Reduction of Compton Background Noise for X-ray Fluorescence Computed Tomography with Deep Learning

Peng Feng,Yan Luo,Pan Huang,Bin Tang,Ruge Zhao,Xiansheng Zhao,Yonghui Li,Peng He

doi:10.3390/photonics9020108

Abstract

For bench-top X-ray fluorescence computed tomography (XFCT), the X-ray tube source will bring extreme Compton background noise, resulting in a low signal-to-noise ratio and low contrast detection limit. In this paper, a noise2noise denoising algorithm based on the UNet deep learning network is proposed. The network can use noise image learning to convert the noise image into a clean image. Two sets of phantoms (high concentration Gd phantom and low concentration Bi phantom) are used for scanning to simulate the imaging process under different noise levels and generate the required data set. Additionally, the data set is generated by Geant4 simulation. In the training process, the L1 loss function is used for its good convergence. The image quality is evaluated according to CNR and pixel profile, which shows that our algorithm is better than BM3D, both visually and quantitatively.

Highlights

Combining X-ray computed tomography (X-CT) with X-ray fluorescence analysis (XRF), X-ray fluorescence computed tomography (XFCT) is a novel method to detect earlystage cancer [1,2,3]
A noise2noise model based on a deep learning framework is used to denoise XFCT images with different noise levels
The model is based on pinhole XFCT imaging modality and does not need a clean image as the ground-truth image

Summary

Introduction

Combining X-ray computed tomography (X-CT) with X-ray fluorescence analysis (XRF), X-ray fluorescence computed tomography (XFCT) is a novel method to detect earlystage cancer [1,2,3]. In 2010, Cheong proposed the first benchtop XFCT imaging system with an X-ray tube source and verified its feasibility in preclinical applications [6]. Deng et al used a conventional X-ray tube source to detect the distribution of GNPs in mouse kidneys [8]. The conventional X-ray tube is bremsstrahlung, causing a lot of Compton background noise, which causes low signal-to-noise-ratio and low detection limits of contrast agents [2]. In order to improve the image quality and detection accuracy of XFCT, Compton background noise removal is an urgent problem to be solved. TThheisinmpeutthaonddotnhlye ntaeregdest oanreenscoaisnyainmdadgoese.s not need to produce clean datasets Both the input and the target are noisy images. TThheenn tthhee ttoottaall fflluuxx rraattee,, II,, ooff tthhee flfluuoorreesscceenntt XX--rraayy rreeaacchhiinngg tthhee ddeetteeccttoorr iiss:: I.

Noise2noise Model

Datasets

Results

Conclusions