Assessing the Equivalence of Brain-Derived Measures from Two 3D T1-Weighted Acquisitions: One Covering the Brain and One Covering the Brain and Spinal Cord =========================================================================================================================================================== * D. Pareto * J.F. Corral * A. Garcia-Vidal * M. Alberich * C. Auger * J. Rio * N. Mongay * J. Sastre-Garriga * À. Rovira ## Abstract **BACKGROUND AND PURPOSE:** In MS, it is common to acquire brain and spinal cord MR imaging sequences separately to assess the extent of the disease. The goal of this study was to see how replacing the traditional brain T1-weighted images (brain-T1) with an acquisition that included both the brain and the cervical spinal cord (cns-T1) affected brain- and spinal cord–derived measures. **MATERIALS AND METHODS:** Thirty-six healthy controls (HC) and 42 patients with MS were included. Of those, 18 HC and 35 patients with MS had baseline and follow-up at 1 year acquired on a 3T magnet. Two 3D T1-weighted images (brain-T1 and cns-T1) were acquired at each time point. Regional cortical thickness and volumes were determined with FastSurfer, and the percentage brain volume change per year was obtained with SIENA. The spinal cord area was estimated with the Spinal Cord Toolbox. Intraclass correlation coefficients (ICC) were calculated to check for consistency of measures obtained from brain-T1 and cns-T1. **RESULTS:** Cortical thickness measures showed an ICC >0.75 in 94% of regions in healthy controls and 80% in patients with MS. Estimated regional volumes had an ICC >0.88, and the percentage brain volume change had an ICC >0.79 for both groups. The spinal cord area measures had an ICC of 0.68 in healthy controls and 0.92 in patients with MS. **CONCLUSIONS:** Brain measurements obtained from 3D cns-T1 are highly equivalent to those obtained from a brain-T1, suggesting that it could be feasible to replace the brain-T1 with cns-T1. ## ABBREVIATIONS: cns : central nervous system EDSS : Expanded Disability Status Scale HC : healthy control ICC : intraclass correlation coefficient PBVC : percentage brain volume change SCA : spinal cord area SSIM : structural similarity index The pathologic characterization of MS includes focal and diffuse areas of inflammation, demyelination, neuroaxonal loss, and gliosis in the central nervous system (cns). Lesions in the brain and spinal cord are identified with MR imaging, an essential and fundamental technique in the diagnosis, prediction of disease progression, and monitoring and prediction of the response to disease-modifying treatments.1 Actually, brain cortical lesions can be identified from routinely acquired T1-weighted MR imaging in conjunction with FLAIR imaging.2 Lesions in the spinal cord can be outlined with just T1-weighted MR imaging.3 Regarding lesion topography and atrophy quantification, both the brain and spinal cord are assessed separately, requiring the acquisition of 2 images,4 although recent studies have highlighted the relevance of acquiring simultaneous brain and spinal cord MR imaging.5,6 Several attempts have been made to assess the estimation of spinal cord area (SCA) from brain acquisitions, instead of from spinal cord MR imaging, showing that the estimated SCA values differ.7,8 Despite SCAs differing, estimation from brain acquisitions is quite common and has been proposed as an alternative in cases in which spinal cord MR imaging has not been acquired, which is a common situation due to the limited availability of the MR imaging scanner. In this context, we propose a study to investigate whether conventional brain T1-weighted MR imaging (brain-T1) can be replaced by an acquisition that encompasses both the brain and the spinal cord (cns-T1). The proposed cns-T1 is a trade-off between having a larger spinal coverage compared with a brain-T1 acquisition, allowing visualization of the presence of lesions, though it is not as complete as conventional spinal cord MR imaging, which covers down to the conus medullaris. The effect on different brain-derived measures such as cortical thickness, regional volumetry, and changes in the brain volume during a 1-year period has been assessed. The differences in the estimated SCA have also been evaluated. ## MATERIALS AND METHODS ### Cohort A group of 36 healthy controls (HCs) and 42 patients with MS was included in the study, with baseline MR imaging. From those, 18 HCs and 35 patients with MS had 12-month follow-up MR imaging. The study was approved by the Vall d’Hebron University Hospital ethics committee, and the participating subjects signed their informed consent (PR(AMI)24/2019). ### MR Imaging Acquisition Images were acquired in a 3T system (Tim Trio; Siemens) with a 12-channel whole-body coil. Acquisition parameters for the 3D T1-weighted images were the following—for brain—T1: TR = 2300 ms, TE = 2.98 ms, TI = 900 ms, 176 slices, FOV = 256 mm, voxel = 1 × 1 × 1 mm3, bandwidth = 240 Hz/px, time = 5 minutes 12 seconds; for cns-T1: TR = 2000 ms, TE = 3.21 ms, TI = 100 ms, 192 slices, FOV = 320 mm, voxel = 1 × 1 × 1 mm3, bandwidth = 150 Hz/px, time = 4 minutes 52 seconds. The parameters for the brain-plus-spine 3D T1-weighted MR imaging were established according to the protocol suggested by Cohen-Adad et al9 and include at least the C7 level. The patient table was moved between the 2 acquisitions so that the center of the image was placed in the isocenter. ### MR Image Analysis Correction for image distortions due to gradient nonlinearity was performed with the Siemens implemented tools on the MR imaging console. The two 3D T1-weighted images were cropped using the robust_fov utility in FSL and bias-field-corrected through the N4 algorithm. The aim of this step was to remove the medulla oblongata, which may affect the performance of brain analysis toolboxes. Then, the brain was segmented with FastSurfer ([https://deep-mi.org/research/fastsurfer/](https://deep-mi.org/research/fastsurfer/)).10 Cortical thickness was determined in 62 regions, 31 per hemisphere,11 and global volumes (GM, WM, thalamus, and total intracranial) were also obtained. The thalamus was chosen because it is a structure highly involved in all MS phenotypes with the presence of clinically relevant volume loss.12 For those subjects with baseline and follow-up MR imaging, the percentage brain volume change (PBVC) was calculated with the SIENA toolbox ([https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/SIENA](https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/SIENA)).13 To ensure that the quality of the baseline and follow-up images was equivalent, we calculated the structural similarity index (SSIM)14 between them. This task can be accomplished by running SIENA with the *-d* option, which does not delete the baseline and follow-up images coregistered to the halfway point of both studies (which were used to calculate the PBVC). The SSIM is based on the product of 3 terms that account for the contribution of the luminance, contrast, and structure, and it was computed as implemented in Matlab (MathWorks). Finally, the SCA was estimated with the Spinal Cord Toolbox15 ([https://github.com/spinalcordtoolbox/spinalcordtoolbox](https://github.com/spinalcordtoolbox/spinalcordtoolbox)) by manually labeling the C2-C3 intervertebral level in the 2 acquisitions. Briefly, the spine was segmented with the DeepSeg algorithm and normalized to the multimodal PAM50 atlas to delineate the SCA. The SCA was calculated as the average of 11 sections, centered at the C2-C3 level. The position of the C2-C3 level was defined manually. ### Statistical Analysis The agreement between the measures obtained from the brain-T1 and cns-T1 was obtained by calculating the intraclass correlation coefficient (ICC). Reproducibility of the cortical thickness, regional brain volumes, and the SCA was assessed by calculating the ICC between the baseline and follow-up measures in the HC group. The Cicchetti criteria16 were taken as a reference: excellent agreement for 0.75