Refined TCGA PRAD Prostate Cancer Pathology Dataset with Enhanced Gleason Grading

A December 2024 release by Codatta provides enhanced Gleason grading annotations for prostate cancer pathology images from the TCGA PRAD cohort. The dataset includes 245 cases identified as having potential for improvement, to be uploaded in two batches of 138 and 107 cases. The second batch is planned for upload around early January 2025.

Use Cases

• Train machine learning models for Gleason grading based on enhanced ROI-level annotations.
• Validate prostate cancer diagnostic algorithms against refined pathology labels.
• Benchmark image analysis models on a dataset with improved ground truth for prostate tissue.

Strengths

• Focuses on 245 TCGA cases specifically identified as having potential for label improvement.
• Provides ROI-level annotations, which likely offer more granular detail than slide-level labels.
• The dataset is described as an enhanced 'second opinion' on existing TCGA PRAD labels.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• The full dataset is not yet available, with 107 cases still in a quality review pipeline.

Provenance

Source

TCGA PRAD cohort, refined by Codatta.

Collection Method

Enhanced annotations applied to existing TCGA pathology images.

Freshness

Last updated 2024-12-21 17:01:15.