Test-retest repeatability of intravoxel incoherent motion (IVIM) measurements in the cervical cord

Abstract

This work aimed at assessing the reliability of intravoxel incoherent motion (IVIM) parameters sensitive to perfusion changes in the cervical cord by determining the test-retest variability across subjects and different post-processing fitting algorithms. IVIM test-retest scans were acquired in the cervical cord (C1-C3) of 10 healthy subjects on a 3T MRI scanner, with a 15-minute break in-between. IVIM parameters, including microvascular volume fraction (F), pseudo-diffusion coefficient (D*), blood flow-related coefficient (FxD*), and diffusion coefficient (D), were derived using voxel-wise and region of interest (ROI)-wise fits. The reliability of each IVIM parameter was determined with coefficients of variation (CV), intraclass correlation coefficients (ICC), Bland-Altman analysis and linear regression. To assess the effects of the different fitting approaches, a two-way repeated-measures analysis of variance (ANOVA) was conducted on the CVs calculated across fitting algorithms. Mean CVs of IVIM parameters calculated across subjects using the voxel-wise fit was lower in the white matter (WM) and grey matter (GM) (2.6% to 15.6%; 2.2% to 16.4%, respectively) compared with those calculated using the ROI-wise fit approach (WM: 4.5%-32.2%; GM: 3.4%- 53.4%). The voxel-wise fit in the WM yielded higher ICC values (good-to-excellent, 0.71 - 0.97) compared to the ROI-wise fit approach (0.49 - 0.90). IVIM parameters, derived using the voxel-wise fitting approach, demonstrated a high reliability in the cervical cord. Robust IVIM metrics, observed across scans and subjects, can facilitate studies targeting perfusion impairment and pave the way to future clinical trials assessing perfusion impairment as a potential quantitative biomarker.