Explainable MRI: Revealing the mechanisms underlying deep learning brain disease classification
Christian Tinauer1, Lukas Pirpamer1, Marc Masana2, Stefan Heber1, Anna Damulina1, Maximilian Sackl1, Martin Soellradl3, Reinhold Schmidt1, Stefan Ropele1, and Christian Langkammer1
1Department of Neurology, Medical University of Graz, Graz, Austria, 2Institute of Computer Graphics and Vision, Graz University of Technology, Graz, Austria, 3Institute of Medical Engineering, Graz University of Technology, Graz, Austria
Increased iron deposition in the basal ganglia is a frequent finding in patients with AD. Using R2* maps we separated Alzheimer's patients (n=115) from healthy controls (n=169) by using a deep neural network and systematically investigated the influence of the learned features using an attached relevance map generator. The highest relevances were found in and adjacent to the basal ganglia, which is in line with established histological findings and additionally confirmed by the conventional ROI-based analysis. This study demonstrates the validity of heat mapping as a means to identify novel areas of pathological tissue changes.
This abstract and the presentation materials are available to members only;
a login is required.