A Digital Tongue Diagnosis Model for High- and Low-risk Esophagogastroduodenal Varices in Cirrhosis

Plain-language summary

The aim of this observational study is to establish an AI deep learning model that can dianosie high-risk varices for patients with cirrhosis effeciently. The main question of this study is to esplore: question 1: Developing a digital tongue diagnosis model, specifically a deep learning model to diagnose high-risk esophageal and gastric varices (HRV) associated with cirrhosis using sublingual vein images. Answering the question of whether the new tongue diagnosis method can accurately diagnose. Question 2: Compare the diagnostic efficacy digital tongue diagnosis model with diagnostic models constructed using other biochemical indicators for HRV in cirrhosis, and answer the question of "how to use it optimally." Question 3: Exploring the correlation between sublingual vein characteristics and Hepatic venous pressure gradient (HVPG). Question 4: Compared with endoscopic examination results, validate the diagnostic performance of the model (AUC ≥ 0.90) and screen for key parameters of sublingual vein characteristics (such as sublingual vein varicosity diameter, vein length, color, etc.). Question 5: Follow-up tongue examination images of patients with cirrhosis who underwent treatment (e.g., endoscopy, splenic embolization, TIPS, etc.) at 1, 2, and 3 years post-treatment were evaluated to assess the efficacy of digital tongue examination models in predicting high-risk esophageal and gastric variceal bleeding at 1, 2, and 3 years post-treatment, as well as the efficacy in predicting endoscopic treatment failure rates and patient mortality associated with bleeding.

Who can participate

What they're measuring

Trial details