Assessing motion-associated tSNR of brainstem and spinal cord fMRI in a post-stroke cohort
Poster No:
1326
Submission Type:
Abstract Submission
Authors:
Neha Reddy1, Kimberly Hemmerling1, Julius Dewald1, Molly Bright1
Institutions:
1Northwestern University, Chicago, IL
First Author:
Co-Author(s):
Introduction:
Motor-task fMRI is a critical modality to study neural changes after a stroke. Although current work has focused on cortical activity patterns in post-stroke cohorts, brainstem and spinal cord fMRI will be equally important to understanding overall adaptation within the central nervous system[7]. However, one challenge of motor-task fMRI in a cohort with stroke is that participants exhibit higher head motion[10], which can decrease data quality. This problem is compounded in the brainstem and spinal cord, which already have lower data quality. Therefore, we aimed to (1) anticipate the degree of head motion in post-stroke participants before an MRI scan, and (2) assess how head motion affects fMRI data quality in the brain, brainstem, and spinal cord.
Methods:
Data collection: 6 individuals (62±7y, 6M) with chronic hemiparetic stroke and a paretic upper limb underwent 3 study sessions: 1 outside the scanner and 2 MRI scans in a Siemens 3T Prisma with a 64-channel head/neck coil. Participants performed a hand-grasp task at 40% maximum force: 10-s 'squeeze', 15-s 'relax', 11 trials/hand. During the first visit, participants lay on an exam table in an out-of-use head coil to simulate the MRI environment, while grip force and head motion data were collected (Fig1).
During MRI sessions, participants performed the hand-grasp task during a cortical-brainstem GRE EPI scan (TR=2.2s, TEs=13.4/39.5/65.6ms, FA=90°, MB factor=2, voxel size=1.731x1.731x4.0mm³) and a spinal cord (~C4-C7) GRE EPI scan with ZOOMit selective excitation (TR=2.13s, TE=30ms, FA=90°, voxel size=1x1x3mm³). Axial slices were aligned perpendicular to the base of the 4th ventricle or longitudinal cord axis, respectively. Structural scans were also acquired.
Lab session analysis: Motion data were downsampled to 0.5Hz to approximate the fMRI TR, and Framewise Displacement (FD) was calculated as the sum of the difference in head motion between samples[8].
Cortical-brainstem fMRI analysis [1,5]: The first 10 fMRI volumes were removed to allow for steady-state magnetization, then scans were distortion-corrected. Head-motion realignment parameters were computed for the first echo with respect to the Single Band reference image, then applied to all echoes. An optimally combined image was calculated[4]. The hand-grasp task and motion parameters were modeled out before mean tSNR was calculated in two ROIs transformed to functional space: gray matter segmented from the T1-w scan, thresholded at 75%; and brainstem, using the Harvard-Oxford subcortical structural atlas brainstem[3] thresholded at 50%. FD was calculated using parameters from volume realignment.
Spinal cord fMRI analysis [1,2,5]: 2D slicewise motion correction was performed and FD calculated using X and Y motion parameters. The spinal cord was manually segmented. Mean cord tSNR was calculated after task and motion regressors were modeled out.
During MRI sessions, participants performed the hand-grasp task during a cortical-brainstem GRE EPI scan (TR=2.2s, TEs=13.4/39.5/65.6ms, FA=90°, MB factor=2, voxel size=1.731x1.731x4.0mm³) and a spinal cord (~C4-C7) GRE EPI scan with ZOOMit selective excitation (TR=2.13s, TE=30ms, FA=90°, voxel size=1x1x3mm³). Axial slices were aligned perpendicular to the base of the 4th ventricle or longitudinal cord axis, respectively. Structural scans were also acquired.
Lab session analysis: Motion data were downsampled to 0.5Hz to approximate the fMRI TR, and Framewise Displacement (FD) was calculated as the sum of the difference in head motion between samples[8].
Cortical-brainstem fMRI analysis [1,5]: The first 10 fMRI volumes were removed to allow for steady-state magnetization, then scans were distortion-corrected. Head-motion realignment parameters were computed for the first echo with respect to the Single Band reference image, then applied to all echoes. An optimally combined image was calculated[4]. The hand-grasp task and motion parameters were modeled out before mean tSNR was calculated in two ROIs transformed to functional space: gray matter segmented from the T1-w scan, thresholded at 75%; and brainstem, using the Harvard-Oxford subcortical structural atlas brainstem[3] thresholded at 50%. FD was calculated using parameters from volume realignment.
