The GNT-ArrhythmiaNet deep learning model detected life-threatening ventricular arrhythmias with 98.27% accuracy and 0.9845 ROC-AUC, outperforming traditional methods in patients monitored by long-term Holter ECG.
Does a deep learning-based hybrid model improve the early prediction of life-threatening ventricular arrhythmias using long-term Holter ECG signals compared to traditional models?
A deep learning-based hybrid model provides an efficient and highly accurate solution for real-time detection of life-threatening ventricular arrhythmias, suitable for wearable healthcare systems.
Effect estimate: Accuracy 98.27%, Precision 98.08%, Recall (Sensitivity) 98.27%, F1-score 97.76%, ROC-AUC up to 0.9845 for ventricular arrhythmias
The hybrid model is efficient compared with the traditional models and offers an extensible solution to wearable healthcare systems that would have the quality of detecting arrhythmia in real-time with a high degree of accuracy.
Wu et al. (Fri,) conducted a other in Patients at risk of life-threatening ventricular arrhythmias undergoing long-term multi-lead Holter ECG monitoring, including data from Sudden Cardiac Death Holter Database representing a high-risk population. GNT-ArrhythmiaNet hybrid deep learning model integrating graph neural networks and transformer architectures vs. Classical rule-based methods, traditional machine learning algorithms (hidden Markov models, CNN-Bi-LSTM, Bi-LSTM) was evaluated on Accuracy of early prediction and detection of life-threatening ventricular arrhythmias (including ventricular tachycardia and ventricular fibrillation) using long-term multi-lead Holter ECG signals (Accuracy 98.27%, Precision 98.08%, Recall (Sensitivity) 98.27%, F1-score 97.76%, ROC-AUC up to 0.9845 for ventricular arrhythmias). The GNT-ArrhythmiaNet deep learning model detected life-threatening ventricular arrhythmias with 98.27% accuracy and 0.9845 ROC-AUC, outperforming traditional methods in patients monitored by long-term Holter ECG.