Research Of4DCBCT Reconstruction Algorithm Based On Compression Sensing

In diagnosis, localization and treatment of cancer, medical imaging is one of thekey links. Image-guided radiation therapy, referred to as IGRT, is a newfour-dimensional radiotherapy technology. IGRT system based on CBCT can pass2d/3dimage registration for patients with preoperative three-dimensional CT volume datasand intraoperative X-ray projection images to maximize the elimination of repeatedpositioning errors, so that the preoperative set can be used in intraoperative treatment.However, for the position of the chest and abdomen tumor, respiratory motion blurringmakes registration reconstruction algorithm in such cases can’t get high image quality,which greatly limits the use of CBCT in IGRT. And with CT radiation damage causedpeople’s concern, the medical community made low-dose CT imaging, high-quality newrequirements. There is a need to improve the existing CBCT image reconstructionalgorithm in IGRT.CS theory is a new theory in signal acquisition and processing field, which has alsoreceived widespread attention in medical imaging field. CS theory has broken theNyquist sampling theorem. When compared with the traditional medical imagereconstruction algorithm, the CS image reconstruction algorithm can use lessprojections to obtain high quality reconstructed image, and that is well positioned tomeet the development needs of high-speed, low-dose, high-quality medical imaging.With in-depth study of the existing reconstruction of medical image registration and CStheory, this article presents a new method of PICCS based on registration, whichintegrated with the advantages of registration and CS reconstruction algorithm. Usingthe prior image obtained from the registration and a sparse priori knowledge of CTimage’s gradient field, the new method can use a small number of projections toreconstruct the high quality images in lung tumor motion.MATLAB simulation using a two-dimensional digital phantom experiments andrespiratory motion time-series images experiments to verify the effectiveness of theproposed algorithm. Experimental results show that in every phase of imagereconstruction using20projection data from different angle, the proposed algorithm cansignificantly improve the reconstruction image of PICCS indicators, when comparingwith CS reconstruction algorithm and the PICCS reconstruction algorithm; the proposedalgorithm is first approaching convergence when these algorithms have iterated10 times. Therefore, this article proposed algorithm can use a small amount ofundersampled projections more accurately reconstruct the original image, and it willhave good clinical application prospects.