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 ± 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 ± 2.28 mm² and the mean WM area was 72.71 ± 7.55 mm² using the automatic method. Reproducibility was high for both methods, while being better for the automatic approach (all mean automatic coefficients of variation, ≤4.77%; all differences, P < .001). The accuracy of the automatic method compared with the manual reference standard was excellent (mean Dice similarity coefficients: 0.86 ± 0.04 for GM and 0.90 ± 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.

Fig 1.

Exemplary axial AMIRA slice of 1 representative volunteer at the C4 level. A–H, Eight images of different tissue contrast acquired by the AMIRA sequence, shown in chronologic order from lowest-to-highest TI. I, Average image from A to E in full view, which delivers a high contrast-to-noise-ratio for GM/WM. J, Average image from F to H, which delivers a high contrast-to-noise ratio for SC/CSF. K, Same average image as in I but histogram-equalized and zoomed.

Fig 2.

Flow chart of the automatic segmentation pipeline. As a first step of the algorithm to align the 12 slices, the images are center-cropped and slice-wise successively coregistered rostral to caudal using translations in pixel-size steps to prevent further interpolation. Then, the algorithm automatically locates and delineates the ring-shaped CSF from its surroundings and extracts the cross-sectional SC surface. Finally, it uses the previously segmented SC surface as a mask for GM/WM differentiation. The iterative steps of CSF segmentation are shown as a zoomed-in view. GM segmentation uses essentially the same steps and is thus not shown in detail.

Fig 3.

Cross-sectional areas of total spinal cord, white matter, and gray matter per axial slice as measured by automatic segmentations. Notice the slight increase of total spinal cord (TSC) and the marked GM cross-sectional area increase caudally, which corresponds to the cervical SC enlargement. The light gray area depicts the limits of ±1 SD.

Fig 4.

Comparison between the reproducibility of manual and automatic measurements (AM) of spinal cord gray matter and white matter per axial slice. Intrasession and intersession reproducibility is assessed in terms of Dice coefficients (graphics on the left) and coefficients of variation (graphics on the right). Manual and automatic intersession reproducibility is shown in dark gray, whereas manual and automatic intrasession reproducibility is shown in light gray. Error bars display mean values ± 0.2 SDs.

Fig 5.

Accuracy measurements in terms of Dice coefficients (graphics on the left) and Hausdorff distances (graphics on the right) of white matter and gray matter per slice. Overlaid boxplots display median values as well as 25th and 75th percentile values. Gray areas depict the mean standard error values ± 1 SD.

Fig 6.

Examples of segmentations of representative patients with MS. The thick continuous line indicates automatic segmentation; the dashed line, manual reference standard. A, A 54-year-old female patient with MS. Rostral cervical SC slices of the C1/C2 level without focal lesions. Automatic segmentation highly corresponds to the manual reference standard. B, A 32-year-old male patient with MS. Rostral cervical SC slice of the C2 level with a focal posterolateral lesion fused with the left posterior gray matter horn. Automatic segmentation misclassifies the focal lesion as SC GM. C, A 33-year-old female patient with MS. A cervical SC slice of the C3/C4 level with a focal posterior lesion fusing with the posterior SC GM horns and the central SC GM commissure. Automatic segmentation misclassifies the focal lesion as SC GM and CSF.