PathBench Logo

A Multi-task, Multi-organ Benchmark for Real-world Clinical Performance Evaluation of Pathology Foundation Models

0+
Foundation Models
0+
Clinical Tasks
0+
Organ Systems
0+
Task Types

Why PathBench?

🔬

Comprehensive Model Coverage

Evaluate 20+ state-of-the-art pathology foundation models including UNI, Virchow, CONCH, Prov-GigaPath, CHIEF, and more.

🏥

Real Clinical Data

Multi-organ coverage across breast, lung, colorectal, prostate, kidney, and other major organs with real-world clinical datasets.

📊

Diverse Task Types

Classification, survival prediction (OS, DFS, DSS), IHC marker prediction, histological grading, and report generation.

🔄

Rigorous Evaluation

10 validation runs with varied seeds, standardized framework, identical hardware and preprocessing for reproducible results.

📈

Interactive Analytics

Dynamic leaderboards, performance heatmaps, comparative charts, and filtering by organ, task type, and metrics.

🌍

Global Collaboration

Collaboration with hospitals worldwide, covering top-10 cancers with strict privacy guardrails and secure evaluation.

Overall Model Rankings

Average rank across all tasks (lower is better)

How to Use PathBench

1

Explore the Leaderboard

Browse comprehensive rankings of pathology foundation models across different tasks and organs. Use filters to focus on specific areas of interest.

2

Analyze Performance

Dive deep into detailed performance metrics, visualizations, and comparative analysis. Examine model behavior across different clinical scenarios.

3

Submit Your Model

Have a new pathology foundation model? Submit a PR to our evaluation repository and we'll benchmark it against our comprehensive test suite.