Evaluation of Intra‐ and Interscanner Reliability of MRI Protocols for Spinal Cord Gray Matter and Total Cross‐Sectional Area Measurements

BACKGROUND: In vivo quantification of spinal cord atrophy in neurological diseases using MRI has attracted increasing attention. PURPOSE: To compare across different platforms the most promising imaging techniques to assess human spinal cord atrophy. STUDY TYPE: Test/retest multiscanner study. SUBJECTS: Twelve healthy volunteers. FIELD STRENGTH/SEQUENCE: Three different 3T scanner platforms (Siemens, Philips, and GE) / optimized phase sensitive inversion recovery (PSIR), T(1)‐weighted (T(1)‐w), and T(2)*‐weighted (T(2)*‐w) protocols. ASSESSMENT: On all images acquired, two operators assessed contrast‐to‐noise ratio (CNR) between gray matter (GM) and white matter (WM), and between WM and cerebrospinal fluid (CSF); one experienced operator measured total cross‐sectional area (TCA) and GM area using JIM and the Spinal Cord Toolbox (SCT). STATISTICAL TESTS: Coefficient of variation (COV); intraclass correlation coefficient (ICC); mixed effect models; analysis of variance (t‐tests). RESULTS: For all the scanners, GM/WM CNR was higher for PSIR than T(2)*‐w (P < 0.0001) and WM/CSF CNR for T(1)‐w was the highest (P < 0.0001). For TCA, using JIM, median COVs were smaller than 1.5% and ICC >0.95, while using SCT, median COVs were in the range 2.2–2.75% and ICC 0.79–0.95. For GM, despite some failures of the automatic segmentation, median COVs using SCT on T(2)*‐w were smaller than using JIM manual PSIR segmentations. In the mixed effect models, the subject was always the main contributor to the variance of area measurements and scanner often contributed to TCA variance (P < 0.05). Using JIM, TCA measurements on T(2)*‐w were different than on PSIR (P = 0.0021) and T(1)‐w (P = 0.0018), while using SCT, no notable differences were found between T(1)‐w and T(2)*‐w (P = 0.18). JIM and SCT‐derived TCA were not different on T(1)‐w (P = 0.66), while they were different for T(2)*‐w (P < 0.0001). GM area derived using SCT/T(2)*‐w versus JIM/PSIR were different (P < 0.0001). DATA CONCLUSION: The present work sets reference values for the magnitude of the contribution of different effects to cord area measurement intra‐ and interscanner variability. Level of Evidence: 1 Technical Efficacy: Stage 4 J. Magn. Reson. Imaging 2019;49:1078–1090.