ECG Feature Importance Rankings: Cardiologists vs. Algorithms

TL;DR

This study evaluates various feature importance methods on real ECG data to compare algorithmic rankings with cardiologists' decision rules, revealing differing performances across methods and pathologies.

Contribution

It provides an empirical comparison of feature importance methods against cardiologists' rules using real-world ECG data, highlighting their strengths and weaknesses.

Findings

Some methods performed well overall

Other methods showed inconsistent results

Performance varied depending on the pathology

Abstract

Feature importance methods promise to provide a ranking of features according to importance for a given classification task. A wide range of methods exist but their rankings often disagree and they are inherently difficult to evaluate due to a lack of ground truth beyond synthetic datasets. In this work, we put feature importance methods to the test on real-world data in the domain of cardiology, where we try to distinguish three specific pathologies from healthy subjects based on ECG features comparing to features used in cardiologists' decision rules as ground truth. Some methods generally performed well and others performed poorly, while some methods did well on some but not all of the problems considered.