Tumor ROIs for: Comprehensive evaluation of cross-cancer generalization in histopathology segmentation models across 21 tumor types

This dataset accompanies the paper "Comprehensive evaluation of cross-cancer generalization in histopathology segmentation models across 21 tumor types."

It contains 7,616 tumor tissue regions of interest (ROIs) extracted from whole-slide images (WSIs) across 21 TCGA cancer types. Rectangular tumor regions were annotated in QuPath and exported as JPEG images. The original ROIs were extracted at native scanner resolution (varying across slides) and used at full size for the segmentation experiments described in the paper. For this deposit the images have been downsampled to a uniform resolution of 0.5 µm/px (approximately 20× equivalent magnification) to reduce the total dataset size while preserving sufficient detail for visual inspection and reuse. These ROIs served as the input for evaluating five organ-specific deep learning segmentation models (trained on breast, colon, lung, kidney, and prostate tissue).

Contents

The dataset is organized as 21 TAR archives, one per TCGA project:

TCGA-BLCA (445 ROIs), TCGA-BRCA (1,007), TCGA-CESC (276), TCGA-CHOL (38), TCGA-COADREAD (590), TCGA-ESCA (157), TCGA-HNSC (456), TCGA-KICH (109), TCGA-KIRC (512), TCGA-KIRP (282), TCGA-LIHC (372), TCGA-LUAD (379), TCGA-LUSC (301), TCGA-MESO (84), TCGA-OV (106), TCGA-PAAD (199), TCGA-PRAD (415), TCGA-SKCM (457), TCGA-STAD (374), TCGA-THCA (504), TCGA-UCEC (553)

Each archive extracts to a directory named after its TCGA project. Individual files follow the naming convention:

{TCGA-ID}_{mpp}_tumor_original.jpg

TCGA-ID — TCGA case and slide identifier

mpp — spatial resolution in microns per pixel (0.500)

tumor_original — indicates an unprocessed tumor tissue ROI

Related datasets

Evaluation data (scoring results, Dice coefficients, clinical metadata): 10.5281/zenodo.18518811

Code (annotation, inference, scoring, and analysis pipeline): 10.5281/zenodo.18520078

Segmentation masks (model prediction masks for all ROIs): 10.5281/zenodo.18669667