The integrity of descending white matter pathways, measured by fractional anisotropy from DW-MRI, is a key prognostic indicator of motor recovery after stroke. Barriers to translation of fractional anisotropy measures into routine clinical practice include the time required for manually delineating volumes of interest (VOIs), and inter-examiner variability in this process. This study investigated whether registering and then editing template volumes of interest 'as required' would improve inter-examiner reliability compared with manual delineation, without compromising validity. MRI was performed with 30 sub-acute stroke patients with motor deficits (mean NIHSS = 11, range 0-17). Four independent examiners manually delineated VOIs for the posterior limbs of the internal capsules on T1 images, or edited template VOIs that had been registered to the T1 images if they encroached on ventricles or basal ganglia. Fractional anisotropy within each VOI and interhemispheric asymmetry were then calculated. We found that 13/30 registered template VOIs required editing. Edited template VOIs were more spatially similar between examiners than the manually delineated VOIs (p = 0.005). Both methods produced similar asymmetry values that correlated with clinical scores with near perfect levels of agreement between examiners. Contralesional fractional anisotropy correlated with age when edited template VOIs were used but not when VOIs were manually delineated. Editing template VOIs as required is reliable, increases the validity of fractional anisotropy measurements in the posterior limb of the internal capsule, and is less time-consuming compared to manual delineation. This approach could support the use of FA asymmetry measures in routine clinical practice.
Keywords: Diffusion-weighted imaging; Magnetic resonance imaging; Prognosis; Stroke.