ABSTRACT
Research and development in mammography over decades has resulted in important improvements in this imaging modality. In addition, the introduction of digital detectors for mammography at the turn of the century allowed for further improvement in clinical performance (Pisano et al. 2005), along with a reduction in radiation dose (Hendrick et al. 2010). However, the performance of mammography is limited by one crucial aspect; it is a two-dimensional imaging modality used to image a three-dimensional pseudo-random pattern of tissue structure, resulting in tissue superposition. This gives rise to the concept of anatomical noise, which has a twofold negative consequence on the detection and diagnosis of breast pathologies. First, normal and abnormal features, such as a tumor, may superimpose on the projected image and make the pathology harder, or impossible, to visualize, resulting in loss of sensitivity. In addition, several normal features may project onto the same area, resulting in a 2D pattern that resembles a tumor, resulting in loss of specificity. Although the introduction of digital detectors to mammography did not directly ameliorate this problem, it allowed for the introduction of advanced breast imaging modalities that could address the issue of tissue superposition. One such modality is digital breast tomosynthesis (DBT), which produces a pseudo-3D image of the breast, consisting of a number of slices parallel to the imaging plane, in which features at certain vertical locations are preferentially in focus, while features at other locations are blurred out, as can be seen in Figure 20.1. Although DBT does not result in real 3D images, the preferential depiction of features located in the focal plane while blurring of off-plane features is enough to decrease the impact of anatomic noise and, therefore, improve the depiction of breast cancer. In large patient trials, it has been shown that DBT, when used as a replacement of or in addition to digital mammography (DM), results in an increase in clinical performance (Ciatto et al. 2013; Houssami and Skaane 2013; Skaane et al. 2013; Friedewald et al. 2014; Lång et al. 2016).
