TCGA Multimodal Oncology Data with Clinical, Imaging, and Molecular Features

Functioning as a multimodal collection of clinical data, pathology reports, slide images, molecular data, and radiology images for cancer patients, created by Lab-Rasool and updated in July 2025. It includes embeddings generated using models like GatorTron, MedGemma, Qwen, Llama, UNI, SeNMo, and REMEDIS. Specific row counts, column counts, and file formats are not provided.

Use Cases

• Train multimodal models to correlate pathology slide images with clinical data for patient outcome prediction.
• Analyze molecular data alongside radiology images to identify biomarkers for specific cancer types.
• Use clinical data and pathology reports to build models for cancer diagnosis or treatment planning.
• Fine-tune language models like GatorTron or Llama on oncology-specific text from pathology reports.

Strengths

• Multimodal scope integrates clinical data, pathology reports, slide images, molecular data, and radiology images.
• Includes embeddings from multiple state-of-the-art models such as GatorTron, MedGemma, Qwen, and Llama.
• Dataset was updated on July 8, 2025, indicating recent maintenance.

Limitations

• The dataset's size, row count, column definitions, and specific file formats are unknown.
• The description is sourced from a single dataset card, and the completeness or representativeness of the data is unclear.
• The license, organization of origin, and detailed data collection methodology are not specified.

Provenance

Source

The Cancer Genome Atlas (TCGA), via Lab-Rasool on Hugging Face.

Collection Method

Collection of clinical data, pathology reports, slide images, molecular data, and radiology images, with embeddings generated using specified models.

Freshness

Last updated on 2025-07-08.