臺灣博碩士論文加值系統

肝癌是一種多種因子誘發且自體細胞產生的病變，因此在治療上十分的繁瑣及複雜，而過去研究也指出癌症照護上的決策是影響病人預後結果的關鍵；從二十一世紀以來，各國研究相繼指出各癌症專科的外科醫師及臨床醫護人員可以組成癌症多專科團隊(Multidisciplinary Team, MDT) ，藉由團隊定期進行多專科癌症會議(Multidisciplinary Cancer Conference, MCC)來討論特定癌症病人的病情並制定合宜的治療準則，而能達到降低醫療成本，以及提高癌症病人診療品質的目的。在2012年，我們在台灣榮民總醫院建置了一套可以輔助癌症多專科團隊定期開會的線上系統(Cancer Multidisciplinary Conference System, CMDTC system)，藉由此套系統，不但可以節省醫護人員對於開會召集、管理、紀錄、追蹤等步驟的時間，也可以方便臨床資料的串聯和比對查詢。
藉由串連CMDTC系統內與榮總癌症登記的資料，我們可以分析所有肝癌病人其各方面的診斷、治療及預後情形，而基於CMDTC系統上，本研究主要分為兩個部分，第一部分為利用Cox比例風險模型來分析肝癌病人是否能藉由此系統結合MCC討論來提升其預後結果，並發現有搭配CMDTC系統和多專科團隊討論的肝癌病人其危險比(Hazard Ratio)顯著地為沒有接受多專科團隊討論病人的0.48；第二部分，本文提出隨機森林的機器學習模型，其達到AUC等於0.82並有0.85的準確率，模型可以從過去醫師的判斷中來學習，哪一類的肝癌病人應該進入多專科團隊會議做討論，藉由系統推薦的方式來提醒醫師，也可由模型找出重要的決策因子，以期可以因此標準化多專科團隊的病人選擇流程。

Liver cancer is a disease caused by multi-factors and abnormal cell growth, so its treatment process is very cumbersome and complicated. The previous study indicated that the clinical decision about cancer care is the key determinant of prognostic outcome of patient. Besides, since 21th century, the research over the world suggested that doctors and other medical professional form a multidisciplinary team (MDT) to hold the multidisciplinary case conference (MCC) regularly to review, discuss the cases and establish the appropriate treatment guideline. These processes not only cut the medical cost, but also improve the quality of cancer care. In 2012, Cancer Multidisciplinary Conference System (CMDTC) system was implemented in Taipei Veterans General Hospital to assist in managing the multidisciplinary cancer conference. With this system, it not only save time for medical professionals to hold and manage the system but also the recording and follow-up of patients. In addition, the connection between the system to other clinical system also help the information join and search for clinical purpose.
With the merge of data collected from CMDTC system and cancer registry, we can have survival analysis on the all the aspect of patients including diagnosis, treatment and prognosis. On the basis of CMDTC system, this research is divided into two part: First part is using Cox proportional hazard model to evaluate the synergic effect of combination of the system and MCC on improvement of prognosis outcome of patients. And the significant result shows that liver cancer patients with MCC got better prognostic outcome (Hazard ratio 0.48; 95% CI 0.26-0.87). And second part of this article proposed a random forest model with AUC 0.82 and accuracy 0.85. It can give prediction about the potential patients with liver cancer requiring the MCC discussion. Also, the importance of clinical factor was given to standardize the patient selection process before MCC.

