![[2602.20994] Multimodal MRI Report Findings Supervised Brain Lesion Segmentation with Substructures](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.20994] Multimodal MRI Report Findings Supervised Brain Lesion Segmentation with Substructures
Summary
This paper presents a novel approach to brain lesion segmentation in MRI scans using report-supervised learning, enhancing accuracy by integrating qualitative and quantitative findings.
Why It Matters
Accurate segmentation of brain lesions is crucial for effective diagnosis and treatment planning in neurology. This study addresses limitations in existing methods by utilizing radiology reports to improve segmentation outcomes, which could lead to better patient care and research advancements in medical imaging.
Key Takeaways
- Introduces a report-supervised learning method (MS-RSuper) for brain lesion segmentation.
- Utilizes both qualitative and quantitative data from MRI reports to enhance segmentation accuracy.
- Demonstrates significant performance improvement over traditional methods on a large dataset.
Electrical Engineering and Systems Science > Image and Video Processing arXiv:2602.20994 (eess) [Submitted on 24 Feb 2026] Title:Multimodal MRI Report Findings Supervised Brain Lesion Segmentation with Substructures Authors:Yubin Ge, Yongsong Huang, Xiaofeng Liu View a PDF of the paper titled Multimodal MRI Report Findings Supervised Brain Lesion Segmentation with Substructures, by Yubin Ge and Yongsong Huang and Xiaofeng Liu View PDF HTML (experimental) Abstract:Report-supervised (RSuper) learning seeks to alleviate the need for dense tumor voxel labels with constraints derived from radiology reports (e.g., volumes, counts, sizes, locations). In MRI studies of brain tumors, however, we often involve multi-parametric scans and substructures. Here, fine-grained modality/parameter-wise reports are usually provided along with global findings and are correlated with different substructures. Moreover, the reports often describe only the largest lesion and provide qualitative or uncertain cues (``mild,'' ``possible''). Classical RSuper losses (e.g., sum volume consistency) can over-constrain or hallucinate unreported findings under such incompleteness, and are unable to utilize these hierarchical findings or exploit the priors of varied lesion types in a merged dataset. We explicitly parse the global quantitative and modality-wise qualitative findings and introduce a unified, one-sided, uncertainty-aware formulation (MS-RSuper) that: (i) aligns modality-specific qualitative cues...