Predicting REM Sleep Fragmentation Decline in OSA Patients Post-Surgery: Analysis of Machine Learning Models

REM fragmentation, characterized by frequent arousal disruptions during REM sleep, is a critical factor influenced by the severity of Obstructive Sleep Apnea (OSA) and its associated health risks. Identifying features that predict the reduction in REM fragmentation after OSA surgery is important for improving patient sleep quality. This study aimed to develop and compare the performance of two machine learning models-Support Vector Machine (SVM) and Logistic Regression (LR)-in predicting the reduction of REM fragmentation in OSA patients post-treatment. Clinical and demographic data, including variables such as the Apnea-Hypopnea Index (AHI), Body Mass Index (BMI), neck circumference, age, and sex, were used to train and evaluate both models. The results demonstrated that LR outperformed SVM in terms of predictive accuracy and sensitivity. Feature importance analysis in the LR model identified REM fragmentation before treatment, age, and neck circumference as significant predictors of REM fragmentation reduction. These findings suggest that LR is a more effective tool than SVM for predicting REM fragmentation reduction in OSA patients following treatment, emphasizing the role of REM fragmentation as a key metric in personalized OSA management strategies.