Development and validation of a theoretical model for optimizing signal-to-noise ratio in x-ray fluorescence imaging

X-ray Fluorescence, Compton scattering, X-ray Physics

Chen, Yunlai ; Di Fulvio, Angela ; Meng ; Lingjian

关键词: X-ray Fluorescence, Compton scattering, X-ray Physics;   

Others  :  https://www.ideals.illinois.edu/bitstream/handle/2142/107943/CHEN-THESIS-2020.pdf?sequence=1&isAllowed=y

美国|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

There has been a growing interest in benchtop X-ray Fluorescence Computed Tomography (XFCT) due to its potential to non-invasively image non-radioactive targets at greater depths than optical techniques. However, this modality suffers from limited sensitivity due to the high scattering background and relatively low fluorescence yield. One direction of improving system sensitivity is by adjusting incident beam spectra shape. The predominant component of scattering noise originates from incident photons whose energy is several keV higher than fluorescence line. After Compton scattering with sample, those photons will lose energy to have the same energy with fluorescence photon, which leads to severe contamination of fluorescence peak on spectrum. This contamination can be suppressed by filtering those incident photons using proper material. This work aims to develop a theoretical model based on X-ray physics, which takes the incident spectrum and system geometry as input, to predict the signal-to-noise ratio (SNR) defined as ratio of fluorescence signal to scattering noise for given source voltage-filter configurations, from where we can pick out the configuration which gives the best SNR. In this model, we consider a small cylindrical volume containing a uniform solution of a given metal-of-interest in a tissue-equivalent solvent, which is irradiated by a narrow pencil-beam of X-rays operated at different voltage and coupled with different filters. We have also performed a series of experimental studies to validate this approach, which demonstrates a relative error less than 5% between analytical relative SNR and experimental relative SNR. The use of this theoretical model could greatly speed up the search for an optimum source-filter configuration for XFCT imaging. Further work includes validating this model in an imaging study, expanding the theoretical model to account for attenuation in thicker samples, and extending it to higher energies.