Automated performance analysis of short-vector versus long-vector electrocardiograms

Abstract

Background: ST-deviation analysis is a cornerstone in the early triage of myocardial ischemia. The current electrocardiogram (ECG) criteria for the diagnosis of occlusive myocardial ischemia focus on ST elevations, but recent studies have proposed to include both ST depression and ST elevation in the diagnosis, hence considering all ST deviations as projections of a single underlying dipole source. In this study, we examined to which extent a dipole ST vector model accounts for the measured ST deviations in acute ischemia patients. Methods: Forty-five patients with acute myocardial infarction were selected retrospectively for the study to have an equal representation of occlusion sites among the 3 major coronary arteries left anterior descending (LAD), right coronary artery (RCA), and left circumflex (LCX). A 12-lead ECGwas recorded from each patient immediately before coronary intervention. All patients showed occlusive single-vessel disease with thrombolysis in myocardial infarction (TIMI) flow 0/1 (15 LAD, 15 RCA, 15 LCX). ST deviations were measured 60 milliseconds after the J-point in all leads using the 12SL algorithm. Three-dimensional lead vectors for each of the 12 standard leads were defined in accordancewith the spatial direction of the leads in the frontal and horizontal planes. The dipole model was established by considering ST deviations as projections of a single-dipole vector onto each of the defined lead vectors. For each recorded ECG, the best-fitting dipole vector was estimated by minimizing the sum of squared errors between measured and projected ST deviations across all leads.Overall goodness-of-fit (R) for the dipolemodelwas evaluated between measured and projected ST deviations across all recordings. Results: The goodness-of-fit across all recordings was R = 0.82-0.87 (95% confidence interval [CI]). When examining recordings grouped by culprit artery, the model fit was R = 0.75-0.85 (95% CI) for the LAD occlusions, R = 0.83-0.90 (95% CI) for the RCA recordings, and R = 0.78-0.87 (95% CI), for the LCX recordings. Analysis of variance showed no significant difference in goodness-of-fit for the different occlusion sites (P = .19). Conclusion: The single-dipole ST vector accounted for approximately 85% of the ST-deviation information in the studied ECG recordings irrespective of the site of occlusion. A simple dipole model may be a useful descriptor of ST-segment deviations, reducing ST-deviation measurements from 12 leads to a single 3-dimensional vector.