臺灣博碩士論文加值系統

腦中風在台灣俗稱中風，美國稱為Stroke，醫學上稱為腦血管疾病。在美國每年有700000人發生腦中風，梗塞性腦中風佔了80%。而台灣1993年至2005年腦中風一直為台灣10大死因前二名，每年約有51000人罹患腦中風。在美國每年在腦中風的照護上需花費將近400億美元，而在台灣每個月要花費1.5-4萬，其住院天數長，這都造成國家、病患及家屬心理上及經濟上的負擔，而不能達到雙贏之狀態，因此若能將資訊科技的技術運用至預測梗塞性腦中風患者的住院天數，並能找出影響住院天數的變項以作為醫護人員臨床照護上之參考，以期能減少腦梗塞病人的住院天數，進而增加急性病房的使用率並減少醫療費用的支出。
本研究利用某醫院因梗塞性腦中風之住院病人共502筆資料，經過病歷回顧後，將Missing data刪除，剩下441筆資料。其採用的變項經參考文獻及與醫師討論後共有14個變項，包括病人的年齡、性別、腦中風的危險因子(糖尿病、高血壓、心臟病、抽菸、喝酒、高血脂)、陳舊的腦中風、住院及出院時NIHSS (National institutes of health stroke scale)與BI(Barthel index)之分數及住院天數。運用此資料來分析影響住院天數之變項，並以資料探勘技術建構一個預測模型。
我們利用C4.5 決策樹及類神經網路分析梗塞性中風病人資料，所分析的結果顯示，影響梗塞性中風病人的住院天數變項包括:年齡、性別、高血壓、高血脂、心臟病、喝酒、剛住院及出院時的BI 分數 和剛住院時的NIHSS分數。如將這些分析結果應用於梗塞性中風病人，以早期針對這些影響住院天數之變項加以注意，並對患有高血壓、高血脂、心臟病、及有喝酒的病人擬定積極的住院照護計畫，其住院天數或許可以縮短。
另外我們比較了C4.5決策樹、類神經網路及邏輯氏回歸在預測梗塞性腦中風住院天數之正確性，發現C4.5決策樹預測的正確性效果最好，其次為類神經網路、最後則是邏輯氏回歸。這也說明運用C4.5決策樹及類神經網路的技術對於預測模型有較高的準確性。期望未來能將此模型實際應用在臨床上以評估其可行性及實用性。

The stroke is a cardiovascular disease. The annual incidence of stroke is approximately 700,000 per year in the United States, and Ischemic stroke accounts for 80% of them. In Taiwan 51,000 people suffer from stroke each year, and 15,000 of them resulted in death. From 1993 to 2005 stroke is the second most common cause of mortality, and 71% was due to ischemic stroke in Taiwan. In America, the annual cost on health care of stroke is near 40 billion USD, while in Taiwan the monthly cost is 15-40 thousand NTD. That was to make the nation, the patients and their family’s financial and psychological burden. So, the win-win situation could not be accomplished to the patient and nation. Therefore, if information techniques can be used to predict the hospitalization length of stay of ischemic stroke patients. And to discover the effect variables of ischemic stroke length of stay. It can provide for the medical personnel to make the medical plan. We hold that can expect the increased usage of acute wards and decrease the cost of medical expenses.
In our study used the 441 patient data records and 14 variables. The variables are included age, sex, Diabetes Mellitus, Hypertension, heart disease, smoking, alcohol, hyperlipemia, old CVA, LOS, admission and discharge BI score and admission of NIHS score. This study applied C4.5 decision tree, ANN and logical regression with past data to establish the prediction model and to predict the patients’ length of stay.
In the analysis from the C4.5 decision tree one can see that the key variables in the LOS of ischemic stroke patients are BI score, age, hyperlipidemia, sex, alcohol, HTN, heart disease, NIHSS at admission hospital, in which the BI during hospitalization is the most important. We found that the C4.5 decision tree model out performs ANN and logistic regression in terms of accuracy. So, it points out if applying the C4.5 decision tree and ANN to predict the ischemic stroke length of stay, it has high of predictive accuracy. In future studies the model should be applied in clinical practice.

