| 摘要: | 研究目的：本研究使用羅吉斯迴歸、支持向量機、決策樹、隨機森林、梯度提升及人工神經網路等機器學習演算法，預測可能造成肝癌復發的危險因子，減少肝癌病人復發的機會。
研究方法：以回溯性世代研究方法，收集2000年至2021年臺北醫學大學臨床研究資料庫中1,201位肝癌病人的相關資料，利用六種機器學習方法將所蒐集之樣本進行隨機分組，建立預測模型，納入人口學變項、生活習慣、過去病史、疾病狀況及治療方式等資?，以準確率、敏感度、特異度、陽性預測值、F1-score及AUC等指標，選擇最準確的演算法，預測可能導致肝癌復發的危險因子。
研究結果：研究結果顯示，1,201位肝癌病人中共有562人發生肝癌復發，復發率為46.79%。在機器學習模型預測肝癌復發危險因子方面，外部驗證以人工神經網路之模型預測結果最佳，其準確率為0.7399、敏感度為0.7857、特異度為0.6905、陽性預測值為0.6688、AUC為0.7399，再根據此模型分析肝癌復發變項重要程度，以喝酒習慣為造成肝癌復發最重要的危險因子，其次為血型，淋巴管或血管侵犯則為危險因子第三名。建議醫療提供者可將本研究機器學習模型的預測結果作為參考，相信有助於提升病人預後效果，為醫病雙方帶來最大的價值，減少肝癌病人復發的機會。
Objective: This study uses machine learning algorithms such as logistic regression, support vector machine, decision tree, random forest, gradient boosting, and artificial neural networks to predict the risk factors for liver cancer recurrence and reduce the chances of relapse in liver cancer patients.
Method: Using a retrospective cohort study design, collected relevant data from 1,201 liver cancer patients in the clinical research database of Taipei Medical University from 2000 to 2021. The collected samples were randomly divided into groups using six machine learning algorithms to build the predictive model. Included data such as demographic variables, lifestyle habits, past medical history, disease condition, and treatment methods. and used indicators like accuracy, sensitivity, specificity, positive predictive value, F1-score, and AUC to select the most accurate algorithm for predicting the risk factors that may lead to liver cancer recurrence.
Results: The research results revealed that out of the 1,201 liver cancer patients, a total of 562 individuals experienced liver cancer recurrence, the recurrence rate is 46.79%. In terms of predicting the risk factors for liver cancer recurrence using machine learning models, the external validation demonstrated that the artificial neural network model had the best predictive performance. It achieved an accuracy of 0.7399, sensitivity of 0.7857, specificity of 0.6905, positive predictive value of 0.6688, and an AUC of 0.7399. Based on this model, the analysis of liver cancer recurrence identified alcohol consumption as the most significant risk factor, followed by blood type, and third risk factor is lymphatic or vascular invasion. It is suggested that healthcare providers can consider the predictive results of this research's machine learning model as a reference, believing that it will contribute to improving patient prognosis and delivering maximum value to both healthcare providers and patients, thereby reducing the chances of liver cancer recurrence. |
