Does intravoxel incoherent motion reliably stage hepatic fibrosis, steatosis, and inflammation?

Objective: To investigate the usefulness of intravoxel incoherent motion (IVIM) in determining the severity of hepatic fibrosis, steatosis, and inflammation in patients with chronic liver disease. Methods: Forty-nine patients who had liver MRI with IVIM sequence and liver biopsy within three months of MRI were enrolled. A reviewer, blinded to histology, placed regions of interest of 1-2 cm(2) in the right liver lobe. In addition, the first twenty patients were assessed with a second reviewer. Perfusion fraction (f), pseudodiffusion coefficient (D-fast), true diffusion coefficient (D-slow), and apparent diffusion coefficient (ADC) were calculated from normalized signal intensities that were fitted into a biexponential model. Errors in the model were minimized with global stochastic optimization using Simulated Annealing. ANOVA with post hoc Tukey-Kramer test and multivariate generalized linear model analysis were performed, using histological findings as the gold standard. Results: The most common etiologies for liver disease were hepatitis C and alcohol, accounting together for 76% (37/49) of patients. Low-grade fibrosis (F0, F1), hepatic steatosis, and inflammation were seen in 24% (12/49), 31% (15/49), and 29% (14/49) of patients, respectively. The interobserver correlation was poor for D-fast and D-slow (0.105, 0.173) and moderate for f and ADC (0.461, 0.418). ANOVA showed a strong inverse association between D-fast and liver fibrosis grade (p = 0.001). A weak inverse association was seen between ADC and hepatic steatosis (p = 0.059). Multivariate general linear model revealed that the only significant association between IVIM parameters and pathological features was between D-fast and fibrosis. On ROC curve analysis, D-fast < 23.4 x 10(-3) mm(2)/s had a sensitivity of 82.8% and a specificity of 64.3% in predicting high-grade fibrosis. Conclusion: D-fast has the strongest association with hepatic fibrosis but has weak interobserver correlation. IVIM parameters were not significantly associated with hepatic inflammation or steatosis.