臺灣博碩士論文加值系統

propofol是一種廣泛使用於麻醉誘導的安眠藥物，醫院中其它的加護病房單位也經常使用propofol來安眠病人。Propofol最大的優點在於它在停藥後病人可以在很快的時間恢復意識，另外它的止吐效果也可以讓病人在手術後減少噁心嘔吐的機會。但propofol有一個令人詬病的缺點：它很容易導致病人發生低血壓的副作用。Propofol 導致低血壓的原因是來自於對心肌收縮力的抑制以及周邊血管阻力的下降。如果對propofol所造成的低血壓沒有作迅速有效的處理，很可能會體內重要的生命器官例如心臟、大腦、腎臟等造成傷害。
我們希望在這研究中建立一個能預測血壓變化的模型，能在使用propofol作麻醉誘導時給麻醉醫師一個有用的決策參考指標。我們使用17種可以從病人身上(例如年齡、性別，病人過去的病史─如糖尿病、氣喘，血壓及血紅素等)得到的參數值來建立這模型。這訓練資料庫中包含著200個以propofol作麻醉誘導的病人。另外為了評估這些模型的預測能力，我們又收集了同樣100個以propofol作麻醉誘導的病人的資料，用建立的模型實際去預測病人血壓的變化。最後我們以area under ROC curve來當作模型預測能力的指標。
我們一共建構了兩組不同結構的模型，第一種模型使用logistic regression的方式。最後一共建構了兩個logistic regression模型，第一個使用了所有的17個參數來預測血壓的變化，第二個則僅使用兩個參數來預測血壓的變化。另一組預測模型其架構是使用類神經網路系統，我們一共建構了五種不同型態的類神經網路。最後為了評估這些預測模型的能力，我們找了三位經驗豐富的麻醉專科醫師，四位麻醉住院醫師，以及15位麻醉技師對同樣的100個病人預測其血壓的變化，以便和我們的模型作比較。
最後我們發現兩個logistic regression models以及5個類神經網路模型和三位麻醉專家比較起來都沒有明顯的統計差異，但我們的預測模型其正確率卻可以明顯的超過四位住院醫師以及15位麻醉技師

Propofol is a popular hypnotic agent used in induction or maintain of anesthesia. Other intensive care units in the hospital also use it as a sedative drug. The most attractive feature of propofol is rapid recovery of patient’s conscious level while terminating the infusion of propofol. Moreover, its antiemetic effect seems attractive to many anesthesia staffs to avoid post operative nausea and vomiting. Unfortunately, propofol can produce hypotension more often than other anesthesia induction agent. The hypotensive effect of propofol comes from direct depress of cardiac muscle and vasodilatation of peripheral vessels. If not treated promptly and properly, hypotension may induce sever damage to vital organs such as kidney, heart and brain.
We want to setup reliable predicting models to forecast the blood pressure change caused by injection of propofol during induction of anesthesia. Seventeen values (including demographic data such as age and gender, patient past histories such as diabetes mellitus and asthma, and laboratory data such as hemoglobin level and blood pressure before induction) from 200 patients who received propofol as their induction agent in a routine operation were collected in about one year. Another data set from 100 patients, for evaluating the performance of the predicting models was collected in the same period by the same induction procedure. Area under ROC (Receiver Operating Characteristic) curve was used as an index tools to evaluate the performance of our predicting models.
Two types of prediction models were built up in our study. The first type is binary logistic regression model. We have make up two logistic regression model using different input variables. The first logistic model contained all the seventeen input variables, and the second logistic regression model had only two input variables. Another type of predicting model used artificial neural network to predict the blood pressure change. We have finally constructed five different neural network models with dissimilar training protocol and network topography. To compare the predicting models with human beings, three anesthesia attending doctors (experts of anesthesia), four anesthesia resident doctors and fifteen anesthesia nurse with different clinical experiences were also enrolled in this study. Their discrimination abilities of blood pressure change caused by injection of propofol in the evaluation group were compared with our models.
