A research project in the field of artificial intelligence (AI) and healthcare, led by ISR-Lisboa and Instituto Superior Técnico researcher Catarina Barata, was published today in Nature Medicine. The research put AI to the service of skin cancer detection, with a method that managed to increase by 12% the rate of correct diagnoses by dermatologists.

AI has shown great accuracy in several fields of diagnostic imaging, leading to the development of decision-support tools and creating possibilities for improvement in healthcare accessibility. This study focused on the use of a specific model called “reinforcement learning”, which was cross-referenced with expert-generated tables with rewards and penalties assigned for different diagnostic errors. When compared to traditional artificial learning this reinforcement learning model produced notable improvements in sensitivity for two skin diseases: melanoma and basal cell carcinoma. The sensitivity for melanoma diagnosis rose from 61.4% to 79.5%, while for basal cell carcinoma, it rose from 79.4% to 87.1%.

The team of researchers also found that this AI model can reduce overconfidence in automated diagnosis while maintaining its accuracy. This could be an important game-changer in the world of medical AI, as overconfidence in such diagnoses has been a cause for concern. Catarina Barata explained that this type of approach is increasingly essential. “The use of AI should be tailored to people and bring them benefit. This model with rewards makes the process simpler and easier to understand for a non-expert, which is very much in line with the direction we want to take with AI.” The researcher is also involved in the international project “Responsible AI”, which aims to implement a series of reflective practices and concerns about the accessibility of this type of technology.