Numerical generation of digital mammograms considering imaging characteristics of an imager

Abstract: Diagnosis and screening of breast cancer using digital mammography, breast tomosynthesis and cone-beam computed tomography are among the most active areas of research in diagnostic radiology. For a better design of imaging systems and the optimization of imaging technologies, numerical phantom studies are essential because of their convenience and the difficulties associated with acquiring clinical data. In this study, we have developed a procedure to numerically generate realistic digital mammograms considering the imaging characteristics of a digital mammography detector, such as resolution and noise. The procedure is composed of the generation of a voxelized breast phantom, projection and image modification. The outer shape of the breast and abnormalities are modeled with geometric functions. Soft tissue is generated by low-pass-filtered white-spectrum noise. Glandular ductal networks are generated by random angles and lengths with some restricted conditions based on anatomical information. A noise-free mammogram with the numerical voxel phantom is achieved by a simple ray-tracing method, and then modified with the known resolution and noise characteristics of an imager. This study will be helpful for the better design of mammography detectors and tomosynthesis algorithms. [Copyright &y& Elsevier]