Anatomical accuracy of hybrid SPECT/spiral CT in the lower spine

Abstract

Aim: The anatomical accuracy of hardware-based registration of skeletal single photon emission computed tomography (SPECT) and X-ray computerized tomography (CT) has as yet not been studied. The aim of this study was to evaluate this variable in the lower spine for a newly introduced hybrid SPECT/spiral-CT camera.

Methods: In 22 patients referred for degenerative joint disease or tumours, whole-body bone scintigraphy including hybrid SPECT/spiral CT of the lower spine was performed. Subsequent analyses were performed on these pairs of images as well as on data sets obtained after using a rigid automated fusion procedure in addition. Two observers independently measured the distances between the visually determined centres of gravity of the CT and SPECT representation of the fourth and fifth lumbar vertebral body in the X-, Y- and Z-directions (X-, Y- and Z-distances).

Results: The distances determined by the two observers for the two vertebral bodies correlated significantly and were averaged for further analysis. For hybrid SPECT/spiral CT without consecutive automated registration, the mean X-, Y- and Z-distances were 1.6+/-1.9 mm, 1.7+/-1.3 mm and 0.9+/-0.5 mm, respectively. Additional automated registration lowered these values to 1.2+/-0.9 mm, 1.1+/-0.7 mm and 0.8+/-0.4 mm, respectively. The difference for the Y-distance proved statistically significant (P<0.05). Additional automated registration significantly reduced the number of subjects in whom at least one of the distances determined was greater than the SPECT pixel size of 4.6 mm from 14% (n=3) to 0% (P<0.05).

Conclusion: Hardware-based fusion between skeletal SPECT and CT offers a nearly perfect data match in the lower spine. The additional use of a tool for automated rigid registration has the potential to reduce the error of alignment even further and may be useful in patients with reduced compliance leading to movements between the two examinations.