Comparison of model-based and expert-rule based electrocardiographic identification of the culprit artery in patients with acute coronary syndrome

- We investigate automated culprit artery assessment in acute coronary syndrome.
- We compare the performance of two methods for assessing the culprit artery.
- The expert-rule based method assesses the correct culprit in 81.1% of the cases.
- The model-based Olson method assesses the correct culprit in 88.7% of the cases.
- The automated Olson method performs at the level of expert cardiologists.

Background and purposeCulprit coronary artery assessment in the triage ECG of patients with suspected acute coronary syndrome (ACS) is relevant a priori knowledge preceding percutaneous coronary intervention (PCI). We compared a model-based automated method (Olson method) with an expert-rule based method for the culprit artery assessment.MethodsIn each of the 53 patients who were admitted with the working diagnosis of suspected ACS, scheduled for emergent angiography with a view on revascularization as initial treatment and subsequently found to have an angiographically documented completely occluded culprit artery, culprit artery location was assessed in the preceding ECG by both the model-based Olson method and the expert-rule based method that considered either visual or computer-measured J-point amplitudes. ECG culprit artery estimations were compared with the angiographic culprit lesion locations. Proportions of correct classifications were compared by a Z test at the 5% significance level.ResultsThe Olson method performed slightly, but not significantly, better, when the expert-rule based method used visual assessment of J-point amplitudes (88.7% versus 81.1% correct; P = 0.28). However, the Olson method performed significantly better when the expert-rule based method used computer-measured J-point amplitudes (88.7% versus 71.7% correct; P < 0.05).ConclusionThe automated model-based Olson method performed at least at the level of expert cardiologists using a manual rule-based method.