Non-invasive score using non-contrast-enhanced MRI for identification of clinically significant portal hypertension (CHESS1802): a prospective, multicentre study

Hepatic venous pressure gradient (HVPG) of 10 mmHg or greater is the gold standard for the assessment of clinically significant portal hypertension, which is a crucial stage in cirrhosis. However, the measurement of HVPG is invasive and not feasible to all patients with cirrhosis. We aimed to investigate whether the non-invasive score on the basis of non-contrast-enhanced MRI (CHESS-CSPHMRI score) can predict the presence of clinically significant portal hypertension in patients with cirrhosis. In this prospective, multicentre, diagnostic study, patients with cirrhosis who were scheduled to undergo HVPG measurements and non-contrast-enhanced MRI were recruited from eight university hospitals in China between Dec 4, 2019, and April 30, 2019. Participants receiving non-contrast-enhanced MRI and without history of chronic liver diseases were recruited as healthy controls. The calculation of CHESS-CSPHMRI score included: (1) region of interest segmentation of the left frontal lobe of the liver on T1-weighted MRI maps; (2) liver tissues identification by histogram analysis; (3) detection of the liver edge; (4) liver surface matching using a least squares method; (5) calculation of the norm of residuals between the edge pixels and the matched surface; and (6) CHESS-CSPHMRI score was calculated on the basis of the norm of residuals and the number of the pixels detected on the edge of the liver. The diagnostic performance of CHESS-CSPHMRI score was assessed using area under the receiver operating characteristic (ROC) curve. This study was approved by all the local institutional review boards with written informed consent obtained from the participants. This study is registered with ClinicalTrials.gov, 03766880. We enrolled 260 participants with cirrhosis, including 110 healthy controls. The calculation of CHESS-CSPHMRI score in participants with or without clinically significant portal hypertension could be done within a few seconds. In the prediction of clinically significant portal hypertension, the area under the ROC curve was 0·81 (95% CI 0·75–0·86), sensitivity was 0·82 (95% CI 0·75–0·88), and specificity of CHESS-CSPHMRI score was 0·70 (95% CI 0·61–0·77). Additionally, we found a positive correlation between CHESS-CSPHMRI score and HVPG (r=0·385, p