CoolFace
Datasetpublic

dellacorte/PANDA-PLUS-Bench

PANDA-PLUS-Bench A benchmark dataset for evaluating WSI-specific feature collapse in pathology foundation models. Dataset Description PANDA-PLUS-Bench contains expert-annotated prostate biopsy patches from 9 whole slide images (9 unique patients) with pixel-level Gleason pattern annotations. Dataset Summary Patches: ~2,770 per augmentation condition Resolution: 224×224 pixels at 20× magnification Classes: Benign (0), GP3 (1), GP4 (2), GP5 (3)… See the full description on the dataset page: https://huggingface.co/datasets/dellacorte/PANDA-PLUS-Bench.

sourceHugging Facecc-by-4.0updated 9mo agoView on Hugging Face
1likes97downloads
Dataset Card

PANDA-PLUS-Bench

A benchmark dataset for evaluating WSI-specific feature collapse in pathology foundation models.

Dataset Description

PANDA-PLUS-Bench contains expert-annotated prostate biopsy patches from 9 whole slide images (9 unique patients) with pixel-level Gleason pattern annotations.

Dataset Summary

  • Patches: ~2,770 per augmentation condition
  • Resolution: 224×224 pixels at 20× magnification
  • Classes: Benign (0), GP3 (1), GP4 (2), GP5 (3)
  • Slides: 9 (one per patient)
  • Augmentations: 8 conditions

Augmentation Conditions

SplitDescription
baselineImageNet normalization only
color_jitterBrightness, contrast, saturation, hue
grayscaleComplete color removal
gaussian_noiseAdditive noise (σ=0.05)
heavy_geometricRotation ±180°, flips
combined_aggressiveAll augmentations combined
macenko_normalizationStain normalization
hedstainaugmentationH/E channel perturbation

Usage

python
from datasets import load_dataset

# Load baseline patches
dataset = load_dataset("dellacortelab/PANDA-PLUS-Bench", split="baseline")

# Access a sample
sample = dataset[0]
image = sample['image']      # PIL Image
label = sample['label']      # 0-3
slide_id = sample['slide_id']  # Slide identifier

Evaluation

See our Colab notebook for standardized evaluation.

Citation

bibtex
@article{ebbert2025pandaplusbench,
  title={PANDA-PLUS-Bench: A Benchmark for Evaluating WSI-Specific Feature Collapse},
  author={Ebbert, Joshua and Della Corte, Dennis},
  year={2025}
}

License

CC-BY-4.0