Refined TCGA PRAD: Enhanced Prostate Cancer Pathology Labels with ROI Annotations

245 TCGA prostate cancer cases identified for label improvement, with a planned two-phase upload in early 2025. This dataset provides enhanced Gleason grading annotations for the TCGA PRAD cohort, created by Humanbased-AI. The dataset page was last updated on 2024-12-21.

Use Cases

• Train machine learning models for Gleason pattern recognition based on enhanced pathology labels.
• Validate automated prostate cancer grading algorithms against refined region-of-interest (ROI) annotations.
• Study inter-observer variability and label quality in cancer pathology datasets.
• Benchmark model performance on a subset of TCGA PRAD cases with potential for improved ground truth.

Strengths

• Provides enhanced annotations for 245 TCGA PRAD cases, aiming to improve upon original labels.
• Annotations are at the region-of-interest (ROI) level, offering granular detail for model training.

Limitations

• Dataset completeness is partial; only 138 cases are in the first batch, with 107 pending review and upload in early 2025.
• Column-level documentation and sample data are unavailable, making detailed assessment difficult prior to download.
• The description references a full page on Hugging Face; core metadata like license, file formats, and size is missing from the provided input.

Provenance

Source

TCGA (The Cancer Genome Atlas) PRAD (Prostate Adenocarcinoma) cohort.

Collection Method

Refinement of existing labels by Humanbased-AI, likely involving expert pathology review.

Freshness

Last updated 2024-12-21 17:01:15; a major update is planned for early 2025.