Metal artifact reduction in kVCT images via L1 sparse regularization of MVCBCT prior images

摘要

Purpose: This study proposes a new metal artifact reduction (MAR) method in kilovoltage computed tomography (kVCT) images from tumor patients under radiotherapy with megavoltage cone beam computed tomography (MVCBCT) images as priors. Methods: MVCBCT images are used as prior images to reduce metal artifacts in kVCT images. The reconstruction of kVCT images is modeled as a total variation that considers the sparsity of difference between the sinogram data gradients of kVCT and MVCBCT images. Results: Experiments show that the proposed approach can suppress metal artifacts. Segmentation errors, which are introduced in the tissue classification step in traditional L1 and normalized MAR methods, are also avoided because MVCBCT images are used as prior images in addition to the segmented kVCT images. The proposed approach achieves lower normalized root mean square errors and higher correction coefficients globally than the values obtained by the existing methods. Conclusions: The proposed approach suppresses the large deviation in computed tomography numbers in the boundary region of metal fillings.