Abstract

Background:

To develop a non-invasive multivariate models based on Three-dimensional MR elastography (3D-MRE) to determine portal hypertension (PH), particularly to diagnose clinically significant portal hypertension (CSPH, HVPG>10mmHg) and severe portal hypertension (SPH, HVPG>12mmHg), using HVPG as the gold standard.

Methods:

This prospective, multicenter study enrolled patients with cirrhosis scheduled for HVPG, and who intended to undergo MR imaging (including 3D-MRE and diffusion-, T1-, and T2-weighted imaging) before the HVPG procedure. A total of 57 patients were recruited from five institutions. Multiple viscoelastic parameters of the liver and spleen, as well as shear stiffness (SS) ratios and subtraction values, were evaluated independently by two radiologists. Univariable and multivariable linear regression analyses were conducted to assess the associations between mechanical parameters and HVPG. Univariable and multivariable logistic regression analyses were used to predict CSPH and SPH, respectively.

Results:

HVPG showed the strongest positive correlation with splenic SS at 60Hz (r = 0.785; P < .001), followed by hepatic SS at 30Hz (r = 0.631; P < .001), splenic SS at 30Hz (r = 0.612; P < .001), hepatic SS at 60Hz (r = 0.547; P < .001). Multivariable linear regression analysis showed splenic SS at 60Hz (β = 1.017; 95% CI: 0.71, 1.325; P < .001), liver SS ratio (β =-2.359; 95% CI: -4.477, -0.241; P = .03), and liver stiffness difference (β =1.169; 95% CI:0.251,2.087; P = .001) was the independent factors determining HVPG. Approximately 56.5% of the total variability in HVPG was explained by these 3 variables (adjusted R 2 = 0.565). Logistic regression analysis showed splenic SS at 60Hz (OR = 2.217 [95% CI:1.292, 3.803]; P = 0.004) and liver SS at 30Hz (OR =6.211 [95% CI: 1.218, 31.669]; P= 0.028) were independently associated with CSPH, establishing a model with excellent performance in diagnosing CSPH(area under the receiver operating characteristic curve [AUC], 0.975 [95% CI: 0.79, 0.99]). As for SPH, it showed splenic SS at 60Hz (OR = 1.65 [95% CI:1.25,2.16]; P = .001) were independently associated with HVPG, also with better performance (AUC, 0.969 [95% CI: 0.81, 0.98]).

Conclusion:

3D-MRE with dual frequencies is a very promising method in both predicting HVPG and diagnosing the presence CSPH and SPH. Splenic stiffness at 60Hz, combining liver stiffness ratio (60Hz/30Hz) and differences (60Hz-30Hz), was the independent parameters associating HVPG.