臺灣博碩士論文加值系統

在影像診斷學的領域中，核子醫學功能性影像對疾病早期偵測的角色日趨重要，尤其是在定量分析上，更有其他構造性影像無所取代的地位，本研究把目標設定在核子醫學之極座標靶心圖影像的部分。本研究的動機乃起因於醫師對於判別靶心圖有所疑惑，目前醫師依其經驗藉由靶心圖來判斷病人的心臟是否有病變，若出現錯誤的診斷（False-Positive），造成醫療資源的浪費、醫療成本的增加與病患的抱怨等，無疑的，也帶給這些病人額外的負擔及時間的浪費。
本研究提出一套結合粗集合理論（Rough Set Theory）與類神經網路兩種不同機制的系統，對靶心圖影像與病患資料做相關定性與定量的處理。透過粗集合理論對於知識解釋的優點彌補類神經網路暗箱模式的處理機制，並且利用訓練後的類神經網路協助診斷粗集合理論所無法定義規則的資料我們利用臨床醫師的觀念與文獻的說明於靶心圖影像做處理，作為特徵值擷取上的方法，以協助醫師診斷靶心圖影像，降低誤診的機率，進而提升整體醫療品質，另外，也可作為新進醫師的診斷依歸。實證部份，分別比較本研究所提出的系統、類神經網路與專業醫師三者的差異，結果顯示本研究所提出的系統在明確性（Specificity）與準確率（Accuracy）的表現上，皆優於類神經網路與專業醫師，對於醫師將來所做診斷有個很好的依據。

Nuclear medicine is a specialty that uses radioactive substance in the diagnosis and treatment of diseases. In contrast to other conventional imaging procedures, nuclear medicine imaging is unique in that it can provide both functional and structural Information of an organ simultaneously.
In this study, we propose a new system that diagnoses the polar bull’s eye images based on nuclear medicine, combining rough set theory and neural network. We can get reduced patients’ textual table, which implies that the number of evaluation criteria is reduced with no information loss through rough set approach. And then, a new table which combines reduced patients’ textual table and image table is used to develop classification rules and train neural network to get rule-base and trained neural network. The effectiveness of our methodology is verified by experiments comparing neural network approach and the physician with our new system. According to the result, the specificity and the accuracy in our new system are better than neural network approach and the physician.

中文摘要 I
英文摘要 II
誌謝 III
目錄 IV
表目錄 VII
圖目錄 IX
第一章 緒論 1
1.1 研究背景與動機 1
1.2 研究目的 5
1.3 研究範圍 5
1.4 研究架構 5
1.5 研究流程 6
第二章 文獻探討 7
2.1 類神經網路於靶心圖之應用 7
2.2 粗集合理論與類神經網路結合之應用 9
第三章 研究方法 14
3.1 前處理－靶心圖與病患資料 14
3.1.1 靶心圖影像之處理 15
3.1.1.1 影像處理步驟一 15
3.1.1.2 影像處理步驟二 16
3.1.1.3 影像處理步驟三 17
3.1.1.4 影像處理步驟四 19
3.1.1.5 影像處理步驟五 20
3.1.1.6 影像處理步驟六 20
3.1.2 病患文字資料之處理 22
3.2 處理－粗集合理論與類神經網路 24
3.2.1 類神經網路 25
3.2.1.1 類神經網路的基本架構 26
3.2.1.1.1 處理單元 26
3.2.1.1.2 層 28
3.1.1.1.3 網路 29
3.2.1.2 類神經網路的能力 29
3.2.1.3 倒傳遞類神經網路 30
3.2.1.4 類神經網路學習法則 31
3.2.1.4.1 倒傳遞演算法（BP） 31
3.2.1.4.2 Levenberg-Marquardt演算法（L-M BP） 34
3.2.2 粗集合理論簡介 36
3.2.2.1 資訊系統 36
3.2.2.2 不可分辨關係 37
3.2.2.3 上、下限近似 38
3.2.2.4 邊界近似的準確性 40
3.2.2.5 屬性的獨立與相依關係 40
3.2.2.6 屬性的核與簡化 41
3.2.2.7 屬性值的核與簡化 43
3.2.2.8 決策表 44
3.2.2.9 規則的產生 48
3.3 新的診斷系統 49
第四章 實驗測試與結果 50
4.1 研究資料 50
4.2 類神經網路之設定 51
4.2.1 類神經網路實驗資料 52
4.2.2 類神經網路模式性能指標 52
4.2.3 類神經網路模式之訓練與測試 52
4.3 規則之找尋 55
4.4 實驗方式 58
4.5 實證結果 59
第五章 結論與建議 65
參考文獻 67
附錄一 百分比紀錄表 71
附錄二 影像紀錄表 73
附錄三 病患文字資料表 76
附錄四 編碼後的病患文字資料表 79
附錄五 另外三組規則庫 82

[1] 北榮核醫，http://www.npcc.gov.tw/，2003。
[2] 陳智宏，「應用類神經網路於電力系統負載之溫度敏感度分析」，國立中山大學電機工程所碩士論文，2001。
[3] 陳逸雯，「肺部腫瘤偵測之電腦輔助診斷系統」，中原大學醫學工程所碩士論文，2001。
