Study On Methods For Low-dose CBCT Reconstruction

Computed Tomography(CT)has been widely used in clinical medicine.CT can obtain the internal structure information without causing damage to the object.CBCT is becoming an important tool in the image guided radiation therapy because of its flexible scanning,high spatial resolution and short acquisition time.Although CBCT offers great help to many doctors,the exposure of normal tissues to high radiation during the acquisition of CBCT projection data increases the lifetime risk of cancer and genetic defect.Low-dose CT reconstruction is deserving of research.Statistical-based iterative CBCT reconstruction algorithms have shown potential to improve low-dose image quality.This paper proposed a novel penalty and a novel approach for CBCT reconstruction.TV penalty has demonstrated meritorious performance in suppressing noise and preserving edges.However,it will lead to the so-called staircase effect.Hessian penalty has been successfully used to replace TV for suppressing the staircase effect in CBCT reconstruction,with extra cost of slightly blurring object edges.In this study,we proposed a new penalty — the TV-H penalty — which combines TV and Hessian for CBCT reconstruction in a structure adaptive way.The proposed penalty retains favorable properties of TV like suppressing noise and preserving edges,and also has the ability in better preserving structures in regions with gradual intensity transition.Typical measurement models ignore blurring effects,and nearly all current approaches make the presumption of independent measurements.In some imaging systems,such as flatpanel-based CBCT,such correlations and blurs can be a dominant factor in limiting the maximum achievable spatial resolution and noise performance.In this work,we propose a novel reconstruction method that includes models for both detector blur and correlated noise in the projection data.We investigate the performance of proposed methods in both simulated data and in CBCT test-bench data.The results demonstrated the superiority of the TV-H penalty in depicting the structure of the images and the potential of the correlated noise model in improving spatial resolution.