Multi Cancer Histopathology Eval

multi-cancer-histopathology-eval

DSVTLA: Deep Swin Vision Transformer-Based Transfer Learning Architecture for Multi-Type Cancer Histopathological Cancer Image Classification — Khan et al. (2026) (arXiv:2604.09468, 2026)

What this evaluates

Evaluates deep learning models on their ability to classify multi-type cancer histopathological images across six distinct cancer categories. It probes the model's capacity to learn discriminative morphological features and generalize across heterogeneous medical imaging conditions.

Datasets

• Multi-Cancer Histopathology Dataset (Kaggle) — total ?; splits: train (-1), val (-1), test (-1)

Metrics

• accuracy (primary) — range: percent
  • Standard classification accuracy: the proportion of correctly classified instances out of the total number of instances in the test set.
• precision — range: percent
  • Ratio of true positive predictions to the total number of positive predictions for each class.
• recall — range: percent
  • Ratio of true positive predictions to the total number of actual positives for each class.
• F1-score — range: percent
  • Harmonic mean of precision and recall, providing a single metric that balances both concerns.

Input / output format

Input: Histopathological images of cancer tissue (e.g., breast, oral, lung, colon, kidney, acute lymphocytic leukemia).

Output: Predicted cancer class label.

Scoring recipe

def compute_metrics(predictions, gold_labels):
    correct = sum(1 for p, g in zip(predictions, gold_labels) if p == g)
    accuracy = correct / len(gold_labels)
    # Precision, Recall, F1 computed per class using standard sklearn metrics
    return accuracy, precision, recall, f1

Common pitfalls

• Dataset size and exact train/val/test split ratios are not reported, making reproducibility difficult.
• Multiple models report 100% accuracy on certain cancer types (e.g., lung, colon, leukemia), which may indicate overly simple splits or data leakage.
• Results are averaged over three runs, but standard deviations or confidence intervals are omitted, obscuring variance.

Evidence (verbatim from paper)

We employed a set of widely accepted classification metrics accuracy, precision, recall, F1-score to quantify the performance of our models. These metrics collectively provide a comprehensive understanding of the model's strengths and limitations in terms of prediction correctness, class-wise balance, and error distribution.

Citation

@misc{khan2026dsvtla,
  title={DSVTLA: Deep Swin Vision Transformer-Based Transfer Learning Architecture for Multi-Type Cancer Histopathological Cancer Image Classification},
  author={Khan et al. (2026)},
  year={2026},
  note={arXiv:2604.09468}
}

• arXiv: 2604.09468