This study firstly demonstrated that the standard ULDCT protocol for PET AC without tin filter maintains adequate PET quantification at both soft tissue and bone equivalent densities (Fig. 3b), for a DLP of 7–8 mGy.cm per 26.3 cm bed position, when scanning a phantom representing a standard adult abdomen (Fig. 4). Given that few facilities have been using such low CT doses for PET AC only protocols [3], it is hoped that the findings of this study will encourage implementation of well-optimised ULDCT protocols when performing CT only for PET AC. Whilst differences in PET quantification with CT tube voltage changes had been previously investigated in soft tissue [15, 16], this study is the first to investigate this in bone equivalent densities, and should assist in optimisation of CT protocols for PET-CT examinations in which bone is a tissue of interest.

Moreover, this is the first study to examine the effect and benefits of the tin filter in PET-CT, demonstrating that it can further reduce dose in ULDCT scans performed only for PET AC by at least 60% (Fig. 4), whilst maintaining adequate PET quantification (Fig. 3b). The minimum required exposure settings, and therefore the dose savings which can be made, depend on the tissue density (Figs. 3b, 4) and accepted changes in PET quantification. At soft tissue equivalent, accurate PET quantification is maintained down to a DLP of 0.9 mGy.cm (Sn100 kV/12 mAs-ref) for the single bed position phantom scan (Fig. 3b), providing 87% dose reduction compared with the standard ULDCT protocol (Fig. 4). However, at bone-equivalent densities, higher exposure settings than this were required to achieve negligible differences (≤ 2%) in PET quantification (Fig. 3b), due to a marked reduction in measured CT HUs (Fig. 3a). This is likely due to a combination of beam hardening artefact from the large thickness of high-density material in the cylindrical insert as shown in Fig. 5, and excessive noise providing a bias in CT data, from absent signal in parts of the sinogram [17, 18]. There is no consensus as to what differences in PET quantification are acceptable, and the lower dose option could be used if differences need not be negligible but clinically acceptable, at ≤ 5% for example.

Sn100 kV provided the best compromise between PET quantification accuracy and dose compared with Sn140 kV (Figs. 3b, 4), given that the beam energies with Sn100 kV provide CT HUs similar to 120 kV with standard filtration (Fig. 3a), and the large effect of increased CT tube voltage on dose. Interestingly, a negative difference in absolute HU corresponded to a positive difference in PET activity concentration for Sn140 kV but not for Sn100 kV. This phenomenon can be explained by the bilinear transformation curve used to derive LAC from CT HU [2]. In this bilinear transformation, at less than around 50 HU LAC is not dependent on tube voltage, but above this HU threshold, LAC is dependent on tube voltage, with a higher kVp giving a higher LAC. Thus, for a given difference in HU compared with the reference standard CT, a higher PET quantification value is seen for Sn140kV compared with Sn100kV, in the bone-equivalent density phantoms.

Whilst this study focused on investigation of ULDCT protocols, data were also gathered for the tin filter at comparable doses to standard localisation CT, with Sn100 kV/400 mAs giving comparable dose to 120 kV/22 mAs. Hence, this work has also validated that PET quantification with the tin filter is also acceptable at higher doses.

PET AC is the primary clinical purpose of the CT scan in cardiac and brain PET-CT examinations [3, 19]; hence, the study findings can be used to enable optimisation of CT radiation dose in these widely performed examinations. In addition, ULDCT scans are also performed when a diagnostic CT scan has already been recently acquired for the patient, or when multiple CT scans are performed in a PET-CT protocol, such as in multiparametric PET [20]. Furthermore, the dose reduction in ULDCT protocols afforded by the tin filter would be particularly beneficial for imaging with long axial field-of-view (LAFOV) PET-CT systems, since the CT scan must always cover the entire PET FOV. Yet, sometimes just a single organ is of interest. Furthermore, owing to their high PET sensitivity, LAFOV systems are used to scan children [21] and pregnant patients [22] allowing PET dose reduction, whom would also benefit from accompanying ULDCT for PET AC. LAFOV systems also scan many research participants and scan patients at multiple timepoints [21], for which an additional effort should be made to keep radiation doses as low as reasonably achievable.

