Abstract
101
Objectives To assess the impact of point spread function (PSF) on the image quality (IQ) of clinical 18FDG scans by comparing the time-of-flight (TOF)-based list-mode ordered subsets expectation maximization (OSEM) reconstructions with and without PSF[1-3].
Methods 14 patients underwent clinical PET/CT (Ingenuity TF PET/CT 128, Philips Healthcare, Cleveland, OH). PET data were reconstructed using OSEM and TOF (OT), OSEM+TOF+PSF (regularization:6; iteration:1) (OTP6), and OSEM+TOF+PSF (regularization:10; iteration:2) (OTP10). Two reviewers examined the effects of using PSF on the IQ of de-identified datasets by visual assessment of the overall IQ, lesions detected <1 cm and conspicuity. Reader agreement (weighted kappa) and the methods’ relative performance (Wilcoxon SR) were assessed. Standard Value Uptake (SUVmax) and the Standard Deviation (SD) were obtained by drawing 2 regions of interest on two axial slices on the liver, also lesions <1cm and >1cm were evaluated in order to assess if there was any significant increase or decrease of the values according to the size of the lesion.
Results Reader agreement for conspicuity and IQ was acceptable (κ=0.62, κ=0.76). OTP6 identified more lesions than OT (p<0.1), OTP10 didn’t offer significant additional improvement. OTP6 was offered a better conspicuity than OT (p<0.01), and OTP10 better yet than OTP6 (p<0.01). OTP6 judged to give better IQ than OT (p<0.01), but OTP10 didn’t offer significant additional benefit. Mean SUV didn’t differ significantly in background regions drawn in the liver, however SD increased progressively with OTP6 and OTP10. SUVmax of lesions <1 cm increased 15-22% with OTP6 and 27-42% with OTP10; SUVmax of lesions >1 cm increased 13-29% with OTP6 and 33-60% with OTP10.
Conclusions Combining TOF and the modeling of PSF significantly improves the image quality. SUVmax value doesn't vary in background organs, but it increases in lesions when using PSF compared to baseline OSEM+TOF reconstruction.
Research Support Philips Healthcare