Optimizing Image Reconstruction Parameters in Time of Flight PET/CT Imaging: a Phantom Study

Purpose- The aim of this study was to determine optimal reconstruction parameters in relation to the image quality and quantitative accuracy for advanced reconstruction algorithms by phantom study.Methods- A house-made image quality phantom, including 6 cylindrical inserts,was filled with an 18F-FDG solution with a 4:1 radioactivity ratio compared to the background. All emission data was acquired in 3D list-mode. The PET data reconstructed with TOF only and TOF+PSF algorithms. The reconstructed images were post-filtered with Gaussian filters with varying FWHM (0 to 10 mm with 0.5 mm increment). All images were reconstructed with different product of iterations and subsets (It×SS) ranging from 3 to 144. Optimal image reconstruction parameters were determined by calculating quantitative parameters including noise, signal to noise ratio (SNR), and recovery coefficient (RC).Results- Our results showed that Gaussian filtering with FWHM greater than 5mm for TOF and greater than 3.5 mm for TOF+PSF algorithms led to an acceptable clinical noise level (<10%). By considering signal to noise ratio of the 10 mm insert (SNR10 mm) and quantitative accuracy of tracer concentration, optimum FWHM of Gaussian filter was 5-6.5 mm for TOF only reconstruction and 3.5-5 mm for TOF+PSF reconstruction. In terms of It×SS, SNR10 mm was maximized for 28 to 48 It×SS. In addition, there was no significant enhancement in RC for It×SS greater than 48.Conclusion- Image quality and quantitative accuracy are strongly influenced byreconstruction parameters. Our findings indicate that the optimization of the reconstruction parameters is necessary to obtain the best performance. Optimal FWHM range was 5-6.5 mm for TOF only reconstruction, and 3.5-5 mm for TOF+PSF reconstruction. Additionally, due to intensifying signal of the focal point by incorporating TOF information, faster SNR convergence can be achieved. Hence smaller It×SS can be applied while using TOF algorithm for image reconstruction.