Transparency in Diagnosis: How XAI Helps Doctors Interpret Deep Learning – Hugo Gävert, DAIN Studios

Session Outline

Deep learning has revolutionized image recognition, with models that can recognize cancer and other diseases as well as or better than human experts. However, the lack of transparency in these models can lead to confusion and even wrong diagnoses and treatments. In this talk, we explore how Explainable AI (XAI) methods can help doctors in making right diagnoses. We will specifically look at a case, where we built a 3D CNN to help diagnose cardiac sarcoidosis and how the added XAI layer can help in the diagnosis of the 3D PET images of the heart.