乳癌手術病患術後焦慮與憂鬱預測模式之探討

研究背景與目的乳癌是全球女性中最常見的癌症，早期乳癌患者中有近50％在確診後一年內有憂鬱、焦慮或兩者兼有的情形，但焦慮、憂慮症合併其它生理疾病時，將其提高死亡率的發生。因此，本研究將藉由以下三種資料探勘技術，類神經網路(ANN)、支持向量機(SVM)與線性複迴歸(Multi Linear Regression, MLR)預測模型，找出乳癌手術病患術後焦慮和憂鬱之最佳的重要影響因子。研究方法本研究採用前瞻性研究設計，選取台灣南部三家醫學中心之乳癌病患，分別在病患術前、術後一年、術後兩年與術後五年進行追蹤，收案期間為2007年6月至2014年12月。使用之問卷為乳癌手術病人、一般健康相關生活品質問卷(SF-36)、貝克焦慮量表(BAI)、貝克憂鬱量表第二版(BDI)，共五種問卷。總樣本數357位，藉著ANN、SVM與MLR 預測模型，並運用殘差分析(Residual Analysis)及全域敏感度分析(Global Sensitivity Analysis)，進行預測模式準確性探討，並找出影響乳癌手術病患術後焦慮和憂鬱的重要預測因子。研究結果研究結果顯示ANN預測模式優於SVM與MLR之預測模式。乳癌手術病患術後焦慮和憂鬱之預測，在焦慮(BAI量表)殘差分析的部份，ANN、SVM、MLR分別為0.15、0.17與0.32；在憂慮(BDI量表) 殘差分析的部份，則分別為0.20、0.23與0.40。整體而言，ANN表現優於SVM與MLR預測模式。結論與建議本研究結果顯示類神經網路(ANN)預測模式優於支持向量機(SVM)、線性複迴歸(MLR)預測模式。故此預測模式能作為日後臨床決策之參考，找出最佳重要預測因子，於最佳時間點介入，欲能提早做好預防乳癌手術病患術後焦慮和憂鬱之發生，顧及患者身心狀態與家屬的照顧品質。

關鍵字

乳癌；類神經網路；支持向量機；資料探勘；焦慮；憂鬱

並列摘要

Abstract Background and purpose of research Breast cancer is the most common form of cancer among women, and nearly 50% of early breast cancer patients have either depression or anxiety. Anxiety and depression are associated with other physiological conditions to increase mortality. Therefore, this study will identify the best important factors for postoperative anxiety and depression in patients after breast cancer surgery by following three data mining techniques, such as Artificial Neural Network (ANN), Support Vector Machine (SVM) and Multi Linear Regression (MLR) predictive models. Research methods Research methods This study adopted a prospective study design, selected three medical centers in southern Taiwan, breast cancer patients, respectively, in patients before surgery, one year, two years and five years after surgery, between June 2007 to December 2014. A total of five questionnaires were used for patients with breast cancer surgery, general health-related quality of Life Questionnaire (SF-36), Baker Anxiety Scale (BAI) and Baker's Depression Scale (BDI). The total sample number is 357 bits, by using ANN, SVM and MLR predictive model, the accuracy of prediction model is discussed by residual analyses and global sensitivity analysis in order to find out the important predictors of postoperative anxiety and depression in patients after breast cancer surgery. Research results The results show that ANN predictive model is better than that of SVM and MLR. Prediction of postoperative anxiety and depression in patients with breast cancer surgery, in the residual analysis of anxiety (BAI scale), ANN, SVM and MLR were 0.15, 0.17 and 0.32 respectively, 0.20,0.23 and 0.40 in the residual analysis of worry (BDI scale). Overall, ANN performs better than SVM and MLR models. Conclusions and recommendations The results show that ANN predictive model is better than SVM and MLR Prediction model. Therefore, predictive model can be used as a reference for future clinical decision making, to find out the best important predictor, to intervene in the best time, to prevent postoperative anxiety and depression of patients with breast cancer surgery, and to take care of the patient's physical and mental state and the quality of the family.