Exploring Best Practices for ECG Pre-Processing in Machine Learning

TL;DR

This study investigates ECG pre-processing techniques for machine learning, finding that lower sampling rates can be effective and that common normalization methods may not improve classifier performance, highlighting the importance of tailored pre-processing.

Contribution

The paper systematically evaluates the impact of various pre-processing steps on ECG classification, revealing that some standard practices may be unnecessary or suboptimal.

Findings

Sampling rates as low as 50Hz perform comparably to 500Hz.

Min-max normalization slightly reduces accuracy.

Band-pass filtering shows no measurable benefit.

Abstract

In this work we search for best practices in pre-processing of Electrocardiogram (ECG) signals in order to train better classifiers for the diagnosis of heart conditions. State of the art machine learning algorithms have achieved remarkable results in classification of some heart conditions using ECG data, yet there appears to be no consensus on pre-processing best practices. Is this lack of consensus due to different conditions and architectures requiring different processing steps for optimal performance? Is it possible that state of the art deep-learning models have rendered pre-processing unnecessary? In this work we apply down-sampling, normalization, and filtering functions to 3 different multi-label ECG datasets and measure their effects on 3 different high-performing time-series classifiers. We find that sampling rates as low as 50Hz can yield comparable results to the commonly used 500Hz. This is significant as smaller sampling rates will result in smaller datasets and models, which require less time and resources to train. Additionally, despite their common usage, we found min-max normalization to be slightly detrimental overall, and band-passing to make no measurable difference. We found the blind approach to pre-processing of ECGs for multi-label classification to be ineffective, with the exception of sample rate reduction which reliably reduces computational resources, but does not increase accuracy.