Qcnmri Tumor Classification Eval

---
name: qcnmri-tumor-classification-eval
description: Evaluates a hybrid quantum-classical convolutional neural network on MRI-based brain tumor detection. It probes the model's ability to classify medical images into binary (tumor vs. non-tumor) and multiclass (specific tumor types) categories under class imbalance and limited resolution constraints. Use when the user wants to benchmark on Brain MRI Tumor Dataset, or asks about evaluating this task. Reports accuracy.
metadata:
  skill_kind: dataset_eval
  source_arxiv: 2509.02582
  bibtex_key: [REDACTED]
  confidence: high
---

# qcnmri-tumor-classification-eval

> Application of Quantum Convolutional Neural Networks for MRI-Based Brain Tumor Detection and Classification — Nugraha et al. (2025) (arXiv:2509.02582, 2025)

## What this evaluates

Evaluates a hybrid quantum-classical convolutional neural network on MRI-based brain tumor detection. It probes the model's ability to classify medical images into binary (tumor vs. non-tumor) and multiclass (specific tumor types) categories under class imbalance and limited resolution constraints.

## Datasets

- **Brain MRI Tumor Dataset** — total 3264; splits: train (2611), test (653)

## Metrics

- `accuracy` **(primary)** — range: percent
  - Fraction of correctly classified instances out of the total number of instances. Calculated as (TP + TN) / (TP + TN + FP + FN).
- `precision` — range: percent
  - Ratio of true positive predictions to the total number of positive predictions. Calculated as TP / (TP + FP).
- `sensitivity` — range: percent
  - Also known as recall or true positive rate. Ratio of true positive predictions to the total number of actual positives. Calculated as TP / (TP + FN).
- `F1-score` — range: percent
  - Harmonic mean of precision and sensitivity. Calculated as 2 * (precision * sensitivity) / (precision + sensitivity).

## Input / output format

**Input**: Grayscale MRI images resized to 12x12 pixels, encoded into qubits using a 2x2 filter size and repeated iteratively.

**Output**: Probability distribution over classes via Softmax (multiclass) or a single probability via Sigmoid (binary), thresholded to output class labels (0/1 for binary, 0-3 for multiclass).

## Scoring recipe

```python
def compute_metrics(y_true, y_pred):
    tp = np.sum((y_true == 1) & (y_pred == 1))
    tn = np.sum((y_true == 0) & (y_pred == 0))
    fp = np.sum((y_true == 0) & (y_pred == 1))
    fn = np.sum((y_true == 1) & (y_pred == 0))
    accuracy = (tp + tn) / (tp + tn + fp + fn)
    precision = tp / (tp + fp) if (tp + fp) > 0 else 0
    sensitivity = tp / (tp + fn) if (tp + fn) > 0 else 0
    f1 = 2 * precision * sensitivity / (precision + sensitivity) if (precision + sensitivity) > 0 else 0
    return accuracy, precision, sensitivity, f1
```

## Common pitfalls

- Oversampling is applied only to the training set, which must be strictly separated from validation/test to prevent data leakage.
- Images are downsampled to 12x12 pixels, discarding fine medical features and capping maximum performance.
- Quantum circuit uses fixed shallow depth (4) and only CX/CZ gates, limiting expressiveness compared to deeper classical baselines.

## Evidence (verbatim from paper)

> The model was evaluated with standard metrics: accuracy, precision, sensitivity, and F1-score. The model trained on the original imbalanced dataset only achieved an accuracy of 52%, sensitivity of 49%, and F1-score equivalent to 50%. Conversely, the model trained on the oversampled dataset improved to 62% in accuracy, 65% in precision, 62% in sensitivity, and 63% in F1-score.

## Citation

```bibtex
@misc{nugraha2025qcnnmri,
  title={Application of Quantum Convolutional Neural Networks for MRI-Based Brain Tumor Detection and Classification},
  author={Nugraha et al. (2025)},
  year={2025},
  note={arXiv:2509.02582}
}
```

- arXiv: 2509.02582