Accurate volume image reconstruction for digital breast tomosynthesis with directional-gradient and pixel sparsity regularization

• Published in: Journal of medical imaging (Bellingham, Wash.) Vol. 12; no. S1; p. S13013
• Main Authors: Sidky, Emil Y., Wu, Xiangyi, Duan, Xiaoyu, Huang, Hailiang, Zhao, Wei, Zhang, Leo Y., Phillips, John Paul, Zhang, Zheng, Chen, Buxin, Xia, Dan, Reiser, Ingrid S., Pan, Xiaochuan
• Format: Journal Article
• Language: English
• Published: United States Society of Photo-Optical Instrumentation Engineers 01.01.2025; SPIE
• Subjects: Adjuvant treatment; Cancer; Peru; digital breast tomosynthesis; sparsity regularization; image reconstruction; dual-energy X-ray tomography
• DOI: 10.1117/1.JMI.12.S1.S13013
• ISSN: 2329-4302

We aim to develop accurate volumetric quantitative imaging of iodinated contrast agent (ICA) in contrast-enhanced digital breast tomosynthesis (DBT). The two main components of the approach are the use of a dual-energy DBT (DE-DBT) scan and the development of an optimization-based algorithm that can yield accurate images with isotropic resolution. The image reconstruction algorithm exploits sparsity in the subject's directional derivative magnitudes, and it also performs direct sparsity regularization to help confine the reconstruction to the true support of the subject. The algorithm is demonstrated with three sets of simulations in 2D and 3D, and a physical DE-DBT scan. The last of the three simulations employs an anthropomorphic phantom derived from the VICTRE project, testing quantitative tumor imaging with ICA. The 2D simulations of the algorithm demonstrate accurate and stable image reconstruction. With the first 3D simulation, the proposed algorithm shows the ability to resolve overlapping objects, and with the anthropomorphic phantom, accurate recovery of the irregular ICA distribution in the shape of a tumor model is demonstrated. Applying the algorithm to DE-DBT transmission data of the CIRS BR3D phantom with solid ICA inserts yields images in which the depth-blurring is greatly reduced and the ICA distribution is accurately reconstructed. The results for the sparsity regularization algorithm applied to DE-DBT show promise, but as the algorithm performance is necessarily subject-dependent, further investigation using subjects with varying complexity in the ICA distribution is required.