BPL reconstruction (also known as Q.Clear in GE Healthcare Imaging) is a relatively newer reconstruction method with the capacity to improve contrast over OSEM, by applying a noise penalty to individual voxels during reconstruction [8,9,10]. From a clinical point-of-view, BPL reconstruction improves the signal-to-noise ratio (SNR) and have been reported to be useful for better localization of the tumour as reported in case reports [12, 15]. However, wider adoption of the method showed it cannot increase the sensitivity and specificity of PSMA PET/CT reporting [13], unable to overcome the day-to-day repeatability of the PSMA PET/CT-derived parameters [8] and may even exceeds the EARL-accredited maximum values in phantom studies which may impact development of the PSMA-based response criteria [16]. Moreover, if different reconstruction methods yield to significant variation on the same PET study, it can have downstream consequences leading to variations in the biomarkers and radiomics features derived from the PET study [17].

To quantify PSMA PET/CT, different thresholding methods including histogram-based and predefined fixed thresholding algorithms have been introduced. We chose fixed thresholding of SUV 3 since it was validated prospectively in a multicentre randomised control trial [1]. Our study shows that the differences in SUVmax between OSEM and Q.Clear are non-significant especially in the lower ranges of SUV (Supplemental Fig. 1). This means that alternative use of the Q.Clear method would not change the quantification of whole body PSMA PET metrics by excluding or adding minimally avid lesions (SUV < 3) to the calculated disease burden.

Total tumoral SUVmean on [68Ga]Ga-PSMA-11 PET/CT, as a reflector of the average concentration of the radiotracer within the whole tumour burden, has been shown to be a valuable predictive biomarker in patients underwent radionuclide therapy. The TheraP trial showed that SUVmean ≥ 10 was predictive of a higher likelihood for favourable response to [177Lu]Lu-PSMA-617 than cabazitaxel with odds ratio of 12.2 vs. 2.2 [3]. Post-hoc analysis of the same study using quartile values of PSMA-PET SUVmean, showed an unfavourable response to radionuclide therapy in subgroup of patients with SUVmean < 6.9 with 29% response to [177Lu]Lu-PSMA compared to 43% response to cabazitaxel. Using these two thresholds, our study showed that 2 patients from SUVmean < 10 subgroup on OSEM changed to SUVmean ≥ 10 on Q.Clear while 1 patients from SUVmean < 7 subgroup on OSEM changed to SUVmean ≥ 7 on Q.Clear reconstructed images, with 90% and 96% of patients remaining in the same subgroups, respectively (Table 1). This result supports that Q.Clear reconstruction can accurately differentiate between the patients with lower and the patients with higher likelihood of response to [177Lu]Lu-PSMA therapy.

Another important PET parameter is SUVmax which is a frequently used metric to describe lesions due to its simplicity and reproducibility [18]. In a study, BPL reconstructions resulted in significantly higher SUVmax of tumour lesions as compared to standard OSEM reconstructions, with significantly higher relative increases in smaller lesions [8]. This effect emphasised as an area requiring harmonisation in the European Association of Nuclear Medicine with their EARL program [19]. On the advice of the committee, image reconstruction techniques which augment the appearance of small lesions through additional processing may lead to inconsistent quantification of common PET biomarkers across sites or equipment. While our study supports the prior observation of higher average SUVmax on Q.Clear compared to OSEM reconstructed images, it shows a strong correlation between produced SUVmax on these two methods with Pearson r value of 0.98. The correlation plots (Fig. 3) demonstrate the differences in SUV become proportionally more marked at higher SUVs. This is to be expected, given the SNR-recovery of the BPL method as compared to OSEM [12]. Given the proportional difference, harmonising SUVs would be feasible as has been demonstrated in other studies [20]. In addition, the subgroup analysis, did not show significant difference between OSEM and Q.Clear acquired parameters (SUVmax & SUVmean) in subgroup of patients with low-volume disease burden.

Lastly, the MTV-PSMA values were lower in the BPL method as compared to OSEM. This is quite unexpected. Given the fixed absolute value thresholding method (i.e., any voxels with SUV above 3) along with discarding small volumes (i.e., < 0.5 mL) to avoid counting the noise into MTV values, we assume that BPL method spuriously led to small islands with volumes less than < 0.5 which were discarded from the analysis. Apart from this minor difference, true three-dimensional tumour volumetric measurement is also feasible with PET/CT by both methods. Barbato et al. showed that volume quantification with PSMA PET (using a 40 ml cut-off) can discriminatebetween low versus high burden metastatic prostate cancer, with additional sub-classification of disease extension critical for guiding targeted or systemic therapy [14]. The current study showed negligible difference (< 5%) between average MTV-PSMA values by applying these two reconstruction methods in total study population as well as both low and high-volume disease subgroups.

The main clinical implication of these results is the referral to radionuclide therapy centres for consideration of [177Lu]Lu-PSMA therapy in patients with BPL reconstructed PSMA PET scans. Although comparative studies should ideally use identical acquisition and processing protocols, this may not always be feasible, particularly in referral centres for radionuclide therapy where new patients with externally-acquired baseline studies are frequently referred. Our results show that the quantified parameters on the PSMA PET studies reconstructed by BPL are reliable for both assessing suitability of [177Lu]Lu-PSMA therapy as well as prognostication goals. Therefore, either reconstruction method can be used interchangeably without significant impact on quantified parameters. For centres using BPL reconstruction, it is worth being aware that any differences, although small, are more profound at higher SUVs.

There are some limitations to the current study. The main limitation is using Q.Clear at a single B value, while other centres may use other B values which have not been tested in the current study. Also, the results are only relevant for the GE cameras described, and we cannot extrapolate to newer cameras with digital detectors or manufactures with different algorithms. Finally, there is potential risk of human error in image processing. To minimize this limitation, all lesions contouring was re-evaluated by a second nuclear medicine specialist. To the extent of our knowledge, this is the first study comparing quantified parameters on OSEM and BPL reconstructed [68Ga]Ga-PSMA-11 PET scans. The key strength of this study is applying these two reconstruction methods to the same set of patients. This provided the ability to compare same images with completely similar characteristics except for reconstruction method.