論文名稱
中文：整合式的卵巢癌預後評估
英文：Integration of ensemble learning and data mining techniques to predict risk factors for recurrent ovarian cancer
作者：婦產部 曾志仁 副院長
| 本篇論文發表於Artificial Intelligence in Medicine期刊 78 (2017) 47–54全文下載 |
研究目的：
Early detection improves a woman’s chances of surviving ovarian cancer, but an early diagnosis of ovarian cancer is difficult, and to date, there are no standardized screening programs. The literature repeatedly emphasizes that a delay in diagnosis increases mortality rates and recurrence rates. Although there is no definitive diagnostic tool at this time, there is enough information regarding the associated signs and symptoms, risk, and protective factors of ovarian cancer to predict the recurrence of ovarian cancer. This paper examines the existing literature on what variables are associated with the risk factors for the recurrence of ovarian cancer. These allow for a better understanding of which variables are more fundamental to the recurrence of ovarian cancer. The purpose of this study was to determine the risk factors for women with ovarian cancer with regard to recurrence.。
研究結果：
這份研究是導入疾病狀態，運用5種資料探勘的決策模式，分別是：支持向量機（Support Vector Machines）、決策樹C5.0算法、極速學習機（Extreme Learning Machine）、多元適應性雲形迥歸（Multivariate Adaptive Regression Splines）、隨機森林（Random Forest），進行卵巢癌復發的危險因子相關性分析比對。總共分析987例個案，研究顯示年齡、FIGO 分期、病理T分期、病理M分期是影響卵巢癌復發的最重要危險因子；根據實驗結果，C5.0演算法在模擬推估是最符合臨床價值。
研究貢獻與臨床應用：
資料探勘技術運用在臨床是目前最熱門的研究方向，尤其隨著大數據時代的來臨，如何導入最趨近於實際結果的模擬驗證是非常重要的一環，我們未來將持續在數據模擬的領域繼續努力。
