Using Data Mining for Identify Patients at High Risk to Hepatocellular Carcinoma in the Cirrhosis Liver: Preliminary Report

Abstract

Background:

Data mining has an interdisciplinary field including various scientific disciplines such as: database systems, statistics, machine learning, artificial intelligence and the others. In the field of medical, data mining algorithms can help physicians to diagnose diseases and chose the best type of treatment. Hepatocellular carcinoma has the most common type of liver cancer. Given the poor prognosis, Hepatocellular carcinoma (HCC) has the fourth leading cause of cancer-related deaths. In this article we aimed to build a decision support system which helps physicians for identify patients at risk to liver cancer.

Materials and Methods:

We analyzed 258 patients with cirrhosis liver. Patients have followed up for four years. We have used decision tree as a data mining tool, for identify patient at high risk to Hepatocellular carcinoma.

Results:

Decision tree determined the importance of attributes such as creatinine, INR and BMI which  could be useful for prediction of cancer. From decision tree model, cirrhosis disease classification rules were extracted and used to improve the prediction of HCC. Decision tree could identify patients at risk to liver cancer with the accuracy of 88% for patients with Sustained virological response (SVR) and the accuracy of 92% for patients with non SVR found.

Conclusions:

According to decision tree results, attributes such as etiology, age, BMI, Platelet, Total Bilirubin, INR, Creatinine, Alfafetoproteina (AFP), and Serum Albumin can predict HCC in patient with cirrhosis. It is suggest that results examine with a greater number of patient.