Generalized filtered back-projection for digital breast tomosynthesis reconstruction

Abstract

Filtered backprojection (FBP) has been commonly used as an efficient and robust reconstruction technique in tomographic X-ray imaging during the last decades. For standard geometries like circle or helix it is known how to efficiently filter the data. However, for geometries with only few projection views or with a limited angular range, the application of FBP algorithms generally provides poor results. In digital breast tomosynthesis (DBT) these limitations give rise to image artifacts due to the limited angular range and the coarse angular sampling. In this work, a generalized FBP algorithm is presented, which uses the filtered projection data of all acquired views for backprojection along one direction. The proposed method yields a computationally efficient generalized FBP algorithm for DBT, which provides similar image quality as iterative reconstruction techniques while preserving the ability for region of interest reconstructions. To demonstrate the excellent performance of this method, examples are given with a simulated breast phantom and the hardware BR3D phantom.

Multigrid reconstruction with block‐iterative updates for breast tomosynthesis

“…We also find that the modified prior strength is not entirely effective for the patient cases. The range of the average noise levels is reduced from 4.1%-6.9% to 3.9%-4.4%, but these differences are clearly visible in the patient images, where the OSTR r (12) and OSTR r (25) reconstructions are considered too noisy for easy diagnosis.…”

“…After the parameter search, we compare the best MLTR pr and OSTR r multigrid sequences to four iterations of the MLTR pr algorithm with under-relaxation in the first two iterations 1 and four full iterations of OSTR r (25) (OSTR r with 25 subsets) or five full iterations of OSTR r (5) or OSTR r (12).…”

“…Following the restrictions listed in Sec. 2.D, we evaluated 8041 multigrid sequences for MLTR pr , 6339 for OSTR r (5), 4383 for OSTR r (12), and 2613 for OSTR r (25). Figures 4-7 show the histograms of the difference between the upper bound of the log-likelihood L max and posterior log-likelihood L, the sum of squared residuals between the reconstruction and the original phantom, the mean contrast of the calcifications in the phantom, and the standard deviation of the mean contrast of calcifications in three locations in the phantom, respectively.…”

“…The patient reconstructions with adjusted β were also evaluated by an experienced radiologist specialized in breast imaging. She found the SG OSTR r (12), SG OSTR r (25), MG OSTR r (12), and MG OSTR r (25) reconstructions to be identical from a diagnostic point of view, but all containing too much noise, making it harder to spot microcalcifications. The multigrid versions of MLTR pr and OSTR r (5) did improve the contrast of soft tissue structures and microcalcifications while maintaining a manageable amount of noise.…”

“…Filtered backprojection is the most common reconstruction method and was adapted for use in DBT by Mertelmeier et al, 9 with further refinement of the filter by Orman et al 10 Work by Ludwig et al 11 and Erhard et al 12 showed that it is possible to design reconstruction filters to include desirable features from iterative methods and more recent developments by Abdurahman et al 13 add statistical artifact reduction and supersampling to further increase image quality. In contrast to the direct inversion in FBP, iterative methods optimize a cost function, typically either to find the least squares solution as in SART (Ref.…”