Vertebral endplate and shell thickness measurement using digital tomosynthesis
Yeni YN, Dix MR, Xiao A, Oravec DJ, and Flynn MJ. Vertebral endplate and shell thickness measurement using digital tomosynthesis. J Orthop Res 2017; 35
Journal of orthopaedic research
INTRODUCTION: Cortical shell of vertebrae has been noted to be thinner in those with osteoporosis  and vertebral deformities  as well as in women not receiving compared to those receiving HRT . In biomechanical studies, shell thickness has been associated with work to failure independently from BMD . Cortical endplates also were implicated in the failure of vertebrae  and in the degenerative disease of the spine . Although useful on a comparative basis [2, 3], the ability of CT to measure vertebral cortical thickness is moderate [7, 8]. This is especially so for endplate thickness, which is oriented in the direction of poor resolution in CT scans. Digital tomosynthesis (DTS) is a clinically available imaging modality that has high image resolution (150-300 mm) in the sagittal (for lateral scans) and coronal (for AP scans) planes. It has been shown that DTS can quantify the surface topography of vertebral endplates  and so it has potential for measuring vertebral cortical thickness. The objective of the current study was to examine the ability of DTS to measure human vertebral shell and endplate thickness in comparison to micro- and high resolution computed tomography (μCT and HRCT). METHODS: Lumbar 1 vertebrae were harvested under local IRB approval from 25 donors (14M/11F, 41-100y). Donors with a history of HIV, hepatitis, diabetes, renal failure, metastatic cancer, osteomalacia, hyperparathyroidism, Paget's disease of bone, spine surgery, cause of death involving trauma, and corticosteroid, anticonvulsant or bisphosphonate use were not included. Specimens were scanned using μCT, and HRCT. DTS scans were performed in AP (producing a stack of coronal plane image) and LM (producing a stack of sagittal plane images) views while aligned axially (0°) or transversely (90°) to the superior-inferior axis of the vertebrae. HRCT images were segmented  and volume masks were produced representing cortical and cancellous bone . A global threshold was applied to delineate bone from soft tissue in both DTS and μCT images . Endplate and cortical shell were separated and local thickness was measured at each voxel for μCT, HRCT and DTS images using Local Thickness plugin for ImageJ . Average and standard deviation of thickness were calculated for superior and inferior endplates (EP.Th.S.Av/SD and EP.Th.I.Av/SD), and the shell (Sh.Th.Av/SD) from each imaging modality (Fig. 1). The relationships of DTS- and HRCT-derived shell and endplate thickness with corresponding (μCT-derived thickness variables were examined using regression. RESULTS: EP.Th measurements from DTS were more strongly correlated to (μCT than did HRCT (Table 1, Fig. 2). The slopes were closer to one with nonsignificant intercepts whereas measurements from HRCT largely overestimated thickness and had nonzero intercepts when regressed against (μCT results (Table 1). Both Sh.Th.Av and Sh.Th.SD from HRCT correlated to those from (μCT, albeit overestimated. For DTS, Sh.Th.Av from LM view only correlated to (μCT (Table 1). When Sh.Th from DTS was correlated to (μCT within a matched central slab volume, rather than the whole volume, Sh.Th.Av and Sh.Th.SD from both views (AP, LM) were significant (R=0.46-0.70, p<0.02 to p<0.0001). DISCUSSION: We have demonstrated that DTS can be used for assessment of vertebral endplate thickness. Nonsignificant intercepts and slopes close to 1 indicate that the measurements are free of bias; however, there was scatter around regression lines resulting in moderate R2 values. This was consistent between AP and LM views (Table 1, Fig. 2). As previously noted, cortical thickness from CT was overestimated and moderately correlated to the reference measurements for the shell [7, 8] but somewhat poorly for the endplate. Sh.Th.SD measured from DTS was not correlated with that measured from (μCT and Sh.Th.Av was correlated with (μCT for the LM view only. When the analysis volume was limited to a central slab for (μCT and DTS, the shell measurements were correlated. This is likely due to the substant al curvature of the shell, rendering the thickness measurable only in a central slab in which the shell is perpendicular to the image plane. The standard image reconstruction of DTS allows for slice by slice analysis of the thickness; however, it may be possible to increase the size of the analysis volume for shell thickness measurements by considering a 3D reconstruction and calculation in DTS. In conclusion, closer agreement and higher correlations of results with (μCT for DTS than for HRCT indicates that DTS offers an improvement over CT for endplate thickness measurements. LM view of DTS may be used to obtain a correlate of Sh.Th.Av, or a central slab of any DTS view may be used to obtain Sh.Th.Av and Sh.Th.SD within that slab. Future work is needed to explore the possibility of increasing the precision of the measurements. This study used (μCT as gold standard. Future work may also consider comparison to histologic measurements. (Figure Presented).