This year, the International Symposium of Biomedical Imaging, IEEE ISBI, took place virtually from April 3 to 7, rather than as planned in Iowa City. The conference is dedicated to mathematical, algorithmic and computational aspects of biological and biomedical imaging.
Dimitrios Lenis and David Major, researchers in our Biomedical Image Informatics research group, presented the paper "Interpreting Medical Image Classifiers by Optimization Based Counterfactual Impact Analysis". The publication introduces the research at VRVis in the field of Explainable AI (XAI) in the field of biomedical imaging. In this area, VRVis is ambitiously researching possibilities of visualizing the classification of medical images in order to make AI-based classification processes more comprehensible and thus optimize them. Experiments with public mammography data based on our findings already show a quantitatively and qualitatively more precise localization and clearer mediation results than the methods available so far.