Computer Vision News - December 2022

50 Medical Imaging Challenge PI-CAI not only focuses on developing and estimating the best AI performance possible today but does it in conjunction with a Reader Study to know the best performance of radiologists overall in the field because one doesn’t exist without the other. With his clinical background, Jasper is more involved in this aspect of the work. The team aims to benchmark the state- of-the-art AI algorithms developed in the challenge against prostate radiologists at the end of the study. The challenge also goes one step further on the technical side. Very often nowadays, “ One big thing that would benefit any team is to really understand the clinical problem and the sort of data they’re dealing with, ” Anindo points out. “ What is prostate MRI good or bad at? What are things in the prostate that lookverysimilar toaggressive cancer ? What are things that are not similar at all but could also be cancer? What are the clinical variables? We’ve included, for example, patient age and a PSA value in blood, which are connected with cancer screening usually. How do these play in, and how can they use this information to reduce false positives or negatives? ”