An Accelerometer-Based Wearable Patch for Robust Respiratory Rate and Wheeze Detection Using Deep Learning.

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Respiratory Medicine

Background:

• Wheezing is a key symptom of respiratory diseases like asthma and COPD.
• Current diagnosis relies on subjective physician auscultation.
• Objective, remote monitoring for respiratory conditions is needed.

Purpose of the Study:

• To develop a low-profile wearable device for objective wheeze detection.
• To enable remote patient monitoring (RPM) of respiratory patterns.
• To utilize deep learning for accurate wheeze and respiration rate analysis.

Main Methods:

• A miniature accelerometer-based wearable patch was developed to capture pulmonary-induced vibrations (PIVs).
• A deep learning model was trained and validated on PIV data from 52 respiratory patients.
• Performance was compared against a deterministic time-frequency method and digital stethoscopes.

Main Results:

• The deep learning model achieved 95% accuracy, 96% sensitivity, and 93% specificity for wheeze detection (AUC 0.99).
• The system accurately captured respiratory wheezes and patterns in diverse patient populations.
• The accelerometer patch outperformed digital stethoscopes and was immune to ambient noise.

Conclusions:

• The wearable accelerometer system with deep learning provides a robust solution for objective wheeze detection.
• This technology enables high-fidelity remote patient monitoring of respiratory conditions.
• It offers a significant advancement over subjective auscultation and existing digital tools.