Benchmark RadarSpaPath-Bench
SpaPath-Bench evaluates pathology foundation models on spatial domain identification using paired whole slide images and spatial transcriptomics data from 42 public slides, measuring partition quality via unsupervised spatial coherence, transcriptomics-referenced agreement, and expert-referenced agreement.
- Released
- 2026-05-25
- Readiness
- Inspectable
- Primary field
- Health & Life Sciences
Why it matters
Standard task-level endpoints obscure what pathology embeddings encode about tissue spatial structure. SpaPath-Bench provides a representation-level diagnostic that isolates spatial understanding, enabling model developers to compare encoders and methods on a fixed protocol and choose architectures suited for spatially aware computational pathology.
Motivation
Pathology foundation models (PFMs) have emerged as a core approach for learning transferable representations from whole slide images (WSIs), and they are typically benchmarked through downstream clinical endpoints.
Primary resources
Benchmark Radar records only publicly supported details and links back to primary sources for verification.