Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

In-ear ECG Signal Enhancement with Denoising Convolutional Autoencoders

Edoardo Occhipinti, Marek Zylinski, Harry J. Davies, Amir Nassibi, Matteo Bermond, Patrik Bachtiger, Nicholas S. Peters, Danilo P. Mandic

TL;DR

A denoising convolutional autoencoder (DCAE) to enhance ECG information from in-ear recordings, producing cleaner ECG outputs and effective removal of noise sources with clinically plausible waveform reconstruction ability is developed.

Abstract

The cardiac dipole has been shown to propagate to the ears, now a common site for consumer wearable electronics, enabling the recording of electrocardiogram (ECG) signals. However, in-ear ECG recordings often suffer from significant noise due to their small amplitude and the presence of other physiological signals, such as electroencephalogram (EEG), which complicates the extraction of cardiovascular features. This study addresses this issue by developing a denoising convolutional autoencoder (DCAE) to enhance ECG information from in-ear recordings, producing cleaner ECG outputs. The model is evaluated using a dataset of in-ear ECGs and corresponding clean Lead I ECGs from 45 healthy participants. The results demonstrate a substantial improvement in signal-to-noise ratio (SNR), with a median increase of 5.9 dB. Additionally, the model significantly improved heart rate estimation accuracy, reducing the mean absolute error by almost 70% and increasing R-peak detection precision to a median value of 90%. We also trained and validated the model using a synthetic dataset, generated from real ECG signals, including abnormal cardiac morphologies, corrupted by pink noise. The results obtained show effective removal of noise sources with clinically plausible waveform reconstruction ability.

In-ear ECG Signal Enhancement with Denoising Convolutional Autoencoders

TL;DR

A denoising convolutional autoencoder (DCAE) to enhance ECG information from in-ear recordings, producing cleaner ECG outputs and effective removal of noise sources with clinically plausible waveform reconstruction ability is developed.

Abstract

The cardiac dipole has been shown to propagate to the ears, now a common site for consumer wearable electronics, enabling the recording of electrocardiogram (ECG) signals. However, in-ear ECG recordings often suffer from significant noise due to their small amplitude and the presence of other physiological signals, such as electroencephalogram (EEG), which complicates the extraction of cardiovascular features. This study addresses this issue by developing a denoising convolutional autoencoder (DCAE) to enhance ECG information from in-ear recordings, producing cleaner ECG outputs. The model is evaluated using a dataset of in-ear ECGs and corresponding clean Lead I ECGs from 45 healthy participants. The results demonstrate a substantial improvement in signal-to-noise ratio (SNR), with a median increase of 5.9 dB. Additionally, the model significantly improved heart rate estimation accuracy, reducing the mean absolute error by almost 70% and increasing R-peak detection precision to a median value of 90%. We also trained and validated the model using a synthetic dataset, generated from real ECG signals, including abnormal cardiac morphologies, corrupted by pink noise. The results obtained show effective removal of noise sources with clinically plausible waveform reconstruction ability.
Paper Structure (12 sections, 5 equations, 9 figures)

This paper contains 12 sections, 5 equations, 9 figures.

Table of Contents

  1. Introduction
  2. Methods
  3. In-ear ECG data collection
  4. Noise modelling and data augmentation
  5. 1D Denoising convolutional autoencoder
  6. Model evaluation
  7. Results
  8. Model Behaviour
  9. Denoising performance evaluation
  10. Heart Rate estimation in-ear ECG
  11. Discussion
  12. Conclusion

Figures (9)

  • Figure 1: The earpiece used in this study consists of a gelled cloth electrode attached to a viscoelastic foam and a PDMS shell to anchor it to the concha, thus preventing the earpiece from coming out from the ear canal
  • Figure 2: Average PSD comparison of in-ear ECG and Lead I ECG across the 10 s signals. We can observe how the in-ear ECG spectrum resembles more closely a 1/f spectrum, typical of EEG signals.
  • Figure 3: SNR distribution of in-ear ECG dataset and synthetic dataset generated by corrupting ECG signals from the PTB-XL database with pink noise.
  • Figure 4: Denoising convolutional autoencoder (DCAE) architecture taking a noisy in-ear ECG as input, and simultaneously recorded clean ECG as output.
  • Figure 5: Top panel: Zoom-in view of the input in-ear ECG showing two successive cardiac cycles, with the corresponding spectrogram in the background. Middle panels: Latent space representation of 2 out of 32 channels, illustrating how different morphological features of the ECG are captured. Bottom panel: Denoised in-ear ECG along with the associated spectrogram, displayed alongside the reference Lead I signal. The spectrogram was computed using a 0.1s window with 50% overlap and smoothed using a Gaussian filter.
  • ...and 4 more figures