A reliability assessment of constrained spherical deconvolution-based diffusion-weighted magnetic resonance imaging in individuals with chronic stroke

Abstract

Background: Diffusion-weighted magnetic resonance imaging (DW-MRI) is commonly used to assess white matter properties after stroke. Novel work is utilizing constrained spherical deconvolution (CSD) to estimate complex intra-voxel fiber architecture unaccounted for with tensor-based fiber tractography. However, the reliability of CSD-based tractography has not been established in people with chronic stroke.

New method: Establishing the reliability of CSD-based DW-MRI in chronic stroke. High-resolution DW-MRI was performed in ten adults with chronic stroke during two separate sessions. Deterministic region of interest-based fiber tractography using CSD was performed by two raters. Mean fractional anisotropy (FA), apparent diffusion coefficient (ADC), tract number, and tract volume were extracted from reconstructed fiber pathways in the corticospinal tract (CST) and superior longitudinal fasciculus (SLF). Callosal fiber pathways connecting the primary motor cortices were also evaluated. Inter-rater and test-retest reliability were determined by intra-class correlation coefficients (ICCs).

Results: ICCs revealed excellent reliability for FA and ADC in ipsilesional (0.86-1.00; p<0.05) and contralesional hemispheres (0.94-1.00; p<0.0001), for CST and SLF fibers; and excellent reliability for all metrics in callosal fibers (0.85-1.00; p<0.05). ICC ranged from poor to excellent for tract number and tract volume in ipsilesional (-0.11 to 0.92; p≤0.57) and contralesional hemispheres (-0.27 to 0.93; p≤0.64), for CST and SLF fibers.

Comparison with existing method: Like other select DW-MRI approaches, CSD-based tractography is a reliable approach to evaluate FA and ADC in major white matter pathways, in chronic stroke.

Conclusion: Future work should address the reproducibility and utility of CSD-based metrics of tract number and tract volume.

Keywords: Constrained spherical deconvolution; Diffusion-weighted magnetic resonance imaging; Intra-class correlation coefficient; Reliability; Stroke.