Page 37 - Craniomaxillofacial Implant Surgery - Jeroen P.J. Dings

CBCT
MDCT
0.39 0.42 0.21 0.26 0.28
P = .42
P = .03
0.53 mm (0.36-0.71) 0.39 mm (0.19-0.59) 0.57 mm (0.45-0.70) 0.59 mm (0.45-0.73)
P < .0001 P = .0003 P < .0001 P < .0001
Caliper (gold standard) 0.99
Reliability and accuracy of image-guided craniofacial implant planning
reference standard. A 2-way analysis of variance (ANOVA) was used to compare the measurements from CBCT and MDCT images at two different contrast settings. The
level of significance was 5%. The intraclass correlation coefficient (ICC) was calculated
with a variance component analysis for the physical and radiological measurements 2 to determine the intra- and inter-observer agreement. Comparisons of interobserver
variability between methods were done using the likelihood ratio test comparing the model with equal variability with the model with different variabilities for the methods. The statistical analyses were performed using SAS system software, 9.4 release.
RESULTS
The reference standard as mean of physical measurements had an average inter- observer error of 0.15 mm. The reliability (ICC) indicates a very good repeatability of the radiological and physical measurements (Table 1)46. The inter- and intraobserver reliability were similar for the CBCT and MDCT measurements and ranged between 0.98 and 0.99.
Table 1. Reliability (ICC), inter- and intra-observer measurement error for CBCT, MDCT measurements at different contrast settings and caliper measurements. P-value indicates significance between two different contrast settings for each imaging modality.
Reliability (ICC)
CBCT 0.3 mm 0.98 (Procera)
CBCT 0.3 mm 0.98 (Maxilim)
MDCT 0.5 mm 0.99 (Procera)
MDCT 0.5 mm 0.99 (Maxilim)
Inter-observer measurement error (mm)
Intra-observer measurement error (mm)
0.35 0.42 0.21 0.22 0.26
Systematic difference with gold standard
(observers pooled; 95% confidence interval)
Overview of the absolute difference between CBCT and MDCT linear measurements versus the gold standard expressed in millimeters. The mean and 95% confidence interval are depicted.
Interobserver measurement error on MDCT images was statistically significant influenced by different brightness and contrast setting (P = .03). After pooling of the CBCT and MDCT images a statistically significant difference in interobserver measurement error is found of 0.40 mm and 0.24 mm, respectively (P <.0001).
35