This study has suggested Nordic NDRL and achievable doses for CT scans performed in 4 PET/CT and 4 SPECT/CT examinations which are specific to the clinical purpose of the CT scan, as shown in Table 1. The data presented in this study demonstrates great variation in CT radiation doses for the same examination and clinical purpose of CT, for all investigated PET/CT and SPECT/CT examinations, as shown in Fig. 1 and Table 2. For instance, up to 9 times difference in DLP75kg was seen for AC only in PET/CT oncology, with some AC only doses exceeding the achievable dose and approaching the suggested NDRL for localisation/characterisation as shown in Fig. 2a. An 18 times difference in DLP75kg was seen for AC only in PET/CT infection/inflammation, with Fig. 2b showing a system giving AC only doses exceeding the suggested NDRL for localisation/characterisation. For PET/CT brain, Fig. 2c shows a system giving localisation doses greater than a system giving diagnostic doses, constituting a 17 times difference in DLP75kg for localisation/characterisation. In the case of SPECT/CT, up to 27 times difference in DLP75kg was seen for AC only CT scans in lung SPECT/CT, with a system giving AC only doses four times greater than the suggested NDRL for localisation/characterisation, as shown in Fig. 2d. On comparing scan protocol settings for the maximum and minimum dose protocols (for the same examination and clinical purpose of CT) in Table 3, variations in kV, effective mAs, scan length, and reconstruction algorithm were found to be key contributors to dose differences. Such findings thus demonstrate the importance of optimising CT scan protocol settings to provide image quality that is specific to the clinical purpose of the CT scan, and applying available dose optimisation features where appropriate.
PET/CT oncology gives one of the greatest CT radiation burdens of all MI examinations, whether performed for diagnostic or localisation/characterisation purposes, as shown in Fig. 1. It is therefore not surprising that it has been the most widely investigated MI examination in terms of CT radiation dosimetry, with several national CT dose surveys published in the literature, namely from France , Bulgaria , USA [12, 13], Korea , UK , Switzerland , and Australia and New Zealand . Excessive variation in dose between facilities was also noted for PET/CT oncology examinations in France , Korea , the UK , and Australia and New Zealand , suggesting great scope for optimisation globally. This Nordic data presents the lowest localisation/characterisation CT doses for PET/CT oncology examinations published in the literature, with third quartile DLP75kg less than half the value published for France . This highlights the importance of using DRLs which are country- or region-specific. The large dose differences with France could reflect that the data for this study was gathered 6 years later, during which time scanners with more sophisticated dose saving technologies may have been utilised, and there may have become a greater awareness of the need for optimisation. This highlights the need to revise NDRLs every 3–5 years . The slightly lower doses compared with the UK may be a result of the Nordic data being weight-derived for a 75-kg person, whilst the UK study did not apply a weight restriction or a weight-derivation due to lack of submitted data on patient weights.
A possible barrier to facilities performing diagnostic CT scans on hybrid systems could be a concern that scanner technologies could be inferior on hybrid systems to those in the radiology department and doses may be higher. Yet, this study demonstrated that diagnostic CT performed as part of PET/CT oncology examinations is common in the Nordics, and that the third quartile CTDIvol,75kg (12.5 mGy), which comprises data mostly from Danish facilities, is within the third quartile NDRL set by the Danish radiation protection authority (Sundhedsstyrelsen, Statens Institute for Strålebeskyttelse (SIS)) in 2015 for thorax/abdomen examinations (17 mGy) . Diagnostic CT doses for PET/CT oncology presented in this study are not suggested NDRL CT doses, as the countries’ already existing NDRLs for standalone diagnostic CT should be used.
Figure 1 shows that other MI examinations can give CT radiation doses in the same range as PET/CT oncology localisation/characterisation scans. Yet, there are still limited CT dose surveys for other PET/CT and SPECT/CT examinations, with national surveys having only been conducted for other examinations in the UK , Switzerland , and Bulgaria . Iball et al. noted large variations in CT doses for other PET/CT and SPECT/CT examinations in the UK  suggesting a need for optimisation. It is therefore important to also survey CT doses in other PET/CT and SPECT/CT examinations, and all ionising radiation exposures should be optimised in keeping with the ALARA principle.
For PET/CT and SPECT/CT cardiac, median and third quartile CTDIvol,75kg values in this Nordic study are comparable with those in UK  and Swiss  studies, as is median DLP75kg. Yet third quartile DLP values in the Nordic study are considerably greater, representing a greater spread in CT scan length where mean and maximum scan lengths were 23 cm and 33 cm as shown in Table 2, compared with 18 cm and 24 cm in the UK study . Presented dose values represent one CT scan, although patients may have two CT scans as part of the complete test (stress and rest).
