TCGA-Reports: A Machine-Readable Pathology Report Resource for Benchmarking Text-Based AI Models. Kefeli et al.

This dataset contains 9,523 machine-readable pathology reports originally published by The Cancer Genome Atlas (TCGA) in PDF format. 32 tissues or cancer-types are represented in the final dataset, and reports were generated at a wide variety of institutions. The reports have undergone data selection, OCR-translation, TCGA-artifact removal, form removal, and site-specific section header removal. The final dataset consists of high-quality, clean text with high information content. Please see Kefeli et al. (2024) for additional details.