---
license: cc-by-4.0
task_categories:
- image-classification
tags:
- pathology
- histopathology
- prostate-cancer
- gleason-grading
- foundation-models
- benchmark
size_categories:
- 1K<n<10K
dataset_info:
  features:
  - name: image
    dtype: image
  - name: label
    dtype: int64
  - name: slide_id
    dtype: string
  splits:
  - name: baseline
    num_bytes: 298170963
    num_examples: 3872
  - name: color_jitter
    num_bytes: 287314003
    num_examples: 3872
  - name: grayscale
    num_bytes: 167680581
    num_examples: 3872
  - name: gaussian_noise
    num_bytes: 458559156
    num_examples: 3872
  - name: heavy_geometric
    num_bytes: 294975593
    num_examples: 3872
  - name: combined_aggressive
    num_bytes: 449136407
    num_examples: 3872
  - name: macenko_normalization
    num_bytes: 303746938
    num_examples: 3872
  - name: hed_stain_augmentation
    num_bytes: 296645574
    num_examples: 3872
  download_size: 2556068549
  dataset_size: 2556229215
configs:
- config_name: default
  data_files:
  - split: baseline
    path: data/baseline-*
  - split: color_jitter
    path: data/color_jitter-*
  - split: grayscale
    path: data/grayscale-*
  - split: gaussian_noise
    path: data/gaussian_noise-*
  - split: heavy_geometric
    path: data/heavy_geometric-*
  - split: combined_aggressive
    path: data/combined_aggressive-*
  - split: macenko_normalization
    path: data/macenko_normalization-*
  - split: hed_stain_augmentation
    path: data/hed_stain_augmentation-*
---

# 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

| Split | Description |
|-------|-------------|
| baseline | ImageNet normalization only |
| color_jitter | Brightness, contrast, saturation, hue |
| grayscale | Complete color removal |
| gaussian_noise | Additive noise (σ=0.05) |
| heavy_geometric | Rotation ±180°, flips |
| combined_aggressive | All augmentations combined |
| macenko_normalization | Stain normalization |
| hed_stain_augmentation | H/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](https://colab.research.google.com/github/dellacortelab/PANDA-PLUS-Bench/blob/main/PANDA_PLUS_Bench_Evaluation.ipynb) 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