Integrating “human” diagnostic choices and evaluations significantly improves the accuracy of artificial intelligence used for skin cancer detection: this is demonstrated by the study A reinforcement learning model for AI-based decision support in skin cancer published in Nature Medicines.
In this study, the team of researchers including Iris Zalaudek, full professor of skin and venereal diseases at the University of Trieste and director of the Dermatological Clinic of Asugi, trained an artificial intelligence system using a method called “reinforcement learning”. .
The computer program – explains the Trieste university – has learned, through trial and error, to make accurate diagnoses considering the pros and cons of the various diagnostic choices. Compared to traditional supervised learning methods, the new approach has increased accuracy in detecting melanoma and basal cell carcinoma, reduced overconfident diagnoses, and improved overall patient care.
In detail, the ability to detect melanoma was improved from 61.4% to 79.5% and, for basal cell carcinoma, from 79.4% to 87.1%. The rate of correct diagnoses made by dermatologists increased by 12% and improved the rate of optimal disease management decisions from 57.4% to 65.3%.
These results – concludes the university – suggest that the integration of human skills and sensitivity in medical AI can lead to better diagnostic and healthcare results.
breaking latest news © Copyright ANSA