Alignment correction during metal artifact reduction for CBCT using mutual information and edge filtering

In cone-beam computed tomography (CBCT) images, much like in computed tomography (CT) images, high-density objects, especially metal parts, generate streak-like artifacts. We present a novel method for metal artifact reduction in CBCT images via virtual replacement of the metal objects in the 3D volume with objects of identical geometry but water like x-ray attenuation coefficient. The algorithm, described in this article, is designed for mobile C-arm systems. As the 3D scan procedure for CBCT of mobile C-arm systems is not completely reproducible in its projection geometry and computed forward projections are elementary for this method an additional correction based on mutual information, using an edge filter as preprocessing step, must account for these misalignments. A reconstruction of the adapted 2D projection images generates a second 3D volume, where the originally metal objects now have a density like water and the streak-like artifacts are clearly reduced. Last step is to transfer the segmented metal parts of the first 3D volume into the metal and artifact free 3D volume of the second reconstruction. The proposed method is applied to clinical images and shows superior performance. The resulting reconstructed images show much reduced streak-like artifacts and related shadows.