Digital PET/CT allows for shorter acquisition protocols or reduced radiopharmaceutical dose in [¹⁸F]-FDG PET/CT

The recent introduction of digital PET/CT scanners for routine clinical use represents a significant milestone for nuclear medicine and molecular imaging. Although PET/CT scanners have always been truly “digital” insofar as their outputs were in the form of digital signals, the replacement of analogue photomultiplier tubes with solid-state detection systems resulted in the first fully “digital” PET/CT, and has surmounted many of the inherent physical limits placed by previous-generation analogue technologies. These new fully digital systems exhibit a plethora of technical advantages, which include a better coupling between the crystal and photodetectors, improved background-to-noise, faster time-of-flight (TOF) and associated advanced TOF reconstruction. In addition, state-of-the-art digital systems often include longer axial coverage, smaller crystals, and more advanced electronics, which lead to higher sensitivity, higher spatial resolution, and shorter deadtime. These improved performance characteristics have been confirmed by a number of publications [1, 2], which correspond to improvements in image quality and lesion detection [2‐6].

Whereas the current iteration of the European Asociation of Nuclear Medicine (EANM) guidelines report examination protocols for analogue scanners and for “step-and-shoot” type acquisitions [7], no current guidelines are yet available for systems with digital acquisition. Although the performance characteristics, and EANM Research Ltd (EARL) compliance of such scanners have been reported [8‐10], the full clinical potential of digital scanners is yet to be fully characterized. In particular, whereas the manufacturers’ literature report shorter-acquisition times and/or reduced activity as potential improvements of such systems, the advantage for the patient needs to be investigated and verified in a clinical setting.

The aim of this present study is to assess the effect of short acquisition times or reduced activity on lesion detectability as assessed by two nuclear medicine physicians, followed by lesion uptake quantitation in a digital scanner.

In n = 21 patients undergoing routine oncological [¹⁸F]-FDG PET/CT on a digital scanner, we find that 98% of lesions (n = 98/100) are detected at a table velocity equivalent to 30 s/bp (25% of standard acquisition time). For the two “missed” lesions, we note that these were in polymetastatic individuals where the extra lesions could be slightly better discriminated, the extra lesion however would not have influenced the patient’s stage. Example images for one such patient is given in Fig. 2.

Only minimal variations in TBR and SUVmax and SUVmean were observed between the three list-mode acquisitions, with almost perfect agreement (Krippendorf’s α = 0.99) between them, suggesting that shorter acquisitions are adequate for routine clinical purposes. We find the shorter acquisitions to be of slightly lower subjective image quality, although such subjective impressions were not the focus of this present study, but rather the aim was to quantify the magnitude of any clinical effect such shorter-acquisition protocols may have. While such acquisitions may be less aesthetically pleasing, we find no evidence of clinical detriment. We contend that while a modestly higher background noise was observed in the 30 s/bp images (as exemplified by Fig. 3), this was not associated with any objective disadvantage.

Given the known relationship between applied radiopharmaceutical dose, acquisition time and count statistics, by analogy, our data can also be interpreted as supportive of a reduction in applied activity in dPET/CT. Indeed, with a mean effective dose (excluding the CT component) of 4.9 mSv (Table 1), the 30 s/bp reconstructions are equivalent to a 75% reduction in dose, or 3.7 mSv.

We find a small variability in TBR (0.79%) when comparing the 30 s to the 2 min/bp reconstructions. Likewise, we find a smaller variability for the parameter SUVmean (0.35%) than for SUVmax (0.77%). In contrast to previous studies comparing the influence of shorter acquisition on SUVmax and SUVmean in a digital system [13] we find no statistically significant differences in SUVmax and TBR between the longer (2 min) and shorter (30 s) acquisitions. In Fig. 4 we show the SUVmax for the different acquisition durations. Shortening image acquisition durations broadens intensity distributions and should shift the supremum (of a set of values) found within a VOI towards higher values. Furthermore, low counts give rise to tail-heavy distributions in images reconstructed using OSEM and PSF-based algorithms [14], and the choice of reconstruction algorithm can have a significant effect on lesion quantification [13] Therefore, SUVmax, by definition the supremum, is expected to rise when lowering image count [14, 15]. However, when considering all SUVmax values found in our study, no clear differences can be seen between images for different list-mode acquisitions.

As described in the introduction, digital PET/CT represents a step forward in molecular imaging, with increasing numbers of publications confirming the favourable performance characteristics for such scans in comparison to previous-generation analogue systems [1, 16], including increased lesion detection and sensitivity [1, 3, 4, 13, 17]. Previous authors have reported head-to-head studies comparing digital and analogue systems, showing both improved image quality and upstaging of individuals [10], as well as improved recovery (as shown by higher SUV values) and lesion sharpness [3, 6]. Traditionally, shorter list-mode acquisitions have been associated with decreased lesion detection, as most recently shown for [⁶⁸ Ga]Ga-PSMA-11 PET/CT by Rauscher et al. (using an analogue scanner) [18] with 1 min/bp reconstructions (equivalent to 1/3^rd of the standard) associated with reduced detection of up to 36% of lesions. We show here that applied dose reduction or reduction in acquisition time of 75% is feasible for the combination of [¹⁸F]-FDG and digital PET/CT without detriment with respect to lesion detection. Whereas extant procedural guidelines report protocols and standards for examinations for analogue scanners, the influence of digital PET/CT on list-mode acquisition length and applied dose has not been adequately addressed [7], increasing further the clinical importance of the topic.

