Direct extraction of respiratory information from pulse waves using a finger-inspired flexible pressure sensor system

Area of Science:

• Biomedical Engineering
• Wearable Technology
• Sensor Technology

Background:

• Continuous respiratory monitoring is vital for respiratory disease management.
• Existing devices are often uncomfortable and obtrusive, limiting long-term use.
• There is a need for user-friendly, unobtrusive respiratory monitoring solutions.

Purpose of the Study:

• To develop a flexible, lightweight, and miniaturized wrist-worn system for long-term respiratory monitoring.
• To create a cost-effective and durable sensor for continuous physiological data collection.
• To utilize deep learning for accurate classification of respiratory states.

Main Methods:

• Fabrication of a highly sensitive flexible pressure sensor via screen printing, tested for durability (>70,000 cycles).
• Integration of signal acquisition circuitry and a mobile client for data collection.
• Development of a bidirectional long short-term memory (BiLSTM) neural network with a residual module for respiratory state classification.

Main Results:

• The flexible pressure sensor demonstrated high sensitivity (11,847.24 kPa⁻¹).
• The BiLSTM neural network achieved over 99.5% accuracy in classifying various respiratory states (slow, normal, fast, simulated breathing).
• The system proved stable, cost-effective, and durable for quantitative respiratory data collection.

Conclusions:

• A novel, wearable respiratory monitoring system has been successfully developed.
• The system offers a comfortable and effective solution for long-term respiratory status tracking.
• This technology has significant potential for widespread use in personal health monitoring and clinical applications.