A pre-clinical evaluation of geometrical stitching errors in a cone beam system: possible implications for weight-bearing CT imaging of the spine

Background Total body cone-beam computed tomography (CBCT) is recently developed for both weight-bearing and non–weight-bearing CT imaging of the spine. Before whole-spine weight-bearing CT is used in clinical practice, potential errors must be addressed, such as the need to stitch multiple volumes due to the field-of-view limitations of CBCT technology. Purpose To determine the geometric error of fused CBCT images of the spine using automatic stitching software. Material and Methods In total, 144 CBCT scans were obtained using three human cadavers. The geometric stitching error was determined in terms of total translation and rotation between vertebrae Th12 and L5, which were positioned in separate image volumes, with a regular spiral CT scan as a reference. The effect of cadaver size, radiation dose, and volume overlap between adjacent CBCT images on the stitching error was determined using Spearman's rank correlation test. Results The median total translation and rotation error were 1.88 mm (interquartile range [IQR] = 1.48–2.42 mm) and 0.54° (IQR = 0.35°–0.63°), respectively. A weak negative correlation between the different volumes of overlap and total translation ( r = −0.396; P < 0.001) and rotation ( r = −0.319; P < 0.001) was found, as well as a weak positive correlation between the cadaver size and total translation ( r = 0.456; P < 0.001). Conclusion The results of this cadaver study showed stitching errors in the order of 2 mm for translation and 0.5° for rotation in fused CBCT volumes of the spine. These findings function as a relevant step towards the clinical and quantitative application of whole-spine weight-bearing CT imaging.