Nordic main body doses for SPECT/CT bone scans are in broad agreement with UK data , although a tendency for a longer CT scan length is noted in the Nordics. Bone SPECT/CT examinations can cover any body region depending on the patient’s clinical indications. Data were therefore categorised into main body, head, and extremities. It would have been ideal to categorise main body data further according to the specific body part as done in the Swiss survey , but the amount of submitted data was insufficient for this. For SPECT/CT parathyroid, Nordic doses are in general agreement with those from the UK  and Switzerland . However, a greater CT scan length is noted for Nordic protocols, giving slightly higher DLP75kg than other published values. This is the first study to suggest NDRL CT doses for PET/CT brain and SPECT/CT lung and thus, there are no reliable published dose values for comparison. It is also the first study to suggest NDRL CT doses for PET/CT infection/inflammation. However, the results from this survey showed that the vast majority of facilities used the same CT protocol as for PET/CT oncology, thus suggesting that facilities outside of the Nordics wishing to evaluate their local DRL CT doses for PET/CT infection/inflammation could use their NDRLs for PET/CT oncology as a reference, in the absence of NDRL CT doses specifically for PET/CT infection/inflammation.
In this study, reported AC only radiation doses had the greatest variation in dose. Good image detail is not required for AC only; therefore, very low dose settings can be used to provide enough information for a reliable CT-based attenuation map. Hence, very low dose scans are used in some facilities for AC only, whereas standard diagnostic scanner protocols which have not been optimised for AC only have been used in other facilities. Thus, dose variations tend to reduce from AC only purposes through to diagnostic purposes, with the highest maximum/minimum dose difference of 27 times being for AC only in lung SPECT/CT, and the lowest maximum/minimum dose difference of 2.1 times being for diagnostic PET/CT oncology.
Collecting information on CT protocol settings allows further investigation of dose differences between systems. Table 3 shows that the factors contributing to the 27 times difference in dose for AC only lung SPECT/CT include a higher tube voltage and effective mAs for the protocol providing the maximum DLP75kg, compared with a low effective mAs (afforded by low mA and high pitch factor) and use of iterative reconstruction in the lowest dose protocol. A 21 times difference in dose for AC only cardiac PET/CT and SPECT/CT protocols was generated by a very high effective mAs and very long scan length for the heart, in the protocol providing the maximum DLP75kg, compared with a low kV, low effective mAs (comprising low mA and high pitch factor), and use of iterative reconstruction in the lowest dose protocol. For PET/CT brain, the main contributor to the 17 times difference in localisation/characterisation dose was an extremely high mAs from the maximum dose protocol compared with other localisation/characterisation protocols, as it was reported that images were originally intended for fully diagnostic purposes, but following a change in circumstance, the images were only being used for localisation in practice. Further protocol comparisons are made in Table 3.
Mattsson et al.  described how dose-saving features such as tube current modulation, choice of x-ray spectra, iterative reconstruction, and new detectors have the potential to reduce dose considerably. As the type and availability of these features differ between systems, there will inevitably be dose differences. However, the technical capabilities of the systems alone cannot account for all differences in dose seen in this study. Table 3 demonstrates that no single parameter is causing the large differences in dose for all examinations, and the large factors of difference are being generated through some facilities having multiple parameters which they have tried to optimise for clinical purpose, and other facilities having multiple parameters which are not optimised for clinical purpose, and for some examinations may have selected standard diagnostic protocols for AC only and localisation/characterisation scans. Yet, even where efforts have been made to optimise CT protocols for clinical purpose, there will inevitably be differences in reader preferences for noise and resolution, causing variability in parameters and thus dose.
Differences in scan length for the same examination and clinical purpose of CT also contribute to differences in DLP. CT scan lengths for PET examinations are generally standardised, due to the technical phenomenon that PET/CT systems require the attenuation information gleaned from the CT data for the full PET FOV, in order to allow AC of the PET images, because recorded PET photons are not collimated when scanning in the conventional 3D mode . However, given that recorded SPECT photons are collimated, the CT scan can be localised to the anatomical area of interest, whilst still allowing AC SPECT reconstructions over that area . Although scan length does contribute to dose differences in all scans, it was not considered a major contributor to DLP75kg differences for PET/CT oncology, PET/CT brain, and SPECT/CT lung, contributing to less than 1.5 times difference in dose. However, Table 2 shows that scan length differed markedly for some SPECT/CT examinations, with up to 2.4, 2.6, and 3.1 times difference in scan length for PET/CT and SPECT/CT cardiac (AC only), SPECT/CT parathyroid (localisation/characterisation), and SPECT/CT bone (localisation/characterisation) scans respectively. The difference in cardiac SPECT/CT scan length can be explained by some facilities restricting the CT scan to the heart, whereas others scan a much greater area. Parathyroid adenomas are most commonly located around the thyroid bed but can occasionally be ectopic (sublingual region down to the heart) . Therefore, some facilities localise the SPECT/CT scan to the thyroid region whilst another scans 2 fields-of-view (FOV) to cover the full possible ectopic area. For SPECT/CT bone, some facilities perform 3 FOV SPECT/CT scans (head to thigh) as standard without planar whole body imaging, as SPECT/CT is known to have greater sensitivity and specificity than planar imaging , whereas other facilities perform planar whole body gamma camera imaging as standard and supplement this with SPECT/CT over areas of particular interest. The wide variations in scan length for these three examinations are also consistent with the tendency for the Nordic scan lengths shown in Table 2 to be greater than corresponding UK scan lengths . These findings suggest that scan length could be a focus area for optimisation efforts in Nordic SPECT/CT examinations.
