Page 292 - ebook HCC

EASL HCC SUMMITHCC SUMMIT
GENEVA, SWITZERLANDA, SWITZERLAND
PROGRAMME
290 PROGRAMME AND ABSTRACTSAND ABSTRACTS GENEV EASL 291
291
290
FEBRUAR
FEBRUARY 13 - 16, 2014Y 13 - 16, 2014
Poster Board Number C60
THE APPLICATION OF DATA MINING Conclusion: Data mining analysis explores data to discover hidden patterns, trends and
TECHNIQUES TO EXPLORE PREDICTORS OF enables the development of models to diagnose HCC utilizing simple laboratory data
as an alternative to liver biopsy avoiding invasive procedures. AFP, cirrhosis, AST, and
HCC BASED ON THE NON-INVASIVE ROUTINE ascites are simple variables that have the prospective to support clinical decisions, without
WORKUP IN EGYPTIAN PATIENTS WITH imposing extra costs for additional examinations.
CHRONIC HEPATITIS What is new: To our knowledge this study has highlighted that a new cutoff value of
AFP≥50.3 ng/ml to diagnose HCC in cirrhotic patients; and that the field of data mining
can be used to solve real health problems that Egypt is currently facing with great success.
Abubakr Awad , Dalia Omran , Mahasen Mabrouk , Ashraf Omar 2
2
2
1
2
1 Computer Science, Faculty of Computers and Information, Cairo University, Endemic
Medicine and Hepatology, Faculty of Medicine, Cairo University, Cairo, Egypt
Corresponding author’s e-mail: bakr.awad@gmail.com

Introduction: Hepatocellular carcinoma (HCC) is the second most common malignancy in
Egypt due to the heavy burden of hepatitis C virus. The stage of hepatocellular carcinoma
(HCC) dictates the therapeutic choice, making early detection a primary objective. These
findings emphasize the need for an innovative, economic, reliable, non-invasive technique
for predicting early HCC diagnosis utilizing simple clinical, and laboratory data. Data
mining is a method of predictive analysis which can explore tremendous volumes of rich
information found in electronic health records to discover hidden patterns and relationships.

Aims: To develop a non-invasive model for early diagnosis of HCC. This model should be
economical, reliable, easy to apply and acceptable by domain experts.

Methodology: This cross sectional study focused on 315 chronic HCV patients (31 chronic
CLINICAL POSTER ABSTRACTS internal validation of 10 folds cross validation for predicting HCC. CLINICAL POSTER ABSTRACTS
hepatitis, 149 cirrhosis, and 135 HCC), between years 2010-2011. Using data mining
analysis, we constructed a C4.5 implementation of decision tree learning algorithms with

Results: Decision tree algorithm was able to diagnose HCC with sensitivity 83.5% and
specificity 83.3% using only routine data. The correctly classified Instances were 263
(83.5%), and the incorrectly classified Instances were 52 (16.5%). Out of 34 attributes, the
decision-tree models showed that Serum level of AFP with an optimal cutoff value of ≥50.3
ng/ml was selected as the best predictor of HCC. To a less extent cirrhosis, AST>64U/L,
and ascites were variables associated with HCC as shown in figure. This was further
confirmed using multivariate logistic regression analysis.