Quadratic regularization design for 3D axial CT: Towards isotropic noise

Abstract

Potential advantages of statistical image reconstruction (SIR) methods over conventional filtered back-projection (FBP) method include reduced patient dose, improved noise and spatial resolution properties. However, the use of statistical weightings and sophisticated modeling of the system that are responsible for these improvements can lead to anisotropic and nonuniform noise characteristics and spatial resolutions. As an extension to our previous work that aimed for more isotropic and uniform spatial resolution, we propose a quadratic regularization design method for 3D axial X-ray computed tomography (CT) to achieve isotropy and uniformity of noise characteristics in the reconstructed image. In practice, noise properties may affect diagnostic quality of the reconstructed image as much as the spatial resolution. Simulations and a phantom experiment demonstrate that the proposed method leads to more isotropic and uniform noise characteristics in 3D axial CT with modest computational cost.