Effect of angular disparity of basis images and projection geometry on caries detection using tuned-aperture computed tomography.

Abstract

The purpose of this study was to determine whether projection geometry and angular disparity of basis images used for tuned-aperture computed tomography (TACT) slice generation influence observer performance in caries detection. Four sets of 8 projections of each of 40 teeth were acquired by using a digital sensor. Each set was radially distributed and subtended angular disparities of 10 degrees, 20 degrees, 30 degrees, and 40 degrees, representing strict projection geometries. A fifth set of images was acquired by using unconstrained geometry. TACT slices were generated from all experimental conditions and presented to 8 observers who viewed the images on a high-resolution monitor. Observers scored the presence/absence of caries with a 5-point confidence scale. Ground truth was achieved by histologic examination of tooth sections. Receiver operating characteristic curves measured observers' diagnostic performance. Analysis of variance was used to test for significant differences among observers and between experimental conditions. No statistically significant difference between angular disparities was found for the detection of either occlusal (P =.105) or proximal (P =.052) caries. No statistically significant difference between unconstrained and stringent projection geometries was found for the detection of either occlusal (P =.879) or proximal (P =.130) caries. Angular disparities ranging from 10 degrees to 40 degrees provide comparable performance in caries detection with TACT. Both unconstrained and stringent projection geometries may be used when reconstructing TACT slices for caries detection tasks.