Computer-Aided Bundle Branch Block Detection Using Symbolic Features of ECG Signal

Bundle Branch Block (BBB) is a cardiac disorder that arises due to the blockage of transmission of electrical impulses in the heart. The detection of BBB is necessary to know the exact condition and accurate diagnosis and treatment of the human heart. An electrocardiogram (ECG) is the most trusted noninvasive tool to reveal heart irregularities. This work extracts the novel set of fourteen morphological and two temporal features from each ECG beat for efficient BBB detection. 89,373 ECG beats from the MIT-BIH Arrhythmia Database (MIT-BIH DB) have been used to validate the proposed technique. The K-nearest neighbor (KNN) classifier has been employed to classify the ECG beats into Normal, Left-BBB, and Right-BBB beats. The performance of the KNN classifier is analyzed for different nearest neighbor values (K), distance metrics, and two different classification strategies, i.e., tenfold cross-validation and hold-out strategy. The best performance is obtained for cityblock distance and nearest neighbor value K = 1 with overall Sen = 99.76 \(\%\) , + P = 99.80 \(\%\) , Spe = 99.88 \(\%\) , and Acc = 99.90 \(\%\) . The ability of proposed symbolic features to characterize the ECG beats enhances the algorithm’s performance. The above results demonstrate that the proposed technique outperforms other existing works.