Multi-modal 3D brain MRI scans and expert-annotated radiology reports are provided for text-guided tumor segmentation. This dataset integrates volumetric imaging data with natural language descriptions to support multi-modal deep learning in neuro-oncology.
Use Cases
- Train multi-modal models to perform 3D brain tumor segmentation using the radiology reports as textual guidance.
- Benchmark vision-language models on their ability to correlate clinical text in radiology reports with volumetric features in MRI scans.
- Develop cross-modal retrieval systems to find relevant MRI volumes based on specific keywords found in the radiology reports.
Strengths
- Includes paired multi-modal brain MRI scans and expert-annotated radiology reports.
- Provides volume-level text-image alignment for 3D brain tumor segmentation.
- Derived from the BraTS challenge, incorporating standardized neuro-oncology imaging protocols.