Related Experiment Video
Updated: Jun 24, 2025
07:40
Multi-Modal Home Sleep Monitoring in Older Adults
Published on: January 26, 2019
7.6K
Machine learning methods for adult OSAHS risk prediction
Shanshan Ge1, Kainan Wu2, Shuhui Li3
1Health Management Center, the First Hospital of Shanxi Medical University, Taiyuan, 030001, China. geshanshan1@163.com.
BMC Health Services Research
|June 5, 2024
Summary
Machine learning accurately predicts obstructive sleep apnea hypopnea syndrome (OSAHS) using patient data. The Multilayer Perceptron (MLP) model demonstrated superior performance in identifying individuals at risk for OSAHS.
Area of Science:
- Medical Informatics
- Machine Learning in Healthcare
- Sleep Medicine
Background:
- Obstructive sleep apnea hypopnea syndrome (OSAHS) is a prevalent condition linked to systemic organ damage.
- Predictive modeling offers a potential avenue for early OSAHS identification and management.
Purpose of the Study:
- To develop and validate a machine learning (ML) model for predicting OSAHS using polysomnography (PSG) data.
- To identify key risk factors contributing to the development of OSAHS.
Main Methods:
- Retrospective analysis of clinical data from 2064 snoring patients.
- Feature importance analysis identified LDL-C, Cr, carotid artery plaque, A1c, and BMI as significant predictors.
- Five ML algorithms (logistic regression, SVM, Boosting, Random Forest, MLP) were trained and evaluated using cross-validation.
Main Results:
- The Multilayer Perceptron (MLP) model achieved the highest performance.
- MLP model metrics: 85.80% accuracy, 0.89 Precision, 0.75 Recall, 0.82 F1-score, and 0.938 AUC.
- Key predictors for OSAHS included LDL-C, Cr, common carotid artery plaque, A1c, and BMI.
Conclusions:
- A robust ML-based risk prediction model for OSAHS was successfully developed.
- The MLP model demonstrated superior predictive capability compared to other evaluated algorithms.
- This model facilitates early diagnosis and personalized treatment strategies for OSAHS patients.
