Establishing a Clinical Prediction Model of Sleep Apnea Syndrome by Adaptive Neuro-Fuzzy Inference System

• Main Authors: Tang-YiWang, 王瑭毅
• Other Authors: Jer-Nan Juang
• Format: Others
• Language: en_US
• Published: 2018
• Online Access: http://ndltd.ncl.edu.tw/handle/evnmr4

碩士 === 國立成功大學 === 工程科學系 === 106 === In recent years, getting a good sleep becomes one of the important issues and there are more researches to study sleep disorders. Obstructive sleep apnea (OSA) is one of the sleep disorders that has attracted much attention. However, the diagnosis of OSA is still limited in the daily clinical practice. Although we can obtain more accurate diagnosis through the examination in the hospital’s sleep lab, it is very time-consuming and expensive to undertake one. Besides, developing an accurate prediction model for OSA is still a difficult task in clinical trial. The purpose of this thesis is to use the machine learning method based on the Adaptive Neuro Fuzzy Network Inference System (ANFIS) to establish a predicting model for the severity of OSA and to explore the correlation between input variables and results divided by genders and age. Therefore, the result will be helpful for medical doctors to make medical decisions, and then the OSA patients could receive the corresponding and suitable follow-up treatment more efficiently. There are three main parts in the research. The first part is the pre-processing of database including anthropometric measurement features, age, questionnaire scores and clinical data of patients, and calculating the single variable correlation between features and the Apnea Hypopnea Index (AHI). The second part is to build up several ANFIS models with different combinations of all input features to describe the distribution of features, and the last part is the validation of model. The results show that the correlation between the characteristic parameters and the results can be found by gender and age, and more accurate prediction results can be obtained. The best sensitivity and specificity in young female is 65.6% and 90% (respectively 74.1% and 61.3% in elder female) for the AHI threshold 15. For male, the result is 79.1% and 77.1% in young male (respectively 82% and 60% in elder male) for the AHI threshold 30. In addition, neck circumference multiplied by waist circumference divided by height (NWH), diastolic blood pressure in the morning and the degree of blood oxygen saturation can be considered to be the influential features for predicting sleep apnea.