Quantification of the impact of TOF and PSF on PET images using the noise-matching concept: clinical and phantom study

This study was to assess quantitatively the accuracy of 18 F-FDG PET/CT images reconstructed by TOF + PSF and TOF only, considering the noise-matching concept to minimize probable bias in evaluating algorithm performance caused by noise. PET images of similar noise level were considered. Measurements were made on an in-house phantom with hot inserts of Φ10–37 mm, and oncological images of 14 patients were analyzed. The PET images were reconstructed using the OSEM, OSEM + TOF and OSEM + TOF + PSF algorithms. Optimal reconstruction parameters including iteration, subset, and FWHM of post-smoothing filter were chosen for both the phantom and patient data. In terms of quantitative accuracy, the recovery coefficient (RC) was calculated for the phantom PET images. The signal-to-noise ratio (SNR), lesion-to-background ratio (LBR), and SUV max were evaluated from the phantom and clinical data. The smallest hot insert (Φ10 mm) with 2:1 activity concentration ratio could be detected in the PET image reconstructed using the TOF and TOF + PSF algorithms, but not the OSEM algorithm. The relative difference for SNR between the TOF + PSF and OSEM showed significantly higher values for smaller sizes, while SNR change was smaller for Φ22–37 mm inserts both 2:1 and 4:1 activity concentration ratio. In the clinical study, SNR gains were 1.6 ± 0.53 and 2.7 ± 0.74 for the TOF and TOF + PSF, while the relative difference of contrast was 17 ± 1.05 and 41.5 ± 1.85% for the TOF only and TOF + PSF, respectively. The impact of TOF + PSF is more significant than that of TOF reconstruction, in smaller inserts with low activity concentration ratio. In the clinical PET/CT images, the use of the TOF + PSF algorithm resulted in better SNR and contrast for lesions, and the highest SUV max was also seen for images reconstructed with the TOF + PSF algorithm.