CT image resolution improvement by geometric modelling in iterative reconstruction

In high resolution CT image reconstruction, a proper geometrical model is beneficial for achieving high quality images. This work aimed to improve CT image resolution with OSC (Ordered Subset Convex) reconstruction algorithm by incorporating a geometric model for the finite-size of CT detector pixels. We propose two different geometric modelling approaches, denoted as R-GM and T-GM to collaborate with OSC reconstruction. R-GM and T-GM methods respectively approximate the voxel response function on the detector pixel by a rectangular function and a trapezoid function. To evaluate the two proposed approaches, we simulated a small animal CT system using an analytical approach. A line-pairs phantom was simulated with multi-sampling approach and Siddon´s algorithm. 360 projections were generated with 1o interval. Image reconstruction results show that the geometric model implemented by R-GM method prominently improves image resolution. No significant improvement is observed in TGM method based image. Future work will focus on further development of T-GM method and further improvement in image resolution b modelling the effect of x-ray focal spot area and x-ray tube motion during exposure.