Respiratory Rate Estimation Using U-Net-Based Cascaded Framework From Electrocardiogram and Seismocardiogram Signals

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Signal Processing

Background:

• At-home respiration monitoring is crucial, especially during pandemics.
• Electrocardiogram (ECG) and seismocardiogram (SCG) offer less cumbersome alternatives to traditional airflow-based methods for respiratory monitoring.
• Estimating respiratory rates (RR) from ECG-derived (EDR) and SCG-derived (SDR) signals is promising but challenged by non-respiratory artifacts.

Purpose of the Study:

• To develop a novel framework for accurate respiratory rate estimation using ECG and SCG signals.
• To address and reduce non-respiratory artifacts present in EDR and SDR signals.
• To validate the performance and generalizability of the proposed method for remote respiratory monitoring.

Main Methods:

• A U-Net-based cascaded framework was developed to process EDR and SDR signals.
• Signals were transformed into the spectro-temporal domain.
• A 2D U-Net was employed for denoising to mitigate non-respiratory artifacts.

Main Results:

• The fused EDR and SDR input U-Net achieved a low mean absolute error of 0.82 bpm and an R² of 0.89.
• Results were obtained using data from a chest-worn wearable patch.
• Qualitative analysis confirmed the complementary nature of EDR and SDR signals and demonstrated framework generalizability.

Conclusions:

• ECG and SCG signals, when collected from a chest-worn patch, complement each other for reliable RR estimation via the proposed cascaded framework.
• The framework facilitates pervasive and accurate respiratory rate measurement using convenient wearable systems.