Akira Yamada MD, Presenter: Nothing to Disclose

Kazuhiko Ueda MD, Abstract Co-Author: Nothing to Disclose

Yasunari Fujinaga MD, Abstract Co-Author: Nothing to Disclose

Masahiro Kurozumi MD, Abstract Co-Author: Nothing to Disclose

Shinichi Miyagawa, Abstract Co-Author: Nothing to Disclose

Masumi Kadoya MD, Abstract Co-Author: Nothing to Disclose

To evaluate value of gadoxetate disodium on noninvasive diagnosis of liver fibrosis by texture analysis of MR images.

Consecutive 46 patients who underwent preoperative gadoxetate disodium-enhanced MR imaging using 3 Tesla MR system were included in this retrospective study. The grade of liver fibrosis (the fibrosis score: F) was histologically diagnosed by surgical specimen in all patients. Pre-contrast respiratory-gated 2D fast spin echo T2-weighted images (voxel size = 0.7 x 0.7 x 5 mm), pre- and post-contrast (20 minutes after venous administration) breath-hold 3D gradient recalled echo T1-weighted images (voxel size = 0.7 x 0.7 x 3 mm) were used for evaluation. Fat-suppression was applied to all images. Region of interests sized 60 x 60 pixels were located in the liver avoiding major vessels and hepatic lesions in each MR image. Four feature values (‘contrast’, ‘correlation’, ‘energy’, and ‘heterogeneity’) of the liver were determined by texture analysis of region of interests. A stepwise liner regression analysis of the fibrosis score on the feature values obtained from texture analysis was performed using 3 different image sets (pre-contrast MR images, post-contrast MR images, and the both). ROC analysis of obtained 3 regression models in differentiation of liver fibrosis (F1-4) from normal liver (F0) was performed.

The area under ROC of obtained 3 regression models in differentiation of liver fibrosis from normal liver was 0.64 for pre-contrast MR images, 0.83 for post-contrast MR images, and 0.85 for the both. Two feature values (x1: ’correlation’ in post-contrast T1-weighted images, P < 0.0001; x2: ’energy’ in pre-contrast T2-weighted images, P = 0.017) were significant predictors for the fibrosis score in eventual regression model (y = -31.232x1 - 10.39x2 + 32.137, R = 0.63, P < 0.0001).

Gadoxetate disodium can add value on noninvasive diagnosis of liver fibrosis by texture analysis of MR images.

The degree of liver fibrosis especially at its early stage can be predicted non-invasively by texture analysis of non-enhanced and gadoxetate disodium-enhanced MR images.

Yamada, A, Ueda, K, Fujinaga, Y, Kurozumi, M, Miyagawa, S, Kadoya, M, Texture Analysis of Non-enhanced and Gadoxetate Disodium-enhanced MR Images of the Liver: A Comparison with Histological Grade of Liver Fibrosis.  Radiological Society of North America 2014 Scientific Assembly and Annual Meeting, - ,Chicago IL. http://archive.rsna.org/2014/14005423.html