Enhanced imaging of microcalcifications in digital breast tomosynthesis through improved image-reconstruction algorithms

PURPOSE: We develop a practical, iterative algorithm for image-reconstruction in under-sampled tomographic systems, such as digital breast tomosynthesis (DBT). METHOD: The algorithm controls image regularity by minimizing the image total $p$-variation (TpV), a function that reduces to the total variation when $p=1.0$ or the image roughness when $p=2.0$. Constraints on the image, such as image positivity and estimated projection-data tolerance, are enforced by projection onto convex sets (POCS). The fact that the tomographic system is under-sampled translates to the mathematical property that many widely varied resultant volumes may correspond to a given data tolerance. Thus the application of image regularity serves two purposes: (1) reduction of the number of resultant volumes out of those allowed by fixing the data tolerance, finding the minimum image TpV for fixed data tolerance, and (2) traditional regularization, sacrificing data fidelity for higher image regularity. The present algorithm allows for this dual role of image regularity in under-sampled tomography. RESULTS: The proposed image-reconstruction algorithm is applied to three clinical DBT data sets. The DBT cases include one with microcalcifications and two with masses. CONCLUSION: Results indicate that there may be a substantial advantage in using the present image-reconstruction algorithm for microcalcification imaging.

💡 Research Summary

The paper addresses a fundamental limitation of digital breast tomosynthesis (DBT): the acquisition of highly under‑sampled projection data due to a limited number of view angles. Conventional one‑pass reconstruction methods such as filtered back‑projection (FBP) assume complete angular sampling and therefore produce artifacts and loss of resolution when applied to DBT. To overcome this, the authors develop a practical iterative reconstruction algorithm that combines two complementary ideas: (1) image regularization based on the total p‑variation (TpV) functional, and (2) enforcement of physical constraints through projection onto convex sets (POCS).

TpV is a flexible regularizer that reduces to total variation when p = 1 and to an image roughness measure when p = 2. By varying p, the algorithm can balance edge preservation against noise suppression. The POCS framework imposes positivity of the reconstructed attenuation map and a data‑consistency constraint defined by a tolerance δ on the norm of the difference between measured and forward‑projected data. Because DBT is under‑sampled, many distinct volume reconstructions satisfy the same data‑consistency tolerance. Consequently, regularization plays a dual role: (i) selection of a unique volume among the infinite set that meets the data tolerance (volume‑selection role), and (ii) the conventional role of smoothing the image at the expense of exact data fidelity (regularization role).

The algorithm is derived from the earlier adaptive steepest‑descent POCS (ASD‑POCS) method but is simplified for clinical use. It alternates between a POCS step that projects the current estimate onto the convex set defined by the data tolerance and positivity, and a steepest‑descent step that reduces the TpV functional. An adaptive step‑size scheme controls the relative contribution of the two steps, allowing rapid convergence within 10–20 iterations and reducing the number of tunable parameters.

System modeling follows a typical DBT geometry: 11 equally spaced projections over a 50° arc, a flat‑panel detector with 100 µm pixel pitch, and a voxel grid of 0.1 mm × 0.1 mm × 1.0 mm (anisotropic voxels with higher in‑plane resolution). The system matrix M is constructed in a ray‑driven fashion, computing ray‑voxel intersections on the fly to avoid storing the massive matrix explicitly.

The algorithm was evaluated on three clinical DBT cases: one containing micro‑calcifications and two containing masses. In the calcification case, the TpV‑based reconstruction markedly improved the signal‑to‑noise ratio of the calcifications and made their morphology clearer compared with standard FBP. In the mass cases, lesion margins were sharper and background texture smoother. Experiments varying the p‑parameter showed that values near p = 1 (close to total variation) best preserve the high‑frequency features of calcifications while still suppressing noise.

Overall, the study demonstrates that controlling data error and image regularity independently is essential for under‑sampled tomographic systems. The proposed TpV‑POCS algorithm provides a practical, fast, and parameter‑light solution that yields superior image quality for DBT, especially for the detection of subtle high‑contrast features such as micro‑calcifications. The authors suggest that this framework could become a new standard for DBT reconstruction and may be extended to other limited‑angle X‑ray imaging modalities.