Robust, accurate and fast automatic segmentation of the spinal cord

Area of Science:

• Medical Imaging
• Neuroscience
• Biomedical Engineering

Background:

• Spinal cord segmentation is crucial for atrophy measurement and group analysis.
• Existing automatic methods struggle with variable image contrast and field-of-view.
• Manual segmentation is time-consuming and prone to user bias.

Purpose of the Study:

• To present PropSeg, a novel automatic spinal cord segmentation algorithm.
• To optimize segmentation for robustness, accuracy, and speed.
• To enable large-scale studies of spinal cord morphology.

Main Methods:

• PropSeg utilizes a deformable model propagation approach.
• Includes initialization via Hough transform, low-resolution propagation with contrast adaptation, and refinement.
• Validated on T1-, T2-, and multiecho T2*-weighted MRI across healthy subjects and spinal cord injury patients.

Main Results:

• Achieved high segmentation precision across various MRI sequences.
• Reported Dice coefficients of 0.9 for T1/T2 and 0.86 for T2* weighted images.
• Segmentation completed in under 1 minute per subject on standard hardware.

Conclusions:

• PropSeg offers a robust, accurate, and rapid solution for spinal cord segmentation.
• Facilitates quantitative analysis of morphological features like cross-sectional area.
• Enables efficient, large-throughput neuroimaging studies of the spinal cord.