Dorsal Root Ganglion Volumetry by MR Gangliography

Abstract

Background and purpose: Dorsal root ganglion MR imaging (MR gangliography) is increasingly gaining clinical-scientific relevance. However, dorsal root ganglion morphometry by MR imaging is typically performed under the assumption of ellipsoid geometry, which remains to be validated.

Materials and methods: Sixty-four healthy volunteers (37 [57.8%] men; mean age, 31.5 [SD, 8.3] years) underwent MR gangliography of the bilateral L4-S2 levels (3D-T2WI TSE spectral attenuated inversion recovery-sampling perfection with application-optimized contrasts by using different flip angle evolution, isotropic voxels = 1.1 mm³, TE = 301 ms). Ground truth dorsal root ganglion volumes were bilaterally determined for 96 dorsal root ganglia (derivation cohort) by expert manual 3D segmentation by 3 independent raters. These ground truth dorsal root ganglion volumes were then compared with geometric ellipsoid dorsal root ganglion approximations as commonly practiced for dorsal root ganglion morphometry. On the basis of the deviations from ellipsoid geometry, improved volume estimation could be derived and was finally applied to a large human validation cohort (510 dorsal root ganglia).

Results: Commonly used equations of ellipsoid geometry underestimate true dorsal root ganglion volume by large degrees (factor = 0.42-0.63). Ground truth segmentation enabled substantially optimizing dorsal root ganglion geometric approximation using its principal axes lengths by deriving the dorsal root ganglion volume term of [Formula: see text]. Using this optimization, the mean volumes of 510 lumbosacral healthy dorsal root ganglia were as follows: L4: 211.3 (SD, 52.5) mm³, L5: 290.7 (SD, 90.9) mm³, S1: 384.2 (SD, 145.0) mm³, and S2: 192.4 (SD, 52.6) mm³. Dorsal root ganglion volume increased from L4 to S1 and decreased from S1 to S2 (P < .001). Dorsal root ganglion volume correlated with subject height (r = . 22, P < .001) and was higher in men (P < .001).

Conclusions: Dorsal root ganglion volumetry by measuring its principal geometric axes on MR gangliography can be substantially optimized. By means of this optimization, dorsal root ganglion volume distribution was estimated in a large healthy cohort for the clinically most relevant lumbosacral levels, L4-S2.

FIG 1.

Evaluation of 6 equations (e1–e6) for estimating DRG volume: Boxplots show deviation of estimated DRG_vol (e1–e6) from ground truth volume (DRG_vol (r) by calculating the quotient of estimated volume and ground truth volume. The closest approximation to ground truth was attained for Equation 6 (DRG_vol [e6], far right): = $\frac{2}{3}\times A\times B\times C+75 \text{mm}³$. The top of the box represents the 75th percentile, the bottom of the box represents the 25th percentile. The line in the middle represents the 50th percentile (median). The whiskers represent the 5th and 95th percentiles, and values beyond the lower and upper bounds represent outliers and extreme values.

FIG 2.

Colored 3D and surface volume-rendered and volume cut visualization of the volumetric results of the 6 different equations for estimation of ipsilateral DRG volume. The green shell represents the ground truth voxelwise segmentation result, DRG_vol (r). The purple shell just at and very close to the outside of the ground truth shell (green) corresponds to the best optimization result, which was obtained from the new Equation 6, including an intercept, only slightly overestimating voxelwise ground truth segmentation. The pink shell just at the inside of the green shell corresponds to the new and second-best-performing Equation 5, without an intercept, slightly underestimating DRG_vol (r). These 2 equations have provided the best results for estimating DRG_vol. For the purpose of visualization, the shells of these 2 optimized equations are slightly offset to allow macroscopic differentiation of the small differences. Significant underestimation of volume was observed for commonly used equations e1–e4; (brown = e1, red = e2, orange = e3, and yellow = e4. The contralateral DRG is surface volume-rendered in gray).

FIG 3.

Comparative DRG volume-rendered visualization of ground truth volume DRG_vol (r) versus best estimated volume by DRG_vol (e6). Left: right-anterior-oblique plane. Middle: frontal plane. Right: left-anterior-oblique. Purple DRGs on the right side with smooth surfaces correspond to estimated volumes using Equation 6 for the most accurate approximation. Green DRGs on the left side with mesh surface correspond to ground truth volumes from expert manual segmentation and contour locations.

FIG 4.

Boxplot of DRG_vol (e6) showing the DRG_vol (cubic millimeters) for the different lumbosacral heights L4–S2. Equation 6 (DRG_vol, e6) provided the best approximation to ground truth. By means of this equation, these boxplots display human MR imaging DRG morphometry over the clinically most relevant segments, L4–S2, in a large validation cohort (n = 510). The top of the box represents the 75th percentile; the bottom of the box represents the 25th percentile. The line in the middle represents the 50th percentile (median). The whiskers represent the 5th and 95th percentiles, and values beyond the lower and upper bounds represent outliers and extreme values. Significances are marked with asterisks (indicating P < . 0001).