Spinal cord (SC) tissue loss is known to occur in some patients with multiple sclerosis (MS), resulting in SC atrophy.
Currently, no measurement tools exist to determine the magnitude of SC atrophy from Magnetic Resonance Images
(MRI). We have developed and implemented a novel semi-automatic method for quantifying the cervical SC volume
(CSCV) from Magnetic Resonance Images (MRI) based on level sets. The image dataset consisted of SC MRI exams
obtained at 1.5 Tesla from 12 MS patients (10 relapsing-remitting and 2 secondary progressive) and 12 age- and gender-matched
healthy volunteers (HVs). 3D high resolution image data were acquired using an IR-FSPGR sequence acquired
in the sagittal plane. The mid-sagittal slice (MSS) was automatically located based on the entropy calculation for each of
the consecutive sagittal slices. The image data were then pre-processed by 3D anisotropic diffusion filtering for noise
reduction and edge enhancement before segmentation with a level set formulation which did not require re-initialization.
The developed method was tested against manual segmentation (considered ground truth) and intra-observer and inter-observer
variability were evaluated.