[4] 張智星，「MATLAB程式設計與應用」，清蔚科技，2000。
[5] 葉怡成，「類神經網路模式 應用與實作」，儒林，1999。
[6] 劉聖文，「調適型銷售互動模式應用於電子商務之研究」，朝陽科技大學企業管理所碩士論文，2001。
[7] 羅華強，「類神經網路－MATLAB的應用」，清蔚科技，2001。
[8] Ahn, B.S., S.S. Cho, and C.Y. Kim, “The Integrated Methodology of Rough Set Theory and Artificial Neural Network for Business Failure Prediction,” Expert Systems with Applications, Vol.18, 2000, 65-74.
[9] Attoh-Okine, N.O., “Combining Use of Rough Set and Artificial Neural Networks in Doweled-Pavement-Performance Modeling-A Hybrid Approach,” Journal of Transportation Engineering, Vol. 128, 2002, 270-275.
[10] Cios, K. J., A, Teresinska, S, Konieczna, J, Potocka, S, Sharma, “A Knowledge Discovery Approach to Diagnosing Myocardial Perfusion,” IEEE Engineering in Medicine and Biology, Vol.19, 2000, 17-25.
[11] Cooke, C.D., C.A. Santana, T.I. Morris, and E.V. Garcia, “Validating Expert System Rule Confidences Using Data Mining of Myocardial Perfusion SPECT Databases,” Computers in Cardiology, Vol.27, 2000, 785-788.
[12] Düntsch, I., and G. Gediga, “The Rough Set Engine GROBIAN,” Proc. of the 15th IMACS World Congress, Vol.4, 1997, 613-618.
[13] Fujita, H., T. Katafuchi, T. Uehara, and T. Nishimura, “Neural Network Approch for the Computer-Aided Diagnosis of Coronary Artery Diseases in Nuclear Medicine,” Neural Networks, Vol.3, 1992, 7-11.
[14] Gu, X.P., S.K. Tso, and Q. Zhang, “Combination of Rough Set Theory and Artificial Neural Networks for Transient Stability Assessment,” Proc. PowerCon 2000. International Conference on Power System Technology, Vol.1, 2000, 19-24.
[15] Hamilton, D., P.J. Riley, U.J. Moila, and A.A. Amro, “A Feed Forward Neural Network for Classificat- ion of Bull’s-Eye Myocardial Perfusion Images,” European Journal of Nuclear Medicine, Vol. 22, No. 2, 1995, 108-115.
[16] Haraldsson, H., and M. Ohlsson, “Value of Exercise Data for the Interpretation of Myocardial Perfusion SPECT,” Journal of Nuclear Cardiology, Vol. 9, 2002, 169-173.
[17] Hashemi, R.R., L.A. Le Blanc, C.T. Rucks, and A. Rajaratnam, “A Hybrid Intelligent System for Predicting Bank Holding Structures,” European Journal of Operational Research, Vol.109, 1998, 390-402.
[18] Jelonek, J., K. Krawiec, R. Slowinski, J. Stefanowski, and J. Szymas, “Neural Networks and Rough Sets - Comparison and Combination for Classification of Histological Pictures, ”Rough Sets, Fuzzy Sets and Knowledge Discovery. Proceedings of the International. Workshop on Rough Sets and Knowledge Discovery ( RSKD''93 ), Banff, Alberta, Canada, Springer-Verlag, Berlin, 1994, 426-433.
[19] Lindahl, D., J. Palmer, M, Ohlsson, C. Peterson, A. Lundin, and L. Edenbrandt, “Automated Interpretation of Myocardial SPECT Perfusion Images Using Artificial Neural Network,” The Journal of Nuclear Medicine, Vol. 38, No.12, 1997, 1870-1875.
