Texture analysis of ultrahigh field T2*-weighted MR images of the brain: application to Huntington's disease

J Magn Reson Imaging. 2014 Mar;39(3):633-40. doi: 10.1002/jmri.24199. Epub 2013 May 30.


Purpose: To develop a framework for quantitative detection of between-group textural differences in ultrahigh field T2*-weighted MR images of the brain.

Materials and methods: MR images were acquired using a three-dimensional (3D) T2*-weighted gradient echo sequence on a 7 Tesla MRI system. The phase images were high-pass filtered to remove phase wraps. Thirteen textural features were computed for both the magnitude and phase images of a region of interest based on 3D Gray-Level Co-occurrence Matrix, and subsequently evaluated to detect between-group differences using a Mann-Whitney U-test. We applied the framework to study textural differences in subcortical structures between premanifest Huntington's disease (HD), manifest HD patients, and controls.

Results: In premanifest HD, four phase-based features showed a difference in the caudate nucleus. In manifest HD, 7 magnitude-based features showed a difference in the pallidum, 6 phase-based features in the caudate nucleus, and 10 phase-based features in the putamen. After multiple comparison correction, significant differences were shown in the putamen in manifest HD by two phase-based features (both adjusted P values=0.04).

Conclusion: This study provides the first evidence of textural heterogeneity of subcortical structures in HD. Texture analysis of ultrahigh field T2*-weighted MR images can be useful for noninvasive monitoring of neurodegenerative diseases.

Keywords: Huntington's disease; MR images; neurodegenerative diseases; subcortical structures; texture analysis; ultrahigh field.

Publication types

  • Comparative Study
  • Research Support, Non-U.S. Gov't

MeSH terms

  • Adult
  • Case-Control Studies
  • Diffusion Magnetic Resonance Imaging / methods*
  • Female
  • Humans
  • Huntington Disease / diagnosis
  • Huntington Disease / pathology*
  • Image Interpretation, Computer-Assisted*
  • Imaging, Three-Dimensional*
  • Male
  • Middle Aged
  • Pattern Recognition, Automated
  • Reference Values
  • Statistics, Nonparametric