# Introduction lthough the rate of mortality from GI bleeding episodes has decreased with improved endoscopic and radiologic techniques together with new pharmacologic therapies, the 13% to 20% mortality implies the clinical importance, whatever most of mortalities due vaiceal rupture or leakage [1]. Concern has been raised about the prediction of Esophageal varices for those expected to develop gastrointestinal bleeding in the future, analyzing data by Data mining programs discovered the significant important leading mortality factors; risky Esophageal varices [2]. Thus the major challenge of biomedical data mining over the next 5-10 years is to make these systems useful to biomedical researchers [3]. This paper discuss the utility usefulness of data mining as a predictor statistical tool used to predict esophageal variceal degrees in cirrhotic patients, through measuring the intra-abdominal esophageal wall thickness, by non invasive ultrasound technique. # Authors : Tropical Medicine Department, (Medical Informatics) Al-Azhar Faculty of Medicine-Egypt. E-mail : ahmadrazek@gmail.com II. [4]. # Clinical View Helicobacter pylori infection, Non Steroidal antiinflammatory drugs (NSAIDs), physiologic stress, and excess gastric acid are major risk factors for bleeding peptic ulcers. Reduction or elimination of these risk factors reduces ulcer recurrence and rebleeding rates, whatever Esophagogastric varices develop as a consequence of portal hypertension Less common causes of upper gastrointestinal bleeding include Hemobilia, Hemosuccus pancreaticus, Aortoenteric fistula and Cameron lesions. [5,6]. # Prediction Analysis of Esophageal Variceal The comparative results accuracy of both two studies Esophageal varices are swollen blood vessels in the esophagus, which is the tube that connects the mouth to the stomach. Esophageal varices often happen in people with serious liver disease, called "cirrhosis." As a result of portal hypertension [Figure1], Cirrhosis is a medical term, usually used to describe a diseased liver that has been severely scarred injury. There are numerous causes of liver cirrhosis, the two most common causes of cirrhosis worldwide are alcoholic liver disease and viral hepatitis C, which together accounted for approximately one-half of patients on the liver transplantation wait list worldwide whatever portal hypertension can also be present in the absence of cirrhosis, a condition referred to as "noncirrhotic portal hypertension". The causes of noncirrhotic portal hypertension can be divided into prehepatic, intrahepatic and post hepatic (presinusoidal, sinusoidal and post sinusoidal causes). Early endoscopy (within 24 hours) is recommended for most patients with acute UGI bleeding, though whether early endoscopy affects outcomes and resource utilization is unsettled [7]. # d) Non-Invasive Diagnoses The identification of Cirrhotic patients with esophageal varices using non-invasive means has been attempted using different clinical, laboratory and radiological approaches, unfortunately many of these approaches still controversial ,recently non-invasive 2 D U/S can detect esophageal variceal degrees in cirrhotic patients with high accuracy, through measuring the intra-abdominal wall thickness of the esophagus, there were proportional relations between esophageal wall thicknesses detected by ultrasound and the esophageal variceal degrees diagnosed by Upper Endoscopy. # III. # Analysis by Data Mining The basic classification is based on supervised algorithm, whatever algorithms are applicable for the input data, the accuracy of each algorithm could be changed according to data nature. In clinical medicine, there are different laboratory, clinical and radiological factors determine the progress of each disease. Identification of important and risky factors is of great importance to predict out come in each disease stage. Analyzing of our data by data mining computed program shed light on the significant important factors for each disease condition. Thus the major challenge of biomedical data mining is to make these systems useful to biomedical researchers. The decision tree analysis was performed using Intelligent Miner software (Rapid Miner, Berlin, ver.4.6, Germany), which can automatically search a data set to find the optimal classification variables leading to the building of a decision tree algorithm . Briefly, all items derived from the patients were evaluated to determine which variables and cutoff points might produce the most significant division into two subgroups; group with risky esophageal varices when esophageal wall thickness >6.5 mm and not risky group when esophageal wall thickness IV. # Validation Accuracy In our previous study the algorithm was selected by evaluating each supervised machine learning algorithms by using supervised learning assessment (10-folds-cross validation) on the training set, we chose the best test applicable to our clinical data, accordingly we used Naïve-base test, and the overall accuracy we obtained was 95%. We followed 59 patients presented with portal hypertension as a result of End stage-liver cirrhosis; the thicknesses of esophageal walls were measured using 2D U/S. All patients underwent diagnostic Esophagogastrodudonoscopy to estimate corresponding degrees of varices. According to the decision tree algorithm we obtained from the previous study, the esophageal wall thickness > 4.2 mm with inner wall irregularities, should be Esophageal varices with variable degrees, esophageal wall thickness <4.2mm measured by U/S without inner wall irregularities, V. # Results Categorical data were compared using the 2 test, where as continuous variables were compared using Student's t test. The overall comparative validation accuracy was 97.9 %. Table [ # Conclusion Based on the available evidence, prediction identification of esophageal variceal degrees in cirrhotic patients presented with manifestations of portal hypertension using non-invasive 2D ultrasound could be a very helpful tool saving time and money. Data mining shed light on the most significant predictors-related esophageal wall thicknesses in 673 patients with very high accuracy; identification high risky groups needed urgent interventional Endoscopy; less complication, better out come and decrease mortality. Data mining would be the coming statistical evolution in clinical medical data statistical analysis with reasonable limitation. # VII. # Limitation of the Study Given the small sample size of 59 patients (8.7%) in comparable to 673 reported in the pervious study, the validation accuracy might be changed if we apply more patients in the future, whatever our clinical experience played a major role in assessing the information mentioned above, in our point of view our results should be confirmed with more evidence-based criteria using independent laboratory and clinical factors all together. 