Performance evaluation of patch group-based denoising algorithm using prior learning stage with K-edge imaging technique in CdTe photon counting spectral X-ray system: A Monte Carlo simulation study

Abstract

Photon counting spectral X-ray imaging system has many advantages compared with the conventional energy integrated X-ray imaging system. The photon counting room temperature semiconductor detector (PCRTSD) splits the energy spectrum into various energy ranges and can use the K-edge imaging technique for elemental composition of objects. However, despite its advantages, this system has limitations with respect to image noise. The aim of this study was to design a patch group-based nonlocal self-similarity prior learning denoising (PGPD) algorithm and to evaluate its image performance with K-edge imaging technique in the CdTe photon counting spectral X-ray imaging system. We simulated CdTe PCRTSD and X-ray source using Monte Carlo simulation using Geant4 Application for Tomographic Emission version 6, and applied the median filter and proposed PGPD denoising algorithms in the acquired phantom image. The denoising algorithm provided significant improvement in image performance (contrast to noise ratio and coefficient of variation) with K-edge imaging technique in PCRTSD.