Detecting multiple sclerosis disease activity – a no-click, AI-based solution for clinicians?
08 December 2021
The detection of disease activity on follow-up brain MR imaging studies critically identifies suboptimal response to disease modifying therapy in patients with MS. For the MS lesion activity MSSEG2 challenge from MICCAI, we presented a novel approach combining self-supervised pre-training to exploit stable voxels, the inclusion of prior knowledge in the form of an attention gate between timepoints and a dual headed Unet architecture. Our technique ranked in the top 3 for segmentation (DSC) and the top 5 for detection (F1 score) from more than 30 teams. Additionally, the preliminary results of our work were recently presented PACTRIMS 2021 by Mariano Cabezas, for which he received the PACTRIMS Young Investigator Award.
For more information please see: A dual headed Unet approach for automatic lesion activity assessment