[20] Lindahl, D., J. Palmer, J. Pettersson, T. White, A. Lundin, and L. Edenbrandt, “Scintigraphic Diagnosis of Coronary Atery Disease:Myocarddial Bull’s Eye Images Contain the Important Information,” Clinical Physiology, Vol. 18, 1998, 554-561.
[21] Lindahl, D., J. Palmer, and L. Edenbrandt, “Myocardial SPET: Artificial Neural Networks Describe Extent and Severity of Perfusion Defects,” Clinical Physiology, Vol. 19, 1999, 497-503.
[22] Lindahl, D., J. Lanke, A Lundin, J. Palmer, and L. Edenbrandt, “Improved Classification of Myocardial Bull’s-Eye Scintigrams with Computer-Based Decision Support System,” The Journal of Nuclear Medicine, Vol. 40, No.1, 1999, 96-101.
[23] Li, Q.D., Z.X. Chi, and W.B. Shi, “Application of Rough Set Theory and Artificial Neural Network for Load Forecasting,” Proc. of the First International Conference on Machine Learning and Cybernetics, Vol.3, 2002, 1148-1152.
[24] Munakata, T., and Z. Pawlak, “Rough Control Application of Rough Set Theory to Control,” Fourth European Congress on Intelligent Techniques and Soft Computing, Vol.1, Aachen, Germany, 1996, 209-218.
[25] Ohlsson, M., “WeAidU — A Decision Support System Using Artificial Neural Networks,” Department of Theoretical Physics, Lund University, 2002.
[26] Pawlak, Z., “Rough Sets: Theoretical aspects of reasoning about data,” Kluwer Academic Publisher, Dordrecht, Netherlands, 1991.
[27] Qin, Z.G., and Z.Y. Mao, “A New Algorithm for Neural Network Architecture Study,” Proc. of the 3rd World Congress on Intelligent Control and Automation, Vol.2, 2000, 795-799.
[28] Saad, P., S.M. Shamsuddin, S. Deris, and D. Mohamad, “Rough Set on Trademark Images for Neural Network Classifier,” International Journal of Computer Mathematics, Vol. 79, 2002, 789-796.
[29] Sarkar, M. and B. Yegnanarayana, “A Review on Merging Some Resent Techniques with Artificial Neural Network,” IEEE International Conference on Systems, Man, and Cybernetics, Vol.2, 1998, 1824-1829.
[30] Swiniarski, R.W., and L. Hargis, “Rough Sets as a Front End of Neural-Networks Texture Classifiers,” Neurocomputing, Vol.36, 2001, 85-102.
[31] Swiniarski, R.W., and A. Skowron, “Rough Set Methods in Feature Selection and Recognition,” Pattern Recognition Letters, Vol.24, 2003, 833-849.
[32] Vellido, A., P.J.G. Lisboa, and J. Vaughan, “Neural Networks in Business: A Survey of Applications(1992-1998),” Expert Systems with Applications, Vol. 17, 1999, 51-70.
[33] Walczak, B., and D.L. Massart, “Rough Sets Theory,” Chemometrics and Intelligent Laboratory Systems, Vol.47, 1999, 1-16.
[34] Wang G.Y., Z. Zheng, and Y. Zhang, “RIDAS-A Rough Set Based Intelligent Data Analysis System,” Proc. of the First International Conference on Machine Learning and Cybernetics, Vol.2, 2002, 646-649.
[35] Wu, Z., “Research on Remote Sensing Image Classification Using Neural Network Based on Rough Sets,” Proc. ICII 2001 - Beijing. 2001 International Conferences on Info-tech and Info-net, Vol.1, 2001, 279-284.
[36] Yahia, M.E., R. Mahmod, N. Sulaiman, and F. Ahmad, “Rough Neural Expert Systems,” Expert Systems with Applications, Vol. 18, 2000, 87-99.
[37] Yasdi, R., “Combining Rough Sets Learning- and Neural Learning-Method to Deal with Uncertain and Imprecise Information,” Neurocomputing, Vol.7, 1995, 61-84.
[38] Zaremba, M.B., and A. Wielgosz, “Predicting Outcome of Hospitalized Cardiac Patients Using a Combined Neural Network and Rough Set Approach,” Proc. Int. Joint IEEE/INNS Conf. On Neural Networks, Washington, DC, 1999, 3665-3668.
[39] Zhang, G., B. E. Patuwo, and M. Y. Hu, “Forecasting with Artificial Neural Network: The State of the Art,” International Journal of Forecasting, Vol. 14, 1998, 35-62.