An explainable algorithm for detecting drug-induced QT-prolongation at risk of torsades de pointes (TdP) regardless of heart rate and T-wave morphology.

摘要: Torsade de points (TdP), a life-threatening arrhythmia that can increase the risk of sudden cardiac death, is associated with drug-induced QT-interval prolongation on electrocardiogram (ECG). While many modern ECG machines provide automated measurements QT-interval, these QT values are usually correct only for noise-free normal sinus rhythm, in which T-wave morphology well defined. As QT-prolonging drugs often affect T-wave, taken under circumstances easily invalidated. An additional challenge QT-value at TdP varies heart rate, slower greater TdP. This paper presents an explainable algorithm uses understanding human visual perception and expert interpretation to automate detection QT-prolongation regardless rate morphology. It was tested large number ECGs (n=5050) variable QT-intervals varying rates, acquired from clinical trial assessed effect four known versus placebo healthy subjects. The yielded balanced accuracy 0.97, sensitivity 0.94, specificity 0.99, F1-score 0.88, ROC (AUC) 0.98, precision-recall Matthews correlation coefficient (MCC) 0.88. results indicate prolonged ventricular repolarisation area be significant predictor TdP, this potentially easier more reliable than measuring distance directly. proposed visualised using pseudo-colour trace, thus intuitively 'explaining' how its decision made, focus group show may help people self-monitor QT-prolongation, as ensuring clinicians validate results.