Biological indexes based reflectional asymmetry for classifying cutaneous lesions.

This paper proposes a novel reflectional asymmetry descriptor to quantize the asymmetry of the cutaneous lesions for the discrimination of malignant melanoma from benign nevi. A pigmentation elevation model of the biological indexes is first constructed, and then the asymmetry descriptor is computed by minimizing the histogram difference of the global point signatures of the pigmentation model. Melanin and Erythema Indexes are used instead of the original intensities in colour space to characterize the pigmentation distribution of the cutaneous lesions. 311 dermoscopy images are used to validate the algorithm performance, where 88.50% sensitivity and 81.92% specificity have been achieved when employing an SVM classifier.