Comparing symbolic representations of cardiac activity to identify patient populations with similar risk profiles

Abstract

This paper proposes electrocardiographic mismatch (ECGM) to quantify differences in the long-term ECG signals for two patients. ECGM compares the symbolic distributions of ECG signals and measures how different patients are electrocardiographically. Using ECGM, we propose a hierarchical clustering scheme that can identify patients in a population with anomalous ECG characteristics. When applied to a population of 686 patients suffering nonST-elevation ACS, our approach was able to identify patients at an increased risk of death and myocardial infarction (HR 2.8, p = 0.003) over a 90 day follow-up period.