| Abstrakt: |
Cholangiocarcinoma (CCA) is a malignant neoplasm originating from biliary epithelial cells. During the early stage, the patients do not show any symptoms, leading to wide and extensive spread of this disease. Nowadays, there has not been a single serum tumor marker which can be used for effective screening of the disease or classification of the patients. This study therefore aims to determine an appropriate serum marker for screening of the patients with early staged CCA by using a technique of data mining. Beginning with the C4.5 Decision tree and Logistic Regression for selection of serum markers for effective screening of the patients with CCA, the selected markers were then used for classification of the patients with CCA from non-CCA patients, and CCA from Benign Biliary Disease (BBD) by C4.5 Decision tree, Logistic Regression, Random Forest, and Artificial Neural Network. In this work, seven serum tumor markers were used, including Carbohydrate Antigen 125 (CA125), Carbohydrate Antigens 19-9 (CA19-9), Carbohydrate Antigen 242 (CA242), Carbohydrate Antigen 50 (CA50), Carbohydrate Antigen 72-4 (CA72-4), Carcinoembryonic Antigen (CEA), Cy-tokeratin-19Fragment (CYFRA 21-1). The model was used to classify the CCA and non-CCA patients and it was discovered that the serum tumor markers which could most efficiently classify the CCA patients from the non-CCA patients were the combination of CA125 and CA242 suggested by the Logistic Regression with C4.5 Decision tree as the classifier, yielding the best performance, with Sensitivity and Specificity being 75.88 % and 86.82%, respectively. In contrast, the classification of CCA patients from BBD patients was best performed by the serum tumor markers CA125 and CA72-4 suggested by C4.5 Decision tree with Logistic Regression or Random Forest as the classifier. [ABSTRACT FROM AUTHOR] |
