Bram van Ginneken
Radboud University Medical Center
Nijmegen, The Netherlands
Some say computer vision is solved. What does this mean for doctors who earn their money looking at images and reporting what they see?
In this lecture, I will show recent results of deep learning on radiology, pathology, and ophthalmology. I will discuss the remarkably slow take-up of these new technologies and ponder upon the longer-term consequences and opportunities for improving our healthcare systems. Open problems and new research directions, such as effective data annotation, image construction, effective detection of rare abnormalities, and systems that can do reasoning will be identified.