Automatic Spinal Cord Gray Matter Quantification: A Novel Approach

Abstract

BACKGROUND AND PURPOSE: Currently, accurate and reproducible spinal cord GM segmentation remains challenging and a noninvasive broadly accepted reference standard for spinal cord GM measurements is still a matter of ongoing discussion. Our aim was to assess the reproducibility and accuracy of cervical spinal cord GM and WM cross-sectional area measurements using averaged magnetization inversion recovery acquisitions images and a fully-automatic postprocessing segmentation algorithm. MATERIALS AND METHODS: The cervical spinal cord of 24 healthy subjects (14 women; mean age, 40 Ϯ 11 years) was scanned in a test-retest fashion on a 3T MR imaging system. Twelve axial averaged magnetization inversion recovery acquisitions slices were acquired over a 48-mm cord segment. GM and WM were both manually segmented by 2 experienced readers and compared with an automatic variational segmentation algorithm with a shape prior modified for 3D data with a slice similarity prior. Precision and accuracy of the automatic method were evaluated using coefficients of variation and Dice similarity coefficients. RESULTS: The mean GM area was 17.20 Ϯ 2.28 mm 2 and the mean WM area was 72.71 Ϯ 7.55 mm 2 using the automatic method. Reproducibility was high for both methods, while being better for the automatic approach (all mean automatic coefficients of variation, Յ4.77%; all differences, P Ͻ .001). The accuracy of the automatic method compared with the manual reference standard was excellent (mean Dice similarity coefficients: 0.86 Ϯ 0.04 for GM and 0.90 Ϯ 0.03 for WM). The automatic approach demonstrated similar coefficients of variation between intra-and intersession reproducibility as well as among all acquired spinal cord slices. CONCLUSIONS: Our novel approach including the averaged magnetization inversion recovery acquisitions sequence and a fully-automated postprocessing segmentation algorithm demonstrated an accurate and reproducible spinal cord GM and WM segmentation. This pipeline is promising for both the exploration of longitudinal structural GM changes and application in clinical settings in disorders affecting the spinal cord. ABBREVIATIONS: AMIRA ϭ averaged magnetization inversion recovery acquisitions; CV ϭ coefficient of variation; DSC ϭ Dice similarity coefficient; HD ϭ Hausdorff distance; SC ϭ spinal cord T he human spinal cord (SC) can be affected by numerous neurologic disorders of variable pathophysiology (eg, genetic, inflammatory, demyelinating, degenerative, and so forth), 1,2 and MR imaging is a valuable part of the diagnostic work-up in patients with suspected intramedullary pathology. 3,4 SC gray matter and white matter can be involved to a various extent not only among different SC disorders but also among patients with the same disease (eg, multiple sclerosis, amyotrophic lateral sclerosis). 5,6 Hence, quantification of SC compartments may add to our understanding of SC pathology 5,6 and hopefully help in the management of individual patients in the future. However, the SC presents additional challenges for MR...

Normalization of Spinal Cord Total Cross-Sectional and Gray Matter Areas as Quantified With Radially Sampled Averaged Magnetization Inversion Recovery Acquisitions

“…With the approval of the first disease-modifying treatments for SMA, the need for biomarkers that allow reliable monitoring of the disease course and the therapeutic response in SMA patients has substantially grown. Current advances in morphometric MRI development allow gray (GM) and white matter (WM) quantification in the spinal cord (SC) ( 1 – 7 ), which may help in improving the in vivo characterization of motor neuron diseases or other neurodegenerative SC diseases. Imaging the cervical and lumbar enlargements could be informative, especially in lower motor neuron diseases, e.g., SMA, or lower motor neuron-predominant amyotrophic lateral sclerosis (ALS).…”

“…More recently, several semi-automated and automated tools ( 10 – 17 ) have been developed for segmentation of the SC GM and WM from MR images, including promising automated segmentation algorithms specifically developed for the AMIRA approach ( 7 , 14 ). The reliability of the segmentation methods in single center studies is in general high, and some methods have been validated in multi-centric settings ( 18 ).…”

Role of artificial intelligence in MS clinical practice

“…2 ). Interestingly, DL algorithms based on CNNs were recently proposed for performing automated segmentation of upper cervical cord GM ( Paugam et al, 2019 , Tsagkas et al, 2019 ). Other studies applied AI on 1 H-MRS and identified N-acetylaspartate as the most significant metabolite differentiating between RRMS and SPMS (83% accuracy) ( Ek et al, 2020 ), although T2 LV and clinical history contributed as well (up to 87%) ( Ion-Margineanu et al, 2017 ).…”

“…Spinal cord segmentation with the proposed open-source method outperformed the compared state-of-the-art method (DSC = 95% vs 88% of PropSeg ), while intramedullary MS lesion segmentation performance were comparable to manual segmentations. Other DL algorithms based on CNN were proposed for performing automated segmentation of upper cervical cord GM ( Paugam et al, 2019 , Tsagkas et al, 2019 ).…”

Analysis of the Curative Effect of Posterior Approach on Lumbar Brucellar Spondylitis with Abscess through Magnetic Resonance Imaging under Improved Watershed Algorithm

“…When Brucella invades the vertebral body or intervertebral disc tissue, it will cause spinal infection, called brucellar spondylitis (BS). Its incidence is also increasing year by year [ 5 ]. The use of medication alone for the treatment of BS is often unsatisfactory.…”