Predictive Identification of Malignant Melanoma from Fractional Differentiation-based Image Enhancement

Abstract

Recent advancements in computer vision and artificial intelligence allow for the automatic detection of some abnormalities in medical photographs. One of these is a skin lesion, and prompt and accurate detection of these conditions substantially aids in treatment. When image processing is combined with the fundamental edge detection technique, there has been potential demonstrated in the automatic identification and delineation of boundaries inside skin lesions. In this work, we investigate how to improve edge-detection in photos using fractional differentiation. We present a fractional differential filter based method for edge detection in images of skin lesions. The original photos are then improved with the help of the derived images. During acquisition, acquired images are classified using deep learning models. In the experiments, a well-researched dataset of HAM10000 is employed. The outcomes demonstrate how well these filters work to cut through noise and pick up minute edge characteristics that can provide useful information while performing recognition tasks.