Until recently, CT images were reconstructed with filtered back projection (FBP), a simplified calculation method based on several assumptions regarding the X-ray source and beam [1, 2]. FBP is a fast reconstruction technique, but its main limitation is that it generates excessive image noise when attempts are made to reduce the radiation dose. Fortunately, over the past few years, the increase in computational processing capabilities has allowed the routine use of iterative reconstruction in CT images.
With iterative reconstruction (IR), a correction loop is introduced: starting from an initial conventional image reconstruction, synthetic projections that exactly represent the reconstructed image are calculated. True projections measured by the CT system and synthetic projections are compared to identify differences generated by noise and to improve the original image. This process can be repeated several times, as predefined by the user, so that image noise can be reduced without losing spatial resolution. At first, a simple algebraic reconstruction technique was used, but during the past 10 years vendors have developed different statistical algorithms (operating in either the raw-data domain or the image domain), including the complex and exhaustive model-based IR (MBIR) which was recently released. General Electric’s latest IR algorithm ASiR-V, Philips’ iDose4, Siemens’ ADMIRE and Toshiba’s AIDR 3D are used today in routine CT image reconstruction.