Applications of both time of flight and point spread function in brain PET image reconstruction

To assess the effects of different high-resolution image reconstruction methods on the quality of images and quantitatively compare their accuracy in human brain PET images.

Brain images were reconstructed using six algorithms including back-projection (B), B+time of flight (TOF), ordered subsets expectation maximization (OSEM), OSEM+TOF, OSEM+point spread function (PSF), and OSEM+PSF+TOF. The qualities of images including the spatial resolution, contrast, and signal-to-noise ratio (SNR) were compared qualitatively and quantitatively, and the mean standardized uptake value (SUVmean) of different brain regions obtained by different reconstruction algorithms were also compared.

The contrast of images obtained by the six reconstruction algorithms was not significantly different (P>0.05). PSF significantly improved the spatial resolution and SNR by 23.8%. Combined PSF significantly increased the SUVmean of all brain regions, except the mesial temporal lobe and part of the basal ganglia. Application of TOF did not obviously increase the resolution and contrast, but improved SNR by up to 11.9%. Compared with OSEM and OSEM+PSF, application of OSEM+TOF and OSEM+PSF+TOF evidently affected the SUVmean of some brain regions and resulted in differentiation phenomena. TOF obviously increased the SUVmean of the temporal lobe, but significantly reduced that of the cingulate and paracingulate gyri, mesial temporal lobe, and cerebellum.

PSF was the most effective reconstruction algorithm to improve the quality of brain images. The combined PSF algorithm could increase the SUVmean and made it closer to the actual standardized uptake value. The combined TOF algorithm should be applied with caution in quantitative analysis.