A non-invasive magnetic resonance imaging-based model predicts portal venous pressure

To establish a non-invasive model for the assessment of portal venous pressure (PVP) based on the magnetic resonance (MR) parameters.

In this prospective study, contrast-enhanced magnetic resonance imaging (MRI) scan was performed in 109 patients indicated for upper abdominal surgeries after their written consents were obtained, and intraoperative PVP measurements were completed in 92 patients. Altogether 17 patients were excluded for not undergoing surgery or unsuccessful catheterization. A linear model was constructed for estimating PVP levels in 56 patients and further validation was conducted in the other 36 patients.

The PVP levels were significantly correlated with MR parameters, including splenic volume (SV), splenic venous diameter (SVD), liver/splenic volume ratio, portal venous diameter, hepatic diameter, portal venous cross-sectional area, ascites, varices and arterial portal shunts. A linear model was established as follows: PVP (mmHg) = 2.529 + 1.572 × SVD (mm) + 0.231 × SV/body mass index (× 10(4) cm(5) /kg) + 3.44 × aspartate aminotransferase-to-platelet ratio index. This model showed excellent accuracy in the detection of portal hypertension, with the area under the receiver operating characteristic curve (AUROC) of 0.945 (95% CI 0.867-1.000), with the sensitivity and specificity of 91.7% and 93.7%, respectively. The agreement analysis revealed that the predictive value using this formula closely reflected the patients' actual PVP level. Moreover, the validation confirmed the accuracy of this model for the assessment of portal hypertension [AUROC 0.935 (95% CI 0.856-1.000)].

The MRI-based formula has great potential for detecting portal hypertension. As a non-invasive measurement, it may be clinically accepted for the replacement of invasive modalities after further refinement.