Towards Trustworthy Automatic Diagnosis Systems by Emulating Doctors' Reasoning with Deep Reinforcement Learning

Explore the development of trustworthy automatic diagnosis systems through deep reinforcement learning in this 30-minute conference talk from the Toronto Machine Learning Series. Delve into the challenges of automating medical evidence acquisition and diagnosis processes, focusing on emulating doctors' reasoning rather than solely improving prediction accuracy. Learn about the importance of generating differential diagnoses, prioritizing severe pathologies, and providing explainable recommendations to gain doctors' trust. Discover a novel approach that models evidence acquisition and automatic diagnosis using deep reinforcement learning, considering three essential aspects of a doctor's reasoning. Examine metrics for evaluating interaction quality and compare the proposed solution's performance against existing models while maintaining competitive pathology prediction accuracy. Gain insights from Arsene Fansi Tchango, Senior Applied Research Scientist at Mila - Quebec Institute in Artificial Intelligence, on advancing the field of automated medical diagnosis systems.

Towards Trustworthy Automatic Diagnosis Systems by Emulating Doctors’ Reasoning with Deep Reinforcem