ELECTROCARDIOGRAM ANALYSIS SYSTEM

Abstract

The present disclosure provides concepts for analyzing electrocardiogram (ECG) data of a subject. Specifically, a traditional prediction unit is configured to process the ECG data with a rules-based ECG interpretation model to generate first diagnostic labels indicating presence or absence of cardiac conditions. A machine learning prediction unit processes the ECG data with a machine learning-based ECG interpretation model to generate second diagnostic labels indicating presence or absence of cardiac conditions. A mapping unit determines a feature mapping describing first and second diagnostic labels associated with similar cardiac conditions. A combination unit analyzes the first diagnostic labels based on the feature mapping and the second diagnostic labels to generate output diagnostic labels indicating presence or absence of cardiac conditions of the subject. The provided concepts therefore leverage the benefits of both the rules-based ECG interpretation model and the machine learning-based ECG interpretation model to provide output diagnostic labels. Specifically, the granularity of the output of the rules-based ECG interpretation model may be maintained whilst ensuring accuracy via the output of the machine learning-based ECG interpretation model.

IPC Classifications