Development of a system for monitoring and validation of proper hand washing using machine learning

Abstract

Background

Hand hygiene is critical in medical settings to prevent infections and ensure patient safety. Despite multiple initiatives to improve adherence to hygiene protocols, compliance rates remain low, posing a persistent challenge.

Methods

We developed an AI-driven application to monitor adherence to hand-rubbing techniques in a clinical setting. The system utilizes machine learning to an alyze hand landmarks extracted from video, incorporating a normalization process based on hand centroids to mitigate biases related to camera distance and hand size. Three machine learning models—Logistic Regression, Support Vector Machine, and Random Forest—were evaluated based on accuracy, inference speed, and memory usage.

Results

Logistic Regression demonstrated the best performance, achieving 99.5 % accuracy and processing each hand-washing step in only 3 ms. The application also tracks the duration of each hygiene step, promoting compliance with recommended hand-washing times. We tested the algorithm in a clinical setting.

Conclusion

This AI-driven solution provides a scalable, real-time method for improving hand hygiene compliance in clinical settings. Its ability to deliver data-supported feedback highlights its potential to enhance patient safety and reduce infection rates.