誌謝 I
中文摘要 II
Abstract IV
Table of Contents VI
List of Figures IX
List of Tables X
List of Abbreviations XI
Chapter I Introduction 1
1.1 Backgrounds 1
1.2 Stroke and risk factor 3
1.3 NIHSS and BI score 6
1.4 Data Mining 8
1.4.1 Decision Trees 10
1.4.2 Logical Regression 12
1.4.3 ANN 12
1.4.4 About WEKA 15
1.5 Research Aims 15
1.6 Organization of this thesis 16
Chapter II Literature Review 17
2.1 Data Mining in Medical Application 18
2.2 Prediction of Hospital Length of Stay and Cost 20
Chapter III Materials and Methods 23
3.1 Procedures of Building Model 23
3.2 Data Source 24
3.3 Data Understanding and Data Preparation 25
3.4 Data Transform 26
Chapter IV Results 30
4.1 Data Description 30
4.2 Variables and Attribute Selection 33
4.3 Establishing Classification Model 36
4.3.1 C4.5 37
4.3.2 ANN 39
4.4 Evaluation 44
Chapter V Discussion and Conclusion 47
5.1 Discussion 47
5.2 Limitations of the study 48
5.3 Conclusions 49
5.4 Future Research 49
Bibliography 51
Appendix 57

[1] 行政院衛生署統計資料http://www.doh.gov.tw/statistic/index.htm
[2] Hu HH, Sheng WY, Chu FL, Lan CF, Chiang BN. “Incidence of stroke in Taiwan”, Stroke; a journal of cerebral circulation. 1992 Sep;23(9):1237-41
[3]陳甫州，“醫學新探索-腦中風之治療與展望”，科學知識 2001 November
[4]World Health Organization http://www.who.int/cardiovascular_diseases/resources/atlas/en/.
[5]腦中風線上完全手冊, http://health.healthonline.com.tw/column/book_1/
[6]王慧儀、張志仲、宋雲通、林永哲. “中風一年後功能恢復的程度與有關因素之關係”,高雄醫學雜誌 2000
[7]Piatetsky-Shapiro G., and W. J. Frawley, “Knowledge Discovery in Databases,” AAAI/MIT Press, 1991
[8]尹相志，“SQL Server 2005 “資料採礦聖經“,學貫行銷股份有限公司，2007
[9] American heart association http://www.americanheart.org/presenter.jhtml?identifier=1200000
[10] 國民健康局 http://www.bhp.doh.gov.tw/BHP/index.jsp
[11] Johnston KC, Connors AF, Jr., Wagner DP, Knaus WA, Wang X, Haley EC, Jr.“A predictive risk model for outcomes of ischemic stroke”, Stroke; a journal of cerebral circulation. 2000 Feb;31(2):448-55
[12] Goldstein LB, Bertels C, Davis JN. “Interrater reliability of the NIH stroke scale. Archives of neurology”; 1989. p. 660-2
[13] Cios KJ, Moore GW. “Uniqueness of medical data mining”, Artificial intelligence in medicine. 2002 Sep-Oct; 26(1-2):1-24
[14] Schlegel D, Kolb SJ, Luciano JM, et al, “Utility of the NIH Stroke Scale as a predictor of hospital disposition “, Stroke; a journal of cerebral circulation. 2003
[15]Johnston KC, Connors AF, Jr., Wagner DP, Knaus WA, Wang X, Haley EC, Jr. “A predictive risk model for outcomes of ischemic stroke.” Stroke; a journal of cerebral circulation, 2000 Feb;31(2):448-55
[16]Lai SM, Duncan PW, Keighley J. “Prediction of functional outcome after stroke: comparison of the Orpington Prognostic Scale and the NIH Stroke Scale”. Stroke;cerebral circulation. 1998 Sep; 29(9):1838-42
[17]Chang KC, Tseng MC, Weng HH, Lin YH, Liou CW, Tan TY. “Prediction of length of stay of first-ever ischemic stroke“, Stroke;cerebral circulation. 2002 Nov;33(11):2670-4
[18]Mahoney FI, Barthel DW. “Functional Evaluation: the Barthel Index”, Maryland state medical journal. 1965 Feb;14:61-5
[19]Lai SM, Duncan PW, Keighley J. “Prediction of functional outcome after stroke: comparison of the Orpington Prognostic Scale and the NIH Stroke Scale”, Stroke; a journal of cerebral circulation. 1998 Sep; 29(9):1838-42
[20]Geert sulter, Christel steen and Jacques De Keyser. “Use of the Barthel Index and Modified Rankin scale in Acute Stroke Trials.” 1999 May; 30:1538-41
[21]羅中平，“簡介NIHSS，Modified Rankin scale，and Barthel Index量表”.
