Measurement of cervical spinal cord cross-sectional area by MRI using edge detection and partial volume correction

Abstract

Purpose: To detail a procedure to accurately measure upper cervical cord cross-sectional area (CSA), using MRI, by correcting for partial volume averaging (PVA), and to assess the usefulness of the procedure for measuring cervical cord atrophy rates in longitudinal studies. Materials and Methods: Analysis of errors associated with measuring CSA in the presence of PVA is given. A numerical phantom image is produced, including simulated acquisition noise, to assess accuracy of the method in idealized conditions, and to verify the results of the error analysis. A phantom, consisting of 11 rods of known CSA, was scanned 10 times and measurement accuracy assessed. A total of 10 normal subjects were scanned twice to assess the reproducibility under experimental conditions. Results: The measurement error for the numerical phantom increased with increased simulated acquisition noise, as predicted by the analysis. Measurement of the plastic phantom revealed a systematic overestimate in CSA due to limited scanner accuracy of 3.15%. The scan-rescan error for the CSA of the cervical spine in the 10 normal subjects was 0.55%. Conclusion: Correcting for PVA allows accurate measurement of the upper cervical cord CSA and accurate measurement of a standard phantom to guard against scanner drift in longitudinal studies of cord CSA. © 2005 Wiley-Liss, Inc.