Finally we have found that the two logistic regression models and five artificial neural network models had the same predicting abilities and were all superior to the three anesthesia experts, although statistical significance did exist between them. But the abilities of our predicting models surpassed the anesthesia resident doctors and the fifteen anesthesia nurses and the statistic significance was also found. The predicting modes can be easily integrated in the hospital information system and can act as a reliable decision supporting system

碩博士論文授權書 ii
誌　謝 v
目　錄 vi
表目錄 xi
圖目錄 xiii
中文摘要 xvii
Abstract xix
Chapter 1 1
Introduction 1
1.1Risk of anesthesia 1
1.2 Introduction of propofol 2
1.3The management of the operation room and goodness of propofol 2
1.4 Introduction of logistic regression 3
1.5 Introduction to artificial neural network. 3
1.6 Area under ROC curve 5
1.7 Purpose of our study. 5
Chapter 2 7
Literature review 7
2.1 Complications propofol 7
2.2 Application of logistic regression and artificial neural network in anesthesia field 8
Chapter 3 9
Methods 9
3.1Data collection 9
3.2 The input and out variables 10
3.2.1 Input variables with categorical nature 11
3.2.2 Input variables with continuous nature 13
3.2.3 Variables derived from other variables 14
3.2.4 The data collecting form 14
3.3 Separating the patient’s database into the training data set and validation data set 16
3.4 The predicting models 17
3.4.1 Logistic regression model 17
3.4.2 Artificial neural network models 18
3.4.3 The human experts 18
3.4.4 The junior anesthesia staffs 19
Chapter 4 20
Results 20
4.1 The histogram of blood pressure change 20
4.2 Comparison of the two data sets 21
4.2 The construction of logistic regression model 22
4.2.1 The 17 variables model 22
4.2.2 The two variables model 26
4.3 The validation of logistic regression model 27
4.3.1 The validation of 17 variables logistic regression model 28
4.3.2 The validation of two variables logistic regression model 29
4.4 Creation of seventeen variables artificial neural network model 30
4.4.1 Creation of the 17 input variables artificial neural network model 30
4.4.2 The three input variables linear artificial neural network model 37
4.4.3 The two input variables linear artificial neural network model 39
4.4.4 The five input variables MLP model 41
4.4.5 The eleven input variables RBF model 43
4.5 Validation the performance of artificial neural network models 46
4.5.1 The seventeen input variables MLP artificial neural network model 46
4.5.2 The three input variables linear artificial neural network model 47
4.5.3The two input variables linear artificial neural network model 48
4.5.4 The five input variables MLP model 49
4.5.5 The eleven input variables RBF model 51
4.6 The human experts 52
4.6.1 Human expert number one 53
4.6.2 Human expert number two 54
4.6.3 Human expert number three 55
4.7 The predicting ability of anesthesia resident doctors 57
4.8 Predicting abilities of anesthesia nurses in validation data set 58
4.9 Comparison of all models 59
4.9.1 Comparison between 17 input variables logistic regression model and two input variables model. 59
4.9.2 Comparison between 17 input variables logistic model and 17 input MLP artificial neural network model. 60
4.9.3 Comparison between 17 input variables logistic model and 3 input variables linear artificial neural network model. 61
4.9.4 Comparison between 17 input variables logistic model and two input variables linear artificial neural network model. 62
4.9.5 Comparison between 17 input variables logistic model and five input variables MLP neural network model. 63
4.9.6 Comparison between 17 input variables logistic model and eleven input variables RBF neural network model. 64
4.9.7 Comparison between 17 input variables logistic model and human expert number one. 64
4.9.8 Comparison between 17 input variables logistic model and human expert number two. 65
4.9.9 Comparison between 17 input variables logistic model and human expert number three. 66
4.9.10 Comparison of the best predicting model with resident doctors 66
4.9.11 Comparison of best predicting model with anesthesia nurses 67
4.10. Summary of performance of all the predicting models in training data set and validation data set 68
Chapter 5 71
Discussion 71
5.1 Detection of blood pressure change during induction of anesthesia with propofol 71
5.2 The factors most important to blood pressure change caused by propofol 72
5.3 The performances of our predicting models. 73
5.4 Comparisons of artificial neural networks and logistic regression 75
5.4 Disadvantages of using neural networks 76
Chapter 6 78
Conclusion 78
6.1 The limitations 78
6.2 The future 79
Reference List 80

(1) Patrick MR, Blair IJ, Feneck RO, Sebel PS. A comparison of the haemodynamic effects of propofol (''Diprivan'') and thiopentone in patients with coronary artery disease. Postgrad Med J 1985; 61 Suppl 3:23-27.
