Geometry-Consistent Discretization of Local Correlation Equations for the Radon Transform in Clinical CT
Authors
Abstract
In this work, we extend the local correlation equation (LCE) framework of the 2D Radon transform to clinical computed tomography (CT) geometry and propose a set of geometry-consistent discretization schemes. Building upon the dual-center-of-rotation formulation, the 1st and 2nd LCEs are reformulated under circular fan-beam geometry and discretized using finite difference methods aligned with clinical acquisition configurations. The numerical validity of the proposed discretized circular LCEs (cLCEs) is evaluated using projection data generated from clinical CT datasets. Quantitative analysis based on residual error, mean absolute error (MAE), and mean percentage error (MPE) demonstrates strong consistency between both sides of the equations, with the 2nd cLCE exhibiting the highest accuracy. Furthermore, the proposed cLCE framework is applied to sparse-view sinogram interpolation. By refining interpolated projections using a gradient-descent-based scheme without additional regularization, improved reconstruction quality is achieved. Experimental results show that the cLCE-refined projections significantly enhance image quality compared to sparse-view reconstruction and can outperform clinical complete view (CCV) reconstruction under certain conditions. These results demonstrate that the proposed cLCE discretization provides an effective tool for exploiting projection redundancy in clinical CT and has strong potential for sparse-view reconstruction applications.