HepatoBench: 87,301 Liver Histopathology Image Patches for Model Evaluation

87,301 liver tissue image patches across seven diagnostic categories, including Tumor and Fibrosis, curated for evaluating pathology foundation models. The dataset was created by author 'xtxx' and last updated on March 16, 2026. It serves as a benchmark for patch-level image classification tasks.

Use Cases

• Benchmarking pathology foundation models based on patch-level classification performance.
• Training image classifiers for liver disease diagnosis based on histopathology patches.
• Evaluating model generalization across seven distinct liver tissue categories.
• Conducting research on automated analysis of liver histopathology slides.

Strengths

• Contains 87,301 annotated image patches, providing a substantial sample size.
• Curated for a specific benchmark task with seven defined diagnostic classes, including Tumor (13,200 images) and Fibrosis (26,364 images).

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

huggingface

Freshness

Last updated 2026-03-16 08:29:33; freshness should be verified.