Normal liver tissue accumulates intravenously infused SPIOs and loses signal (SI) on gradient recalled echoes (GREs); fibrotic tissue does not accumulate SPIOs and does not lose SI. On delayed images after Gd administration, fibrotic tissue enhances, further increasing its conspicuity. Thus, on double-contrast enhanced GREs (DEGREs), fibrosis appears as hyperintense reticulations against a dark liver background. Quantitative analysis of liver reticulation is necessary to assess fibrosis objectively. The aim of this study was to identify image texture features that reliably differentiate fibrotic from normal liver on DEGRE images.

45 patients, 23 with advanced fibrosis (METAVIR fibrosis score F3 or F4) and 22 histologically confirmed normals, underwent DEGRE imaging at 1.5T (TR 140/TE 4.5/FA 70/8-10 mm slice thickness/no gaps/32-40 mm FOV/176x256 matrix). Ninety comparably sized regions of interest (ROIs) were placed in representative areas of the 45 livers (2 ROIs/liver) without knowledge of truth standard, avoiding blood vessels. On each ROI, 73 quantitative texture variables were analyzed, including descriptive statistics (mean SI, standard deviation, coefficient of variation, etc.); pixel SI distribution (kurtosis, skewness); image gradient and edge detection; and image information (angular momentum and entropy). Mean values in fibrotic and normal livers were compared (unpaired Student t-tests with Bonferroni–corrected two-tailed alpha levels of 0.0007). Receiver operator characteristics (ROC) and diagnostic performance for differentiation of fibrotic from normal liver were assessed for each variable.

For all 73 texture variables, mean values were higher in fibrotic than normal livers, 48/73 (66%) significantly. Areas under the ROC curve (AUC) exceeded 0.9 for 38/73 (52%) variables. On these 38 variables, optimal sensitivity/specificity ranged from 90%/85% to 98%/98%.

Liver texture can be assessed quantitatively on DEGRE images to accurately differentiate advanced fibrosis from normal liver. This capability shows promise for noninvasive grading of liver fibrosis. Validation in a large cohort of patients with a spectrum of fibrosis severity is needed.