A novel method for melanocytic skin lesion extraction and analysis

Abstract

Identification and early detection of skin melanoma is a crucial step towards fatalities caused due to skin melanoma. Computer vision techniques play an important role for detection and diagnosis of the aforesaid disease. In this article, a new unsupervised skin melanoma extraction along with analysis approach is presented. The proposed method initially removes the hairs present in dermoscopic images and further uses a simple thresholding based approach for skin lesion extraction. Further, the extracted skin lesions are analysed for validation of melanocytic cells using a color scoring method. The proposed method is compared both quantitatively with the state of art existing skin lesion extraction approaches. Experimental results reveal the superiority of the proposed approach by achieving an average accuracy of 93% and 96% for ISIC 2016 and ISIC 2018 datasets respectively.

Subject Classification:

• Primary 68T45
• Secondary 92C55

Keywords:

• Skin lesion extraction
• Otsu thresholding
• Gaussian blurring
• Color score