RT Journal Article ID 559bfc8c7d227396 A1 Hopkins, Christie B. A1 Suleman, Jawal A1 Cook, Carl T1 An Artificial Neural Network for the Electrocardiographic Diagnosis of Left Ventricular Hypertrophy JF Critical Reviews™ in Biomedical Engineering JO CRB YR 2000 FD 2000-08-01 VO 28 IS 3&4 SP 435 OP 438 AB Objective: A neural network was constructed to predict the presence of left ventricular hypertrophy (LVH) using both clinical information and the electrocardiogram (ECG).
Design and setting: In this retrospective study of 317 adult male patients, clinical parameters were age and history/physical examination: normal, heart failure, LV outflow obstruction, mitral regurgitation or aortic regurgitation. Multiple ECG parameters were used. A back-propagation neural network was constructed. The network was trained on 217 patients. A test set of 100 patients was then evaluated. The network was used to predict both LV mass and LVH by the criterion of LV mass index > 132 g/m2.
Results: LV mass was predicted with an accuracy of 79%. In predicting LVH, the network showed 82% correct diagnosis, sensitivity 94%, and specificity 65%. Positive predictive accuracy was 81% and negative predictive accuracy was 89%.
Conclusions: The neural network integrates clinical and ECG data and its resultant prediction of LVH is superior to that obtained using conventional ECG diagnostic criteria. PB Begell House LK https://www.dl.begellhouse.com/journals/4b27cbfc562e21b8,657c94ee3fd00c46,559bfc8c7d227396.html