Combination of Stokes polarized light imaging, roughness metrics and morphological features for the detection of melanoma

Skin cancer is the most common and most rapidly increasing form of cancer in the world. Optimal treatment of skin cancer before it reaches metastasis depends critically on early diagnosis. Usually physicians will measure some outward features to diagnose malignancy of pigmented skin lesion. These are mostly morphological features like border irregularity, size, shape, and color. Valuable information can be obtained from the analysis of skin roughness. Previously, we developed a hemispherical imaging Stokes polarimeter to monitor skin cancer based on a roughness assessment of the epidermis. In this study, Stokes images were analyzed to measure polarization properties of skin samples such as the principal angle of the polarization ellipse and the degree of polarization. A processing algorithm based on morphological operators was also developed and applied on Stokes images to extract shape information. Finally, an appropriate classifier was designed to determine the type of lesion based on morphological features as well as the roughness information. Clinical evaluation of the technique was performed on patients with benign nevi, melanocytic nevi, melanoma, and normal skin.