Recommendations for Cord Atrophy Assessment

Below are a few general principles for improving the reproducibility of atrophy assessment, based on our experience.

• Use the minimum number of shape coefficients that is necessary to accurately define the outline of the cord. In general 16 may be sufficient, and will help to make the measurement process more robust in case some your images are of lower quality (poorer signal-to-noise ratio or with motion artefacts).
• Use the minimum order of longitudinal variation that is consistent with the changes in cord outline shape along the section of cord that you are interested in. For short sections of cord you may only need an order of 2 or 3.
• If you are performing a follow-up study, use the Follow-Up Tool. This should make the assessment of change more robust.
• We have found (unpublished data) that the section of the cord which is most sensitive to the MS disease process is C2/C3 - C5/C6. Assessing atrophy on longer sections of cord does not appear to improve reproducibility if a global measure of atrophy is required.
• Including the section of the cord up the odontoid process of the epistropheus (OPE) may make reproducibility worse, since identifying the exact slice for the OPE is problematic. Curvature of the cord, and its rapidly changing cross-sectional area mean that slight changes in the Landmark position will result in significant changes to the average cord area.
• Always use the exact same settings for all subjects and time-points when analysing a cohort. This means that the settings must be such that analysis is robust to the poorest quality image in the whole cohort. Exclusion of images based on image quality risks biasing the outcome measures, since poor image quality may be associated with greater disability.