Diagnosing Left Bundle Branch Block in 12-lead Electrocardiogram using Self-Attention Convolutional Neural Networks

ABSTRACT The left bundle branch block is a cardiac conduction disorder that affects the heart’s electrical system. It causes the left ventricle of the heart to contract later than the right ventricle, leading to an irregular heartbeat. The diagnosis of left bundle branch block is crucial in determining the most effective treatment for heart irregularities, including cardiac resynchronization therapy. Cardiac resynchronization therapy uses a pacemaker-like device to resynchronize the heart’s contractions and improve its function. However, diagnosing left bundle branch block accurately can be challenging using traditional diagnostic methods that rely on electrocardiograms. This study introduces Self-Attention Convolutional Neural Networks for the detection of left bundle branch block from 12-lead electrocardiograms data using SE-Residual blocks and a self-attention mechanism to highlight important input data features for more accurate diagnosis of left bundle branch block. The model is trained and validated on a database of 10,344 12-lead electrocardiogram samples using a 10-fold cross-validation approach. The results demonstrate an accuracy of 98.91% ± 0.1%, specificity of 99.28% ± 0.15%, precision of 73.09% ± 3.5%, recall or sensitivity of 82.83% ± 4.34%, F1 score of 77.5% ± 1.59%, and area under the receiver operating characteristics curve of 0.91 ± 0.02. The experimental results demonstrate that the proposed deep learning model achieves high accuracy, specificity, and F1 score. These findings suggest that the proposed model can serve as an effective diagnostic tool for identifying left bundle branch block with a high level of efficiency, improving the diagnostic process, and promoting early treatment in medical settings.