口試委員會審定書 #
誌謝 i
中文摘要 iii
ABSTRACT iv
CONTENTS v
LIST OF FIGURES vii
LIST OF TABLES viii
Chapter 1 Introduction 1
1.1 Background 1
1.1.1 Multidisciplinary cancer conference (MCC) 1
1.1.2 Cancer Multidisciplinary Team Conference (CMDTC) System 2
1.1.3 Liver cancer 4
1.1.4 Machine Learning in clinical decision support 5
1.2 Motivation and Objective 6
Chapter 2 Material and Method 8
2.1 Study workflow 8
2.2 Data collection 9
2.2.1 Cancer Registry 9
2.3 Survival Analysis 10
2.3.1 Study Design 10
2.3.2 Overview of the Statistics Analysis 11
2.3.3 Propensity Score matching 12
2.3.4 Cox Proportional Hazards model 13
2.4 Machine Learning 14
2.4.1 Study Design 14
2.4.2 Input Feature and Feature Engineering 15
2.4.3 Imputing missing data 17
2.4.4 Imbalance Data Problem 18
2.4.5 Cross validation 20
2.4.6 Random Forest 21
2.4.7 Hyper-parameter Tuning 26
2.4.8 Performance Evaluation 27
Chapter 3 Result 29
3.1.1 Survival Analysis 29
3.1.2 Machine Learning 35
3.1.3 Summary 42
Chapter 4 Conclusion and Future Works 44
REFERENCE 46

1.Fleissig, A., et al., Multidisciplinary teams in cancer care: are they effective in the UK? The lancet oncology, 2006. 7(11): p. 935-943.
2.Ferlay, J., et al., Cancer incidence and mortality worldwide: sources, methods and major patterns in GLOBOCAN 2012. International journal of cancer, 2015. 136(5): p. E359-E386.
3.Jemal, A., et al., Global cancer statistics. CA: a cancer journal for clinicians, 2011. 61(2): p. 69-90.
4.Cruz, J.A. and D.S. Wishart, Applications of machine learning in cancer prediction and prognosis. Cancer informatics, 2006. 2.
5.Kourou, K., et al., Machine learning applications in cancer prognosis and prediction. Computational and structural biotechnology journal, 2015. 13: p. 8-17.
6.Chiang, C.-J., et al., Cancer trends in Taiwan. Japanese journal of clinical oncology, 2010. 40(10): p. 897-904.
7.Caliendo, M. and S. Kopeinig, Some practical guidance for the implementation of propensity score matching. Journal of economic surveys, 2008. 22(1): p. 31-72.
8., D., Regression Models and Life-Tables. Journal of, 1972.
9.Grambsch, P.M. and T.M. Therneau, Proportional hazards tests and diagnostics based on weighted residuals. Biometrika, 1994. 81(3): p. 515-526.
10.Buuren, S. and K. Groothuis-Oudshoorn, mice: Multivariate imputation by chained equations in R. Journal of statistical software, 2011. 45(3).
11.White, I.R., P. Royston, and A.M. Wood, Multiple imputation using chained equations: issues and guidance for practice. Statistics in medicine, 2011. 30(4): p. 377-399.
12.Chawla, N.V., et al., SMOTE: synthetic minority over-sampling technique. Journal of artificial intelligence research, 2002. 16: p. 321-357.
13.Kohavi, R. A study of cross-validation and bootstrap for accuracy estimation and model selection. in Ijcai. 1995.
14.Breiman, L., Random forests. Machine learning, 2001. 45(1): p. 5-32.
15.Chen, C., A. Liaw, and L. Breiman, Using random forest to learn imbalanced data. University of California, Berkeley, 2004: p. 1-12.
16.Segal, M.R., Machine learning benchmarks and random forest regression. Center for Bioinformatics & Molecular Biostatistics, 2004.
17.Breiman, L., et al., Classification and regression trees. 1984: CRC press.
18.Breiman, L., Bagging predictors. Machine learning, 1996. 24(2): p. 123-140.
19.Wolpert, D.H. and W.G. Macready, An efficient method to estimate bagging''s generalization error. Machine Learning, 1999. 35(1): p. 41-55.
20.Tibshirani, R., Bias, variance and prediction error for classification rules. 1996: Citeseer.
21.Bergstra, J.S., et al. Algorithms for hyper-parameter optimization. in Advances in Neural Information Processing Systems. 2011.
22.Bergstra, J. and Y. Bengio, Random search for hyper-parameter optimization. Journal of Machine Learning Research, 2012. 13(Feb): p. 281-305.