以類神經網路及分類迴歸樹輔助肝癌病患預測存活情形

Prediction of Survival in Patients with Liver Cancer Using Artificial Neural Networks and Classification and Regression Trees

陳正美(Cheng-Mei Chen) ； 徐建業(Chien-Yeh Hsu) ； 邱泓文(Hung-Wen Chiu) ； 白其卉(Chyi-Huey Bai) ； 吳柏動(Po-Husn Wu)

《台灣公共衛生雜誌》 30卷5期 (2011/10) Pp. 481-493

https://doi.org/10.6288/TJPH2011-30-05-08

摘要

This study created a survival prediction model for liver cancer using data mining algorithms. Methods: The data were collected from the cancer registry of a medical center in Northern Taiwan between 2004 and 2008. A total of 227 patients were newly diagnosed with liver cancer during this time. Following a literature review, expert consultation, and collection of patients' data, nine variables pertaining to liver cancer survival rates were analyzed using t-tests and chi-square tests. Six variables were significant. An artificial neural network (ANN) and a classification and regression tree (CART) algorithm were adopted as prediction models. The models were tested in three conditions: one variable (clinical stage alone), six significant variables, and all nine variables (significant and non-significant). Five-year survival was the output prediction. Results: The ANN model with nine input variables was a superior predictor of survival (p＜0.001). The area under the receiver operating characteristic (ROC) was 0.843, and 0.78, 0.76, and 0.80 for accuracy, sensitivity, and specificity respectively. Conclusions: An artificial neural network was more accurate than a CART system in predicting liver cancer survival. In the future, we suggest developing a computer system using the nine input variables in the ANN prediction model to predict liver cancer survival. The system would use an ANN algorithm to automatically calculate the prediction result and assist patients in understanding their potential treatment outcomes and survival.

關鍵字

肝癌 ； 類神經網路 ； 分類迴歸樹 ； 預測模型

並列摘要

This study created a survival prediction model for liver cancer using data mining algorithms. Methods: The data were collected from the cancer registry of a medical center in Northern Taiwan between 2004 and 2008. A total of 227 patients were newly diagnosed with liver cancer during this time. Following a literature review, expert consultation, and collection of patients' data, nine variables pertaining to liver cancer survival rates were analyzed using t-tests and chi-square tests. Six variables were significant. An artificial neural network (ANN) and a classification and regression tree (CART) algorithm were adopted as prediction models. The models were tested in three conditions: one variable (clinical stage alone), six significant variables, and all nine variables (significant and non-significant). Five-year survival was the output prediction. Results: The ANN model with nine input variables was a superior predictor of survival (p＜0.001). The area under the receiver operating characteristic (ROC) was 0.843, and 0.78, 0.76, and 0.80 for accuracy, sensitivity, and specificity respectively. Conclusions: An artificial neural network was more accurate than a CART system in predicting liver cancer survival. In the future, we suggest developing a computer system using the nine input variables in the ANN prediction model to predict liver cancer survival. The system would use an ANN algorithm to automatically calculate the prediction result and assist patients in understanding their potential treatment outcomes and survival.

並列關鍵字

Liver Cancer ； Artificial Neural Networks ； Classification and Regression Trees ； Prediction Model

參考文獻

林志陵、高嘉宏(2008)。肝癌的流行病學。中華癌醫會誌。24，277-81。

Lin, T. M.,Chen, C. J.,Tsai, S. F.,Tsai, T. H.(1988).Hepatoma in Taiwan.J Natl Public Health Assoc(ROC).8,91-100.

Okuda, K.,Ohtsuki, T.,Obata, H.(1985).Natural history of hepatocellular carcinoma and prognosis in relation to treatment. Study of 850 patients..Cancer.56,918-28.

徐千彝、霍德義(2010)。肝癌的預後評估及分期系統。臨床醫學。66，216-24。

Lau, H.,Fan, S. T.,Ng, I. O.,Wong, J.(1998).Long term prognosis after hepatectomy for hepatocellular carcinoma: a survival analysis of 204 consecutive patients.Cancer.83,2302-11.