Evaluation of spatial precision and accuracy of cone-beam CT using an in vitro phantom model

High accuracy and precision are essential in stereotactic neurosurgery, as targeting errors can significantly affect clinical outcomes. Image registration is a vital step in stereotaxis, and understanding the error associated with different image registration methods is important to inform the choice of equipment and techniques in stereotactic neurosurgery. The authors aimed to quantify the test-retest reliability and stereotactic accuracy of cone-beam CT (CBCT) compared with the current clinical gold-standard technique (i.e., CT).

Two anthropomorphic phantom models with 40 independent unique steel spheres were developed to compare CBCT frame and stereotactic space registration with the clinical gold standard (CT). The cartesian coordinates of each sphere were compared between the imaging modalities for test-retest reliability and overall accuracy.

Both imaging modalities showed similar levels of fiducial deviation from the expected geometry. The equivalence test demonstrated mean differences between CT and CBCT registration of -0.082 mm (90% CI -0.27 to 0.11), -0.045 mm (90% CI -0.43 to 0.34), and -0.041 mm (90% CI -0.064 to 0.018) for coordinates in the x-, y-, and z-axes, respectively. The mean euclidean distance difference between the two modalities was 0.28 mm (90% CI 0.27-0.29).

Accuracy and precision were comparable between CBCT and CT image registrations. These findings suggest that CBCT registration can be used as a clinically equivalent substitute to gold-standard CT acquisition.