Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

DCAE-SR: Design of a Denoising Convolutional Autoencoder for reconstructing Electrocardiograms signals at Super Resolution

Ugo Lomoio, Pierangelo Veltri, Pietro Hiram Guzzi, Pietro Lio'

TL;DR

The paper tackles the challenge of obtaining high-resolution ECG signals from noisy, low-resolution recordings. It introduces DCAE-SR, a denoising convolutional autoencoder with a dedicated Decoder SR to perform both reconstruction and tenfold upsampling from 50 Hz to 500 Hz, trained with a double-MSE loss to enforce both LR reconstruction and HR denoising SR. Using the PTB-XL dataset with synthetic artifacts, the method achieves state-of-the-art SR metrics and demonstrates robustness to artifacts and missing channels, outperforming several baselines. The work emphasizes explainability through latent-space analysis and activation maps, suggesting practical impact for improved diagnosis and cardiovascular care, with potential extensions to other biomedical signals.

Abstract

Electrocardiogram (ECG) signals play a pivotal role in cardiovascular diagnostics, providing essential information on the electrical activity of the heart. However, the inherent noise and limited resolution in ECG recordings can hinder accurate interpretation and diagnosis. In this paper, we propose a novel model for ECG super resolution (SR) that uses a DNAE to enhance temporal and frequency information inside ECG signals. Our approach addresses the limitations of traditional ECG signal processing techniques. Our model takes in input 5-second length ECG windows sampled at 50 Hz (very low resolution) and it is able to reconstruct a denoised super-resolution signal with an x10 upsampling rate (sampled at 500 Hz). We trained the proposed DCAE-SR on public available myocardial infraction ECG signals. Our method demonstrates superior performance in reconstructing high-resolution ECG signals from very low-resolution signals with a sampling rate of 50 Hz. We compared our results with the current deep-learning literature approaches for ECG super-resolution and some non-deep learning reproducible methods that can perform both super-resolution and denoising. We obtained current state-of-the-art performances in super-resolution of very low resolution ECG signals frequently corrupted by ECG artifacts. We were able to obtain a signal-to-noise ratio of 12.20 dB (outperforms previous 4.68 dB), mean squared error of 0.0044 (outperforms previous 0.0154) and root mean squared error of 4.86% (outperforms previous 12.40%). In conclusion, our DCAE-SR model offers a robust (to artefact presence), versatile and explainable solution to enhance the quality of ECG signals. This advancement holds promise in advancing the field of cardiovascular diagnostics, paving the way for improved patient care and high-quality clinical decisions

DCAE-SR: Design of a Denoising Convolutional Autoencoder for reconstructing Electrocardiograms signals at Super Resolution

TL;DR

The paper tackles the challenge of obtaining high-resolution ECG signals from noisy, low-resolution recordings. It introduces DCAE-SR, a denoising convolutional autoencoder with a dedicated Decoder SR to perform both reconstruction and tenfold upsampling from 50 Hz to 500 Hz, trained with a double-MSE loss to enforce both LR reconstruction and HR denoising SR. Using the PTB-XL dataset with synthetic artifacts, the method achieves state-of-the-art SR metrics and demonstrates robustness to artifacts and missing channels, outperforming several baselines. The work emphasizes explainability through latent-space analysis and activation maps, suggesting practical impact for improved diagnosis and cardiovascular care, with potential extensions to other biomedical signals.

Abstract

Electrocardiogram (ECG) signals play a pivotal role in cardiovascular diagnostics, providing essential information on the electrical activity of the heart. However, the inherent noise and limited resolution in ECG recordings can hinder accurate interpretation and diagnosis. In this paper, we propose a novel model for ECG super resolution (SR) that uses a DNAE to enhance temporal and frequency information inside ECG signals. Our approach addresses the limitations of traditional ECG signal processing techniques. Our model takes in input 5-second length ECG windows sampled at 50 Hz (very low resolution) and it is able to reconstruct a denoised super-resolution signal with an x10 upsampling rate (sampled at 500 Hz). We trained the proposed DCAE-SR on public available myocardial infraction ECG signals. Our method demonstrates superior performance in reconstructing high-resolution ECG signals from very low-resolution signals with a sampling rate of 50 Hz. We compared our results with the current deep-learning literature approaches for ECG super-resolution and some non-deep learning reproducible methods that can perform both super-resolution and denoising. We obtained current state-of-the-art performances in super-resolution of very low resolution ECG signals frequently corrupted by ECG artifacts. We were able to obtain a signal-to-noise ratio of 12.20 dB (outperforms previous 4.68 dB), mean squared error of 0.0044 (outperforms previous 0.0154) and root mean squared error of 4.86% (outperforms previous 12.40%). In conclusion, our DCAE-SR model offers a robust (to artefact presence), versatile and explainable solution to enhance the quality of ECG signals. This advancement holds promise in advancing the field of cardiovascular diagnostics, paving the way for improved patient care and high-quality clinical decisions
Paper Structure (9 sections, 15 figures, 9 tables)

This paper contains 9 sections, 15 figures, 9 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Material and Methods
  4. Dataset Preprocessing
  5. The architecture of the DCAE-SR
  6. Metrics
  7. Results and Discussion
  8. Ablation studies
  9. Conclusion

Figures (15)

  • Figure 1: Our proposed DCAE-SR: a modified DCAE Encoder-Decoder architecture with a second Decoder for denoising and super-resolution (the DECODER SR).
  • Figure 2: Input: LR corrupted signal sampled at 50 Hz (blue), Target: HR clean signal sampled at 500 Hz (green), Output: predicted SR denoised signal sampled at 500 Hz. Lead I only reported for simplicity.
  • Figure 3: Corrupted LR 50 Hz in input of our DCAE-SR (up), 500 Hz cleaned super-resolution (bottom)
  • Figure 4: An example of ECG lead I channel LR data for each diagnostic superclass available in the PTB-XL dataset: Conduction Disturbance (CD), Hypertrophy (HYP), Myocardial Infraction (MI), Normal (NORM) and ST-T change (STTC)
  • Figure 5: Example of not-corrupted lead I Myocardial Infraction LR ECG window and the three versions corrupted by respiration BW, EMG and EDA artifact respectively.
  • ...and 10 more figures