Digital Dermoscopy Analysis and Artificial Neural Network for the Differentiation of Clinically Atypical Pigmented Skin Lesions: A Retrospective Study

Abstract

Noninvasive diagnostic methods such as dermoscopy or epiluminescence light microscopy have been developed in an attempt to improve diagnostic accuracy of pigmented skin lesions. The evaluation of the many morphologic characteristics of pigmented skin lesions observable by epiluminescence light microscopy, however, is often extremely complex and subjective. With the aim of obviating these problems of qualitative interpretation, methods based on mathematical analysis of pigmented skin lesions have recently been designed. These methods are based on computerized analysis of digital images obtained by epiluminescence light microscopy. In this study we used a digital dermoscopy analyzer with 147 clinically atypical pigmented skin lesions (90 nevi and 57 melanomas) to determine its discriminating power with respect to histologic diagnosis. The system evaluated 48 objective parameters used to train an artificial neural network. Using the artificial neural network with 10 variables selected by a stepwise procedure, we obtained a maximum accuracy in distinguishing melanoma from benign lesions of about 93%. Comparing this result with those of the many studies using classical epiluminescence light microscopy, it emerges that the method proposed is equal or even superior in diagnostic accuracy and has the advantage of not depending on the expertise of the clinician who examines the lesion.

Automated Extraction and Description of Dark Areas in Surface Microscopy Melanocytic Lesion Images

“…Because melanomas are often variegated, the ‘ratio of dark to light regions’ [12], the ‘variance of gray intensity’ [16]and the ‘color variegation’, defined as the standard deviation of the lesion reflectance [14], have turned out to be characteristic parameters for melanoma diagnosis, but they only give information about the overall brightness of the lesion. The extension and the distribution of darkly pigmented structures, based on the calculation of polar moments of inertia, appeared useful for the distinction between melanomas and nevi, comprising Spitz/Reed nevi, too [15, 17, 18, 19, 21]. However, since the concept of ‘dark area’ is ambiguous and influenced by human perception, it is essential to define and describe the method of dark area identification employed for image analysis.…”

“…Programs for image analysis enable the numerical description of some aspects of pigmented skin lesions, providing a reproducible quantification of several features and an aid for clinical diagnosis [8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21]. Different parameters, based on the measurement of brightness values, have been employed for the quantification of the overall darkness of the lesion.…”

“…Different parameters, based on the measurement of brightness values, have been employed for the quantification of the overall darkness of the lesion. Great attention was paid to the automatic identification and the description of ‘dark areas’ inside the lesion employing the DBDermoMIPS system, but the method has not been clearly described [15, 17, 18, 19, 21]. Moreover, the concept of ‘darkness’ is not ‘absolute’, but it is related to human perception which can be influenced by the overall pigmentation of the lesion as well as by the color of the skin.…”

Search for Resolution Invariant Wavelet Features of Melanoma Learned by a Limited ANN Classifier

“…In [35] authors analyzed 48 parameters belonging to four categories: geometry, color, texture, and color clusters inside the lesion and performed step-wise feature selection to identify an optimal subset of 10 variables (starting from the most significant: red multicomponent, decile of red, border homogeneity, mean value of red, grey-blue areas, contrast, interruptions of the border, mean skin-lesion gradient, background regions imbalance, variance of the border gradient). The clinical/dermoscopic equivalents of those variables are: multicomponent pattern/homogeneity, lesion darkness, border cleanliness, mean color of the lesion, grey-blue areas, network analysis, variation in the border cleanliness, grading of the border, color asymmetry, intensity in the border interruptions.…”

Utilizing Soft Computing for Determining Protein Deficiency

“…The ability of ANN is to learn trends from sample data and then apply this knowledge for future classification. It is a good technique for many clinical pattern recognition problems and being increasingly used in medical diagnoses [11,2]. Thus, BPNN should be properly designed and trained to produce reliable and accurate result.…”