Interpreting Wide-Complex Tachycardia using Artificial Intelligence

Background Adopting artificial intelligence in medicine may improve speed and accuracy in patient diagnosis. We sought to develop an artificial intelligence (AI) algorithm to interpret wide complex tachycardia (WCT) electrocardiograms (ECG) and compare its diagnostic accuracy to cardiologists. Methods Using 3330 WCT ECGs (2906 SVT and 424 VT), we created a training/validation (3131) and test set (199 ECGs). A convolutional neural network (CNN) structure using a modification of differentiable architecture search (DARTS), ZeroLess-DARTS, was developed to differentiate between SVT and VT. Results The mean accuracy of electrophysiology (EP) cardiologists was 92.5% with a sensitivity of 91.7%, specificity of 93.4%, positive predictive value of 93.7%, negative predictive value of 91.7%. NonEP cardiologists had an accuracy of 73.2 ± 14.4% with a sensitivity, specificity, positive and negative predictive value of 59.8 ± 18.2%, 93.8 ± 3.7%, 93.6 ± 2.3%, and 73.2 ± 14.4%, respectively. AI had superior sensitivity and accuracy (91.9% and 93.0%, respectively) than NonEP cardiologists, and had similar performance of EP cardiologists. Mean time to interpret each ECG varied between 10.1-13.8 seconds for EP cardiologists and 3.1 -16.6 seconds for NonEP cardiologists. Conversely AI required a mean of 0.0092 ± 0.0035 seconds for each ECG interpretation. Conclusions AI appears to diagnose WCT with superior accuracy than Cardiologists and similar to those of Electrophysiologists. Using AI to assist with ECG interpretations may improve patient care.