A unified FLAIR hyperintensity segmentation model for various CNS tumor types and acquisition time points

2025-12-19

AI Summary

Clinical demand exists for automated segmentation of edema and tumor volumes in FLAIR MRI across multiple CNS tumor types (e.g., meningioma, metastasis, glioma) and timepoints (pre-/post-operative), yet existing methods are typically task-specific and lack generalizability. Method: We propose the first unified, cross-disease and cross-timepoint FLAIR hyperintensity segmentation model, built upon an Attention U-Net architecture and trained end-to-end on ~5,000 multi-center, multi-disease, multi-timepoint FLAIR images. The model is integrated into the open-source platform Raidionics. Results: It achieves Dice scores of 61.27%–90.92% on multi-source benchmarks including BraTS, matching specialized models and substantially outperforming conventional single-task approaches. Its core innovation lies in enabling robust FLAIR segmentation without disease- or timepoint-specific modeling—marking the first such general-purpose solution—and thereby significantly enhancing clinical deployability and cross-domain generalization.

Technology Category

• Computer Vision: Segmentation
• Machine Learning: Multimodal Learning
• Knowledge Representation and Reasoning: Diagnosis and Abductive Reasoning

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• User Modeling, Personalization and Recommendation: Fairness-aware retrieval and ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

T2-weighted fluid-attenuated inversion recovery (FLAIR) magnetic resonance imaging (MRI) scans are important for diagnosis, treatment planning and monitoring of brain tumors. Depending on the brain tumor type, the FLAIR hyperintensity volume is an important measure to asses the tumor volume or surrounding edema, and an automatic segmentation of this would be useful in the clinic. In this study, around 5000 FLAIR images of various tumors types and acquisition time points from different centers were used to train a unified FLAIR hyperintensity segmentation model using an Attention U-Net architecture. The performance was compared against dataset specific models, and was validated on different tumor types, acquisition time points and against BraTS. The unified model achieved an average Dice score of 88.65% for pre-operative meningiomas, 80.08% for pre-operative metastasis, 90.92% for pre-operative and 84.60% for post-operative gliomas from BraTS, and 84.47% for pre-operative and 61.27% for post-operative lower grade gliomas. In addition, the results showed that the unified model achieved comparable segmentation performance to the dataset specific models on their respective datasets, and enables generalization across tumor types and acquisition time points, which facilitates the deployment in a clinical setting. The model is integrated into Raidionics, an open-source software for CNS tumor analysis.

Problem

• Develops a unified model for FLAIR hyperintensity segmentation across diverse CNS tumors
• Enables automatic segmentation to assess tumor volume and edema in clinical settings
• Generalizes across different tumor types and acquisition time points for clinical deployment

Innovation

• Unified Attention U-Net model for FLAIR segmentation
• Generalizes across tumor types and time points
• Integrated into open-source Raidionics software