1![Figure 1 : Showing the esophageal varices interlacing the Esophagus and Upper part of the stomach, these varices can burst and cause internal bleeding, leading to death in many institutions. With a permission from UpToDate®; Graphic 63611, Version 4.0 c) Diagnostic Techniques Upper gastrointestinal endoscopy remains the gold standard for the diagnosis of esophagogastroduodenal lesions, despite its limitations.Early endoscopy (within 24 hours) is recommended for most patients with acute UGI bleeding, though whether early endoscopy affects outcomes and resource utilization is unsettled[7].](image-2.png "Figure 1 :") ![2013 Global Journals Inc. (US) Global Journal of Computer Science and Technology Volume XIII Issue X Version I](image-3.png "©") 1Esophageal wall ThicknessCasesResultComparative< 4.2 ; No varices66100%>4.2 mm*; Varices5353100%>6.5 mm; Risky varices484593.75%Over All Accuracy97.9%VI. * Reductions in 28-day mortality following hospital admission for upper gastrointestinal hemorrhage CCrooks TCard JWest Gastroenterology 141 62 2011 * Detection of Risky Esophageal varices by 2D U/S: when to perform Endoscopy AbdEl Razek Am J Med Sci 2012 Dec 21 Epub ahead of print * A survey of current work in biomedical text mining AMCohen WRHersh Briefings in Bioinformatics 6 2005 * A nationwide study of mortality associated with hospital admission due to severe gastrointestinal events and those associated with nonsteroidal antiinflammatory drug use ALanas MAPerez-Aisa FFeu Am J Gastroenterol 100 1685 2005 * Association of aspirin use with major bleeding in patients with and without diabetes DeBerardis GLucisano G D'ettorre A JAMA 307 2286 2012 * A prospective characterization of upper gastrointestinal hemorrhage presenting with hematochezia CMWilcox LNAlexander GCotsonis Am J Gastroenterol 92 231 1997 * The hematocrit level in upper gastrointestinal hemorrhage: safety of endoscopy and outcomes VBalderas RBhore LFLara Am J Med 124 970 2011 * Self-organizing maps TKohonen 2001 Springer Berlin * PVC discrimination using the QRS power spectrum and self-organizing maps MLTalbi ACharef 10.1016/j.cmpb.2008.12.009 Comput Methods Programs Biomed 94 2009 * Community health assessment using self-organizing maps and geographic information systems HGBasara MYuan 10.1186/1476-072X-7-67 Int. J Health Geogr 30 7 67 2008 * Algorithm to determine the outcome of patients with acute liver failure: a data-mining analysis using decision trees NobuakiNakayama MakotoOketani YoshihiroKawamura MieInao SumikoNagoshi KenjiFujiwara HirohitoTsubouchi SatoshiMochida J Gastroenterol 47 6 2012 June * The hematocrit level in upper gastrointestinal hemorrhage: safety of endoscopy and outcomes VBalderas RBhore LFLara Am J Med 124 970 2011 * Patients whose first episode of bleeding occurs while taking a ?-blocker have high long-term risks of rebleeding and death ARDe Souza LaMura VReverter E Clin Gastroenterol Hepatol 10 670 2012 * Underestimation of liver-related mortality in the United States SKAsrani JJLarson BYawn Gastroenterology 145 375 2013 * Betablockers to prevent gastroesophageal varices in patients with cirrhosis RJGroszmann GGarcia-Tsao JBosch N Engl J Med 353 2254 2005 * Diagnosis and treatment of severe hematochezia. The role of urgent colonoscopy after purge DMJensen GAMachicado Gastroenterology 95 1569 1988 * An objective measure of stool color for differentiating upper from lower gastrointestinal bleeding GRZuckerman DRTrellis TMSherman REClouse Dig Dis Sci 40 1614 1995 * Helicobacter pylori infection and gastric lymphoma JParsonnet SHansen LRodriguez N Engl J Med 330 1267 1994 * pylori infection: its role in chronic gastritis, carcinoma and peptic ulcer JmHPajares Hepatogastroenterology 42 827 1995 * Helicobacter pylori infection in patients with gastric carcinoma in biopsy and surgical resection specimens TShibata IImoto YOhuchi Cancer 77 1044 1996 * 10 Challenging Problems in Data Mining Research QYang XindongWu International Journal of Information Technology & Decision Making 5 4 2006 * From Data towards Knowledge: Revealing the Architecture of Signaling Systems by Unifying Knowledge Mining and Data Mining of Systematic Perturbation Data SLu JinBCowart LALu X Plos One 8 4 e61134 2013 Apr 23 * Predictive data mining in clinical medicine: current issues and guidelines RBellazzi BZupan 10.1016/j.ijmedinf.2006.11.006 Int J Med Inform 77 2008 * Does this patient with liver disease have cirrhosis? JAUdell CSWang JTinmouth JAMA 307 832 2012 * An update on treatment of genotype 1 chronic hepatitis C virus infection: 2011 practice guideline by the MGGhany DRNelson DBStrader Hepatology 54 1433 2011 American Association for the Study of Liver Diseases * US Burden of Disease Collaborators. The state of US health, 1990-2010: burden of diseases, injuries, and risk factors JAMA 310 591 2013 * Prevalence and mechanisms of malnutrition in patients with advanced liver disease, and nutrition management strategies KCheung SSLee MRaman Clin Gastroenterol Hepatol 10 117 2012 * Endoscopic management of Dieulafoy's lesion using Isoamyl-2-cyanoacrylate 23951399 World Journal of Gastrointestinal Endoscopy 5 8 2013 Aug 16