Meeting Banner
Abstract #2430

Robust and automatic spinal cord detection on multiple MRI contrasts using machine learning

Charley Gros1, Benjamin De Leener1, Allan R. Martin2, Michael G. Fehlings2, Virginie Callot3,4, Nikola Stikov1,5, and Julien Cohen-Adad1,6

1NeuroPoly Lab, Institute of Biomedical Engineering, Polytechnique Montreal, Montréal, QC, Canada, 2Division of Neurosurgery, Department of Surgery, University of Toronto, Toronto, ON, Canada, 3Aix Marseille Univ, CNRS, CRMBM, Marseille, France, 4APHM, Hôpital de la Timone, Hôpital de la Timone, Pôle d’imagerie médicale, Marseille, France, 5Montreal Heart Institute, Montréal, QC, Canada, 6Functional Neuroimaging Unit, CRIUGM, Université de Montréal, Montreal, QC, Canada

Detecting the spinal cord on a large variety of MRI data is challenging but essential for the automation of quantitative analysis pipelines. For the past few years, machine learning algorithms have outperformed most unsupervised image processing methods. The present study investigates the performance of two different machine learning algorithms, Convolutional Neural Networks (CNN) and Support Vector Machine (SVM), on MRI data from different vendors, with a variety of pathology, contrast, resolution and FOV. Results suggest strong performance of the CNN approach, opening the door to application in multi-center analysis pipelines.

This abstract and the presentation materials are available to members only; a login is required.

Join Here