Screening of Finger Pulse Oximeter Pre-calibration Function Test with Response Time Analysis using Machine Learning Model

The assessment of pulse oximeter Response Time (RT) during the screening process for pulse oximeter function calibration still relies on manual Excel calculations. This study aims to enhance the efficiency and effectiveness of the screening process by employing a machine learning (ML) model. The goal is to reduce the risk of passing a pulse oximeter with slow response time. We categorize RT data into three classes: monitoring, diagnostic, and preventive. Labeled datasets are fed into various ML models to identify the most effective classifier. The analysis considers RT measurements obtained from finger-attached devices, encompassing various conditions, device types, body temperature, and environmental humidity. Among the evaluated algorithms (SVM, Random Forests, Decision Tree, and KNN), the random forest and decision tree models exhibit superior precision and accuracy compared to other algorithms. © 2024 Owner/Author.