Effect of Multipeak Spectral Modeling of Fat for Liver Iron and Fat Quantification: Correlation of Biopsy with MR Imaging Results

Abstract

To investigate the effect of the multipeak spectral modeling of fat on R2* values as measures of liver iron and on the quantification of liver fat fraction, with biopsy as the reference standard. Institutional review board approval and informed consent were obtained. Patients with liver disease (n = 95; 50 men, 45 women; mean age, 57.2 years±14.1 [standard deviation]) underwent a nontargeted liver biopsy, and 97 biopsy samples were reviewed for steatosis and iron grades. MR imaging at 1.5 T was performed 24-72 hours after biopsy by using a three-echo three-dimensional gradient-echo sequence for water and fat separation. Data were reconstructed off-line, correcting for T1 and T2* effects. Fat fraction and R2* maps (1/T2*) were reconstructed and differences in R2* and steatosis grades with and without multipeak modeling of fat were tested by using the Kruskal-Wallis test. Spearman rank correlation coefficient was used to assess fat fractions and steatosis grades. Linear regression analysis was performed to compare the fat fraction for both models. Mean steatosis grade at biopsy ranged from 0% to 95%. Biopsy specimens in 26 of 97 patients (27%) showed liver iron (15 mild, six moderate, and five severe). In all 71 samples without iron, a strong increase in the apparent R2* was observed with increasing steatosis grade when single-peak modeling of fat was used (P=.001). When multipeak modeling was used, there were no differences in the apparent R2* as a function of steatosis grading (P=.645), and R2* values agreed closely with those reported in the literature. Good correlation between fat fraction and steatosis grade was observed (rS=0.85) both without and with spectral modeling. In the presence of fat, multipeak spectral modeling of fat improves the agreement between R2* and liver iron. Single-peak modeling of fat leads to underestimation of liver fat.