Exploring the Association Pattern of Rheumatoid Arthritis Hand Rehabilitation Posture with Nutritional Supplements and Drugs by Using Deep Learning Mechanism

View Statistics Email Alert RSS Feed

Information

Details

Project title

Code/計畫編號

MOST108-2221-E019-046-MY2

Translated Name/計畫中文名

利用深度學習機制探討類風溼性關節炎手部復健姿勢與其營養補充品和藥品之關聯模式

Project Coordinator/計畫主持人

Hao-Hsiang Ku

Funding Organization/主管機關

National Science and Technology Council

Co-Investigator(s)/共同執行人

曹彥博
洪欣儀

Department/Unit

Institute of Food Safety and Risk Management

Website

https://www.grb.gov.tw/search/planDetail?id=13334080

Year

2020

Start date/計畫起

01-08-2020

Expected Completion/計畫迄

31-07-2020

Bugetid/研究經費

848千元

ResearchField/研究領域

資訊科學--軟體

本計畫將針對類風濕性關節炎病患設計一利用深度學習機制探討類風溼性關節炎手部復健姿勢與其營養補充品和藥品之關聯模式(Exploring the Association Pattern of Rheumatoid Arthritis Hand Rehabilitation Posture with Nutritional Supplements and Drugs by Using Deep Learning Mechanism, APRON)，由所提出的深度學習導向之類風溼性關節炎診斷因子分析模型(Deep Learning-Oriented Rheumatoid Arthritis Criteria Factor Analysis Model, DRAM)配合卷積神經網路(Convolutional Neural Networks)與深度信念網路(Deep Belief Networks, DBNs)將病患手部復建數據與攝取營養品等數據做為訓練因子進行深度學習訓練，產出診斷因子。診斷因子會以倒傳遞類神經(Back Propagation Neural Network)整合本體論(Ontology)技術，建構一大眾本體論推論模式(Mass Behavior Model)，以此做為六種手部復健姿勢驗證模組來判斷病患復健姿勢是否需要調整的依據。而日後，將依照患者數次的復健狀況與相關攝取資訊反覆進行深度學習訓練與推導，使其具有適合個人的輕量化本體論(Ontology-lite)法則。最後，以混合實境(Mixed Reality)技術將病患復健數據重新建模，提供醫生觀察復建情形，以此作為診斷依據並給予病患建議。有鑑於上述內容，本計畫所設計之APRON將建置(1)具輕量化本體論與深度感知互動性之使用者端、(2)具六種復健姿勢分析之個人化評估應用伺服器、及(3)具分散式深度學習之異質資料整合雲端伺服器。最後所產出之APRON除可當作一未來研究的參考模型外，亦與臺北榮民總醫院合作進行測試及使用。 This project is exploring the Association Pattern of Rheumatoid Arthritis Hand Rehabilitation Posture with Nutritional Supplements and Drugs by Using Deep Learning Mechanism, which is called APRON. APRON designs and defines personalized ontology lite by six kinds of hand-based rehabilitation postures to evaluate injuries of Rheumatoid Arthritis (RA) patients. The proposed APRON is a three-tier architecture, which is composed of rheumatoid arthritis patients with IOT and wearable devices, rehabilitation assessment application servers and a cloud-based distributed deep learning server. The reference engine of APRON is Deep Learning-oriented Rheumatoid Arthritis Criteria Factor Analysis Model (DRAM). DRAM is composed of Convolutional Neural Networks and Deep Belief Networks to reason out influence factors and a mass behavior model for a new RA patient. Mass Behavior model is based on the Hadoop database to manipulate patients’ raw data and influence factors. Doctors and APRON can evaluate each patient by the patient’s hand rehabilitation status, nutritional supplements and drugs intake patterns. Furthermore, APRON will modify mass behavior model and combine influence factors to construct ontology-lite for each patient. Finally, APRON provides Mix Reality (MR) to re-construct and reproduce patients’ hand-based rehabilitation postures and food intake patterns for doctors and patients. Hence, this project defines and designs the mass behavior model, the ontology-lite model, Deep Learning-oriented Rheumatoid Arthritis Criteria Factor Analysis Model (DRAM) and six kinds of hand-based rehabilitation postures of APRON. Finally, APRON can be a reference model for researchers and engineers.

Keyword(s)

手部姿勢分析模式
輕量化本體論
大眾模式
深度學習
混合實境
類風濕性關節炎
Hand Posture Analysis Model
Ontology-lite
Mass Behavior Model
Deep Learning
Mixed Reality
Rheumatoid Arthritis

DSpace CRIS

Exploring the Association Pattern of Rheumatoid Arthritis Hand Rehabilitation Posture with Nutritional Supplements and Drugs by Using Deep Learning Mechanism

Details

Description