ImmunoBench: Pre-Extracted Features from Pathology Whole Slide Images

ImmunoBench provides pre-extracted feature representations from whole slide images for evaluating pathology foundation models. The underlying benchmark integrates data from 13,657 patients, 26,058 WSIs, and over 9.6 million TMA-derived IHC images. The dataset was created by AI4Pathology and was last updated on HuggingFace in May 2026.

Use Cases

• Benchmarking pathology foundation models based on pre-extracted features from whole slide images.
• Training models for IHC-centered pathology analysis based on feature representations.
• Evaluating model performance on a large-scale, multi-source pathology dataset.
• Developing feature-based classifiers or similarity search for pathology image retrieval.

Strengths

• Large-scale integration from 13,657 patients and 26,058 whole slide images.
• Includes over 9.6 million TMA-derived IHC images and 464,294 annotated items.
• Specifically designed for benchmarking pathology foundation models.

Limitations

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

Provenance

Source

ImmunoBench project, aggregated by AI4Pathology.

Collection Method

Pre-extracted feature representations from whole slide images.

Freshness

Last updated 2026-05-25 12:05:30; freshness should be verified.