Hyperspectral X-ray imaging for scatter removal in mammography

The purpose of this study is to investigate the use of hyperspectral X-ray imaging i.e, an imaging modality where spectral information on detected X-rays is available, for the removal of scatter in mammography. Features that suggest the presence of cancer are often low in contrast or small in size and therefore a good image quality is required in order to locate them. Current mammography systems use an anti-scatter grid to remove scatter from the image. However, this device also absorbs a proportion of the primary beam and therefore a rise in the dose is required in order to compensate for the loss in statistics. Two alternative methods of scatter removal were investigated in this study. Compton scattered X-rays lose energy in the scattering process and therefore appear at a lower energy in the detected spectrum. Imaging using a monochromatic X-ray beam and a spectroscopic detector means that these lower energy scattered X-rays can be removed from the image through spectral windowing. An alternative method of removing scatter is to simulate the scatter using Monte Carlo modelling. Once a scatter spectrum has been obtained using an input spectrum similar to the experimental spectrum it can be subtracted from the detected spectrum, creating a scatter-free image. This work presents imaging work carried out with a pixellated spectroscopic CdTe detector. A first approach involved the use of a mosaic crystal monochromator producing a quasimonochromatic spectrum from which the scattered component can be removed. The second approach involved the subtraction of the scattered spectrum, as obtained from Monte-Carlo modelling, from a full polychromatic spectrum. Both approaches were tested on a customdesigned low contrast test object. Results showed that in the monochromatic approach scatter removal gave a 40% increase in contrast. It was also found that removing scatter using a simulated scatter spectrum and a polychromatic beam produced a contrast improvement of around 15% when compared to full spectrum imaging.