9 Classification of Electrocardiogram Arrhythmias Using Neural Networks
Download citation file:
- Ris (Zotero)
- Reference Manager
The heart will beat over a billion times throughout a lifetime; however, a series of abnormal beats could result in death. Premature ventricular contraction (PVC), left bundle branch block (LBBB), and right bundle branch block (RBBB), are three types of abnormal heart beats which can lead to or indicate the risk of heart failure. A neural network has been trained and tested using signals from the MIT-BIH Arrhythmia Database to detect PVC, LBBB, RBBB, normal, and other arrhythmias. Wavelet, frequency, and temporal analysis of the QRS complex have been used to create neural network feature vectors used for classification. The network showed a specificity of 98.12% and a sensitivity of 93.35% for normal beats, 83.72% and 90.06% for PVC beats, 88.18% and 89.09% for LBBB beats, 92.85% and 96.14% for RBBB beats, and 77.21% and 89.78% for other arrhythmias. Overall, the network classified 97,545 true positives of the 105,287 testing beats, for an overall positive predictability of 92.65%.