Self-supervised inter-intra period-aware ECG representation learning for detecting atrial fibrillation

TL;DR

Considering ECGs of atrial fibrillation patients exhibit the irregularity in RR intervals and the absence of P-waves, this work develops specific pre-training tasks for interperiod and intraperiod representations, aiming to learn the single-period stable morphology representation while retaining crucial interperiod features.

Abstract

Atrial fibrillation is a commonly encountered clinical arrhythmia associated with stroke and increased mortality. Since professional medical knowledge is required for annotation, exploiting a large corpus of ECGs to develop accurate supervised learning-based atrial fibrillation algorithms remains challenging. Self-supervised learning (SSL) is a promising recipe for generalized ECG representation learning, eliminating the dependence on expensive labeling. However, without well-designed incorporations of knowledge related to atrial fibrillation, existing SSL approaches typically suffer from unsatisfactory capture of robust ECG representations. In this paper, we propose an inter-intra period-aware ECG representation learning approach. Considering ECGs of atrial fibrillation patients exhibit the irregularity in RR intervals and the absence of P-waves, we develop specific pre-training tasks for interperiod and intraperiod representations, aiming to learn the single-period stable morphology representation while retaining crucial interperiod features. After further fine-tuning, our approach demonstrates remarkable AUC performances on the BTCH dataset, \textit{i.e.}, 0.953/0.996 for paroxysmal/persistent atrial fibrillation detection. On commonly used benchmarks of CinC2017 and CPSC2021, the generalization capability and effectiveness of our methodology are substantiated with competitive results.

Figures (8)

• Figure 1: The classical ECG curve displays common waveforms and crucial intervals with measurement points.
• Figure 2: Schematic overview of the proposed approach. Inspired by medical knowledge that ECGs of AF patients are related to the irregularity in RR intervals and the absence of P-waves, we propose a self-supervised pre-training method to learn representations with interperiod and intraperiod awareness.
• Figure 3: Denoising of an ECG signal. The blue line represents the original signal, while the red line represents the signal after denoising. Only 1,000 samples of the ECG signal (representing 2 seconds at a sampling rate of 500 Hz) are displayed here to better distinguish between the original and denoised signals.
• Figure 4: An overview of the pre-training architecture for our self-supervised method. The interperiod task primarily involves predicting RR intervals information, while the intraperiod task focuses on aligning single-period morphologies through contrastive learning in this experiment.
• Figure 5: The self-supervised fine-tuning process for downstream tasks, specifically atrial fibrillation detection in this experiment.