Speaker
Description
Gain variations of detector pixels in cone-beam computed tomography (CBCT) may lead to streaking artifacts in sinogram and ring artifacts in reconstructed CBCT images. Such gain variations can be mainly caused by inconsistent response of detector pixels owing to their defects and aging. This study presents an effective method to identify and correct streaking artifacts in sinogram using a specific-artifact detection and correction scheme for obtaining artifact-free projections, followed by the filtered-backprojection reconstruction. In the proposed scheme, the total variation of the artifact-contaminated sinogram is minimized with two l1-norm-based sparsity constraints of the artifacts using a primal-dual hybrid gradient-based optimization process. To verify the efficacy of the proposed approach, we conducted a semi-experiment on a clinical CBCT projections emulated with various detector pixel gain variations in the range of 5‒98% of normal detector gain and investigated the image quality. According to our preliminary results, the proposed method significantly reduced streaking artifacts in sinogram and thus ring artifacts in CBCT images, demonstrating its effectiveness. More systematic and quantitative results will be discussed in the presentation.
