Regularized reconstruction of digital time-of-flight 68Ga-PSMA-11 PET/CT for the detection of recurrent disease in prostate cancer patients

Abstract

Accurate localization of recurrent prostate cancer (PCa) is critical, especially if curative therapy is intended. With the aim to optimize target-to-background uptake ratio in 68Ga-PSMA-11 PET, we investigated the image quality and quantitative measures of regularized reconstruction by block-sequential regularized expectation maximization (BSREM). Methods: The study encompassed retrospective reconstruction and analysis of 20 digital time-of-flight (TOF) PET/CT examinations acquired 60 min post injection of 2 MBq/kg of 68Ga-PSMA-11 in PCa patients with biochemical relapse after primary treatment. Reconstruction by ordered-subsets expectation maximization (OSEM; 3 iterations, 16 subsets, 5 mm gaussian postprocessing filter) and BSREM (β-values of 100-1600) were used, both including TOF and point spread function (PSF) recovery. Background variability (BV) was measured by placing a spherical volume of interest in the right liver lobe and defined as the standard deviation divided by the mean standardized uptake value (SUV). The image quality was evaluated in terms of signal-to-noise ratio (SNR) and signal-to-background ratio (SBR), using SUVmax of the lesions. A visual assessment was performed by four observers. Results: OSEM reconstruction produced images with a BV of 15%, whereas BSREM with a β-value above 300 resulted in lower BVs than OSEM (36% with β 100, 8% with β 1300). Decreasing the acquisition duration from 2 to 1 and 0.5 min per bed position increased BV for both reconstruction methods, although BSREM with β-values equal to or higher than 800 and 1200, respectively, kept the BV below 15%. In comparison of BSREM with OSEM, the mean SNR improved by 25 to 66% with an increasing β-value in the range of 200-1300, whereas the mean SBR decreased with an increasing β-value, ranging from 0 to 125% with a β-value of 100 and 900, respectively. Decreased acquisition duration resulted in β-values of 800 to 1000 and 1200 to 1400 for 1 and 0.5 min per bed position, respectively, producing improved image quality measures compared with OSEM at a full acquisition duration of 2 min per bed position. The observer study showed a slight overall preference for BSREM β 900 although the interobserver variability was high. Conclusion: BSREM image reconstruction with β-values in the range of 400-900 resulted in lower BV and similar or improved SNR and SBR in comparison with OSEM.