Discriminating electrocardiographic responses to His-bundle pacing using machine learning.

Area of Science:

• Cardiology
• Artificial Intelligence
• Medical Imaging

Background:

• His-bundle pacing (HBP) offers physiological ventricular activation compared to conventional pacing.
• Distinguishing between selective His-bundle pacing (S-HBP), non-selective His bundle pacing (NS-HBP), and myocardium-only capture (MOC) on ECGs is challenging.
• Accurate interpretation of HBP ECGs is crucial for effective device management.

Purpose of the Study:

• To develop and validate an AI-driven tool for automated interpretation of HBP ECGs.
• To utilize supervised machine learning, specifically a convolutional neural network (CNN), for this task.
• To automate the differentiation of S-HBP, NS-HBP, and MOC.

Main Methods:

• A convolutional neural network (CNN) was trained on 1297 segmented QRS complexes from 59 patients.
• The dataset included ECG data from S-HBP, NS-HBP, and MOC.
• A 3-fold cross-validation approach was used, with 75% of data for training and 25% for testing.

Main Results:

• The CNN achieved an overall accuracy of 75% on the testing set (17 patients).
• Performance metrics included Cohen kappa of 0.59 (95% CI 0.30 to 0.88; P <.0001).
• Specific accuracies were 67% for S-HBP, 71% for NS-HBP, and 84% for MOC.

Conclusions:

• A neural network can be successfully trained to automate the discrimination of HBP ECG responses.
• This AI approach demonstrates proof of concept for automated HBP ECG analysis.
• Further development with larger datasets could enhance accuracy and clinical utility, aiding HBP implantation and follow-up.