A feature selection for detection of non ST elevation myocardial infarction

A feature selection (FS) process of biomarkers for detecting Acute Myocardial Infarction (AMI) for Non ST Elevation patients (NSTEMI) is presented. FS has been applied by a retrospective analysis of biomarkers - high sensitivity cardiac troponin T (hs-cTnT), heart-type fatty acid-binding protein (H-FABP), creatine kinase-MB (CK-MB), myoglobin, GPBB, CA3 and NTproBNP which are measured at different times from presentation. ECG anomalies at presentation which are key for diagnosis were not considered for FS since they are routinely assessed in the emergency department (ED). Biomarkers measurements and additional data were collected at the ED from patients with chest pain of suspected cardiac origin comprising 478 cases (97 NSTEMI). hs-cTnT, H-FABP and CK-MB are statistically significant biomarkers to detect AMI according to ROC curve analysis and logistic regressions using data at different time windows. Overall, hs-cTnT as a sole marker is superior for AMI detection. However, H-FABP can be detected earlier and it demonstrates net gains in classification for non-AMI that makes it relevant for AMI rule-out approaches.