Pre-clinical evaluation and optimization of image quality for a Next Generation Tomosynthesis prototype

Abstract

A next generation tomosynthesis (NGT) prototype has been developed to investigate alternative scanning geometries for digital breast tomosynthesis (DBT). The NGT system uses a 2D plane as an address space for the x-ray source to define an acquisition geometry. In previous work, tests of physics have been used as objective metrics to evaluate image quality for NGT. In this work, the performance of custom NGT acquisition geometries is evaluated for mastectomy specimens to validate previous phantom experiments. Two custom acquisition geometries - incorporating T- and K-shaped source motion paths in the posteroanterior direction - were compared with a conventional DBT acquisition geometry. Noise power spectra (NPS) are calculated using 3D image reconstructions of the three acquisition geometries to evaluate the degradation of image quality due to noise and to visualize NGT sampling properties in the Fourier domain. NPS are used to describe features of the specimen image reconstructions and compare acquisition geometries. NGT acquisition geometries were found to improve high-frequency performance with isotropic super resolution, reduce out-of-plane reconstruction artifacts, and improve overall image reconstruction quality. The T-geometry combines the benefits of narrow- and wide-angle tomosynthesis in a single scan improving high-frequency spatial resolution and out-of-plane blurring, respectively.

Keywords: acquisition geometry; cone beam artifact; digital breast tomosynthesis; mastectomy specimen; noise power spectra; out-of-plane blurring; super resolution.

Figure 1:

Acquisition geometries of the NGT prototype.

Figure 2:

NPS of the three NGT acquisition geometries.

Figure 3:

Examples of different slices through the NPS volumes for each geometry. Zero-frequency NPS slice of the f_x – f_z plane (left). NPS slice of the f_y – f_z plane at f_x = 0 (middle). High-frequency NPS slice in the f_x – f_z plane at f_y = 2/3 • f_N (right).

Figure 4:

Example of the entire specimen and ROI for the conventional (A), T (B), and K (C) acquisition geometries. Image contrast settings are equivalent for all images.

Figure 5:

Clinical mastectomy reconstruction slices comparing conventional and NGT (T geometry). Specific features are highlighted in each image. NGT reconstructions show fewer out-of-plane reconstruction artifacts and improved demarcation of finer structures (e.g. Cooper’s ligaments and calcifications). NGT also improves contrast of structures within the glandular signal (top left arrow).