is performed on the basis of measures taken between landmarks defined on surface features of the head, face, and ears [12]. Stereophotogrammetry refers to combining multiple views of photos to form a 3D image [13].

population of comparable age and sex [7]. A most recent study of Hammond [7] has re‐ viewed the surface-based image capture in 3D in syndrome delineation and discrimination, in the categorization of individual facial dysmorphology and in phenotype–genotype stud‐ ies which is a very complex implementation of dysmorphological diagnosis process in a clinical environment. For dysmorphic syndromes with known genetic causes, molecular and/or cytogenetic analysis is the appropriate route of investigation in order to confirm a di‐ agnosis. However, applying right analysis method throughout many probable analyses is very much dependent to the accurate diagnosis considered before genotyping is undertak‐ en. The aim of this chapter is to propose a new methodology by observing the characteristic key components of facial dysmorphology associated with genetic disorder to indicate the right diagnostic criteria.

[1]: first, it influences patient management because awareness of the pattern of anomalies as‐ sociated with a particular syndrome highlights the investigations that need to be undertak‐ en; second, diagnosis provides information about the long-term prognosis and may help to identify options for treatment: for example, bone marrow transplantation or enzyme re‐ placement therapy can now be offered for some inborn errors of metabolism (e.g. in Fabry disease); third, the diagnosis determines what genetic advice can be given, including an esti‐ mation of genetic risks and possible means of prenatal diagnosis. Thus, delays in early treat‐ ment can have a miserable impact on the patient's health and can dramatically influence the chances of the child catching up to his/her peers and leading to a normal life. Therefore, reaching a thorough genetic diagnosis at an early stage is crucial. Not only does this affect the choice of cure for the patient, but it also enables the proper guidance, counseling and support for the parents to ensure an overall improvement in the patient's quality of life and care. Most physicians, neurologists or pediatricians are capable of noticing these early signs in a child, but are not equipped to perform a precise genetic diagnosis on their own especial‐ ly for very rare diseases. Since there are thousands of possible genetic conditions to be taken into account and each condition is in itself very rare, a specialist's evaluation generally seems the best path to ensure a proper genetic diagnosis is reached though it is sometimes

The face is widely recognized as an attribute which best distinguishes a person, even at first glance. Facial features give lots of clues about the identity, gender, age and ethnicity. The face develops under the influence of many genes and in many cases a face provides impor‐ tant information to diagnose a syndrome. Thus, facial appearance can be a significant hint in the initial identification of genetic anomalies generally associated with cognitive impair‐ ments. There are specific properties, especially for facial dysmorphology caused by genetic syndromes and these properties are used by geneticists to pre-diagnose even before a clini‐ cal examination and genotyping are undertaken. Analyzing of properties in faces is some‐ times sufficient to diagnose for some cases, however, it is necessary to analyze other specific properties of the body such as the structure of the skeleton and the characteristics of speech produced for some other cases. Diagnosing of genotype-phenotype correlations correctly among many syndromes seems beyond the capability of human especially for very rare dis‐ eases. Most of the genetic diagnostic decision support systems (DDSS) have been applied in terms of the anthropometry and more specifically craniofacial anthropometry including

dromes correctly by using computer-assisted face analysis DSS as asserted by some scien‐ tists such as Farkas [4], Loos [5], Boehringer [2], and Hammond [6, 7]. Farkas [4] pioneered techniques for studying facial morphology using direct anthropometry for nearly 40 years ago [4]. His approach, using a ruler, calipers, tape measure and protractor, has been applied widely in the analysis of facial dysmorphology [7]. Many clinicians undertake such a man‐ ual craniofacial assessment and compare a patient's phenotype to the norms of a control

1 Anthropometry is defined as the biological science of measuring the size, weight, and proportions of the human body [4]. Craniofacial anthropometry is performed on the basis of measures taken between landmarks defined on surface features of the head, face, and ears [12]. Stereophotogrammetry refers to combining multiple views of photos to form a

so far. It might be possible to diagnose a good number of syn‐

very difficult even for geneticists to diagnose these orphan diseases.

stereophotogrammetry1

68 Decision Support System

3D image [13].

Douglas [8] has stated that classification of faces based on facial patterns in isolation is un‐ likely to be accepted by dysmorphologists unless the mathematical features extracted and identified by feature selection algorithms to be discriminatory can be related to facial ap‐ pearance. Principal component analysis (PCA), independent component analysis (ICA), ker‐ nel principal component analysis (KPCA), local feature analysis (LFA), probability density estimation (PDA), multi-linear analysis (MLA), elastic graph matching (EGM), kernel dis‐ criminant analysis (KDA), support vector machine (SVM), Gabor wavelet and Fischer's line‐ ar discriminant analysis (LDA) exist to analyze the features of a face. Among them, the methods of PCA using eigenface, elastic graph matching, Gabor wavelet and Fischer's LDA are popular for face recognition. PCA can be computed as an optimal compression scheme that minimizes the mean squared error between an image and its reconstruction [9] as well as it may achieve good results of up to 96% recognition under ideal conditions [10] which can be provided with either capturing images in ideal environments by having a good illu‐ mination and using a good capturing device or preprocessing of images with several image processing techniques before extracting features. PCA using eigenfaces have been identified as computational efficient on even reduced hardware [11]. Additionally it is used for dimen‐ sion reduction as in our study from 2D to 1D to ease and speed up the calculations. That is to say, the dimensionality of the required space for all images can be reduced to the number of input images instead of the pixel count that all images have in total.

