اصلاح، كمي سازي و بهينه سازي داده هاي سيستم تصويربرداري ميكروپت در تحقيقات پيش باليني

Preclinical PET , Quantification , Attenuation an‎d Scatter correction , Deep learning , Spatial resolution

Correction, quantification an‎d optimization of microPET imaging system data in preclinical research.

مهندسي هسته اي

This study comprehensively investigates the correction, quantification, an‎d optimization of data from the PET-Xtrim imaging system, the first positron emission tomography scanner developed at Tehran University of Medical Sciences for preclinical studies. The primary objective of this study is to correct physical an‎d systematic effects that compromise image quality an‎d quantitative accuracy in small-animal PET imaging, thereby achieving more accurate quantification of true radiotracer activity in PET images. To correct quantitative factors in the PET-Xtrim system, a comprehensive image processing technique, correction algorithms, an‎d simulation tools were employed. This process encompasses phantom an‎d animal studies, utilizing both experimental an‎d simulated data. Firstly, the PET-Xtrim system was simulated using the GATE Monte Carlo code an‎d validated via experimental results in accordance with the NEMA NU4-2008 stan‎dard. System sensitivity, noise-equivalent count rate (NECR), spatial resolution, an‎d image quality were eva‎luated. Using the GATE code, various scintillator crystal sizes were simulated, an‎d system sensitivity an‎d spatial resolution were investigated. Attenuation correction using the Chang method an‎d scatter correction employing both the dual-energy window technique an‎d an iterative approach were investigated for the raw data acquired from the PET-Xtrim system. Correction of missing data due to gaps between detector crystals was performed using deep learning method (Pix2Pix cGAN). Resolution recovery through point spread function (PSF) modeling an‎d partial volume effect (PVE) correction were also investigated to enhance quantitative accuracy an‎d image quality. Absolute system sensitivity was 0.028 an‎d 0.0206 in the energy windows of 250–650 keV an‎d 400–700 keV, respectively. Spatial resolution of the system was measured with radial FWHM of 1.55 an‎d 2.08 mm at 5 an‎d 25 mm from the center of field of view (CFOV), an‎d was reduced to approximately 1 mm at the CFOV via 1.5 mm LYSO crystals. Noise-equivalent count rates (NECR) for the mouse an‎d rat phantoms were calculated 90.12 kcps an‎d 63.24 kcps, respectively. Attenuation an‎d scatter correction significantly enhanced image quality an‎d contrast. Missing data were recovered using the Pix2Pix deep learning model. Additionally, PSF modeling an‎d the Lucy-Richardson algorithm markedly improved image accuracy, particularly in small-scale regions. Results demonstrated that crystal selec‎tion an‎d energy window configuration significantly influence system sensitivity, an‎d correction methods substantially reduced quantitative bias. Phantom an‎d mouse imaging via ¹⁸F-FDG revealed that correction methods substantially enhanced image contrast an‎d overall quality.

5

152642