Tomoko Hyodo MD, Presenter: Nothing to Disclose

Norihisa Yada, Abstract Co-Author: Nothing to Disclose

Osamu Maenishi MD, Abstract Co-Author: Nothing to Disclose

Peter Lamb, Abstract Co-Author: Employee, General Electric Company

Kosuke Sasaki MS, Abstract Co-Author: Employee, General Electric Company

Takamichi Murakami MD, PhD, Abstract Co-Author: Nothing to Disclose

Masakatsu Tsurusaki MD, PhD, Abstract Co-Author: Nothing to Disclose

Kazunari Ishii MD, Abstract Co-Author: Nothing to Disclose

Mitsuru Matsuki, Abstract Co-Author: Nothing to Disclose

Seishi Kumano MD, Abstract Co-Author: Nothing to Disclose

Teruhito Mochizuki MD, Abstract Co-Author: Nothing to Disclose

To assess the clinical accuracy and reproducibility of a new multi-material decomposition (MMD) algorithm developed for the quantification of hepatic fat content in fast kVp-switching dual-energy CT (DECT).

Our institutional review board approved this prospective study. Thirty-three patients with suspected hepatic steatosis (BMI, 19-34) underwent unenhanced DECT scans and single-voxel 1H-MR spectroscopy (MRS) within 4 weeks prior to liver biopsy. Histologically, steatosis was graded as 0 (<5% of hepatocytes), 1 (5-33%), 2 (33-66%) and 3 (>66%) by the nonalcoholic fatty liver disease activity score (NAS). Hepatic fat volume fraction (FVFDECT, %) images of 5 mm thickness were generated from DECT data using MMD. FVFDECT was measured in the region-of-interest (ROI; size, 300mm2) corresponding to the biopsy site. The ROI analysis was repeated by two observers. Inter- and intra-observer agreements of FVFDECT were evaluated and agreement between FVFDECT and MRS-determined fat volume fractions (FVFMRS, %; voxel size, 25 x 25 x 25 mm) was assessed using Bland-Altman analysis. FVFDECT and FVFMRS were compared regarding histological grade of steatosis using one-way analysis of variance with Tukey-Kramer and Spearman correlations.

NAS steatosis scores were score 0 in 5 patients; 1 in 14; 2 in 11; and 3 in 3. Intra- and interobserver agreement of the FVFDECT were very good (mean differences < 0.1%). There were good correlations between NAS steatosis score and both FVFDECT (r = 0.72; P < .0001) and FVFMRS (r = 0.78; P < .0001). In pairwise comparisons, no statistical significant difference was found between the scores 0 and 1 for both FVFDECT and DECTMRS. Significant differences were found between NAS steatosis scores 2 and 3 only for FVFMRS (P = .019) and between the other pairwise comparisons for both FVFDECT and FVFMRS. Bland-Altman analysis of FVFDECT and FVFMRS showed significant proportional bias (r = 0.66; P < .001).

MMD algorithm for DECT is feasible for the quantification of hepatic fat content with comparable accuracy to MRS, and with excellent reproducibility.

 The liver fat volume fraction quantification using the MMD algorithm for DECT can be used as a non-invasive method in screening and follow up of patients with fatty liver disease. Considering the proportional bias between DECT and MRS, the follow-up studies should preferably be performed with the same method.

Hyodo, T, Yada, N, Maenishi, O, Lamb, P, Sasaki, K, Murakami, T, Tsurusaki, M, Ishii, K, Matsuki, M, Kumano, S, Mochizuki, T, A Multi-material Decomposition Algorithm for Liver Fat Quantification in Dual-energy CT: Reproducibility of the Method, and Comparison with MR Spectroscopy.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14001969.html