Menu

Javascript is not activated in your browser. This website needs javascript activated to work properly.
You are here

Variability in diffusion kurtosis imaging: Impact on study design, statistical power and interpretation.

Author:
  • Filip Szczepankiewicz
  • Jimmy Lätt
  • Ronnie Wirestam
  • Alexander Leemans
  • Pia Sundgren
  • Danielle van Westen
  • Freddy Ståhlberg
  • Markus Nilsson
Publishing year: 2013
Language: English
Pages: 145-154
Publication/Series: NeuroImage
Volume: 76
Issue: 1
Document type: Journal article
Publisher: Elsevier

Abstract english

Diffusion kurtosis imaging (DKI) is an emerging technique with the potential to quantify properties of tissue microstructure that may not be observable using diffusion tensor imaging (DTI). In order to help design DKI studies and improve interpretation of DKI results, we employed statistical power analysis to characterize three aspects of variability in four DKI parameters; the mean diffusivity, fractional anisotropy, mean kurtosis, and radial kurtosis. First, we quantified the variability in terms of the group size required to obtain a statistical power of 0.9. Second, we investigated the relative contribution of imaging and post-processing noise to the total variance, in order to estimate the benefits of longer scan times versus the inclusion of more subjects. Third, we evaluated the potential benefit of including additional covariates such as the size of the structure when testing for differences in group means. The analysis was performed in three major white matter structures of the brain: the superior cingulum, the corticospinal tract, and the mid-sagittal corpus callosum, extracted using diffusion tensor tractography and DKI data acquired in a healthy cohort. The results showed heterogeneous variability across and within the white matter structures. Thus, the statistical power varies depending on parameter and location, which is important to consider if a pathogenesis pattern is inferred from DKI data. In the data presented, inter-subject differences contributed more than imaging noise to the total variability, making it more efficient to include more subjects rather than extending the scan-time per subject. Finally, strong correlations between DKI parameters and the structure size were found for the cingulum and corpus callosum. Structure size should thus be considered when quantifying DKI parameters, either to control for its potentially confounding effect, or as a means of reducing unexplained variance.

Keywords

  • Radiology, Nuclear Medicine and Medical Imaging
  • Diffusion kurtosis imaging
  • Diffusion tensor imaging
  • DKI
  • DTI
  • Statistical power
  • Study design
  • Group size
  • Tractography
  • Statistics
  • Effect size
  • Random effects model

Other

Published
  • ISSN: 1095-9572
Freddy Ståhlberg
E-mail: freddy.stahlberg [at] med.lu.se

Professor

Medical Radiation Physics, Lund

+46 46 17 31 19

+46 70 688 31 19

32

Professor

Diagnostic Radiology, (Lund)

+46 46 17 70 30

32