Sparse Angular X-ray Cone Beam CT Image Iterative Reconstruction Using Normal-dose Scan Induced Nonlocal Prior

Repeated X-ray cone beam computed tomography (CBCT) scans are frequently entailed in some clinical cases, such as CT-guided lung lesions puncture examination, image-guided intervention and radiotherapy, which result in individual patient doses soaring due to the associative volume scanning. Sparse-views based image reconstruction, as an effective method for reducing the radiation dose in CBCT scans, has been extensively studied recently. Due to the huge anatomical information similarity of images from repeated scans, if given a full-views scan, the associative reconstructed images can be used as important priori information for image reconstruction from sparse-views data. With above observation, in this paper, we propose a normal-dose scan induced nonlocal prior (ndiNLM-Prior) for yielding accurate image from the sparse-views data with an iterative image reconstruction process. The present ndiNLM-prior can exploit the similar information from the images reconstructed from the full-views data without needing accurate image registration. Evaluations with the physical and digital phantom data clearly demonstrate that the presented method achieves higher image reconstruction accuracy in terms of streak artifacts suppression.