Robust, accurate and fast automatic segmentation of the spinal cord

Spinal cord segmentation provides measures of atrophy and facilitates group analysis via inter-subject cone spondence. Automatizing this procedure enables studies with large throughput and minimizes user bias. P though several automatic segmentation methods exist, they are often restricted in terms of image contrast am field-of-view. This paper presents a new automatic segmentation method (PropSeg) optimized for robustness, at_ curacy and speed. The algorithm is based on the propagation of a deformable model and is divided into thre parts: firstly, an initialization step detects the spinal cord position and orientation using a circular Hough trans form on multiple axial slices rostral and caudal to the starting plane and builds an initial elliptical tubular mer: Secondly, a low-resolution deformable model is propagated along the spinal cord. To deal with highly variab contrast levels between the spinal cord and the cerebrospinal fluid, the deformation is coupled with a loccontrast-to-noise adaptation at each iteration. Thirdly, a refinement process and a global deformation are applie on the propagated mesh to provide an accurate segmentation of the spinal cord. Validation was performed in healthy subjects and two patients with spinal cord injury, using and T-2-weighted images of the entire spini cord and on multiecho T-2*-weighted images. Our method was compared against manual segmentation am against an active surface method. Results show high precision for all the MR sequences. Dice coefficients we 0.9 for the T-1- and T-2-weighted cohorts and 0.86 for the T-2*-weighted images. The proposed method runs less than 1 min on a normal computer and can be used to quantify morphological features such as crosr sectional area along the whole spinal cord. 2014 Elsevier Inc. All rights reserve