Related Experiment Video
Updated: Aug 5, 2026
07:00
Diffusion Tensor Magnetic Resonance Imaging in Chronic Spinal Cord Compression
Published on: May 7, 2019
Magnetic Resonance Imaging Preprocessing for Robust Spinal Cord Segmentation in Cervical Myelopathy
Hediyeh Toufani1,2, Richard M Dansereau3, Philippe Phan4
1Department of Mechanical Engineering, University of Ottawa, Ottawa, ON K1N 1A2, Canada.
Journal of Imaging
|July 27, 2026
Summary
This study introduces a new preprocessing method to improve spinal cord segmentation in magnetic resonance imaging (MRI) for patients with cervical myelopathy. The pathology-aware framework enhances accuracy and reliability in challenging compressed regions.
Area of Science:
- Medical Imaging
- Neuroscience
- Biomedical Engineering
Background:
- Accurate spinal cord segmentation is crucial for quantitative analysis in MRI, including cross-sectional area and microstructural characterization.
- Extreme shape deformation in cervical myelopathy due to cord compression challenges automated segmentation methods, especially deep learning models.
- Current deep learning methods struggle with reliability when anatomical assumptions are violated by severe compression.
Purpose of the Study:
- To develop and evaluate a pathology-aware, boundary-focused preprocessing framework to improve spinal cord segmentation in cervical myelopathy.
- To mitigate segmentation failure modes caused by extreme cord compression.
- To enhance the accuracy, robustness, and stability of automated spinal cord segmentation in pathological conditions.
Main Methods:
- Introduced a novel preprocessing framework with components designed to enhance intensity homogeneity, suppress noise, and improve boundary visibility.
- Utilized a multi-representation input strategy, feeding intensity-, contrast-, and edge-enhanced views derived from a single T2*-weighted scan into a U-Net model.
- Evaluated the framework on spinal cord MRI data from 194 cervical myelopathy cases across three clinical centers.
Main Results:
- The proposed preprocessing framework significantly improved spinal cord segmentation accuracy, robustness, and stability.
- Improvements were particularly notable in anatomically challenging regions affected by severe cord compression.
- The pathology-aware approach demonstrated superior performance compared to generic preprocessing methods in the context of cervical myelopathy.
Conclusions:
- Pathology-aware preprocessing is essential for reliable spinal cord segmentation in cervical myelopathy.
- The developed boundary-focused framework effectively addresses the limitations of current methods in handling severe cord compression.
- This approach holds promise for advancing quantitative MRI analysis in spinal cord pathologies.
