Applications for Machine Learning in Mohs Micrographic Surgery: Increased Efficiency and Accuracy.

Abstract

Mohs micrographic surgery (MMS) is a precise method of skin cancer treatment via removal in stages for complete resection of malignancy. 1 Machine learning (ML) offers multiple potential applications to the procedure, some of which are discussed here. The first step in MMS is identifying patients who meet criteria for referral, which often is completed via the histologic confirmation of skin cancer. ML may accelerate referral to a Moh’s surgeon by automatically categorizing histologic findings. For example, an image classification system was developed using a cascade of three independently-trained convolutional neural networks (CNN) to sort digitized dermatopathol-ogy slides into categories of basaloid, squamous, melanocytic, and other; this system demonstrated an accuracy of up to 98%. 2 A system such as this would allow a dermatologist who interprets biopsies to review cases of a certain category (i.e., basaloid or squamous) and refer other cases. 2 Clinical dermatologists may identify patients who meet criteria for MMS and direct them to Mohs surgeons in a timelier manner with the assistance of ML.