vpasx/lgg-mri-segmentation-research

LGG Brain MRI Segmentation with Genomic Clusters

This repository provides a Patient-Centric version of the Lower-Grade Glioma (LGG) Segmentation dataset. While other versions of this data exist, they often treat slices as independent images. This version preserves the 3D patient volume and integrates all genomic/clinical labels directly into a multimodal-ready format.

🌟 Why This Version?

Developed for Multimodal AI Research, this dataset addresses several limitations found in previous releases:

1. 1.Patient-Level Grouping: All slices belonging to a single patient are grouped into one record.
2. 2.Preserved Raw Data: Images are kept in their original 3-channel .tif format (16-bit) to prevent data loss from 8-bit PNG/JPG conversion.
3. 3.Integrated Metadata: Genomic clusters (RNASeq, miRNA, Methylation) and clinical outcomes are linked directly to each image stack.
4. 4.LLM Optimized: Structured via metadata.jsonl to allow seamless loading for Vision-Language Models (VLM).

🩺 Dataset Description

This dataset contains brain MR images together with manual FLAIR abnormality segmentation masks for 110 patients included in The Cancer Genome Atlas (TCGA) lower-grade glioma collection.

• —Total Patients: 110
• —Total Files: more than 10,000 images (Slices + Masks)
• —Source: The Cancer Imaging Archive (TCIA) and The Cancer Genome Atlas (TCGA).
• —Imaging Modalities: 3-channel images:
• —Channel 1: Pre-contrast
• —Channel 2: FLAIR (Fluid-attenuated inversion recovery)
• —Channel 3: Post-contrast
• —(Note: Missing sequences were replaced with the FLAIR sequence to maintain 3-channel consistency).

🧬 Metadata & Labels

The data.csv (and metadata.jsonl) includes 18 columns of clinical and molecular data:

• —Genomic Subtyping: RNASeqCluster, MethylationCluster, miRNACluster, CNCluster, RPPACluster, COCCluster.
• —Pathology: histological_type, neoplasm_histologic_grade (WHO II vs III).
• —Demographics: age_at_initial_pathologic, gender, race, ethnicity.
• —Outcomes: death01 (Survival status).

📂 Data Structure

The repository is organized to keep images and masks paired within patient-specific directories:

.
├── data.csv
├── metadata.jsonl
└── data/
    └── TCGA_<institution>_<patient_id>_<date>/
        ├── TCGA_..._<slice_number>.tif      (MRI Image)
        └── TCGA_..._<slice_number>_mask.tif (Binary Mask)

Citation

If you use this dataset in your research, please cite the original publications:

@article{buda2019association,
  title={Association of genomic subtypes of lower-grade gliomas with shape features automatically extracted by a deep learning algorithm},
  author={Buda, Mateusz and Saha, Ashirbani and Mazurowski, Maciej A},
  journal={Computers in Biology and Medicine},
  volume={109},
  pages={218--225},
  year={2019},
  publisher={Elsevier},
  doi={10.1016/j.compbiomed.2019.05.002}
}

@article{mazurowski2017radiogenomics,
  title={Radiogenomics of lower-grade glioma: algorithmically-assessed tumor shape is associated with tumor genomic subtypes and patient outcomes in a multi-institutional study with The Cancer Genome Atlas data},
  author={Mazurowski, Maciej A and Clark, Kal and Czarnek, Nicholas M and Shamsesfandabadi, Parisa and Peters, Katherine B and Saha, Ashirbani},
  journal={Journal of Neuro-Oncology},
  volume={133},
  number={1},
  pages={27--35},
  year={2017},
  publisher={Springer},
  doi={10.1007/s11060-017-2420-1}
}