Spinal Cord Cross Sectional Area normalization in MR Images: Case of healthy population

Spinal Cord (SC) Cross-Sectional Area (CSA)(SCCSA) is a recognized indicator for monitoring SC atrophy in neurodegenerative disorders such as Devic Neuro-Myelitis Optica (DNMO). CSA is traditionally measured with reference to a vertebral level to predict the spinal level, which can be inaccurate as the measurement location varies between subjects. An automated model for normalizing CSA based on the detection of Ponto-Medullary Junction (PMJ) instead of a vertebral level has been tested on a public dataset. Our objective is to examine the inter-individual variability of CSA and to evaluate the correlations of CSA with age, gender, and manufacturer. As well, we aim to compare the results of CSA measured from the PMJ to those measured from vertebral levels. Cervical CSA was calculated on T2-w Magnetic Resonance Imaging (MRI) scans for 260 healthy subjects from the spine generic multi subject database. It was computed at 64 mm from the PMJ and at the C2-C3 vertebral level. We explored the influence of factors by calculating Pearson correlation coefficients. Subsequently, significant predictors were identified using a stepwise linear regression. Finally, we used the coefficients from the most suitable model for the CSA normalizing.