SNR efficient 3D reconstruction algorithm for multi-source Inverse-Geometry CT system

The multi-source Inverse-Geometry CT(MS-IGCT) system uses a 2D array of sources opposite a smaller 2D detector array. One sample system design uses 3 rows of 21 sources each. Because the MS-IGCT system provides sufficient sampling in the axial direction, cone beam artifacts can be reduced. Projection data from the 21 sources at the same z-location can be rebinned into one cone beam projection, therefore, we can have 3 different cone beam projection data sets after rebinning, and reconstruction can be performed by using the FDK algorithm. However, if FDK is used, each of the three data sets by itself produces different cone beam artifacts. For example, the upper sources can provide artifact free images in the upper reconstruction volume, but cone beam artifacts can be observed in the central and lower reconstruction volume. The central and lower sources also provide artifact free image at different z-locations. We can achieve an artifact free volume by using artifact free images at different z-locations. However, if we could use all the data, the SNR can be improved. In this study, we develop a method to combine reconstructed volumes in Fourier space, and the main goal is to keep the exactness and improve the SNR in the combined image. The method was tested with a simulation of a Defrise phantom and the proposed method did not show cone beam artifacts. A noise simulation was also performed by using ideal bowtie filter so that all projection data had the same noise level. A noise simulation showed that the noise variance was ∼1/3 of that in a single FDK reconstruction.