A Novel Imaging Based Quantitative Model-aided Detection of Portal Hypertension in Patients With Cirrhosis (CHESS2104): A Prospective, Multicenter Study

China suffers the heaviest burden of liver disease in the world. The number of chronic liver disease is more than 400 million. Either viral-related hepatitis, alcoholic hepatitis, or metabolic-related fatty hepatitis, etc. may progress to cirrhosis, which greatly threatens public health. Portal hypertension is a critical risk factor that correlates with clinical prognosis of patients with cirrhosis. According to the Consensus on clinical application of hepatic venous pressure gradient in China (2018), hepatic venous pressure gradient (HVPG) greater than 10,12,16,20 mmHg correspondingly predicts different outcomes of patients with cirrhosis portal hypertension. It is of great significance to establish a risk stratification system and perform tailored management for portal hypertension in cirrhosis. As a universal gold standard for diagnosing and monitoring portal hypertension, HVPG remains limitation for clinical application due to its invasiveness. How to construct a novel, non-invasive, accurate, and convenient method to achieve prediction of HVPG is an important general problem in the management of portal hypertension in cirrhosis.

The development of radiomics technique provides an approach to solve abovementioned clinical issues. Based on artificial intelligence algorithms, radiomics harnesses mineable, high-resolution, and quantitative features from encrypted medical images, along with clinical or genetic data to produce evidence-based decision support system, to achieve the clinical targets including diagnosis, treatment effect evaluation, and prognosis prediction. In this project, aiming at development of a risk stratification system for hypertension management in cirrhosis, we will construct a standard-of-care database and utilize radiomics tool to construct the decision making system. We will take responsibility for achievement of organ and vessel segmentation, radiomic feature selection, and signature construction for prediction of hypertension classification, and accomplish the development of prototype system which would integrate four modules including database management, HVPG risk stratification application module, predicted outcome presentation module, and prognostic information curation module. This project will focus on two aspects which are correspondingly machine learning algorithms optimization and prototype system development, so as to promote the precision medicine in liver disease.