Investigating the extent and localization of DISH
A total of 89 embalmed human cadaveric thoracic spines, stripped from their surrounding soft tissues, were obtained from the Dept. of Anatomy of our institution and screened for DISH, by two independent observers, by macroscopical inspection and fluoroscopy (OmnidiagnostEleva; Philips Medical Systems, Best, the Netherlands) using the criteria of Resnick et al. . The absence of sacroiliac erosion was not considered an absolute requirement for this part of the study since these joints were not available for screening in the specimens obtained. A total of ten specimens (five male/five female donors, on average 80.4 years old, ranging from 66 to 91 years) met the criteria and underwent CT to:
establish a definite diagnosis of DISH (excluding AS and other osseous abnormalities); and
perform quantitative analysis of the anterolateral ossifications.
Scanning was performed in supine position with a 64-slice scanner (Philips Brilliance; PhilipsMedical Systems, Best, the Netherlands), using 0.625-cm axial helical scans with high resolution (120 kV, 200 mAs, slice thickness 0.9 mm). On sagittal and coronal reconstructions, the level of interest was determined by identifying and recording the first four contiguous spinal levels connected by an anterolateral ossification mass (ALOM), counting from the twelfth thoracic vertebra and moving cranially. From all four-segment specimens thus identified, the mid-vertebral body level; intervertebral disc space level and in-between levels were visualized. See also Fig. 2 for a detailed illustration of the location of the levels under study. A total of 130 CT images (10 specimens × 13 transverse planes) were captured in .bmp format (768 × 765 pixels) and subsequently imported to, and analyzed by, 2D/3D modeling software (Rhinoceros® version 3.0, Seattle, USA). First, the anteroposterior (AP) axis was defined by the line drawn from the most ventral to the most dorsal part of the vertebral body/disc. This AP line was bisected to create a vertebral body center node (VBC). Multiple nodes (up to 50) were subsequently drawn manually to outline the ALOM and a connecting line was drawn through these nodes (see Fig. 3). The centroid node (geometrical center; CN) of the ALOM was calculated and automatically drawn by the software and a line was drawn from the CN to VBC in order to measure the angle of this line relative to the AP axis, called the anteroposterior-centroid node angle (AP–CN angle). Finally, a line was drawn over the CT-generated ruler to convert to a real-world dimension (mm). Total ALOM area was calculated by the software (in mm2) as was the AP–CN angle (in °, where a positive value was assigned to all orientations pointing to the right side of the body, i.e. counter-clockwise from the AP axis, and a negative value assigned to orientations pointing to the left side of the body, i.e. clockwise from the AP axis).
Localization of vertebral segmental arteries in DISH and controls
The results from the first experiment were verified using computed tomography angiography data (using the same scanner and scanning protocol described above) obtained from 20 patients examined previously for aneurysms of the abdominal aorta, ten of whom also had DISH according to the full set of Resnick criteria (all male, age on average 71.4 years, ranging from 57 to 85 years) while the diagnosis DISH was definitely ruled out for the other ten patients (all male, age on average 68.9 years, ranging from 58 to 80 years) again by two independent observers. The tenth thoracic vertebra, a level frequently involved in DISH and present in the majority of the cadaveric specimens used, was identified on the CTA scans and coronal images were obtained from the following three locations: the anterior vertebral body wall; the center of the vertebral body and the plane in-between (see Fig. 4). On these images the right-sided VSA was identified and its cranial-caudal location was recorded using the same levels as for the cadaveric experiment: at the mid-vertebral body level; intervertebral disc space level and in-between adjacent levels.
A repeated measures analysis with two within factors [affected vertebral level (T
) and measurement localization] was used to examine the main effects of these two factors but also to assess whether these effects depend on each other (interaction between vertebral level and measurement localization). If there was a significant main effect, a post hoc analysis with Bonferroni adjustment was performed. For the measurement localization as well as for the vertebral levels, the most cranial location/level respectively, was compared with all others. Statistical significance was set at p < 0.05.