智慧型醫療診斷模式建構於肝病診斷應用

Intelligence Medical Diagnosis Modeling in Liver Disease Application

陳憶萱(Yi-Hsuan Chen)

指導教授 ： 林榮禾

國立臺北科技大學/管理學院/商業自動化與管理研究所/碩士(2006年)

https://doi.org/10.6841/NTUT.2006.00361

摘要

肝病是台灣地區最主要的本土病，根據衛生署統計資料顯示，慢性肝病及肝硬化歷年位居十大死因排名之內，估計國人每年約有一萬人以上死於肝炎、肝硬化、肝癌，其主要與個人特質、日常生活習性、及生理狀況習習相關，加上肝病在初期症狀不明顯，即使是肝硬化或肝癌等嚴重的肝病，恐怕也要等到病情相當嚴重時才會出現症狀，因此如何建立一個早期診斷肝病機制將有其必要性。過去針對醫療診斷所作的相關研究當中，以傳統統計方法來建立診斷模式時必須符合相關之假設與條件，所建立之方程式亦多為線性模式，使得建構出的模式的準確度無法達到理想的標準，應用上有其侷限性。因此，本研究運用資料探勘技術，發展一套智慧型醫療診斷模式，收集病患之生化數據檢查，建立兩階段診斷流程，第一階段運用分類迴歸樹（CART）建構肝病分類預測模式，其主要目的則是判定是否罹患肝病，第二階段運用案例式推理（CBR）診斷罹患何種肝病類型，並利用CART探究肝病病患與無肝病病患在生活行為模式的診斷規則，以更完整找出肝病之藥品採用、治療方法以及導致肝病的關鍵生活行為模式指標，研究結果顯示，CART的分類準確度達88%，CBR的比對準確度達94%，而生活行為模式的關鍵指標為年齡、喝酒、抽煙、嚼檳榔、血型、捐血習慣、每週吃蔬菜天數、每週吃水果天數、身體不舒服其處理方式，即可有效輔助醫師在疾病診斷給予協助，藉以減少醫師主觀的判斷所產生的偏差，降低診斷過程所需花費的時間以及病患的就診成本浪費，達到更正確且完善的醫療與預防，進而提昇醫療品質。

關鍵字

肝病診斷 ； 決策樹 ； 案例式推理 ； CART

並列摘要

Liver disease is the most common disease inTaiwan. According to the Statistics of Department of Health, liver disease is one of the top ten fatal diseases in Taiwan. Every years, around one thousand people die of liver cirrhosis, liver cancer and other liver diseases. The most important reasons for people to have liver diseases are long-term drinking, hepatitis B virus (HBV), and hepatitis C virus (HCV), which are much related to modern people’s daily life. Therefore, the early diagnosis model is very worth to study. The past research of disease diagnosis is aimed at using statistical methods for modeling. However, statistic model require a few assumptions and is normally constrained by the linearity, thus it’s quietly difficult to deal with the mass and complicated data. In order to solve statistical constraints, we tried to establish an intelligence medical diagnosis system for early diagnosis of prelivers by means of computerized analyzing the laboratory data in the patients during medical visits. There are two main stages included of the diagnostic system. In the first step, we used classification and regression tree（CART）to diagnose whether suffer from liver disease. In the second step, we used case-based reasoning（CBR） to diagnose the types of liver diseases and to recommend the appropriate prognosis and treatment. Simultaneously, we used CART to investigate human life behaviour custom and induce liver risk factors.
The research results indicate that CART rate of accuracy is 88% and CBR diagnostic accuracy rate is 94%. In the case of liver disease key risk factors are age, drinking, smoking, chew betel nut, blood, blood donor habit, how many days in a week of eat vegetable, how many days in a week of eat fruit, how to deal with when you are sick. All of the above, these two forecast methods are seen to yield considerably accurate results, and are most likely to reach high forecasting quality. The proposed intelligence diagnosis model can be of great assistance to physicians with the diagnosis of liver diseases, to reducing diagnostic errors and improving the medical treatment quality.

並列關鍵字

Liver diagnosis ； Decision Tree ； Cased-Based Reasoning ； CART

參考文獻

[6]高嘉宏、陳定信，C型肝炎在台灣，中華衛誌，第十七卷，第三

國立台灣科技大學電子工程系，台北，2002。

[18]姚志成，運用資料探勘技術建構脂肪肝預測模式，碩士論文，

[19]行政院衛生署，http://www.doh.gov.tw, 2005

[20]肝病防治學術基金會，http://www.liver.org.tw, 2005