AI Augmented Training for Skin Specialists

Melanoma, Skin Cancer

Artificial Intelligence, Education, Diagnostic Accuracy, Skin Cancer, Melanoma, Feedback loop

Background: The worldwide incidence of skin cancer has been rising for 50 years, in particular the incidence of malignant melanoma has increased approx. 2-7% annually and is the most common cancer amongst Danes aged 15-34. Currently there is a significant amount of misdiagnosis of skin cancer and mole cancer, and most excised skin lesions are benign. Previous studies have shown that there is no significant increase in doctors diagnostic accuracy during the first 6 years of clinical work. The resources spend on healthy people could be put to better use, if the Benign-Malignant Ratio could be lowered. This could potentially be done by better educating the doctors during their everyday clinical practice. Aim: The aim of this study is to investigate the dose/response effect of an AI augmented training and clinical feedback on the diagnostic accuracy of skin cancer and clinical decisions among doctors from specialized skin cancer centers. Research question: How much specialized doctors need to train before their diagnostic accuracy and clinical decisions change?

Design: This study is a superiority trial designed as an international multicenter randomized controlled trial of doctors in highly specialized centers that diagnose and/or treat skin- and mole cancer. Randomization Eligible participants will be randomized into either the intervention or control group, ratio 1:1. Intervention: The participants of group A are given access to a digital educational online system developed by the research group, are asked to register all skin lesions seen with a registration app (clinical and dermoscopic photos and clinical data), also developed by the research group, and will be given clinical feedback on every registered skin lesion. Participants in group B are also asked from day one to register all skin lesions and will receive feedback on these as the participants of group A, but are withheld their access to the digital educational online system for 2 months. Feedback on removed/biopsied skin lesions is given directly from the pathologist, who in turn are given easy access to photographs and clinical data of the patient and skin lesion in question. Statistics: The average increase in diagnostic accuracy for the population of participating doctors as an effect of the hours spent with the digital educational online system is calculated using Generalized Estimating Equations (GEE). As benign lesions can be excised/treated for other reasons than suspicion of malignancy we will analyze correctly diagnosed benign lesions treated for different reasons (cosmetic or functional complaints etc.) separately. We expect a majority of registered lesions to be benign, despite most of the patients already having been seen by GPs before referral. Ethical considerations: Patient participation contains no immediate strain or discomfort for the patient, and no change to current clinical practice, as dermoscopic evaluation is part of the clinical examination of skin lesions. The images captured are stored safely and anonymously with no risk for the patient. With the current low diagnostic accuracy of young doctors the educational nature of the intervention justifies the study for the sake of all future patients with skin lesions that are less likely to be misdiagnosed. Educational interventions on doctors do not require approval by The National Committee on Health Research Ethics in Denmark. However ethical considerations have been made and the project is in concordance with the Helsinki Declaration II.

No related papers found yet.

Denmark