To evaluate the predictive value of intra-voxel incoherent motion diffusion-weighted imaging (IVIM-DWI) multi-parameter MRI in microvessel invasion (MVI) of hepatocellular carcinoma (HCC).

Clinical and imaging data of 96 patients with HCC admitted to the Third Affiliated Hospital of Sun Yat-sen University from March 2014 to May 2020 were retrospectively analyzed. The informed consents of all patients were obtained and the local ethical committee approval was received. Among them, 81 patients were male and 15 female, aged (51±3) years old on average. According to pathological results, all patients were divided into the positive MVI group (n=33) and negative MVI group (n=63). All the patients received abdominal contrast-enhanced MRI and IVIM-DWI within preoperative 1 month. Imaging features of HCC, such as capsule integrity, tumor halo sign and mosaic sign, etc. were assessed by conventional MRI. The tumor apparent diffusion coefficient (ADC), pseudo apparent diffusion coefficient (D*) and perfusion fraction (F) values were measured by IVIM-DWI. The imaging features of positive MVI patients were analyzed by t test or Chi-square test. The correlation between imaging parameters and MVI was determined by Logistic regression analysis. The predictive value of imaging parameters on MVI was assessed by diagnostic test.

In positive MVI patients, the incidence of tumor mosaic sign and tumor halo sign was significantly increased (χ²=15.492, 9.412; P<0.05), the ADC and F values were remarkably decreased, whereas the D* value was significantly elevated (t=-2.252, -3.034, 2.977; P<0.05). Multivariate Logistic regression analysis demonstrated that mosaic sign, tumor halo sign and F value were the independent influencing factors of MVI in HCC (HR=5.589, 5.408, 0.324; P<0.05). The sensitivity and specificity of mosaic sign, tumor halo sign and F value in predicting MVI were 0.42, 0.79, 0.79, and 0.94, 0.91, 0.49, respectively. The specificity of these three signs combined in predicting MVI was increased to 0.97.

Multi-parameter MRI of IVIM possesses certain predictive value for MVI of HCC. The combination of mosaic sign, tumor halo sign and F value yields high specificity in predicting MVI of HCC.