Ecg Arrhythmia Detection Eval

ecg-arrhythmia-detection-eval

rECGnition_v1.0: Arrhythmia detection using cardiologist-inspired multi-modal architecture incorporating demographic attributes in ECG — Shreya Srivastava et al. (2024) (arXiv:2410.18985, 2024)

What this evaluates

Evaluates a deep learning model's ability to classify cardiac arrhythmias from ECG signals, testing both intra-dataset performance and cross-dataset generalization using demographic attributes.

Datasets

• MITDB — total ?; splits: train (-1), test (-1)
• INCARTDB — total ?; splits: train (-1), test (-1)
• EDB — total ?; splits: train (-1), test (-1)

Metrics

• F1-score (primary) — range: [0, 1]
  • Harmonic mean of Precision and Sensitivity: 2 * (Precision * Sensitivity) / (Precision + Sensitivity). Reported per-class and macro-averaged.
• Accuracy — range: [0, 1]
  • Ratio of correctly classified instances to total instances.
• Precision — range: [0, 1]
  • Ratio of true positive predictions to all positive predictions (TP / (TP + FP)).
• Sensitivity — range: [0, 1]
  • Ratio of true positive predictions to all actual positives (TP / (TP + FN)).

Input / output format

Input: 2D image representations of ECG beats concatenated with patient demographic attributes (age, gender, height, weight, BMI).

Output: Predicted class label for each heartbeat (e.g., N, L, R, V, /, A, f, F, j, a, AB) or demographic category.

Scoring recipe

def compute_f1(predictions, gold, classes):
    precisions = []
    recalls = []
    for cls in classes:
        tp = sum(1 for p, g in zip(predictions, gold) if p == cls and g == cls)
        fp = sum(1 for p, g in zip(predictions, gold) if p == cls and g != cls)
        fn = sum(1 for p, g in zip(predictions, gold) if p != cls and g == cls)
        prec = tp / (tp + fp) if (tp + fp) > 0 else 0
        rec = tp / (tp + fn) if (tp + fn) > 0 else 0
        precisions.append(prec)
        recalls.append(rec)
    f1s = [2 * p * r / (p + r) if (p + r) > 0 else 0 for p, r in zip(precisions, recalls)]
    return sum(f1s) / len(f1s)

Common pitfalls

• Class imbalance significantly impacts F1-scores for rare beats (e.g., Fusion beats f/F, Nodal escape j).
• Transferability tests vary ECG leads (V1 vs II), causing minor performance shifts that must be accounted for.
• Demographic prediction tasks (UCIDB) use a separate XGBoost classifier, not the main rECGnition_v1.0 architecture.

Evidence (verbatim from paper)

rECGnition_v1.0 achieved an overall F1-score of 0.9855 with a prediction accuracy of 98.56% (Table 5).

Citation

@misc{srivastava2024recgnition,
  title={rECGnition_v1.0: Arrhythmia detection using cardiologist-inspired multi-modal architecture incorporating demographic attributes in ECG},
  author={Shreya Srivastava et al. (2024)},
  year={2024},
  note={arXiv:2410.18985}
}

• arXiv: 2410.18985