Abstract

Liver cirrhosis poses a significant global health burden, with clinically significant portal hypertension (CSPH), defined by a hepatic venous pressure gradient (HVPG) ≥10 mmHg, being a critical prognostic determinant. While HVPG measurement remains the gold standard, its invasive nature limits widespread clinical application. This study aimed to develop and validate a novel non-invasive virtual HVPG (vHVPG) model by integrating multi-modal data through biofluid mechanics, deep convolutional neural networks (DCNNs), and mixed reality (MR) technologies. This is a prospective, multicenter trial involving 248 patients with cirrhosis scheduled for HVPG measurement. Participants were randomly assigned 1:1 to an original cohort (model development and parameter refinement, n=124) or a validation cohort (independent model assessment, n=124). The vHVPG model was constructed using patient-specific CT, Doppler ultrasound and blood test results. DCNNs and MR technologies were employed for three-dimensional geometric reconstruction and computational fluid dynamics (CFD) analysis was used to solve the Navier-Stokes equations for blood flow simulation. The primary objective was to determine the numerical correlation between vHVPG and invasive HVPG. Secondary objectives included assessing the diagnostic accuracy of vHVPG for portal hypertension (HVPG ≥5 mmHg) and CSPH (HVPG ≥10 mmHg). Study findings will be disseminated through peer-reviewed publications and conference presentations.

Keywords

Portal hypertension, liver cirrhosis, hepatic venous pressure gradient, biofluid mechanics, deep convolutional neural network, mixed reality; noninvasive