Machine learning for improved medical device management: A focus on infant incubators

Area of Science:

• Biomedical Engineering
• Artificial Intelligence in Healthcare
• Medical Device Regulation

Background:

• Medical devices (MDs) require stringent maintenance for patient safety and effective treatment.
• Infant incubators are critical for premature infants, necessitating robust post-market surveillance.
• Current surveillance methods may miss critical device failures between scheduled inspections.

Purpose of the Study:

• To develop an automated machine learning system for predicting infant incubator performance.
• To address the risk of undetected faulty infant incubators between surveillance periods.

Main Methods:

• Collected 1997 samples during infant incubator inspections in Bosnia and Herzegovina.
• Evaluated machine learning algorithms: Decision Tree (DT), Random Forest (RF), Naïve Bayes (NB), and Logistic Regression (LR).
• Developed an automated system for predicting infant incubator performance status.

Main Results:

• Naïve Bayes achieved a 0.93 AUC, demonstrating strong predictive capabilities.
• The study compared the predictive accuracy of NB against DT and RF.
• Machine learning algorithms were effective in handling large datasets for performance prediction.

Conclusions:

• Machine learning effectively predicts infant incubator performance using post-market surveillance data.
• AI-driven automated systems can overcome challenges in ensuring the quality of in-use medical devices.
• Implementing these systems enhances the safety and reliability of infant incubators in healthcare settings.