Constrained Bayesian streak artifact reduction approach for contrast enhanced computed tomography imaging of the intervertebral disc

A promising approach for the study of progressive herniation damage of the intervertebral discs under flexion/extension motions as well as compressive loads is the use of contrast-enhanced computed tomography (CECT). One of the biggest limitations of using CECT is the presence of significant streak artifacts in the acquired tomograms, due primarily to the contrast agent injected into the intervertebral disc. To address this issue, a novel constrained Bayesian approach to streak artifact reduction in CECT imagery is introduced in this paper. The problem of artifact reduction is formulated as a constrained Bayesian estimation problem in projection space, and a non-parametric Parzen window estimation approach is employed to estimate the underlying posterior distributions. Experimental results show that the proposed approach provides significant artifact reduction while preserving the intervertebral disc regions to allow for clear visualization of progressive intervertebral disc damage.