A Novel Detection Method of Bundle Branch Block from Multi-lead ECG

Area of Science:

• Cardiology
• Biomedical Engineering
• Signal Processing

Background:

• Bundle branch block (BBB) is a prevalent cardiac conduction disorder.
• Electrocardiogram (ECG) is a standard diagnostic tool for BBB.
• Accurate differentiation of right BBB (RBBB) and left BBB (LBBB) is clinically significant.

Purpose of the Study:

• To develop and validate a novel method for detecting RBBB and LBBB.
• To improve the accuracy of BBB detection using a hybrid feature approach.
• To evaluate the proposed method on established arrhythmia databases.

Main Methods:

• A novel detection method combining deep learning features and expert-defined features was proposed.
• The method was evaluated using the MIT-BIH Arrhythmia database (AR) and the China Physiological Signal Challenge 2018 database (CPSC).
• Performance was assessed through class-oriented and subject-oriented evaluations.

Main Results:

• The proposed method achieved high accuracy: 99.96% (AR, class-oriented) and 98.76% (AR, subject-oriented).
• Excellent performance was also demonstrated on the CPSC database with 97.88% accuracy (subject-oriented).
• The method outperformed existing baseline approaches in BBB detection.

Conclusions:

• The novel hybrid feature method offers a highly accurate and reliable approach for detecting RBBB and LBBB.
• This technique shows significant potential for clinical application in diagnosing BBB.
• The study validates the effectiveness of combining deep and expert features for ECG-based cardiac condition detection.