loading page

A High Precision Algorithm to Classify Left and Right Outflow Tract Ventricular Tachycardia
  • +14
  • Jianwei Zheng,
  • Guohua Fu,
  • Islam Abudayyeh,
  • Magdi Yacoub,
  • Anthony Chang,
  • William Feaster,
  • Louis Ehwerhemuepha,
  • Hesham El-Askary,
  • Xianfeng Du,
  • Bin He,
  • Mingjun Feng,
  • Yibo Yu,
  • Binhao Wang,
  • Jing Liu,
  • Hai Yao,
  • Huimin Chu,
  • Cyril Rakovski
Jianwei Zheng
Chapman University
Author Profile
Guohua Fu
Ningbo First Hospital of Zhejiang University
Author Profile
Islam Abudayyeh
Loma Linda University Health
Author Profile
Magdi Yacoub
Imperial College London
Author Profile
Anthony Chang
CHOC
Author Profile
William Feaster
CHOC
Author Profile
Louis Ehwerhemuepha
CHOC
Author Profile
Hesham El-Askary
Chapman University
Author Profile
Xianfeng Du
Ningbo First Hospital of Zhejiang University
Author Profile
Bin He
Ningbo First Hospital, Zhejiang University
Author Profile
Mingjun Feng
Ningbo First Hospital of Zhejiang University
Author Profile
Yibo Yu
Ningbo First Hospital of Zhejiang University
Author Profile
Binhao Wang
Ningbo First Hospital of Zhejiang University
Author Profile
Jing Liu
Ningbo First Hospital of Zhejiang University
Author Profile
Hai Yao
Zhejiang Cachet Jetboom Medical Devices CO.LTD
Author Profile
Huimin Chu
Ningbo First Hospital of Zhejiang University
Author Profile
Cyril Rakovski
Chapman University
Author Profile

Abstract

Several algorithms based on 12-lead ECG measurements have been proposed to identify right ventricular outflow tract (RVOT) and left ventricular outflow tract (LVOT) locations from which ventricular tachycardia (VT) and frequent premature ventricular complex (PVC) originated. However, a clinical-grade artificial intelligence algorithm is not available yet, which can automatically analyze characteristics of 12-lead ECGs and predict RVOT to LVOT origins of VT and PVC. We randomly sampled training, validation, and testing datasets from 420 patients who underwent successful catheter ablation (CA) to treat VT or PVCs, containing (340, 80%), (38, 9%), and (42, 10%) patients, respectively. We iteratively trained an AI algorithm that was supplied with 1,600,800 features extracted from 12-lead ECGs of the patients in the training cohort. The area under the curve (AUC) of the receiver operating characteristic (ROC) curve was calculated from the internal validation dataset to choose an optimal discretization cutoff threshold. After running on the testing dataset, the proposed approach attained the following performance metrics and 95% CIs (confidence intervals), accuracy (ACC) of 97.62 (87.44 -99.99), weighted F1-score of 98.46 (90-100), AUC of 98.99 (96.89-100), sensitivity (SE) of 96.97 (82.54-99.89), and specificity (SP) of 100 (62.97-100). The proposed multi-stage diagnostic scheme attained clinical-grade precision of prediction for LVOT and RVOT locations of VT origin with fewer applicability restrictions than prior studies.