Noise level and no-reference-based evaluations for gamma-ray image using non-local means algorithm with CZT photon counting semiconductor detector.

The purpose of this study was to propose and evaluate an algorithm that maximizes the image quality of gamma-ray images using a cadmium zinc telluride (CZT) photon-counting semiconductor detector (PCSD) under thin detector thickness conditions. In addition to the CZT PCSD, a pixel-matched parallel-hole collimator that can optimize the spatial resolution of gamma-ray images was modeled. A non-local mean (NLM) noise reduction algorithm was applied to the acquired images using Geant4 Application for Tomographic Emission platform to quantitatively evaluate the overall image quality improvement. When the proposed source-to-pixel-matched collimator distance was shortened, a thin CZT PCSD (1 mm) was selected, and the NLM algorithm was applied to the acquired image to obtain a full width at a half maximum value of 0.957 mm. We demonstrated that the spatial resolution was improved by approximately 40.89% compared to when using a 3-mm-thick PCSD at the same source-to-collimator distance. In addition, the contrast-to-noise ratio and coefficient of variation of the image acquired from the system applying the proposed NLM algorithm were derived to be almost similar to those of the 3-mm-thick detector system. We demonstrated that the proposed approach based on the NLM algorithm is a PCSD gamma-ray imaging technology that is capable of reducing costs and improving image quality.