Diffusion-Based Electrocardiography Noise Quantification via Anomaly Detection

TL;DR

This paper introduces a diffusion-based anomaly detection method for quantifying noise in ECG signals, improving robustness and generalizability over traditional artifact classification approaches.

Contribution

The study presents a novel diffusion model for ECG noise quantification that does not rely on artifact labels and employs a Wasserstein-1 distance metric for evaluation.

Findings

Achieved a macro-average $W_1$ score of 1.308, surpassing previous methods by over 48%.

Demonstrated strong generalizability in external validation.

Enabled effective exclusion of noisy segments to enhance diagnostic accuracy.

Abstract

Electrocardiography (ECG) signals are frequently degraded by noise, limiting their clinical reliability in both conventional and wearable settings. Existing methods for addressing ECG noise, relying on artifact classification or denoising, are constrained by annotation inconsistencies and poor generalizability. Here, we address these limitations by reframing ECG noise quantification as an anomaly detection task. We propose a diffusion-based framework trained to model the normative distribution of clean ECG signals, identifying deviations as noise without requiring explicit artifact labels. To robustly evaluate performance and mitigate label inconsistencies, we introduce a distribution-based metric using the Wasserstein-1 distance ($W_1$). Our model achieved a macro-average $W_1$ score of 1.308, outperforming the next-best method by over 48\%. External validation confirmed strong generalizability, facilitating the exclusion of noisy segments to improve diagnostic accuracy and support timely clinical intervention. This approach enhances real-time ECG monitoring and broadens ECG applicability in digital health technologies.