[22]Neves J, Alves V, Nelas L, Romeu A, and Basto S. “An information system that supports knowledge discovery and data mining in medical imaging. Machine learning an applications: machine learning in medical applications”. Chania, Greece 1999 37-42
[23]Sit fatimah MD saad an Rogayah Ghazali. “Data mining for medical database. Proceedings of the first national conference on artificial intelligence application in industry. Kuala Lumpur 1999 72-79
[24]Siti Nurul Huda Sheikh Abdulah and Miswan Surip. Satu metofologi perlombongan data Untuk pesakit AIDS. Proccedings of the first national conference on Artificial intelligence Application in Industry. Kuala Lumpur. 1999 57-71
[25]Ian HW, Eibe F. “data mining practical machine learning tools and techniques”, CA: Morgan Kaufmann; 2005
[26]Quinlan J. “Induction of decision trees”, Mach learn 1986; 1:81-106
[27]Quinlan J. “C4.5: programs for machine learning”, San Mateo, CA: Morgan Kaufmann; 1993
[28]Software For Predictive Modeling and Forecasting http://www.dtreg.com/dtintro.htm
[29]Salvatore Ruggieri, “Efficient C4.5,” IEEE transactions on knowledge and data engineering, vol 14,no2, p438-444,2002
[30]Breiman L, Friedman JH, Olshen RA, Stone CJ. “Classification and regression tree”, Monterey ,CA: Wadsworth Brooks/Cole Advanced Books software; 1984
[31]Gerard E. “Logistic Regression”, http://www.tufts.edu/~gdallal/logistic.htm
[32] Wikipedia. http://en.wikipedia.org/wiki/Wiki
[33] Dursun D., Glenn W., Amit Kadam. “Predicting breast cancer survivability: a comparison of three data mining methods”, Artificial intelligence in medicine(2005) 34, 113-127
[34]Haykin S.”neural network:a comprehensive foundation”, New Jersey:prentice Hall;1998
[35]Alexopoulos E., Dounias G.D., Vemmos K,. “medical diagnosis of stroke using inductive machine learning”
[36]Milan Zorman a, Vili Podgorelec a, Peter Kokol a, Margaret Peterson b, Matej S_ progar a, Milan Ojstersˇek, “Finding The Right Decision Tree's Induction Strategy For A Hard Real World Problem”
[37]陳瑞杰, “探討台灣外傷系統之設計-資料探勘分析法於健康資料庫之應用”, 台北醫學大學醫學資訊研究所碩士論,2003
[38]Min HH, Yu CL, Wen TC, Ju CY,“Outcome prediction after moderate and severe head injury using an artificial neural network”, European federation for medical informatics. 2005, 241-245
[39]Jorgensen HS, Nakayama H, Raaschou Ho, Olsen TS, “Acute stroke care and rehabilitation: an analysis of the direct cost an its clinical and social determinants: the Copenhagen stroke study”,Stroke. 1997; 28:1138-41
[40]Holloway RG. Witter DM Jr, Lawton KB,Lipscomb J, Samsa G. “inpatient costs of specific cerebrovascular events at five academic medical centers”, Neurology.1996;46:854-860
[41]Mamoli A, Censori B, Casto L. Stileo C, Cesana B, Camerlingo M,“An analysis of the costs of ischemic stroke in an Italian stroke unit”, Neurology. 199; 53:112-116
[42]張英明、陳品玲、胡朝榮、游家銘、吳柏瑞、陳俊亨,“某區域醫院急性缺血性腦中風之醫療成本”,北市醫學雜誌 2005;2(3):286-292
[43]袁繼銓,“以類神經網路預測燒傷病患住院日之研究”， 國立中山大學資訊管理研究所碩士論文,2003
[44]梁忠詔, “影響復健科患者住院天數長短因素之探討—以花東某醫學中心為例”，國立東華大學企業管理研究所碩士論文,2000
[45]Burke HB, Goodman PH, Rosen DB, “Artificial neural networks improve the accuracy of cancer survival prediction”, Cancer. 1997 Feb 15; 79(4):857-62.