Artificial Intelligence as a Tool for Population Screening of Skin Tumors

Abstract

BCC/SCC and melanoma are the most common types of skin cancer. Low public awareness, unavailability of local professional dermatological expertise, and ineffective screening programs prevent early detection critical for successful treatment. Aim. Evaluate capabilities of AI-powered mobile application for population screening. Materials and methods. We proposed a free mobile application for regular skin self-examination based on AI analysis of photos taken using a smartphone enriched with demographic data and risk factors. Our model provides a binary output (malignant/benign), encouraging people to see dermatologists for an in-depth examination if a risk of a malignant skin tumor is detected. Results. We received and processed 500,000+ images of skin neoplasms taken by 290,000 users from 86 regions of the Russian Federation from 01.2021 to 12.2022. The images were accompanied by demographic data and responses to a questionnaire on cancer risk factors. 5,957 cases of BCC/SCC and 7,622 cases of melanoma were detected. Analysis of the campaign results revealed a significantly higher prevalence of malignant skin tumors per 100,000 population. Our results exposed a much younger average age of skin cancer patients in both men and women populations compared to official statistics, with a difference of 15 years. Conclusion. The AI-based mobile application proved to be a feasible vehicle for skin cancer mass screening campaigns requiring no significant investment from the public health authorities. The proposed tool provides an efficient way to increase public awareness about skin cancer and associated risk factors as well as encourages people to seek dermatologists’ help in case of skin cancer risk.