Deep learning-Based Fit Level Detection for Industrial Respirators with Embedded Breath Sensors

Area of Science:

• Occupational Health and Safety
• Biomedical Engineering
• Artificial Intelligence in Healthcare

Background:

• Industrial respirators are crucial for worker safety, but ensuring a proper fit is challenging.
• Current methods for respirator fit testing are often time-consuming and may not reflect real-world conditions.
• Continuous, real-time monitoring of respirator performance is needed to enhance occupational safety.

Purpose of the Study:

• To develop and evaluate a deep learning approach for real-time monitoring of respirator fit and breathing conditions.
• To assess the accuracy of an embedded breath sensor system in detecting respirator fit under various conditions.
• To enhance occupational safety and health through continuous, non-intrusive monitoring in industrial environments.

Main Methods:

• An embedded breath sensor system was developed for industrial respirators.
• Controlled experiments were conducted with 21 healthy subjects, measuring breathing frequency and fit under OSHA protocols.
• Data from proper fitted and unsealed mask conditions were analyzed using a deep learning model.

Main Results:

• The developed deep learning model achieved up to 80% accuracy in detecting the fit level of industrial respirators in real-time.
• The system demonstrated the ability to estimate fit quality under different operating conditions.
• Continuous, non-intrusive monitoring of working conditions was enabled.

Conclusions:

• The proposed deep learning approach with an embedded breath sensor system offers a promising solution for real-time respirator fit monitoring.
• This technology can significantly enhance occupational safety and health by providing continuous feedback on respirator performance.
• The system has the potential to detect impairing occupational safety and health conditions and related illnesses early.