Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

RespDiff: An End-to-End Multi-scale RNN Diffusion Model for Respiratory Waveform Estimation from PPG Signals

Yuyang Miao, Zehua Chen, Chang Li, Danilo Mandic

TL;DR

Experiments demonstrate that RespDiff outperforms notable previous works, achieving a mean absolute error of 1.18 bpm for RR estimation while others range from 1.66 to 2.15 bpm, showing its potential for robust and accurate respiratory monitoring in real-world applications.

Abstract

Respiratory rate (RR) is a critical health indicator often monitored under inconvenient scenarios, limiting its practicality for continuous monitoring. Photoplethysmography (PPG) sensors, increasingly integrated into wearable devices, offer a chance to continuously estimate RR in a portable manner. In this paper, we propose RespDiff, an end-to-end multi-scale RNN diffusion model for respiratory waveform estimation from PPG signals. RespDiff does not require hand-crafted features or the exclusion of low-quality signal segments, making it suitable for real-world scenarios. The model employs multi-scale encoders, to extract features at different resolutions, and a bidirectional RNN to process PPG signals and extract respiratory waveform. Additionally, a spectral loss term is introduced to optimize the model further. Experiments conducted on the BIDMC dataset demonstrate that RespDiff outperforms notable previous works, achieving a mean absolute error (MAE) of 1.18 bpm for RR estimation while others range from 1.66 to 2.15 bpm, showing its potential for robust and accurate respiratory monitoring in real-world applications.

RespDiff: An End-to-End Multi-scale RNN Diffusion Model for Respiratory Waveform Estimation from PPG Signals

TL;DR

Experiments demonstrate that RespDiff outperforms notable previous works, achieving a mean absolute error of 1.18 bpm for RR estimation while others range from 1.66 to 2.15 bpm, showing its potential for robust and accurate respiratory monitoring in real-world applications.

Abstract

Respiratory rate (RR) is a critical health indicator often monitored under inconvenient scenarios, limiting its practicality for continuous monitoring. Photoplethysmography (PPG) sensors, increasingly integrated into wearable devices, offer a chance to continuously estimate RR in a portable manner. In this paper, we propose RespDiff, an end-to-end multi-scale RNN diffusion model for respiratory waveform estimation from PPG signals. RespDiff does not require hand-crafted features or the exclusion of low-quality signal segments, making it suitable for real-world scenarios. The model employs multi-scale encoders, to extract features at different resolutions, and a bidirectional RNN to process PPG signals and extract respiratory waveform. Additionally, a spectral loss term is introduced to optimize the model further. Experiments conducted on the BIDMC dataset demonstrate that RespDiff outperforms notable previous works, achieving a mean absolute error (MAE) of 1.18 bpm for RR estimation while others range from 1.66 to 2.15 bpm, showing its potential for robust and accurate respiratory monitoring in real-world applications.
Paper Structure (15 sections, 10 equations, 2 figures, 2 tables)

This paper contains 15 sections, 10 equations, 2 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Works
  3. Methods
  4. Conditional Diffusion Models
  5. Network Architecture
  6. Multi-scale fine-grained feature encoder
  7. Multi-scale coarse-grained feature encoder
  8. Bidirectional RNN
  9. Spectral Loss
  10. Experiments
  11. Experimental Setup
  12. Results Analysis
  13. Ablation Study on Spectral Loss
  14. Ablation Study on Multi-Scale Encoder
  15. Conclusion

Figures (2)

  • Figure 1: The overall architecture of the proposed RespDiff model. The respiratory waveform is gradually corrupted in the forward process and iteratively denoised in the backward diffusion. During training, features are extracted from the PPG signal and corrupted respiratory waveform to estimate the noise. Losses in the diffusion domain and frequency domain are calculated and added together.
  • Figure 2: Examples of respiratory waveform estimation. (a): RespDiff with 50 diffusion steps and spectral loss. (b): RespDiff with 50 diffusion steps and without spectral loss. (c): RespDiff with 6 diffusion steps and spectral loss. (d): RespDiff with 6 diffusion steps and without spectral loss.