Four-dimensional flow MRI-based prediction of minimal hepatic encephalopathy in patients with liver cirrhosis

To develop a noninvasive predictive model for minimal hepatic encephalopathy (MHE) according to the blood flow indices of spontaneous portosystemic shunt (SPSS) on four-dimensional (4D) flow MRI in patients with liver cirrhosis.

A consecutive cohort of patients with liver cirrhosis complicated with enhanced abdominal CT-confirmed SPSS from April 2022 to October 2023 was prospectively recruited and divided into MHE and non-MHE groups. All patients underwent portal vascular imaging with 4D flow MRI, the blood flow indices of SPSSs (mainly including gastroesophageal shunts (F-GES) and splenorenal/gastrorenal shunts (F-SRS/F-GRS)), the splenic vein (F-SV), and the superior mesenteric vein (F-SMV) were measured. Additionally, the total crosssectional area of the SPSSs (TA-S) on enhanced CT, demographic characteristics and clinical factors were collected. Clinical, morphological, hemodynamic, and combined predictive models (Model_C, Model_M, Model_H, and Model_C+H) were developed. Their discrimination, calibration and net clinical benefit were evaluated to determine the value of 4D flow MRI-based hemodynamic assessment and to identify the optimal predictive model for MHE.

A total of 72 patients (mean age 53.9 ± 8.0 years, 43 males) were included, with 27 in the MHE group and 45 in the non-MHE group. Child-Pugh class C, TA-S, F-GES, F-SRS/F-GRS and F-SV were factors influencing the presence of MHE. The combined Model_C+H, integrating clinical factors and 4D flow parameters, substantially outperformed models based on either data type alone (AUC 0.97 vs. 0.77 and 0.90), demonstrating the complementary value of clinical and hemodynamic information. The model also exhibited good calibration (χ² = 14.80, p = 0.35) and the highest net clinical benefit, confirming its excellent predictive performance for MHE.

4D flow MRI can reveal the hemodynamic characteristics of SPSSs in patients with liver cirrhosis, and a prediction model based on its hemodynamic index and clinical factors can provide valuable insights for early detection and serving as a clinical foundation for timely intervention in MHE.