A new H&E benchmark with 14 fine-grained cell types and biological domain splits shows pathology foundation models scoring around 0.20 to 0.28 macro F1, far below their near-saturated performance on older benchmarks.
In: Medical Imaging with Deep Learning (2024)
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CV 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
PhenoBench: A Comprehensive Benchmark for Cell Phenotyping
A new H&E benchmark with 14 fine-grained cell types and biological domain splits shows pathology foundation models scoring around 0.20 to 0.28 macro F1, far below their near-saturated performance on older benchmarks.