Linear Measurement Accuracy of Eight Cone Beam Computed Tomography Scanners

Information regarding linear accuracy is necessary for efficient treatment evaluation, especially for maxillofacial reconstruction or implants.

To investigate the accuracy of linear measurements from multiple cone beam computed tomography (CBCT) devices.

A RANDO® phantom was scanned with eight CBCT scanners (11 modes). The viewing software accompanying each scanner was employed for measurements in mediolateral, anteroposterior, and supero-inferior dimensions by two dentomaxillofacial radiologists. Digital caliper measurements were used as a "gold standard." ANOVA with Scheffé post hoc analysis and intraclass correlation coefficient (ICC) were utilized for statistical analyses. The level of confidence was 95%.

Differences from the gold standard among 11 acquisition modes were statistically significant (p < .001). Measurements from one unit were always underestimated compared with all others (p < .001). The range of absolute measurement errors for tested units was -2.56 to 0.54 mm (mean ± SD 0.45 ± 0.71) including the outlier and -0.34 to 0.54 mm (0.16 ± 0.11) excluding the outlier. Slightly more values were underestimated than overestimated (41 of 66 measurements, 7 out of 11 CBCT modes). ICC scores for inter- and intraobserver agreement were perfect (1.000).

Treatment planning from large-volume CBCT was found to be reliable in all except one of the investigated scanners. New CBCT scanners should always be tested for geometric accuracy.