Experimental phantom lesion detectability study using a digital breast tomosynthesis prototype system

To compare the sensitivity of conventional two-dimensional (2D) projection imaging with tomosynthesis with respect to the detectability of mammographic phantom lesions.

Using a breast tomosynthesis prototype based on a commercial FFDM system (Siemens MAMMOMAT Novation), but modified for a wide angle tube motion and equipped with a fast read-out amorphous selenium detector, we acquired standard 2D images and tomosynthesis series of projection views. We used the Wisconsin mammographic random phantom, model RMI 152A. The anode filter combinations Mo/Mo and W/Rh at two different doses were used as typical radiographic techniques. Slice images through the phantom parallel to the detector were reconstructed with a distance of 1 mm employing a filtered back-projection algorithm. The image data sets were read by five radiologists and evaluated with respect to the detectability of the phantom details.

For all studied radiographic techniques, the detection rate in the tomosynthesis mode was 100 %, i. e. 75 true positive findings out of 75 possible hits. In contrast, the conventional projection mode yielded a detection rate between 80 and 93 % (corresponding to 60 and 70 detected details) depending on the dose and X-ray spectrum.

Tomosynthesis has the potential to increase the sensitivity of digital mammography. Overlapping structures from out-of-plane tissue can be removed in the tomosynthesis reconstruction process, thereby enhancing the diagnostic accuracy.