Spinal cord fMRI analysis [1,2,5]: 2D slicewise motion correction was performed and FD calculated using X and Y motion parameters. The spinal cord was manually segmented. Mean cord tSNR was calculated after task and motion regressors were modeled out.
Results:
Head motion in the lab was significantly correlated with both head and spinal cord motion (FD) during fMRI (Fig2A). Head and spinal cord motion was significantly negatively correlated with mean tSNR in cortical, brainstem, and spinal cord ROIs (Fig2D).
Conclusions:
In cohorts with increased movement during motor-task fMRI, motion can be evaluated in a mock-MRI environment before scanning. However, while the lab session is designed to simulate the MRI session, minor differences in setup still exist, causing variability in the lab vs. MRI motion relationship.
Even after motion correction and denoising, spinal cord tSNR was lower in this stroke cohort compared to a similar spinal cord dataset in younger controls[6], indicating the importance of other physiological denoising techniques for brainstem and spinal cord fMRI. We anticipate that multi-echo independent component analysis for cortical-brainstem scans will improve tSNR[9]. Overall, our findings demonstrate that a motion-capture lab session can help anticipate and plan for potential decreases in tSNR throughout the CNS in a clinical population.
Even after motion correction and denoising, spinal cord tSNR was lower in this stroke cohort compared to a similar spinal cord dataset in younger controls[6], indicating the importance of other physiological denoising techniques for brainstem and spinal cord fMRI. We anticipate that multi-echo independent component analysis for cortical-brainstem scans will improve tSNR[9]. Overall, our findings demonstrate that a motion-capture lab session can help anticipate and plan for potential decreases in tSNR throughout the CNS in a clinical population.
Disorders of the Nervous System:
Neurodegenerative/ Late Life (eg. Parkinson’s, Alzheimer’s) 2
Modeling and Analysis Methods:
Activation (eg. BOLD task-fMRI) 1
Motion Correction and Preprocessing
Motor Behavior:
Motor Planning and Execution
Novel Imaging Acquisition Methods:
BOLD fMRI
Keywords:
Brainstem
FUNCTIONAL MRI
Motor
Movement Disorder
Spinal Cord
Other - stroke
1|2Indicates the priority used for review
Provide references using author date format
[2] Deshpande, R. (2022). Neptune: A toolbox for spinal cord functional MRI data processing and quality assurance [Abstract]. International Society for Magnetic Resonance in Medicine.
[3] Desikan, R. S. (2006). An automated labeling system for subdividing the human cerebral cortex on MRI scans into gyral based regions of interest. NeuroImage, 31(3), 968–980. https://doi.org/10.1016/j.neuroimage.2006.01.021
[4] Dupre, E. (2021). TE-dependent analysis of multi-echo fMRI with tedana. The Journal of Open Source Software, 6(59), 1–5. https://doi.org/10.21105/joss.03103
[5] Jenkinson, M. (2012). FSL. NeuroImage, 62, 782–790. https://doi.org/10.1016/j.neuroimage.2011.09.015
[6] Hemmerling, K. J. (2023). Spatial distribution of hand-grasp motor task activity in spinal cord functional magnetic resonance imaging. Human Brain Mapping, 44(17), 5567–5581. https://doi.org/10.1002/hbm.26458
[7] Li, S. (2019). A unifying pathophysiological account for post-stroke spasticity and disordered motor control. In Frontiers in Neurology (Vol. 10, Issue MAY). https://doi.org/10.3389/fneur.2019.00468
[8] Power, J. D. (2012). Spurious but systematic correlations in functional connectivity MRI networks arise from subject motion. NeuroImage, 59(3), 2142–2154. https://doi.org/10.1016/j.neuroimage.2011.10.018
[9] Reddy, N. A. (2023). Multi-echo ICA denoising increases tSNR in combined cortical and brainstem task-fMRI [Abstract]. Organization for Human Brain Mapping Annual Meeting.
[10] Seto, E. (2001). Quantifying head motion associated with motor tasks used in fMRI. NeuroImage, 14(2), 284–297. https://doi.org/10.1006/nimg.2001.0829