This study indicates that a reduction of the scan time duration or administered 68Ga-PSMA activity produces results comparable to the reference acquisition protocol on a digital Biograph Vision PET/CT system both for detectability (98% of regions correctly identified) and image quantification (mean absolute deviation ≤10%) for all reconstruction algorithms but OSEM-TOF (4i).
In our cohort of 20 prostate cancer patients across a variety of miTNM stages undergoing 68Ga-PSMA PET, only two small nodal lesions (short-axis diameter of ≤4 mm) were missed (Figs. 1 und 2) leading to miTNM downstaging in one case. The first lesion was located close to the right common iliac artery, the second in the mediastinum. In one of these cases (lesion 1) this downstaging would have possibly impacted patient management negatively, in the other (lesion 2) the reduced emission time would have been unlikely to cause major changes in management as the patient also had bone metastases.
However, low sensitivities of PSMA PET performed with “conventional” PET/CT systems have previously been reported for the detection of small lesions (<5mm) due to partial volume effects [20]. Both of the missed lesions in our cohort showed moderate to faint tracer uptake (SUVmax 5.7 and SUVmax 3.8), which is also known to negatively affect detectability [21]. Additionally, high background due to unspecific small intestinal or mediastinal uptake considerably hampered lesion detection (Figures 1 and 2). A further possible explanation would be motion artifacts, among others caused by gastrointestinal peristaltic and arterial pulsation. Administered activities for these patients were above average (129 MBq, 135 MBq) and uptake time within one standard deviation of the mean (48 and 55 min); therefore, both factors are unlikely to be causal.
The main drawback of our reduced acquisition protocol is that small nodal lesions could be missed leading to false-negative 68Ga-PSMA PET reports, especially in the non-prostate lesion, where the acquisition time is particularly short. This could be addressed by optimizing patient selection, and performing reduced activity protocols in patients, in whom missing small nodal lesions would not impact management. One example could be the imaging of patients with known remote metastases (although the appearance of small new lesions could be missed) and/or those before PSMA-directed radioligand therapy. On the other hand, patients with suspected low tumor burden and/or patients at initial diagnosis would not be ideal candidates.
Recently, two studies have been published that tried to optimize the administered activity and to reduce the emission time duration. First, our findings differ from a previous trial by Rauscher et al. [5], who showed unsatisfactory results for list-mode reconstructed images simulating the administration of one-third and two-third of the standard activity. In contrast to their methodology, whole-body PET list-mode reconstruction was not applicable in our study due to the use of continuous-bed-motion mode, which should preferably be used if available [22]. Additionally, patients enrolled in the trial by Rauscher et al.[5] underwent 68Ga-PSMA PET on a Siemens Biograph mCT, so the discrepancy in findings might potentially be explained by the different imaging characteristics when compared to its successor, the Siemens Biograph Vision. Second, van Sluis et al.[10] showed an improvement in visually assessed image quality, tumor lesion demarcation, and overall image quality in oncological patients undergoing 18F-FDG PET/CT [7]. In agreement with our short acquisition protocol the same group [10] also found that a threefold reduction of administered activity in oncological patients was feasible, with TNM down-staging only occurring in 1/30 patient cases.
No differences with regards to the detectability were observed for the different reconstruction algorithms. Of note, additional PSF reconstruction did not provide additional value in terms of detectability. This can be largely explained by the implementation of a 4-mm Gaussian filter, producing similar PET reconstructed spatial resolutions for TOF- and TOF+PSF-reconstructed images (6.2 mm vs. 5.6 mm) [9]. In addition, under reduced statistical conditions, PET images will inevitably display higher noise [23]. To compensate this loss in image quality, a careful adaptation of iteration number could be considered without compromising lesion detectability by insufficient iterative convergence. In fact, our data suggests that with TOF and TOF+PSF modelling, image noise in the liver can be reduced by applying 2 iterations instead of 4 (Figure 6). This results confirm previous investigations, underlining the fast convergence capability of TOF [24]. Additionally, recent publications suggest that the implementation of machine learning approaches might enable the image reconstruction of standard activity images even when very low activities are used [25].
There are several limitations in this study. A limitation of our study is the relatively small and heterogeneous sample size, encompassing patients with a wide variety of miTNM stages. Additionally, the reduced acquisition protocol was applied after the reference acquisition protocol and uptake intervals were quite heterogeneous. On the one hand, by doing so the radionuclide decay occurring in the meantime as well as the better alignment with the CT scan favor the standard protocol. On the other hand, metabolic activity changes can occur between both scan acquisitions. As prior studies have observed an increase in tumoral PSMA uptake between images acquired 3 hours after tracer administration vs 1 hour after tracer administration, this might have contributed to differences in image quantification (i.e., the observed overestimation for a few lesions) [26]. This leads to the alternative hypothesis that lesions inapparent on the images acquired (later) with the reduced acquisition protocol, might have been non-neoplastic lesions with decreasing PSMA-uptake over time. Therefore, the reason for missing them might rather be the later uptake interval than the reduced acquisition time.
Furthermore, detectability was performed on a per-region level instead of a per-lesion level. Since the per-region analysis does not account for the identification of additional lesions in the standard acquisition protocol in a region that is already rated positive in both acquisition protocols, the per-lesion detectability is potentially lower. As the scan time duration of the non-prostate region in the clinical protocol was slightly below the 2-4 min recommended by the EANM guideline, patients might also have potentially been understaged by the clinical protocol. A further limitation of this study is that lesion validation was not performed. However, the current literature suggests a high positive predictive value of 68Ga-PSMA PET making this a minor issue [3].