Statistical iterative reconstruction using adaptive fractional order regularization.

Yi Zhang,Feng Lin,Yan Wang,Weihua Zhang,Jiliu Zhou,Yifei Pu

doi:10.1364/boe.7.001015

Abstract

In order to reduce the radiation dose of the X-ray computed tomography (CT), low-dose CT has drawn much attention in both clinical and industrial fields. A fractional order model based on statistical iterative reconstruction framework was proposed in this study. To further enhance the performance of the proposed model, an adaptive order selection strategy, determining the fractional order pixel-by-pixel, was given. Experiments, including numerical and clinical cases, illustrated better results than several existing methods, especially, in structure and texture preservation.

Highlights

X-ray computed tomography has been widely used for clinical and industrial purposes
The parameters in total variation (TV)-SIR and edge preserving TV (EPTV) were set according to the suggestions in the original references [11, 19]
The results further showed the advantage of three iterative algorithms over the filtered backprojection (FBP) on noise suppression, as well as the advantage of the AFTV-SIR over the EPTV and the TV-SIR on edge/contrast preservation at the matched noise level

Summary

Introduction

X-ray computed tomography has been widely used for clinical and industrial purposes. the radiation risk accompanying with the increasing number of scans draws a lot of attention [1]. The famous as low as reasonably achievable (ALARA) principle is encouraged to evade excessive radiation dose in the clinical field. The major methods to reduce the radiation dose can be divided into two categories: the first one is to lower the x-ray tube current or shorten the exposure time (mAs) and the second is to reduce the photon number penetrating human body. The former will introduce quantum noise into projection data and the later will cause sparse view, limited angle, and interior CT. We only focused on the first category of methods and the proposed model is natural to extend to other topics

Methods

Results

Conclusion