City block distance, Euclidean distance, sub/space method, multiple similarity method, Bayes decision method and Mahalanobis distance are known typical distance functions [9]. Kapoor [9] highlights that the Mahalanobis distance is the most effective of the typical eval‐ uation distance-based approaches that calculate the distance from a point to a particular point in the data set. We have tested the Mahalanobis distance and Euclidean distance in our study to find the genotype-phenotype correlations. The well/known Mahalanobis dis‐ tance classifier2 is based on the assumption that the underlying probability distributions are Gaussian [9]. Kapoor [9] indicates the difference such that Mahalanobis distance is a dis‐ tance measure based on correlations between variables by which different patterns can be identified and analyzed. It is a useful way of determining similarity of an unknown sample

<sup>2</sup> Interested reader may found detailed information about Mahalanobis distance from Gul's thesis [14].

set to a known one. It differs from Euclidean distance in that it takes into account the corre‐ lations of the data set and is scale-invariant, i.e. not dependent on the scale of measure‐ ments. In our study, the well-known Euclidean distance matching process outperformed the process of Mahalanobis distance. Thus, we chose this matching technique in our study.

**2. Methods**

benefited OpenCV4

are explained in following subsections in detail.

The FaceGP DDSS methodology has been established in C++ programming language. We

quiring much CPU and memory while processing thanks to the easy implementation of PCA. The methodology comprises several main modules, namely *face detection and image ac‐ quisition, image processing, training and diagnosis/recognition module,* and these main modules are divided into several sub modules as illustrated in Figure1. Functions of these modules

**Figure 1.** 1. Overall architecture of the methodology: the system consists of four main modules; face detection and image acquisition, image processing, training and diagnosis/recognition module. These modules are divided into sev‐

4 OpenCV (Open Source Computer Vision) library can be reached from the site, http://opencv.willowgarage.com/

eral sub modules that are delineated in the specified sections of the modules.

wiki/.

library. The application can function on any present computer, not re‐

Diagnostic Decision Support System in Dysmorphology

http://dx.doi.org/10.5772/51118

71

Hammond [7] pointed out that in terms of future technological support, two (2D) or threedimensional (3D) models of facial morphology are showing potential in syndrome delinea‐ tion and discrimination, in analyzing individual dysmorphology, and in contributing to multi-disciplinary and multi-species studies of genotype–phenotype correlations. Our study is an example of substantiating this potential. We have developed a real-time computer sys‐ tem that can locate and track a patient's head, and then recognize the patient by comparing characteristics of the face to those of trained individuals with classified dysmorphic diseas‐ es. In this study, in terms of both the feature extraction and helping non experienced practi‐ tioners in diagnosis process as well as to support experts in their decisions, we established an application to ease the process and we refer to our method as "Facial Genotype-Pheno‐ type Diagnostic Decision Support System (FaceGP DDSS) in Dysmorphology". Up to date, no complete solution has been proposed that allow the automatic diagnosis of dysmorphic diseases from the raw data (live camera, video or frontal photographs) without human in‐ tervention. The FaceGP DDSS aims not only to ease the required on-site expertise, but also to eliminate the time consuming catalog search of practitioners and geneticists to diagnose facial dysmorphic diseases through approximately 4.700 known dysmorphic diseases3 auto‐ matically, no intervention from the user such as preprocessing of images. The FaceGP DDSS methodology can be implemented on any site easily. In the methodology, reference images or reference patients on live subjects having the specific dysmorphic diseases are used as a guide for identifying the facial phenotypes (the outward physical manifestation of the geno‐ types) to train the system. Digital facial image processing methods are employed to reveal facial features with disorders indicating dysmorphic genotype-phenotype interrelation. A great number of genetic disorders indicating a characteristic pattern of facial anomalies can be classified by analyzing specific features (eigenfaces) with the aid of facial image process‐ ing methods such as PCA. Distance algorithms such as Euclidean, Mahalanobis are used to construct the correlation of the input image with the trained images in matching. Some im‐ age enhancement methods such as histogram equalization and median filter are implement‐ ed on detected images to capture better features and compensate for lighting differences. This study proposes a novel and robust composite computer-assisted and cost-effective method by merging several methods in the characterization of the facial dysmorphic pheno‐ type associated with genotype, in particular a method relying primary on face image cap‐ ture (acquisition from either camera, video or frontal face images) and manipulation to help medical professionals to diagnose syndromes efficiently.

<sup>3</sup> Many new dysmorphic diseases are described each year. London Dysmorphology Database (http://www.lmdatabas‐ es.com).