Design of a suitable method for reporting CT NDRL doses for CT in MI examinations is essential to enable accurate data comparisons. Many methodological questions arose during the preprocessing of collected data which were difficult to predefine before starting the study. One such source of variability is the clinical purpose of the CT scan. The UK survey grouped data into 3 clinical purposes of CT (attenuation correction, localisation, and fully diagnostic) . This study included a fourth category of characterisation, which should in theory give more detail and thus a higher dose than localisation, but less than diagnostic. However, some facilities communicated that they were not familiar with this term and the data showed no clear distinction in dose between localisation and characterisation. Thus, localisation and characterisation purposes were combined, thereby allowing a greater data pool for generating suggested NDRL CT doses. Furthermore, the validity of these survey results is reliant on the correct clinical purpose of CT being recorded on the data capture form. It is expected that this has been discussed between the relevant health professionals for each facility.
Alkhybari et al. published recommendations for establishing PET/CT and SPECT/CT NDRLs in 2018  after this study had commenced, explaining that future PET/CT and SPECT/CT NDRL data should include a minimum of 50 patients without weight restriction, based on the current ICRP publication . However, this study, similar to that by Iball et al. , is limited by the quantity of data submitted. In both studies, data submissions from a facility for a scanner-examination combination were included if there were data for ten or more patients. This is less than the number recommended by the ICRP . However, given that hybrid examinations have a much longer examination time than standalone CT and submitted data must be further subdivided according clinical purpose, it is difficult to obtain as high a number of data submissions compared with standalone CT. Given these limitations on patient numbers, data were acquired in this study without weight restriction to obtain as much data as possible, but since the data included a maximum of 30 patients per system, doses were then interpolated to a 75-kg person to get a more fair comparison. Alkhybari et al. explained that less than 2% difference in dose has been found between weight-restricted or non-weight-restricted methods, meaning that non-weight-restricted methods are still valid . Data were commonly excluded in this study due to insufficient numbers of patients (less than 10 for systems utilising tube current modulation) due to limited throughput during the data collection period. Other reasons for exclusion included absence of patient body mass data meaning that the data could not be weight-derived, diagnostic CT datasets which were additional to a low dose CT scan or where combined low dose and diagnostic CT data were submitted and could not be separated, and cardiac PET/CT datasets which were not for assessment of myocardial perfusion, such as multiple PET FOV localisation scans for sarcoidosis.
Studies proposing NDRLs for MI examinations have either analysed data according to dose information gathered from a population of systems (scanner types per facility) [2, 11, 15, 17], or per facility (using a dose average across all scanners at a given facility) . This study analysed data according to systems as opposed to facilities, in keeping with the methodology of the other studies covering a broad range of PET/CT and SPECT/CT examinations [2, 11, 15, 17]. Lima et al. identified a possible limitation to this approach, whereby there could be a bias towards facilities with a large number of scanners, but on investigation, they found no significant influence on the distribution of doses .
This study has some recognised limitations. Despite this study gathering a large amount of data from 83 systems across 43 facilities, collected data were not sufficient to suggest NDRL and achievable doses for PET/CT bone and SPECT/CT sentinel node, thyroid post ablation, and octreotide/mIBG examinations. Furthermore, data were insufficient to suggest NDRLs for all clinical purposes for all examinations. Details of all CT acquisition parameters for PET/CT and SPECT/CT were collected with the intention of exploring the protocol settings contributing to the greatest dose variations, which in turn could inform dose optimisation strategies. A basic evaluation was undertaken where possible, as shown in Table 3, but a full evaluation was not feasible, due in part to differences in how scanner vendors define the reference image quality where tube current modulation is applied. This study provides suggested Nordic NDRL data for PET/CT and SPECT/CT scans. However, it is important to note that the data presented in this study are not official NDRLs, as they must first be ratified by the relevant local radiation protection authorities.