Repeatability and effect of different voxel sizes on linear and volumetric tooth and pulp measurements using cone-beam computed tomography

Abstract

Objective In cone-beam computed tomography (CBCT), voxel resolution and size directly affect image quality. This study aimed to analyze the effect of voxel resolution on the linear and volumetric measurements of tooth and pulp in three-dimensional images obtained with different voxel sizes and to test the repeatability of these measurements. Materials and methods CBCT images of 30 extracted single-rooted permanent teeth were obtained using 3 different voxel sizes (0.1-, 0.2-, 0.4-mm). Two observers performed volumetric and linear measurements of pulp and teeth using semi-automatic segmentation. The data were analyzed using descriptive and nonparametric tests. For interobserver reliability, the Wilcoxon signed test and Kendall’s tau correlation coefficient were used. The intraobserver repeatability was analyzed using the intraclass correlation coefficient. The significance level was set at 0.05. Results Friedman test showed no significant difference ( p > 0.05) in volumetric measurements between different voxels in one of the observers, but a significant difference was found in the pulp volume measurement in the other ( p < 0.001). The voxel sizes did not affect the linear measurements of both observers ( p > 0.05). The interclass reliability (Wilcoxon signed test and Kendall’s tau correlation coefficient) and intraobserver repeatability (intraclass correlation coefficient) decreased as the voxel size increased for linear and volumetric measurements of tooth and pulp ( p > 0.05). While the values indicated “good” and “excellent” reliability for all parameters measured on the 0.1- and 0.2-mm voxels, a few “moderate” reliability values were found on the 0.4-mm voxel. Conclusion Voxel sizes of 0.1- and 0.2-mm did not significantly affect linear and volumetric tooth measurements. When the voxel size increased, the reliability and repeatability decreased. To increase measurement reliability, the voxel size should be as small as possible. Supplementary Information The online version contains supplementary material available at 10.1186/s12903-024-05259-6.