Only a small number of publications considering these issues are available in the literature, and only few consider the impact of recent developments in PET technology. For example, Sonni et al. consider the influence of shorter acquisitions with a digital system on image quality (measured on a subjective 5-point scale), but do not report the influence on lesion detection [19]. In the context of combined PET/MRI scanners, multiple publications demonstrate the feasibility of reducing the [¹⁸F]-FDG dose using solid-state (silicon) photomultipliers [20, 21]. Van Sluis et al. consider the influence of shorter acquisitions in a digital PET/CT reporting both phantom [1] and clinical data, albeit in a small cohort of patients (n = 30) with no estimate of statistical power [13]. Although the authors of this latter study conclude that dose reduction is acceptable, they nevertheless report downstaging for one patient in this small cohort, albeit without any reported influence on therapeutic choice. Any putative benefit deriving from a small reduction in overall radiation exposure in an examination must be balanced against any potential clinical detriment arising from the misstaging of an established cancer—for example the harms caused by the missing of a contralateral lung nodule in a non-small cell lung cancer patient (NSCLC) far outweigh the benefit of a modest reduction in radiation exposure [22]. In contrast we find no detriment in a similarly small cohort, albeit with statistically defined endpoints and adequate statistical power to test our hypotheses. Larger studies must be performed to confirm this for all cancer types and in all patient groups. In their totality, we urge caution when interpreting these data. The equivalency of shorter acquisitions cannot be established with semi-quantiative lesion uptake or qualitative impression of image quality as the sole basis, and further work using clinically defined endpoints are required. The issue is therefore far from being exhaustively investigated, and prospective, randomized studies with larger cohorts are required to determine which patients can safely receive these shorter acquisitions.

Radiological protection concerns, however, are not the sole motivation of applied radioactivity dose reduction. Given the increasing utilisation of PET/CT for monitoring of therapeutics or treatment response, patients are increasingly re-referred for multiple restaging examinations, making accumulated radiation burden of potential relevance, particularly in neonates and infants. Given the limited supply of some radiopharmaceuticals, particulary those with short half-lives [23], we consider that reduction of dose/patient and faster examinations to be of considerable importance in increasing access to nuclear medicine examinations, where average post-referral waiting times for oncological PET/CT can be as high as 14 weeks in some jurisdications [24].

While longer acquisition times traditionally represent a balance between acquisition time and reduction in applied radiation dose, longer acquisitions can be associated with increased motion artefact, and point toward another key advantage in modern scanner design using proprietary end-expiratory PET acquisition for the reduction in respiratory motion artefact [25]. Likewise, while not the subject of this enquiry, the ability to increase numbers of patients examined through shorter-acquisition time is of considerable economic importance in an era of rapidly increasing utilisation of nuclear and molecular medicine imaging techniques.

We note some weaknesses with our study. Firstly, these initial experiences with a new scanner type are in n = 21 patients, although, in contrast to similar studies, a sample size calculation confirmed the adequacy of this cohort size to answer our hypothesis with adequate statistical power. Nevertheless, these findings must be considered preliminary and interpreted alongside the other similarly small cohorts hitherto published. We use clinically routine and vendor recommended reconstruction methods, EARL-accredited protocols may be better suited to quantitative PET studies in a multicentre setting [9]. However, the aim of our study was to show differences in vision detectability, and our chosen acquisition protocol was optimized for this in a clinical setting. We did not assess the influence of lesion size and note that in one case a small lesion was missed (Fig. 2). Previous studies report increased signal recovery from small structures in digital PET/CT systems [26], and further studies are needed to investigate the impact of shorter examinations or low dose protocols on small or low-uptake lesions. Reader bias was reduced using reviewers who had not seen the cases prior to the study. Furthermore, scans were read in randomized and blinded fashion. Few studies confirm the magnitude of any recall bias effect or confirm the optimum time between scan reading [27], although longer waiting periods may be beneficial. Instead, we highlight that scans were read starting with 30 s/bp (the images of lowest quality) and in randomized order with a pragmatic 48 h waiting period in between, limiting the impact of any recall bias inanycase. Other methods of reducing bias, such as using different readers for each list-mode reconstruction are not applicable to this study, where the magnitude of the interobserver agreement is likely larger than the magnitude of the absolute difference between scans (Krippendorf’s α = 0.99 in this study, inter-observer agreement in FDG PET κ = 0.69–0.79 [28]). Finally, institutional ethics board permission afforded anonymous and retrospective data acquisition: further studies, ideally of prospective design, are required to confirm the clinical non-inferiority of images acquired either with shorter-acquisition times, or by analogy lower radiopharmaceutical doses. Previous studies demonstrate that dose reduction is only feasible in non-obese individuals [20]. Our study included only two individuals with BMI > 30 kg/m², and none with class II obesity (BMI > 35 kg/m²). Further studies in these population groups are required. Caution must be taken when generalising these results to other radiotracers, and further studies for non-oncological PET/CT (for example in PET for neurodegeneration) and using other tracers are required.

The authors report no conflict of interest. A waiver was granted by the regional ethics committee for the retrospective, anonymous analysis of patient data (Req-2020–00678). The study was performed in accordance with the declaration of Helsinki. No funding was obtained.