2021 CMSC Annual Meeting

Accurate Cervical Spinal Cord Area Measurements Can be Extracted from Brain Images

IMG05

Background: Multiple Sclerosis (MS) is an inflammatory, demyelinating disease with neurodegeneration in the central nervous system. Spinal cord atrophy has been associated with disability in MS.The C2/3 segment of the cervical cord is often used to assess atrophy; however, the spinal cord is imaged far less frequently compared to the brain due to additional cost and scanning time including repositioning and coil changes. 3DT1 brain images often capture a few upper cervical spinal cord levels.
Objectives: If C1 cross-sectional areas (CSA) from brain imaging could accurately reflects C2/C3 CSA from cord imaging, spinal cord area measures could be obtained using only a standard T1-weighted brain image.
Methods: We collected data from 28 healthy controls and 73 MS patients. A For brain imaging, a 3D T1- weighted magnetization-prepared rapid gradient-echo sequence was acquired. For spinal cord imaging, a T2*-weighted multi-echo rapid gradient echo scan was collected. Spinal Cord Toolbox (SCT) was used to extract the mean cord CSA. Spearman correlations and Bland-Altman analyses (BA) were assessed between: (1) C1 CSAcord and C2/3 CSAcord, (2) C1 CSAbrain and C1 CSAcord, and (3) C1 CSAbrain and C2/3 CSAcord.
Results: C1 CSAcord was correlated with C2/C3 CSAcord in both controls (r=0.94, p<0.0001) and MS patients (r=0.85, p<0.0001). BA analysis showed a bias of 0.50 and -0.38 mm2 for controls and MS patients respectively with no significant non-zero slope (p=0.052; p=0.37). C1 CSAcord and C1 CSAbrain showed a correlation in both controls (r=0.89, p<0.0001) and MS patients (r=0.91, p<0.0001). BA analysis showed a bias of 2.4 and 0.43mm2 for controls and MS patients with no significant non-zero slope (p=0.16; p=0.78). C1 CSAbrain was correlated with C2/3 CSAcord for both controls (r=0.84, p<0.0001) and MS patients (r=0.81, p<0.0001). BA analysis showed a bias of -1.9 and -0.81mm2 for controls and MS patients with no significant non-zero slope (p=0.095; p=0.59). Conclusions: A significant, positive correlation was seen between C1 and C2/3 CSA using spinal cord scans. The BA further supported the use of C1 as a proxy for C2/3 CSA in cord atrophy analysis. There was a strong correlation between C1 CSA from brain and cord images, supporting acquisition of an accurate C1 CSA using 3DT1 brain images. Direct comparison of C1 CSAbrain to C2/3 CSAcord displayed a strong relationship in both healthy controls and MS patients.