Exploring the limits of limited-angle computed tomography complemented with surface data

Computed tomography investigations are challenging for planar objects, such as printed circuit boards (PCB), microprocessors packaging, etc. A conventional scan with a full rotation around a fixed axis may be impossible for such samples due to potential collisions with the system’s components or due to the limited penetration capability of the source and the finite dynamic range of the detect or. Analytical reconstruction techniques are proved unsuitable given the incomplete projection data resulting from limited source trajectory coverage. Iterative algorithms with prior information offer a solution to this reconstruction problem. In this contribution, we demonstrate that, by incorporating surface data in the reconstruction, better contrast homogeneity and more reliable sample boundaries can be reconstructed. A maximum likelihood expectation maximization (MLEM) algorithm is employed in this work. The surface data of the object, which is preregistered to the reconstructed object, is added into the objective function of the reconstruction as prior information. The values of the pixels outside the object ’s surface, extracted from the surface scanning, are penalized during the iteration process of the optimization. The proposed algorithm is applied to the limited angle projection data of simulated benchmark phantoms. The contrast of the reconstructed image, and the rooted mean square error between the phantom and the reconstructed image, are calculated to evaluate the performance of the proposed algorithm. Compared to the reconstruction without integrating the surface data, the reconstruction with complementary data gives an 80% higher homogeneity of the contrast, which is important to distinguish different materials and resolve fine features.