Physiologic Interindividual Variability of Volume and Atrophy in Central Nervous System Structures With Focus on Spinal Cord as Measured by Quantitative Magnetic Resonance Imaging.

Multiple sclerosis (MS) is a chronic immune-mediated inflammatory and neurodegenerative disease that affects central nervous system. Magnetic resonance imaging (MRI) is the main paraclinical examination used to monitor disease activity and response to treatment. However, there is only limited correlation between the clinical symptoms and findings seen on the conventional MRI - a phenomena called clinical-radiological paradox. Measuring whole and regional brain atrophy provides additional information to MRI and seems to correlate better with clinical course and prognosis. In the last two decades, brain atrophy in MS has been exten sively studied, cut-offs for brain volume loss over time has been established and it is likely that it will become an important outcome measure in both clinical studies and routine practice.

Compared to brain atrophy, spinal cord (SC) atrophy and its relation to MS-related disability has been given less attention. This is partly due to technical challenges such as inhomogeneous magnetic field in this region, small physical dimensions of SC and artefacts caused by motion of the SC within the spinal canal together with the flow of cerebrospinal fluid and periodic motion due to respiratory and cardiac cycles. Moreover, focal spinal cord MS lesions may cause both swelling and shrinking of SC that influence its absolute volume, resulting in problems with interpretation of absolute SC volume in MS. Despite all these challenges, it is possible to identify SC lesions and reliably measure SC volume, but there is no "gold standard" (standardized) software for SC volume measurement and no agreement on the SC level (segments) that is most suitable for SC volume loss measurement.

During 2016 we have developed an in-house semiautomatic pipeline for measurement of cervical spinal cord volume that is a part of Scanview program. The segmentation is performed on T2-weighted images. In the first step, a marker is manually placed in a center of an intervertebral disc C3-4 (sagittal plane). Subsequently a cord straightening (manual rotation of the cord to achieve a perpendicular orientation of spinal cord to dorsal part of C3 vertebral body), that enable to reduce partial volume due to the cord orientation. After this centering and straightening,transformation matrix is saved and all subsequent steps are fully automatized. These steps include: 1. Sub-pixel division, 2. Reversing the contrast of T2-weighted images and smoothing by applying a set of median, Gaussian and edge-enhancing filters and 4. Cubic spline interpolation to find a curve that represents a border of a spinal cord with highest probability. Finally, a sum of mean areas of 21 1-mm slices is calculated (1 center slice fixed at center of a intervertebral disc C3-4 and 2 x 10 slices in cranial and caudal direction. Using this new software we have assessed 1,036 MS patients during 2016 and 2017. The intra- and inter-rater variability of SC volume assessment was done by using the intraclass correlation coefficient (ICC) with a two-way mixed absolute agreement and single-measures design. The analysis confirmed very good consistency (> 98%) of the method. The preliminary results showed a significant correlation between spinal cord volume and different clinical phenotypes.