ABSTRACT
X-ray tomosynthesis is a pseudo-tomographic imaging method that can partially address the problem of structural overlap encountered in planar X-ray imaging (e.g., rib-lung overlap in chest X-ray, or normal breast tissue-focal lesion overlap in mammography) (see Section II, Chapter 20). In a typical X-ray tomosynthesis image acquisition, sequential two-dimensional (2D) projection images are recorded at different view angles within a limited angular span (e.g., ±30°), and the projection image dataset is processed with a limited-angle tomographic reconstruction algorithm to synthesize three-dimensional (3D) images of the image object (Figure 52.1). These tomosynthesis images can be considered as a stack of in-focus planes cutting through the object at different depths, thus alleviating the structural overlap issue. Compared with X-ray computed tomography (CT) that can record hundreds of projections over 360° in less than a second (see Section III, Chapter 32), the total angular range, the image acquisition speed, and the detector readout speed are greatly reduced in tomosynthesis imaging, which helps to lower its system cost. In addition, a single tomosynthesis image acquisition does not require multiple full tube rotations as in modern CT, and the detector can be kept stationary (see Section II, Chapter 23). Therefore, a typical tomosynthesis imaging system does not require a costly slip-ring gantry, as in CT. Instead, a tomosynthesis capable gantry can be quite similar to a standard planar X-ray imaging system with the addition of limited angular motion. Another major benefit of tomosynthesis is that the flat panel detectors used in tomosynthesis systems typically have spatial resolution about an order of magnitude better than a typical CT detector, leading to very high in-plane spatial resolution. On the other hand, due to the limited angular range of tomosynthesis, its spatial resolution along the depth (z) direction is generally inferior to that of true 3D imaging such as CT. As a result, tomosynthesis images are usually reconstructed with small in-plane pixel size (e.g., 100 μm), but a much thicker slice (e.g., 1 mm).
