Abstract
BACKGROUND: Positron emission tomography (PET) imaging of (90)Y following selective internal radiation therapy (SIRT) is possible, but image quality is poor, and therefore, accurate quantification and dosimetry are challenging. This study aimed to quantitatively optimise (90)Y PET imaging using a new Bayesian penalised likelihood (BPL) reconstruction algorithm (Q.Clear, GE Healthcare). The length of time per bed was also investigated to study its impact on quantification accuracy. METHODS: A NEMA IQ phantom with an 8:1 sphere-to-background ratio was scanned overnight on a GE Discovery 710 PET/CT scanner. Datasets were rebinned into varying lengths of time (5-60 min); the 15-min rebins were reconstructed using BPL reconstruction with a range of noise penalisation weighting factors (beta values). The metrics of contrast recovery (CR), background variability (BV), and recovered activity percentage (RAP) were calculated in order to identify the optimum beta value. Reconstructions were then carried out on the rest of the timing datasets using the optimised beta value; the same metrics were used to assess the quantification accuracy of the reconstructed images. RESULTS: A beta value of 1000 produced the highest CR and RAP (76% and 73%, 37 mm sphere) without overly accentuating the noise (BV) in the image. There was no statistically significant increase (p < 0.05) in either the CR or RAP for scan times of > 15 min. For the 5-min acquisitions, there was a statistically significant decrease in RAP (28 mm sphere, p < 0.01) when compared to the 15-min acquisition. CONCLUSION: Our results indicate that an acquisition length of 15 min and beta value of 1000 (when using Q.Clear reconstruction) are optimum for quantitative (90)Y PET imaging. Increasing the acquisition time to more than 15 min reduces the image noise but has no significant impact on image quantification.