Fully automated grey and white matter spinal cord segmentation

Area of Science:

• Neuroimaging
• Medical image analysis
• Neurology

Background:

• Spinal cord atrophy, indicated by reduced cervical cross-sectional area (CSA), is a key factor in irreversible disability in multiple sclerosis (MS).
• Magnetic resonance imaging (MRI) and image segmentation are used to indirectly assess in vivo axonal loss.
• Accurate segmentation is vital for reliable CSA measurements and understanding MS progression.

Purpose of the Study:

• To develop and validate a fully automated spinal cord segmentation technique for accurate CSA measurements in MS.
• To evaluate the performance of the proposed method against inter-rater variability.

Main Methods:

• The proposed technique integrates two multi-atlas segmentation methods: Optimized PatchMatch Label fusion (OPAL) for initial segmentation and Similarity and Truth Estimation for Propagated Segmentations (STEPS) for detailed white and grey matter segmentation.
• A retrospective analysis of MRI data was performed to assess the method's accuracy.

Main Results:

• The automated method achieved high accuracy in CSA measurements, comparable to inter-rater variability, with a Dice score (DSC) of 0.967 at the C2/C3 level.
• Grey matter segmentation at C2/C3 showed accuracy close to inter-rater levels (DSC of 0.826 for healthy subjects, 0.835 for clinically isolated syndrome MS patients).

Conclusions:

• The novel automated spinal cord segmentation technique provides accurate and reliable measurements of spinal cord atrophy.
• This method holds promise for improved monitoring of MS progression and clinical disability assessment.