Using artificial intelligence and medical health big data to reverse the medical care and management of chronic diseases project (NSTC 110－2634－F－002－049), sub-plan8, model3
These models were built by Yu-Ting Huang (Professional Master Program in Artificial Intelligence in Medicine, College of Medicine, Taipei Medical University)
Background: Obstructive sleep apnea is recurrent collapse or obstruction of the upper airway, resulting in strenuous breathing and shallow airflow, which may lead to suffocation in severe cases. Assisting in the early detection of OSA is an important goal.
Objective: To analyze the correlation between clinical test data and obstructive sleep apnea, and to construct a machine learning model for predicting the risk of obstructive sleep apnea.
Methods: The data were collected from the Sleep Center and Medical Laboratory Department of Shuang-Ho Hospital of the Ministry of Health and Welfare (entrusted to build and operate by Taipei Medical University). The date of acceptance is from 2016 to 2021, including adults and seniors aged 20 to 90, using SPSS software for statistical data analysis, using machine learning to construct a predictive model, and providing one to those in need (such as: health check customers) A tool that can predict the risk of obstructive sleep apnea.
Results: For AHI 15 without MiniO2 prediction model, AUC, (95% CI) reached 0.848 (0.836 - 0.860); for AHI 15 with MiniO2 prediction model, AUC, (95% CI) reached 0.888 (0.882 - 0.894). In AHI 30 without MiniO2 prediction model, AUC, (95% CI) reached 0.809 (0.797 - 0.821); while in AHI 30 with MiniO2 prediction model, AUC, (95% CI) reached 0.857 (0.851 - 0.863). Physical parameters combined with clinical test data and the lowest blood oxygen saturation can effectively predict the risk of obstructive sleep apnea.
Keywords: sleep center, medical laboratory, machine learning, obstructive sleep apnea, physical parameters, clinical laboratory data