Model-Based Image Processing Algorithms for CT Image Reconstruction, Artifact Reduction and Segmentation

X-ray Computed Tomography (CT) has been widely used in medical diagnosis and security inspection. However, the more complex and advanced system hardware, as well as the higher demand of the low dose scan requirement, pose challenges for image reconstructions. For example, unlike the parallel-beam CT scanner, the recent developed multi-slice helical CT scanner has the more complicated 3D scanner geometry, and as cone angles become wider, there is an increasing need to use true 3D reconstruction methods in order to avoid the image artifacts introduced by 2D approximations.Model-based iterative reconstruction (MBIR) is a family of reconstruction methods under the model-based image processing framework, which has been shown to be effective in many imaging modalities, such as CT [1], PET, and MRI. These algorithms have the advantage that they can incorporate more detailed models of both the scanner and the objects being reconstructed. In addition, they offer more flexibility in various new applications since they allow for more accurate reconstruction for non-traditional geometries. Model-based algorithms have the potential to more accurately account for a wide array of scanner characteristics including photon counting and electronic noise, beam hardening, metal attenuation ans scanner, and the detector point spread function. More accurate modeling of the scanner can be used to reduce streak artifacts from high density objects. In addition, MBIR methods incorporates a prior model that can be tuned to the characteristics of typical objects and the performance metrics of interest.