Influence of the number of basis images and projection array on caries detection using tuned aperture computed tomography (TACT)

to determine if the number of basis images and spatial distribution of the projection array used for TACT slice generation influence observer performance in caries detection. In the first experiment, 2, 4, 8 and 12 basis projections of each of 40 teeth were acquired using a CMOS digital radiography sensor. Projections were distributed radially in space using a 20 degree angular disparity. TACT slices were generated from the four subgroups of images, presented to eight observers, and viewed on a high-resolution monitor. Observers scored the presence/absence of caries using a 5-point confidence scale. Gold standard was histological examination of tooth sections. ROC curves measured observer diagnostic performance. ANOVA tested for significant differences between observers and experimental conditions. In the second experiment, the number of basis projections judged to be satisfactory for TACT slice generation was used. Horizontal and vertical linear arrays of projections were compared to the circular projection array. There was a statistically significant difference between the numbers of basis projections in the detection of both occlusal (P=0.006) and proximal caries (P=0.005). No significant difference was found between projection arrays in the detection of either occlusal (P=0.065) or proximal (P=0.515) caries. The number of TACT basis projections significantly influences caries detection. Eight or more images should be used. Either linear-vertical, linear-horizontal or circular arrays of basis projections may be used for TACT slice generation in caries detection tasks.