Continuous Breathing Monitoring From Heart And Pulse Signals

Jenn Hoskins
18th June, 2025

Continuous Breathing Monitoring From Heart And Pulse Signals

The proposed method improves respiratory rate estimation by extracting multiple respiratory modulated components from ECG and PPG signals (1), which are then denoised and screened for quality (2) before being fused using principal component analysis (PCA) into a single waveform for final analysis (3-4).

Image adapted from: Lin et al. / CC BY (Source)

Key Findings

  • Researchers in China and Marche developed a new method to accurately monitor breathing rate by combining subtle breathing-related changes from both ECG and PPG signals
  • This novel approach significantly improved accuracy, reducing measurement errors by over 11% compared to existing methods, making continuous, non-invasive monitoring more reliable
Respiratory rate is a fundamental indicator of a person's health, offering crucial insights into various physiological changes. It is highly sensitive to conditions like adverse cardiac events, pneumonia, and overall clinical deterioration, as well as stressors such as emotional strain, cognitive load, and physical exertion[2]. In fact, its sensitivity often surpasses that of many other vital signs. Despite its recognized importance and the availability of technology, respiratory rate is still not routinely monitored in many healthcare, occupational, or sports settings[2]. Traditional methods for measuring it can be intrusive or inconvenient, making continuous, widespread monitoring a challenge. Existing efforts to non-invasively estimate respiratory rate often rely on signals from wearable devices, such as the photoplethysmogram (PPG). A PPG sensor, commonly found in smartwatches and pulse oximeters, measures changes in blood volume in the skin, which can reflect heart rate and oxygen saturation. However, extracting respiratory rate from PPG signals can be difficult because these signals are susceptible to external noise and motion artifacts[3]. For instance, body movements can introduce distortions, making it hard to accurately interpret the signal. While advancements in signal processing, such as specialized filtering techniques, have been developed to reduce these motion artifacts from PPG[4], many single-input methods still face challenges in real-world scenarios, sometimes introducing phase distortions. Furthermore, many algorithms for estimating respiratory rate from PPG alone require careful parameter selection, often through trial and error, and can struggle with low-quality signals[3]. To address these limitations and improve the accuracy and robustness of continuous respiratory rate monitoring, researchers from Electronic Science and Technology of China and Marche have proposed a novel approach[1]. Their study focuses on combining information from two different non-invasive cardiovascular signals: the electrocardiogram (ECG) and the photoplethysmogram (PPG). While previous studies have explored estimating respiratory rate from either ECG or PPG signals individually, or by fusing different components derived from a single signal, there has been limited research on robust continuous estimation by combining all types of respiratory-modulated components (RMCs) from both ECG and PPG in the time domain. The core idea behind this new method is the "temporal fusion" of these respiratory-modulated components. Respiratory-modulated components are subtle variations within the ECG and PPG signals that are caused by breathing. For example, as we inhale and exhale, changes in chest pressure affect blood flow and heart activity, leaving a faint "signature" on these cardiovascular signals. The researchers extracted six different RMCs from both the ECG and PPG signals. To ensure accuracy, they then identified which of these extracted RMCs were of high quality, using a measure called the respiratory quality index. This step is crucial because not all parts of the signal will be equally clear or useful for respiratory rate estimation, especially given the challenges of noise and signal quality that methods like those described in[3] aim to overcome. Once the high-quality RMCs were identified, the researchers fused them into a single, comprehensive respiratory signal. This fusion process utilized a technique called Principal Component Analysis (PCA). PCA is a statistical method that helps to reduce the complexity of data while retaining its most important features. By combining the best parts of the different RMCs from both ECG and PPG, PCA helps to create a more robust and clearer representation of the respiratory signal, effectively minimizing the impact of noise or poor quality from any single source. From this newly fused signal, the respiratory rate was then estimated. The effectiveness of this temporal fusion method was validated using two publicly available datasets: the Capnobase dataset (containing data from 42 subjects) and the BIDMC dataset (with data from 53 subjects). The results demonstrated significant improvements in accuracy. The proposed method achieved a Mean Absolute Error (MAE) of 1.39 breaths per minute for the Capnobase dataset and 3.29 breaths per minute for the BIDMC dataset. Mean Absolute Error (MAE) is a common metric used to measure the average magnitude of the errors in a set of predictions, indicating how close the estimated values are to the true values. Lower MAE values signify higher accuracy. These results represent a substantial advancement, showing an average 11.61% reduction in MAE compared to existing state-of-the-art approaches. For context, a previous algorithm specifically designed to estimate respiratory rate using only PPG signals, which focused on robustness to noise and low-quality signals, reported an MAE of 2.05 breaths per minute on a publicly available dataset, and comparable results on the BIDMC dataset[3]. The new fusion approach, by combining the strengths of both ECG and PPG signals, appears to surpass the performance of such single-sensor methods, particularly on datasets like Capnobase. This indicates that integrating information from multiple physiological signals can lead to more reliable and precise respiratory rate measurements, expanding the functionality of wearable devices beyond just heart rate and oxygen saturation to continuous respiratory rate monitoring at home and in clinical settings. The study thus builds upon the understanding of respiratory rate's critical role[2] and addresses the inherent limitations of single-sensor approaches[3] by leveraging a multi-modal data fusion strategy.

MedicineHealthBiotech

References

Main Study

1) Continuous respiratory rate monitoring through temporal fusion of ECG and PPG signals

Published 17th June, 2025

https://doi.org/10.1371/journal.pone.0325307

Related Studies

2) The Importance of Respiratory Rate Monitoring: From Healthcare to Sport and Exercise.

https://doi.org/10.3390/s20216396

3) Photoplethysmography-Based Respiratory Rate Estimation Algorithm for Health Monitoring Applications.

https://doi.org/10.1007/s40846-022-00700-z

4) Blockwise PPG Enhancement Based on Time-Variant Zero-Phase Harmonic Notch Filtering.

https://doi.org/10.3390/s17040860


Related Articles

An unhandled error has occurred. Reload 🗙