Evaluation of dermatological asymmetry measure of shape by histogram energy measure.

INTRODUCTION: This study addresses automatic detection of skin changes especially asymmetry and symmetry of the lesions and their crucial role in the definition of dermatological asymmetry measure. METHOD: The present research employs an integrated methodology, combining systematic evaluation of skin lesions. In the paper, we use unsupervised learning methods to assess the asymmetry of the lesions for dermatology as tool for computer-aided diagnostic systems. The asymmetry features are extracted from the images as vectors and then calculated using two dermatological shape asymmetry measure (DASMShape) methods. In the next step, to evaluate the clusters and their clusters to classes assignment, we have used classification methods: Support Vector Machines (SVM), k-nearest neighbors (kNN), Random Forrest, Multilayer Perceptron, and C4.5 to assess defined above dermatological asymmetry procedure. RESULTS: This multi-analytical approach generates comprehensive data on skin lesion analysis. The proposed approach effectively combines shape asymmetry (DASMShape) with texture analysis (Gabor filter energy) and intensity distribution to create a comprehensive diagnostic profile The high classification results --reaching up to 100% in certain configurations --confirm that unsupervised learning can accurately group lesions into diagnostic categories that align with expert. DISCUSSION: This system represents a significant step toward bridging the gap between general practitioners and specialists, providing a tool that prioritizes high sensitivity while maintaining the analytical depth required to reduce false negatives in melanoma screening.