A controlled evaluation of tuned-aperture computed tomography applied to digital spot mammography

The purpose of this work was to compare the detection accuracy of 3-dimensional (3D) modalities of tuned-aperture computed tomography (TACT) with that of conventional 2-dimensional (2D) digital spot mammograms. A standardized mammographic phantom was placed beneath cadaveric breast tissues of varying densities. Five radiologists were asked to detect as many objects (specks, fibers, and low-contrast masses) as possible from 90 displays in a controlled and factorially balanced multivariate experiment. Radiographic exposure was varied systematically, and projections were averaged to ensure stochastic comparability. Scores were weighted to eliminate task-specific bias and were analyzed by multivariate analyses of variance. All display modalities based on the linear application of the 3D TACT reconstruction method yielded significantly higher detection scores for all tasks than did conventional 2D digital spot mammography, which served as the scientific control modality. This effect was found to be statistically significant (P < .001) in spite of significant variations between tissues (P < .001), observers (P < .001), and exposures (P < .01). TACT may be a promising alternative or enhancement to conventional 2D digital mammography for tasks well simulated by this experimental design.