Purpose: In breast imaging, microcalcifications detected in a screening exam can be characterized with better detail in magnification mammography. In this exam, the breast is elevated closer to the x-ray source, and a 2D image is acquired. Unlike clinical systems, the next-generation tomosynthesis (NGT) system built in our lab is capable of 3D magnification mammography. We are also designing the system to generate super-resolution (SR) reconstructions. In our earlier work modeling the tomosynthesis acquisition in screening mammography, we found that there are anisotropies in SR at various positions in the image. This study analyzes how the anisotropies in SR vary with the use of magnification.
Methods: A high-frequency sinusoidal test object was analyzed with theoretical modeling. The input frequency was oriented in the direction of x-ray source motion. Image quality in the reconstruction was quantified using the Fourier transform. The amplitude of low-frequency information was measured relative to the input frequency; this ratio should be as small as possible to achieve SR. We used these calculations to identify anisotropies in SR at various magnifications. For experimental validation of the anisotropies, a bar pattern phantom was analyzed with the NGT system.
Results: In screening mode, there are anisotropies in SR in the direction perpendicular to the breast support; i.e., in various slices of the reconstruction. With theoretical modeling, we demonstrated that the spacing between anisotropies is reduced with increasing magnification; however, the relative amplitude of low-frequency (aliasing) signal at the anisotropies is not as pronounced as in screening mode. The anisotropies were validated experimentally with the NGT system.
Conclusion: SR can be achieved in tomosynthesis reconstructions. Using the conventional acquisition geometry, anisotropies in SR exist in both screening and magnification tomosynthesis. Our future work will use this model of SR to optimize the acquisition geometry for magnification tomosynthesis.
Funding Support, Disclosures, and Conflict of Interest: A.M. is the spouse of a shareholder of Real Time Tomography (RTT). P.N. receives support from Philips Healthcare and is consultant to Stryker. Support was provided by grants W81XWH-18-1-0082 (DoD), IRSA 1016451 (Burroughs Wellcome), R01CA196528 (NIH), and IIR13264610 (Komen). Equipment support was provided by Analogic Inc., Barco NV, and RTT.