In the framework of breast imaging, in contrast to mammography, commonly used to diagnose breast cancer, novel three-dimensional x-ray based imaging technologies, such as Digital Breast Tomosynthesis and breast-dedicated Computed tomography eliminate the problem of overlying tissues in conventional 2D images. Thus, they significantly improved the efficiency of large scale screening programs. Despite that, the small difference in x-ray attenuation, in particular between the glandular and tumour tissue, still constitutes a major problem in these methods. In recent years, the development of x-ray phase-contrast imaging techniques, which are able to measure the effects of x-rays refraction in the body, have shown promising results for refining breast cancer diagnosis. These techniques permit the visualization of soft-tissue structures that are not detectable by use of conventional x-ray radiographic methods, and also hold the potential to reduce the radiation dose delivered to the patient. Among the different phase-contrast techniques, the propagation based phase-contrast CT (PB-CT) method is the one analysed in this work. PB-CT does not require the use of any additional X-ray optical element, it is therefore the easiest method to implement.
Up to now the required level of spatial coherence of the incident x-ray beam used in PB-CT has practically limited its application to synchrotron facilities. However, the optimization of PB-CT for breast imaging will provide practical guidelines to evaluate novel generator-based phase-contrast setups (including analyser-based imaging, edge-illumination, and grating-based imaging) or compact sources.
The goal of this work is to evaluate how experimental conditions and reconstruction parameters affect the image quality in breast PB-CT and establish, from a quantitative point of view, the optimum combination of acquisition (distance, energy and detector) and reconstruction (δ∕β ratio and algorithm) parameters. Image quality of reconstructed images was quantitatively assessed measuring objective characteristics such as noise, Contrast-to-Noise ratio, edge sharpness and other indices considering also the delivered dose. The influence of each parameter was systematically evaluated, while the others were kept fixed.
We showed that the best images are obtained where the phase contrast effects are better exploited, using larger propagation distances, applying a phase retrieval algorithm and using iterative reconstruction algorithms rather than the conventional FBP one.
In general we showed that, with an adequate parameters optimization it is possible to obtain a good image quality even with very low radiation doses.