資料探勘應用於乳癌治療評估

Abstract

在現今的乳癌治療中，以手術摘除乳房腫瘤為主。摘除腫瘤後，為避免癌症復發，會施以其他後續治療，如化學治療(chemotherapy)、放射治療(radiation therapy)以及賀爾蒙治療(hormonal replacement therapy)，不同的治療方式對患者的狀況會造成不同的影響。而相同的治療方式對不同的患者也會有不同的結果。故難以確切的知道何種治療方式對患者的預後狀況有最好的影響。因此，本研究利用資料探勘演算法中的決策樹演算法分析乳癌患者的身體檢查資料建立乳癌治療方式評估模型。利用本研究所建立的模型，能預測出乳癌患者在不同治療方式下的未來十年的存活曲線及不復發曲線，幫助醫生評估治療方式以及協助了解患者未來狀況。 本研究建立的乳癌治療方式評估模型也可用以探討影響治療效益的特徵。透過結合先驗演算法及平均值分析的方式對探勘出來的結果進行延伸應用，能從資料中找出可能會影響乳癌狀況的生理特徵。由此，醫生可在本乳癌治療方式評估模型的幫助下決定該對已進行手術的乳癌患者採用何種後續治療，提升患者的未來存活機率。

Nowadays, the most effective way to cure breast cancer is to remove the tumor in patients’ breasts. After the surgery, patients must take other treatments to prevent the relapse of the cancer, such as chemotherapy, radiation therapy, and hormonal replacement therapy. Different treatments will have different influence on a patient’s prognosis. Besides, a treatment might have different influence on different patients. As a result, it is difficult to determine what the best treatment for a patient is. Therefore, this study uses decision tree algorithm, which is one of the most popular data mining algorithms, to analyze physiological features of the patients, and then establish a prediction model for breast cancer. With this model, we can predict the survival curve and no recurrence curve in the next ten years of the breast cancer patients with different treatments. The model can help doctors to evaluate which treatment is the best treatment for a patient. The model established in this study can also be used to explore which physiological feature has influence on the treatment. By using Apriori algorithm and feature average analysis, we can find out the physiological features which might influence the patient’s condition. As a result, doctors can use the model from this study to decide which treatment is the best for a patient.