The Use of AI in the Identification of Cancerous Skin Lesions: A Comparison of Commercially Available Means and Clinically Available Tools.

Abstract

Skin cancer is a broad term, used to refer to a varied set of distinct diseases with diverse yet potentially similar presentations that can make consistent and reliable identification all but impossible for the untrained eye. Yet, despite the difficulty, distinguishing the differing forms of the disease from a benign nevus or other skin blemish is vitally important. Skin cancer is often dangerous, especially when not treated early, as the early-stage detection and identification of malignant skin cancer is critically linked to its successful treatment and the long-term prognosis of those affected. Moreover, the occurrence rate for skin cancer has been increasing and is expected to continue to rise. Given the difficulty in diagnosis and prevalence, it is not surprising that a number of smart device based applications intended to aid in the identification of cancerous skin lesions have been made available for download to a personal device. This work sought to determine the accuracy and overall utility of some of these applications by directly testing them on images of known cancerous lesions as well as benign skin features and marks. The achieved results were then analyzed and compared to the established results for similar tools that are intended for use in clinical settings. The clinical device results were obtained for this work via a literature review and results comparison.