Effects of signal averaging, gradient encoding scheme, and spatial resolution on diffusion kurtosis imaging: An empirical study using 7T MRI

Chia-Wen Chiang; Shih-Yen Lin; Kuan-Hung Cho; Kuo-Jen Wu; Yun Wang; Li-Wei Kuo

doi:10.1002/jmri.26755

Effects of signal averaging, gradient encoding scheme, and spatial resolution on diffusion kurtosis imaging: An empirical study using 7T MRI

J Magn Reson Imaging. 2019 Nov;50(5):1593-1603. doi: 10.1002/jmri.26755. Epub 2019 Apr 16.

Authors

Chia-Wen Chiang¹, Shih-Yen Lin^{1

2}, Kuan-Hung Cho¹, Kuo-Jen Wu³, Yun Wang³, Li-Wei Kuo^{1

4}

Affiliations

¹ Institute of Biomedical Engineering and Nanomedicine, National Health Research Institutes (NHRI), Miaoli, Taiwan.
² Department of Computer Science, National Chiao Tung University, Hsinchu, Taiwan.
³ Center for Neuropsychiatric Research, NHRI, Miaoli, Taiwan.
⁴ Institute of Medical Device and Imaging, National Taiwan University College of Medicine, Taipei, Taiwan.

PMID: 30990956
DOI: 10.1002/jmri.26755

Abstract

Background: Although diffusion gradient directions and b-values have been optimized for diffusion kurtosis imaging (DKI), little is known about the effect of signal averaging on DKI reliability.

Purpose: To evaluate how signal averaging influences the reliability of DKI indices using two gradient encoding schemes with three spatial resolutions.

Study type: Prospective.

Animal model: Fifteen naïve Sprague-Dawley rats.

Field strength/sequence: DKI was performed at 7T using two schemes (30 directions with three b-values [30d-3b] and six directions with 15 b-values [6d-15b]), three resolutions, and eight repetitions.

Assessment: DKI reliability was assessed using voxelwise relative error (σ) and test-retest error of fractional anisotropy (FA), mean diffusivity (MD), and mean kurtosis (MK) within gray matter (GM) and white matter (WM). The number of excitations (NEX) was optimized by considering DKI reliability. The influence of the partial volume effect (PVE) was also assessed.

Statistical test: One-way analysis of variance.

Results: The 30d-3b scheme, compared with the 6d-15b scheme, exhibited apparently smaller σ_FA and σ_MK (eg, at NEX 1, in GM, for three resolutions, σ_FA : 19.9-38.2% vs. 34.2-61.4%, σ_MK : 6.9-11.4% vs. 14.1-15.4%) and similar σ_MD (all differences between two schemes <1.6%). The optimal NEX was determined as 2 for enabling a reliable measurement of DKI-derived indices. The PVE at the lowest resolution apparently increased σ_FA for both schemes (19.9% for 30d-3b and 34.2% for 6d-15b) and σ_MK for the 6d-15b scheme (14.7%) in GM, and exerted lower effects on MK values for the 30d-3b scheme (P > 0.05).

Data conclusion: A higher number of diffusion directions would benefit FA and MK estimation. A higher spatial resolution helps to reduce PVE. By using the 30d-3b scheme, MK is considered a robust index to reflect microstructural changes in GM and WM. We propose a systematic approach to determine the optimal DKI protocols for appropriate preclinical settings.

Level of evidence: 2 Technical Efficacy: Stage 1 J. Magn. Reson. Imaging 2019;50:1593-1603.

Keywords: Diffusion kurtosis imaging; diffusion gradient encoding scheme; signal averaging; spatial resolution.

Publication types

Research Support, Non-U.S. Gov't

MeSH terms

Animals
Image Processing, Computer-Assisted / methods*
Magnetic Resonance Imaging*
Rats
Rats, Sprague-Dawley
Reference Values
Reproducibility of Results
Signal-To-Noise Ratio