Hi All, Hope you all doing good. When analyzing the training data, I observed that there is an segmentation label overlap between left hippocampus and left lentiform nucleus for training data pair 33 (LISA_0033_HF_hipp.nii.gz, LISA_0033_HF_baga.nii.gz) in the following voxels: [73, 73, 73], [119, 120, 121], [72, 71, 71]. It might not be a problem when building independent models for Subtask 2a, Subtask 2b. But for an integrated model, which label should we consider for these locations? Regards, Sharat

Created by Sharat Embrandiri sharat.s.e
Hi @LISA_mri_challenge, Thanks for the clarification! Best, Vicent
Hi @sharat.s.e and @vcasellesb, Thanks for raising the question on hippocampus/basal ganglia overlap. After an internal review and discussion with our expert labeler, the voxels in question should be considered basal ganglia (lentiform nucleus). For consistency, please assign such overlaps to the basal ganglia in your models. Best, The LISA 2025 Challenge Organizers
Hi! I have also detected this overlap, and was also building a single model to segment everything, I am wondering what would be the best approach to solve this.. Vicent

Segmentation overlap in training GT page is loading…