基於身體組成分析構建預測阻塞性睡眠呼吸中止症之深度學習模型

Taipei Medical University Institutional Repository > 醫學院 > 人工智慧醫療碩士在職專班 > 博碩士論文 > Item 987654321/62460

請使用永久網址來引用或連結此文件: http://libir.tmu.edu.tw/handle/987654321/62460

題名:	基於身體組成分析構建預測阻塞性睡眠呼吸中止症之深度學習模型 Constructing a Deep Leaning Model for Predicting Obstructive Sleep Apnea based on Body Composition Analysis
作者:	郭文蕙 KUO, WEN-HUI
貢獻者:	醫學院人工智慧醫療碩士在職專班劉文德
關鍵詞:	阻塞性睡眠呼吸中止症;身體組成分析;深度學習 Obstructive Sleep Apnea;Body Composition Analysis;Deep Leaning
日期:	2022-07-01
上傳時間:	2023-01-17 14:52:51 (UTC+8)
摘要:	研究背景：睡眠呼吸中止症是常被人們忽視的一種隱形疾病。病友常因病床數的不足，錯失及早診治的時機，再者病友在醫院檢查時，安裝在身上量測儀器的不適，導致檢驗數據失真而誤判病情，無形中造成醫療資源的浪費。如何讓病友在的穩定環境中，方便取得預測睡眠呼吸中止症的特徵因子並建構預測模型是國內外專家學者努力的目標。目前現有研究中多採用單次檢驗數據，且使用規則式（Rule-based）的演算法。近年曾有研究使用機器學習中支援向量機(Support Vector Machine，SVM)建立模型正確率達到83%，特異性與敏感度可達八成，卻鮮少人使用深度學習中的神經網路建立阻塞型睡眠呼吸中止症的預測模型。研究目的：採用身體組成分析的各項特徵值，運用深度學習框架方法，構建阻塞型睡眠呼吸中止症預測模型研究方法：本研究採用日常可量測的生理參數（頸、腰圍及血氧等）與睡覺前後身體組成各項生理變化數據，運用AI深度神經網路學習方法(Deep Neural Network，DNN），以性別與年齡分群，構建阻塞型睡眠呼吸中止(Severe Obstructive Sleep Apnea Syndrome, OSA)預測模型研究結果：性別與年齡分群所建構的預測模型在正確率可達到87%之阻塞型睡眠呼吸中止(Severe Obstructive Sleep Apnea Syndrome, OSA)預測模型，其結果正確率是優於之前研究建立的預測模型模型。研究結論：期待所構建的阻塞型睡眠呼吸中止症預測模型，可以真正落實幫助患者能夠早期發現，進而早期治療，減少國家勞健保費用支出，輔助醫生做前期病情篩選機制，減輕醫生工作量的負荷。 Research background: Sleep apnea is an invisible disease that is often overlooked. Patients often miss the opportunity for early diagnosis and treatment due to the lack of hospital beds. Furthermore, patients are uncomfortable with measuring instruments installed on their bodies during examinations in the hospital, resulting in distorted test data and misjudgment of the condition, which invisibly results in a waste of medical resources. How to make patients in a stable environment to easily obtain the characteristic factors for predicting sleep apnea and build a predictive model is the goal of domestic and foreign experts and scholars. At present, most of the existing researches use single test data and use a rule-based algorithm. In recent years, there have been studies using the Support Vector Machine (SVM) in machine learning to build a model with an accuracy rate of 83% and a specificity and sensitivity of 80%. However, few people use the neural network in deep learning to build blocking models. A predictive model for sleep apnea. Research purposes: To construct a predictive model for obstructive sleep apnea by using the eigenvalues of body composition analysis and the deep learning framework method. Research methods: This study uses daily measurable physiological parameters (neck, waist circumference, blood oxygen, etc.) and various physiological changes in body composition before and after sleeping, using AI deep neural network learning method (Deep Neural Network, DNN). Gender and age grouping to construct a predictive model for Severe Obstructive Sleep Apnea Syndrome (OSA) Research results: The prediction model constructed by gender and age group has an accuracy rate of 87% for the Severe Obstructive Sleep Apnea Syndrome (OSA) prediction model. The accuracy rate of the results is better than the prediction model established by previous studies Model. Research conclusions: It is expected that the constructed prediction model of obstructive sleep apnea can truly help patients to be detected and treated early, reduce national labor and health insurance expenses, assist doctors in the early disease screening mechanism, and reduce the workload of doctors.
描述:	碩士指導教授：劉文德委員：何淑娟委員：彭徐鈞委員：劉文德
資料類型:	thesis
顯示於類別:	[人工智慧醫療碩士在職專班] 博碩士論文

文件中的檔案:

檔案	描述	大小	格式	瀏覽次數
index.html		0Kb	HTML	363	檢視/開啟

在TMUIR中所有的資料項目都受到原著作權保護.

TAIR相關文章

資料載入中.....