The Research About 4D-CBCT Spare-view Reconstruction Based On Deformable Registration

With the rapid development of medical imaging technology, image-guided radiation therapy(IGRT) is one of the main means in tumor radiotherapy. CBCT based IGRT technology combines CBCT imaging with radiotherapy technology, obtaining real-time information of target area while achieving treatment, according to the CBCT images, making registration between CBCT and planning CT, calculating the motion error of the organization in inter-fraction and intra-fraction, making correction to the change of the target location, the shape, and the dose distribution timely. Therefore, 4D-CBCT imaging technology is an important factor to ensure the accuracy of IGRT.Since CBCT utilizes X-ray imaging technology, in order to reduce the harm to patients, lowering CT radiation dose is the important part in the medical imaging field. A effective method of reducing the radiation dose is to scan tumor with sparse-view, but it brings a problem of sparse-view reconstructing. Recent years, compressed sensing(CS) theory causes widespread concerns in the medical field, the nature of the CS theory is different from the traditional signal reconstruction algorithm, it breaks the requirements of Nyquist theorem, it can reconstruct high quality image accurately only with a small amount of data, not only the reconstruction of sparse-view problem is solved but also the high quality and low dose requirements of modern medical imaging is satisfied.This paper takes advantage of compressed sensing, combines it with the registration algorithm, proposes a 4D-CBCT image reconstruction algorithm based on deformable registration: all the projections are grouped into different respiratory phase bins according to the breathing signal tagged on every projection image, a new CBCT image for each breathing phase is then reconstructed by using registration of planing CT and CBCT images, using the total variation minimization method based on compressed sensing to reconstruct CBCT. The image registration of CT and CBCT is the initial images, using NEATA iterative reconstruction algorithm, obtaining the require image for each breathing phase, merging them into a 4D-CBCT image finally.The validity of the proposed algorithm is proved through two dimensional Shepp-Logan images and clinical cases simulation experiments separately. Experimental results show that when the CBCT projection data noise is serious, compared with traditional FDK reconstruction algorithm and CS reconstruction algorithm, R-CS reconstruction method can reduce image noise and reduce artifacts better, achieving accurate reconstruction.Therefore, the proposed method based on registration and compressed sensing could use the sparse-view projection data reconstruct the original image accurately, improving the efficiency and the accuracy of IGRT.