A comparison of the effect of task-specific with fixed beam alignment in tuned-aperture computed tomography

To test the hypothesis that fixed beam alignment, representative of conventional, fixed-projection three-dimensional (3D) imaging systems, yields significantly lower detectability of approximal radiolucencies than does a task-specific strategy that 'opens the contacts'.

Eleven pairs of extracted teeth containing systematically varying numbers of induced approximal low-contrast defects were evaluated from 3D displays produced by means of tuned-aperture computed tomography (TACT) under controlled conditions. Each of the resulting tomographic presentations was displayed in two random sequences to seven trained clinicians. The task-specific (test) series was generated from eight different projections varying in vertical disparity but constrained to open the contacts. The fixed-aiming (control) series was generated from eight projections distributed uniformly on a circle.

Accuracy of defect detection averaged 5%, significantly higher with task-specific aiming than with fixed alignment projection geometry (P = 0.02). All observers unequivocally preferred the displays produced from projections with open contacts; however, they demonstrated considerable variation in their approach to the task, including the time required to interpret the displays.

The improvement in observer accuracy obtained from radiographs displayed in three dimensions results from an imaging strategy that selectively increases the signal-to-noise ratio in regions of interest.