Artificial Intelligence to Predict Esophageal Varices in Patients with Cirrhosis

Background: Screening for varices remains as the best strategy to decrease associated mortality that reaches 25%. Diagnostic endoscopy is gold standard but invasive routine screening. Non-invasive stiffness measurements with elastography costly and impractical. Non-elastogarphic tests use available laboratory clinical variables are feasible their performance inferior elastography. Non-invasive, accessible accurate test needed. Machine learning methods can be used in this sense provide better diagnostic performances. We aimed ability of a machine model predict esophageal patients cirrhosis. Materials methods: retrospectively evaluated cirrhosis at time screening upper endoscopies from our institutional database. Demographic, clinical, radiologic, endoscopic data was collected. Child-Pugh, APRI, FIB-4, AAR, PCSD were calculated each patient. Gradient boosted algorithm constructed problem. A logistic regression well tests’ model’s performances areas under ROCs compared detect presence varices. Results: Study population consisted 201 whom 105 had esopheageal which 33 higher risk. Patients older, advanced Child stages, larger splenic diameters MELD-Na scores. Composite scores’ follows: FIB-4 0.57 (0.49-0.65), APRI 0.47 (0.38-0.55), 0.511 (0.42-0.59), AAR 0.481 (0.39-0.56). mean AUC 0.68(0.060), F1- score 0.7 accuracy 63%. Conclusions: outperformed non-invasive cirrhotic patients.