Dli Path Eval

---
name: dli-path-eval
description: Evaluates the ability of multiple-instance learning models to classify six key pathological indicators (cholestasis, portal fibrosis, inflammation, steatosis, macrovesicular steatosis, hepatocellular ballooning) from donor liver whole slide histopathological images. Use when the user wants to benchmark on DLiPath, or asks about evaluating this task. Reports AUC, Accuracy, Precision, Recall, F1-Score.
metadata:
  skill_kind: dataset_eval
  source_arxiv: 2506.03185
  bibtex_key: pan2025dli_path
  confidence: high
---

# dli-path-eval

> DLiPath: A Benchmark for the Comprehensive Assessment of Donor Liver Based on Histopathological Image Dataset — Pan et al. (2025) (arXiv:2506.03185, 2025)

## What this evaluates

Evaluates the ability of multiple-instance learning models to classify six key pathological indicators (cholestasis, portal fibrosis, inflammation, steatosis, macrovesicular steatosis, hepatocellular ballooning) from donor liver whole slide histopathological images.

## Datasets

- **DLiPath** — total 636; splits: test (-1); repo https://github.com/panliangrui/ACM_MM_2025

## Metrics

- `AUC, Accuracy, Precision, Recall, F1-Score` **(primary)** — range: [0, 1]
  - Standard classification metrics. AUC is the area under the receiver operating characteristic curve. Accuracy is the proportion of correct predictions. Precision, Recall, and F1-Score are computed using standard true/false positive/negative counts. All are reported as macro-averaged or per-class means over five-fold cross-validation.

## Input / output format

**Input**: Whole slide histopathological images (WSIs) of donor liver tissue, represented as bags of image patches for multiple-instance learning.

**Output**: Softmax probability distribution over six pathological indicators (cholestasis, portal fibrosis, inflammation, steatosis, macrovesicular steatosis, hepatocellular ballooning).

## Scoring recipe

```python
# For each fold in 5-fold CV:
y_true = one_hot_encode(labels)
y_pred_proba = model.predict(X)
y_pred = (y_pred_proba >= 0.5).astype(int)
acc = accuracy_score(y_true, y_pred)
prec = precision_score(y_true, y_pred, average='macro')
rec = recall_score(y_true, y_pred, average='macro')
f1 = f1_score(y_true, y_pred, average='macro')
auc = roc_auc_score(y_true, y_pred_proba, multi_class='ovr')
# Average metrics across 5 folds
```

## Common pitfalls

- Metrics are averaged over five-fold cross-validation, not a single fixed train/test split.
- The task is multi-class classification with one-hot labels; metrics must be computed per indicator or macro-averaged.
- Input consists of whole slide images processed via Multiple Instance Learning (bags of patches), not single cropped images.

## Evidence (verbatim from paper)

> In our experiments, we used AUC, Accuracy, Precision, Recall, and F1-Score as evaluation metrics for the model’s predictions of cholestasis, Portal tract fibrosis, portal inflammation, total steatosis, macrovesicular steatosis, and hepatocellular ballooning. Higher values of AUC, Accuracy, Precision, Recall, and F1-Score indicate better model performance. All evaluation metrics are reported as the mean over five-fold cross-validation to evaluate model robustness.

## Citation

```bibtex
@misc{pan2025dli_path,
  title={DLiPath: A Benchmark for the Comprehensive Assessment of Donor Liver Based on Histopathological Image Dataset},
  author={Pan et al. (2025)},
  year={2025},
  note={arXiv:2506.03185}
}
```

- arXiv: 2506.03185