SAFE: A Smart Adherence Detection Framework for Monitoring Personal Protective Equipment in Healthcare Settings

Area of Science:

• Computer Vision
• Healthcare Technology
• Infection Control

Background:

• Personal protective equipment (PPE) is crucial for healthcare infection control, but adherence to guidelines is inconsistent.
• Manual monitoring of PPE use is impractical due to labor intensity and infection risks.
• Automated PPE assessment in clinical videos is challenging due to occlusions and subtle adherence variations.

Purpose of the Study:

• To develop and evaluate a computer vision system, SAFE, for real-time monitoring of PPE adherence.
• To address challenges in automated PPE assessment, focusing on gowns and masks.
• To provide a scalable solution for improving healthcare infection control.

Main Methods:

• Proposed a cascaded computer vision system (SAFE) with a two-stage design for PPE status detection.
• Stage 1: Gown status detection and head region localization.
• Stage 2: Mask status classification from head crops. Evaluated on R2PPE dataset and real-world trauma-room data.

Main Results:

• SAFE improved overall average precision from 0.48 to 0.67 on the R2PPE dataset.
• Increased mask-class average precision by 0.33 compared to a baseline detector.
• Validated on real-case data, improving class-level mask accuracy from 0.59 to 0.65 with 0.98 alarm-level recall.

Conclusions:

• SAFE offers a reliable automated solution for real-time PPE monitoring in healthcare.
• The system enhances infection control by ensuring consistent PPE adherence with minimal human intervention.
• SAFE demonstrates potential for scalable deployment in diverse clinical settings.