- 學位論文
中西醫結合於消化性潰瘍證型診斷專家系統之建立
A Diagnostic Expert System of Peptic Ulcer Disease in Chinese and Western Medicine
摘要
摘 要 本研究在根據消化性潰瘍病患之症狀，依據中醫典籍的辨證論治加以科學化定量分型，藉此開發一套結合西方醫學與傳統中國醫學之電腦化專家診斷系統，以提供疾病診斷證型分類及相關之中藥處方和患者特殊注意事項，輔助醫療人員進行可能疾病症候的篩選，並藉此求得特徵值，作為未來消化性潰瘍中西醫辨證中重要的參考因素。 系統設計上，主要是利用類神經網路並輔以人機操作介面來建立一個電腦化中西醫整合系統；並將建立資料庫儲存相關的臨床資料，以便後續相關研究之使用。 在資料分析系統設計方面，主要核心是利用類神經網路、規則萃取與決策串列等技術所架構出來。在類神經網路方面，採用倒傳遞類神經網路(Back-Propagation)的學習法則來建立，網路共分為輸入層、隱藏層與輸出層三層。輸入層為中醫症狀與西醫診斷因子，輸出層分別為中西醫證型以及處方用藥。在規則萃取部份，採用分解型(decompositional)演算法。最後，再以決策串列形式的診斷順序輸出系統推論結果。 本研究共收集107名消化性潰瘍患者來作為本次研究的對象，採用的症狀共60種，證型有7種，以及45味中藥為研究主軸。將所建立的系統在經過中西醫五種模式資料測試後，分成訓練組72例，測試組35例，得知系統的平均診斷準確率為80％。此外，根據所獲得之規則，可發現某些中醫證型與西醫的診斷因素彼此之間有相關性。然而由於系統在推論處方用藥的部份並無法有良好的分析結果，所以在處方用藥的部份僅供醫師參考之用。 期望本系統的建立，能輔助使用者觀摩其它資訊，並可以成為中西醫師很好的輔助工具。
並列摘要
Abstract In this study, we proposed to build a quantitative expert system to classify Peptic Ulcer (P.U.) patients according to the symptom-based pattern system of traditional Chinese medicine. The proposed expert system will be based on a model combining the information both from the diagnostics of western medicine and the symptom-asking techniques from Chinese medicine. The neural network and database techniques will be utilized as the core of this system and provide related information for patient diagnosis, prescriptions and particular things for attention. This system will assist to filter out possible diseases and relieve the burden of medical staff. In addition, statistical analysis was applied to study the relation of syndrome of Chinese medicine and physiological parameter. The result could be helpful for Chinese and Western medicine about P.U. in the future. The expert system was built by neural network and inference engine and to set up the database of expert system to store relevant clinical data. In the analysis system, it was established by neural network, rule extraction and decision list. In the neural network, it was built by Back-Propagation neural network. There were three layers in this neural network. The input layer was the layer for symptoms and the output layer was the layer for syndromes of Chinese medicine output and drugs output. In the rule extraction, a decompositional algorithm was used. Finally, we could get the diagnosis sequence of the decision list from the expert system. There were 107 P.U. patients, 60 symptoms, 7 syndromes, and 45 prescriptions of Chinese medicine were included in this study. In the test of diagnosis system, the expert system reached a average diagnostic accuracy rate of 80%. According to the diagnosis rules we found that some syndromes of Chinese medicine and diagnosis factors of Western medicine were related. Finally, this study is hopefully to achieve the goal of helping the users. Abstract In this study, we proposed to build a quantitative expert system to classify Peptic Ulcer (P.U.) patients according to the symptom-based pattern system of traditional Chinese medicine. The proposed expert system will be based on a model combining the information both from the diagnostics of western medicine and the symptom-asking techniques from Chinese medicine. The neural network and database techniques will be utilized as the core of this system and provide related information for patient diagnosis, prescriptions and particular things for attention. This system will assist to filter out possible diseases and relieve the burden of medical staff. In addition, statistical analysis was applied to study the relation of syndrome of Chinese medicine and physiological parameter. The result could be helpful for Chinese and Western medicine about P.U. in the future. The expert system was built by neural network and inference engine and to set up the database of expert system to store relevant clinical data. In the analysis system, it was established by neural network, rule extraction and decision list. In the neural network, it was built by Back-Propagation neural network. There were three layers in this neural network. The input layer was the layer for symptoms and the output layer was the layer for syndromes of Chinese medicine output and drugs output. In the rule extraction, a decompositional algorithm was used. Finally, we could get the diagnosis sequence of the decision list from the expert system. There were 107 P.U. patients, 60 symptoms, 7 syndromes, and 45 prescriptions of Chinese medicine were included in this study. In the test of diagnosis system, the expert system reached a average diagnostic accuracy rate of 80%. According to the diagnosis rules we found that some syndromes of Chinese medicine and diagnosis factors of Western medicine were related. Finally, this study is hopefully to achieve the goal of helping the users.