When implementing ULDCT protocols, it is important to use a fast rotation time and high pitch factor, to allow the system to provide a lower effective mAs when delivering the lowest available tube current [23], since the lowest possible effective mAs is determined by the product of the tube current and rotation time (mAs) divided by the pitch value. An additional benefit of tin filter imaging for ULDCT is that a higher reference mAs is needed compared with standard filtration. This means that when tube current modulation is applied, for small patients and low-density body regions, the mAs can go lower than the reference as required, whereas the standard ULDCT protocol is already set to deliver just 7mAs-ref, and the tube cannot use a much lower current.

A ± 2% difference in PET quantification compared with the reference standard was considered negligible by the authors in this study, although there are no guidelines to inform this decision. However, since it had been reported that variation in standardised uptake value (SUV) maximum, SUVmax, can exceed 15–20% in clinical practice [24] and its value had thus been debated for many years [25, 26], this seemed like a reasonable compromise between allowing some additional error on the SUV measurement which is already subject to considerable inaccuracy, whilst not increasing the error so much that its clinical or research utility is brought further into question. However, whilst we considered a ± 2% difference in PET quantification to be negligible, this does not mean that larger differences in PET quantification would not be clinically acceptable. This topic should be discussed further in relation to the specific clinical or research circumstances under which the data are used, for example, the trade-off between absolute effective dose saving and quantification accuracy, different patient groups, and how the quantitative value will be used.

The CT acquisitions undertaken in this study used the spine organ characteristic, since that was in clinical use for bone PET-CT examinations and the phantoms with bone-equivalent densities in the cylindrical insert represented imaging of a spine. It should be borne in mind that use of other organ characteristics will deliver slightly different effective mAs for a given reference, hence to deliver the same dose and image quality, a slightly different quality reference mAs may be required.

The CT images were reconstructed with ADMIRE, which is the best available CT reconstruction on Siemens Healthineers PET-CT systems, providing superior artefact reduction to the other types. Other types of CT reconstruction may provide slightly different artefact severities and CT HUs, thereby yielding slightly different PET quantification results. However, Definition Edge CT systems with tin filter are usually equipped with ADMIRE.

This study evaluated ULDCT scans in a standard abdominal-sized phantom. CARE Dose 4D tube current modulation was utilised for this phantom investigation, since this is normally applied to patient scans in clinical practice, to ensure consistent image quality for all patient sizes, as well as in all slices across the z-axis despite differing organ densities. Henceforth, the findings in this study should also be applicable to obese patients. However, a greater number and severity of artefacts are expected in larger patients, especially in slices where highly attenuating structures are present, for example between the shoulders and hips, which are particularly susceptible to beam hardening artefacts. This could change CT HUs dramatically in some locations, potentially affecting PET quantification. An additional iterative beam hardening correction has recently been added to ADMIRE, which may yield slightly different results, although this has not been investigated in this study. On the other hand, the thickness of the high-density sand in the cylindrical insert might overestimate the extent of beam hardening in a patient. Thus, to help determine the most appropriate exposure settings for tin filter ULDCT for PET AC, future work should examine the extent and location of CT artefacts in standard and obese-sized patient phantoms at the different exposure settings, and also quantify the effect of CT HU differences on PET quantification on the Biograph Vision.

Since TOF performance influences how robust PET reconstructions are to data inconsistencies [27], it should be considered that these data were gathered on a Biograph Vision silicon photomultiplier (SiPM) PET system, with 249 ps timing resolution [28]. It could also be investigated as to how PET quantification is affected with the examined protocols on conventional photomultiplier PET systems such as the Biograph mCT Edge, with slower timing resolution at 540 ps [29], if equipped with the tin filter, or if the standard ULDCT protocol is used for AC only.

Whilst this study has focused on the tin filter in ULDCT for PET AC only, CT scans are also commonly carried out for lesion localisation and characterisation, requiring superior image quality and therefore increased dose. Future work should evaluate how much CT dose reduction is afforded with the tin filter for localisation/characterisation CT for the different exam types, and which exposure settings are required.

This study has only considered CT tin filtration on PET-CT systems, yet the tin filter has also been recently introduced to single-photon emission computed tomography (SPECT)-CT. Although we would still expect to see marked dose savings with the tin filter in SPECT-CT, the magnitude of dose saving may be different, since these systems use different tube voltages, and the effect of HU changes may affect SPECT quantification to a different extent. Hence, a similar study should also be conducted for SPECT-CT to enable optimisation of tin filter protocols on those systems.