Quantitative comparison between aMRI and DENSE for the assessment of brain tissue motion

PurposeAmplified MRI (aMRI) holds potential for assessing brain tissue motion and strain, using images acquired from readily-available sequences. However, image registration is necessary to extract displacements from the motion-amplified images, which may limit its accuracy. We aimed to separately assess the errors from imperfections in the aMRI amplification, and errors from the registration algorithm, using a semi-synthetic approach. MethodsGround truth brain tissue motion was derived from smoothed Displacement Encoding with Stimulated Echoes (DENSE) measurements acquired at 7T (8 subjects). Those were then applied to a still sagittal anatomical balanced-SSFP image to obtain a DENSE-animated MRI series to which aMRI (amplification factor 10) was applied. DENSE-amplified MRI series served as a reference (Damp-MRI; amplification factor, 10). Amplified displacements were extracted from aMRI and Damp-MRI using a common registration algorithm. Linear regression was used to estimate the amplification and r2 agreement of the amplified displacements relative to the ground truth. ResultsThe estimated amplification was consistently lower for aMRI-derived displacements (range: [4.9{+/-}0.3 5.7{+/-}0.3]) than for Damp-MRI measurements (range: [6.7{+/-}0.5 7.7{+/-}0.5]). Nevertheless, the spatial, temporal and average characteristics of brain tissue motion derived from aMRI were comparable to the ground truth for Anterior-Posterior and Feet-Head displacements: (group averaged r2[≥]0.84), as were the Damp-MRI derived displacements (r2[≥]0.88). When aMRI was applied to in-vivo cine-bSSFP images and compared to the ground truth, the results were less favorable, highlighting the need for artefact-free images. ConclusionThese results strengthen the potential of aMRI as a tool for semi-quantitative assessment of brain tissue motion in disease.