This page presents our work on spinal cord centerline localization.
Our main result is a very robust and efficient method to localize the spinal cord on 3D images.
A localisation map is first computed using a HOG+SVM spinal cord detector. The centerline curve is then found as the solution of a nonlinear optimization problem. An original algorithm is proposed to find the global solution of this problem in a efficient and non iterative manner.
The method has been presented at RITS 2017[1], MICCAI 2017 [2] and has been published in Medical Image Analysis[3].
Details of the method can be found here.
Software
The binaries are availaible here for linux-x86_64, macos 10.7, macos 10.11 and windows.
It is also included and packaged in the Spinal Cord Toolbox as sct_detect_spinalcord along with a lot of spinal cord dedicated tools. It is now used for example to initialize the spinal cord segmentation tool.
Please reference the MEDIA journal article [3].
Author
M. Sdika
Collaborations
V. Callot at CRMBM
The Neuropoly team at the Polytechnique Montréal.
References
- SDIK-17. . Automatic Detection of the Spinal Cord Centerline using Machine Learning and Global Nonlinear Optimization. Dans: Recherche en Imagerie et Technologies pour la Santé (RITS). Recherche en Imagerie et Technologies pour la Santé (RITS). ; 2017.
- GROS-17a. OptiC: Robust and Automatic Spinal Cord Localization on a Large Variety of MRI Data Using a Distance Transform Based Global Optimization. Dans: International Conference on Medical Image Computing and Computer-Assisted Intervention. International Conference on Medical Image Computing and Computer-Assisted Intervention. Springer; 2017. p. 712–719.
- GROS-18. Automatic spinal cord localization, robust to MRI contrasts using global curve optimization. Med Image Anal. 2018 ;44:215-227.