Deep Learning-Based Electrocardiograph in Evaluating Radiofrequency Ablation for Rapid Arrhythmia

Area of Science:

• Cardiology
• Medical Imaging
• Artificial Intelligence

Background:

• Tachyarrhythmia treatment often involves radiofrequency ablation.
• Evaluating the efficacy of radiofrequency ablation can be challenging.
• Electrocardiography (ECG) provides valuable diagnostic information.

Purpose of the Study:

• To analyze the role of deep learning-based ECG in assessing radiofrequency ablation efficacy for tachyarrhythmia.
• To compare ECG indicators between effective and ineffective treatment groups.

Main Methods:

• 158 patients undergoing radiofrequency ablation for tachyarrhythmia were studied.
• A deep learning convolutional neural network model quantified ECG indicators.
• ECG parameters were compared between effective (142 patients) and ineffective (16 patients) treatment groups.

Main Results:

• Effective treatment showed significantly decreased end-systolic volume (ESV), end-diastolic volume (EDV), ESV index (ESVI), EDV index (EDVI), and increased left ventricular ejection fraction (LVEF).
• Ventricular rate was significantly lower at 12h and 24h post-ablation in the effective group.
• Higher QT dispersion was observed in the effective treatment group.
• ECG evaluation achieved 86.81% accuracy, 84.29% specificity, and 77.27% sensitivity, with an ROC AUC of 0.798.

Conclusions:

• Deep learning-based ECG analysis provides reliable auxiliary reference information for radiofrequency ablation efficacy evaluation in tachyarrhythmia.
• Quantified ECG indicators can differentiate treatment success.
• This approach aids in optimizing tachyarrhythmia management.