(2) Vermeyen KM, Erpels FA, Janssen LA, Beeckman CP, Hanegreefs GH. Propofol-fentanyl anaesthesia for coronary bypass surgery in patients with good left ventricular function. Br J Anaesth 1987; 59(9):1115-1120.
(3) Stephan H, Sonntag H, Schenk HD, Kettler D, Khambatta HJ. Effects of propofol on cardiovascular dynamics, myocardial blood flow and myocardial metabolism in patients with coronary artery disease. Br J Anaesth 1986; 58(9):969-975.
(4) Al Khudhairi D, Gordon G, Morgan M, Whitwam JG. Acute cardiovascular changes following disoprofol. Effects in heavily sedated patients with coronary artery disease. Anaesthesia 1982; 37(10):1007-1010.
(5) Monk CR, Coates DP, Prys-Roberts C, Turtle MJ, Spelina K. Haemodynamic effects of a prolonged infusion of propofol as a supplement to nitrous oxide anaesthesia. Studies in association with peripheral arterial surgery. Br J Anaesth 1987; 59(8):954-960.
(6) Claeys MA, Gepts E, Camu F. Haemodynamic changes during anaesthesia induced and maintained with propofol. Br J Anaesth 1988; 60(1):3-9.
(7) Coates DP, Prys-Roberts C, Spelina KR, Monk CR, Norley I. Propofol (''Diprivan'') by intravenous infusion with nitrous oxide: dose requirements and haemodynamic effects. Postgrad Med J 1985; 61 Suppl 3:76-79.
(8) Coates DP, Monk CR, Prys-Roberts C, Turtle M. Hemodynamic effects of infusions of the emulsion formulation of propofol during nitrous oxide anesthesia in humans. Anesth Analg 1987; 66(1):64-70.
(9) Grounds RM, Twigley AJ, Carli F, Whitwam JG, Morgan M. The haemodynamic effects of intravenous induction. Comparison of the effects of thiopentone and propofol. Anaesthesia 1985; 40(8):735-740.
(10) M.Palazoo RE. Logistic regression analysis of fixed patient factors for postoperative sickness: A model for risk assessment. British Journal of Anaesthesia 1993; 70:135-140.
(11) Toner CC, Broomhead CJ, Littlejohn IH, Samra GS, Powney JG, Palazzo MG et al. Prediction of postoperative nausea and vomiting using a logistic regression model. Br J Anaesth 1996; 76(3):347-351.
(12) Vefghi L, Linkens DA. Internal representation in neural networks used for classification of patient anaesthetic states and dosage. Comput Methods Programs Biomed 1999; 59(2):75-89.
(13) Leon MA, Lorini FL. Ventilation mode recognition using artificial neural networks. Comput Biomed Res 1997; 30(5):373-378.
(14) Kim WO, Kil HK, Kang JW, Park HR. Prediction on lengths of stay in the postanesthesia care unit following general anesthesia: preliminary study of the neural network and logistic regression modelling. J Korean Med Sci 2000; 15(1):25-30.
(15) Sharma A, Roy RJ. Design of a recognition system to predict movement during anesthesia. IEEE Trans Biomed Eng 1997; 44(6):505-511.
(16) Hug CC, Jr., McLeskey CH, Nahrwold ML, Roizen MF, Stanley TH, Thisted RA et al. Hemodynamic effects of propofol: data from over 25,000 patients. Anesth Analg 1993; 77(4 Suppl):S21-S29.