Making AI Fit for Ethnic Diversity in Dermatology
Prof. Alexander Navarini and Philippe Gottfrois, PhD student in Navarini’s research group "Digital Dermatology" at the DBE, are training algorithms for the diagnosis of skin diseases in Africa. Collaborating with researchers in Madagascar, Malawi and Guinea. They presented their finding at the ‹International Conference on Medical Image Computing and Computer Assisted Intervention MICCAI 2024› in Marrakesh.
Within the PASSION project, Navarini and his colleagues are addressing the crucial aspect of bias in artificial intelligence: since it can only work on what it has been trained with, analyzing skin diseases in Africa with models trained on European skin types leads to wrong diagnoses. This is why Navarini and Gottfrois have teamed up with researchers from Antananarivo (Madagascar), Konakry (Guniea), Lilongwe (Malawi), and Moshi (Tanzania), as well as Geneva, Lucerne, and Basel, to gather 4,901 images of 1,653 African patients and provide a baseline machine learning model for the image-based diagnosis eczema, fungals, scabies, and impetigo in different subpopulations.
More: