Deriving a reduced lead system from the 80-lead body surface map in the electrocardiographic determination of acute myocardial infarction

BackgroundBody surface mapping (BSM) involves the application of extra electrodes or “nonstandard lead sets” over a larger area of the thorax than the routine 12-lead electrocardiogram (ECG). The number of electrodes can range from 15 to more than 250, and the optimum number and placement of leads is a major topic of research. We used a novel retrospective approach, analyzing the records from patients admitted to our center with acute ischemic-type chest pain, to determine optimum lead number and placement.MethodsWe analyzed 576 patients with acute myocardial infarction (MI) (242 anterior, 173 inferior, 69 lateral, 69 posterior, and 23 right ventricular; peak cardiac troponin T > 0.09 nmol/L) and 279 controls (normal ECG and cardiac markers). Infarct location was verified by 2 cardiologists not involved in the study. Initially, an 80-lead (64 anterior, 16 posterior) BSM (Prime ECG) was recorded on all patients prior to revascularization. A combined method of analysis for a reduced lead set was used—analysis of variance (ANOVA) determined which leads had the greatest mean ST elevation (ST0 [mm]), logistic regression analysis of the ST0 ranked each lead in terms of MI diagnosis, and a computer algorithm determined which leads in combination produced the greatest sensitivity and specificity for MI diagnosis.ResultsThus, a 30-lead set (20 anterior, 10 posterior) was derived from the combined method of analysis. The sensitivity and specificity of the 30-lead set for MI was 84% and 97%, respectively, compared with the 80-lead BSM of 91% and 93%. Receiver operating characteristic (ROC) c statistic for the 30 lead set was 0.825 (95% confidence interval, 0.797-0.852), which was within the 95% confidence intervals for the 80 lead ROC c statistic of 0.850 (0.824-0.875). The data set was split equally into training and validation set. The training set ROC c statistic for the 30-lead set was 0.860 (0.825-0.896) and 0.908 (0.881-0.935) for the 80-lead set. The validation set ROC c statistic was 0.821 (0.782-0.860) for the 30-lead set and 0.782 (0.739-0.825) for the 80-lead set.ConclusionUsing the ROC c statistic, we showed that the 30-lead BSM was noninferior to the 80-lead model.