Towards Artificial Intelligence Powered Medical Imaging

Machine learning in medical image computing.

Medical image computing, powered by machine learning, started to yield algorithmic solutions for interpreting radiological images that can perform as well as experts in a number of tasks. In this talk, I will first discuss the need for algorithmic solutions for interpreting medical images and extracting clinically relevant information. Specifically, I will argue that we need algorithms to help human experts overcome the increasing work-load in the clinics and improve accuracy. Then I will talk about the advances the field of medical image computing has made over the last two decades to this end. Despite the advances, there still remains a number of critical tasks where algorithms’ performance is far from human quality. I will discuss some such tasks and conclude with an outlook for the future and the potential role we can algorithms to play.