Classification of skin moles from optical spectropolarimetric images: a pilot study

A new skin-imaging method, which was developed recently, gives an informative image data about skin moles by collecting scattered polarized light reflected from their tissue area at different wavelengths and polarizations. This method is called as optical rotational spectropolarimetric imaging (ORSI), which scans the polarization states by continuously rotating a linearly polarized light incident on the lesion and collecting the reflected sequence of images. A novel method to distinguish cancerous from benign moles by analyzing the images obtained using this imaging system is proposed. The proposed method performs an automatic examination of the polarized images according to characteristics such as their cross-image local contrasts, large-scale homogeneity, border disorder, and asymmetry. The pilot study was conducted with 10 subjects, in which two ORSI image sequences at two different wavelengths were taken for each subject. Results show good separation between cancerous and benign moles and between benign moles with high dysplasia.