NeuroWeave-3D: Uncertainty-Aware Evidence Completion for Brain Tumor Segmentation with Missing MRI Modalities

NeuroWeave-3D enables 3D glioma segmentation under missing MRI modalities by treating absent sequences as uncertain evidence, with voxel-level uncertainty estimation.

• Why it matters: Multi-parametric MRI models can fail when required sequences are missing or corrupted, creating a need for segmentation that remains reliable under arbitrary modality availability.
• What they did: The framework uses spatial modality routing, probabilistic latent feature completion, complete-to-missing consistency, and a curriculum training one network across all non-empty modality subsets.
• The result: A BraTS-centered evaluation reports complete-input accuracy plus macro and worst-case missing-modality performance, corruption robustness, calibration, and failure detection.

AI summary of the abstract · may contain errors · not medical advice · read the paper.