Optimization of breast tomosynthesis: Computer simulations of image acquisition and glandular dose

Breast tomosynthesis (BT) is an X-ray imaging technique recently introduced as an alternative or complement to standard digital mammography (DM) in breast imaging and breast cancer screening. In BT, a set of projection images is acquired over a limited angular range and reconstructed into a volume of slice images. The method includes many possible combinations of acquisition parameters that have to be optimized for best possible clinical performance and outcome. The visibility of breast cancer lesions is important in this context. Compared to DM images, the reconstructed BT volume provides additional information on depth, reducing the superposition of breast tissue, which may hide true lesions or appear as false positive findings. Thus, the BT volume also contains information about the distribution of dense tissue within the breast, which is of interest when estimating the radiation dose from DM and BT exposure.
In this thesis, a simulation procedure was developed for the optimization of image acquisition and estimation of individual glandular dose in BT. The procedure was shown to be useful in generating BT images with realistic sharpness, though with higher image noise and contrast than experimentally acquired images (Paper I). The procedure was used to investigate the influence of angular range, distribution of projection images, and dose distribution on simulated microcalcifications in reconstructed BT volumes. Image acquisitions with very high central dose yielded significantly lower visibility than acquisitions with more uniform dose distributions, and the depth resolution increased with wider angular range (Paper II).
A method for localizing dense tissue from reconstructed BT volumes was verified using the simulation procedure (Paper III). A prototype software program was used for automatic and objective estimation of breast density in BT, with similar performance as DM (Paper IV). Using software breast phantoms recreated from reconstructed BT volumes, the glandular dose could be estimated with good overall accuracy for breast phantoms with different amounts and distributions of dense tissue (Paper V).
The developed simulation procedure has been a useful tool for optimizing acquisition parameters and estimating glandular dose in BT. The procedure could potentially be developed for further evaluation of the imaging chain and estimation of individual glandular dose in human cases.