EANM practice guideline for quantitative SPECT-CT

https://doi.org/10.1007/s00259-022-06028-9 GUIDELINES EANM practice guideline forquantitative SPECT‑CT JohnC.Dickson1· IanS.Armstrong2· PabloMinguezGabiña3,4· AnaM.Denis‑Bacelar5· AronK.Krizsan6· JonathanM.Gear7· TimVandenWyngaert8,9· Lioe‑FeedeGeus‑Oei10,11· KenHerrmann12 Received: 19 August 2022 / Accepted: 30 October 2022 © The Author(s) 2022 Abstract Purpose Quantitative SPECT-CT is a modality of growing importance with initial developments in post radionuclide therapy dosimetry, and more recent expansion into bone, cardiac and brain imaging together with the concept of theranostics more generally. The aim of this document is to provide guidelines for nuclear medicine departments setting up and developing their quantitative SPECT-CT service with guidance on protocols, harmonisation and clinical use cases. Methods These practice guidelines were written by members of the European Association of Nuclear Medicine Physics, Dosimetry, Oncology and Bone committees representing the current major stakeholders in Quantitative SPECT-CT. The guidelines have also been reviewed and approved by all EANM committees and have been endorsed by the European Association of Nuclear Medicine. Conclusion The present practice guidelines will help practitioners, scientists and researchers perform high-quality quantitative SPECT-CT and will provide a framework for the continuing development of quantitative SPECT-CT as an established modality. Keywords SPECT-CT· Quantification· Dosimetry· Bone· Neurology· Cardiology Preamble The European Association of Nuclear Medicine (EANM) is a professional non-profit medical association that facilitates communication worldwide among individuals pursuing clinical and research excellence in nuclear medicine. The EANM was founded in 1985. These guidelines are intended to assist practitioners in providing appropriate nuclear medicine care for patients. They are not inflexible rules or requirements of practice and are not intended, nor should they be used, to establish a legal standard of care. The ultimate judgment regarding the propriety of any specific procedure or course of action must be made by medical professionals taking into account the unique circumstances of each case. Thus, there is no implication that an approach differing from the guidelines, standing alone, is below the standard of care. On the This article is part of the Topical Collection on Miscellanea * Ken Herrmann [email protected] 1 Institute ofNuclear Medicine, University College London Hospitals Foundation Trust, London, UK 2 Nuclear Medicine, Manchester University NHS Foundation Trust, Manchester, UK 3 Department ofMedical Physics andRadiation Protection, Gurutzeta-Cruces University Hospital/Biocruces Health Research Institute, Barakaldo, Spain 4 Department ofApplied Physics, Faculty ofEngineering, UPV/EHU, Bilbao, Spain 5 National Physical Laboratory, Teddington, UK 6 ScanoMed Nuclear Medicine Centers, Debrecen, Hungary 7 Joint Department ofPhysics Institute ofCancer Research andRoyal Marsden, NHS Foundation Trust, Sutton, Surrey, UK 8 Department ofNuclear Medicine, Antwerp University Hospital, Edegem, Belgium 9 Faculty ofMedicine andHealth Sciences (MICA – IPPON), , University ofAntwerp, Wilrijk, Belgium 10 Department ofRadiology, Section ofNuclear Medicine, Leiden University Medical Center, Leiden, TheNetherlands 11 Biomedical Photonic Imaging Group, University ofTwente, Enschede, TheNetherlands 12 Department ofNuclear Medicine, University ofDuisburg-Essen, andGerman Cancer Consortium (DKTK)-University Hospital Essen, Essen, Germany / Published online: 5 December 2022 European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 contrary, a conscientious practitioner may responsibly adopt a course of action different from that set out in the guidelines when, in the reasonable judgment of the practitioner, such course of action is indicated by the condition of the patient, limitations of available resources or advances in knowledge or technology subsequent to publication of the guidelines. The practice of medicine involves not only the science but also the art of dealing with the prevention, diagnosis, alleviation and treatment of disease. The variety and complexity of human conditions make it impossible to always reach the most appropriate diagnosis or to predict with certainty a particular response to treatment. Therefore, it should be recognised that adherence to these guidelines will not ensure an accurate diagnosis or a successful outcome. All that should be expected is that the practitioner will follow a reasonable course of action based on current knowledge, available resources and the needs of the patient to deliver effective and safe medical care. The sole purpose of these guidelines is to assist practitioners in achieving this objective. Introduction In nuclear medicine, gamma cameras and SPECT-CT systems are routinely used for quantitative imaging. From determining relative kidney performance, to binding ratios in the brain, one of the strengths of gamma camera imaging is its ability to quantify in-vivo physiology for a wide range of conditions and applications. While traditional gamma camera imaging has focused mostly on relative uptake of the radiopharmaceutical, for example to the opposing kidney, or specific structures in the brain, PET imaging has focused on absolute quantification in organs or features, producing images with quantitative units such as kBq/mL or standardised uptake value (SUV). This has brought advantages. The ability to perform absolute quantification, with 18F-FDG for example, provides the opportunity to assess the metabolic status of the disease for diagnosis, staging, treatment monitoring and disease progression. For many years, SPECT-CT imaging has been perceived to be the poor relative of PET-CT for quantitative imaging. However, equipment and software developments including the incorporation of measured CT-based attenuation correction, scatter correction and correction for partial volume effects, have made significant progress in improving SPECTCT quantification [1]. Using modern techniques, quantification of SPECT-CT data is now possible in a similar way as it is in PET-CT. However, quantitative SPECT-CT offers many advantages over PET-CT, which has the potential to lead to a wider range of applications. SPECT uses longer physical half-life radiopharmaceuticals that can better match biological processes. While longer half-life radionuclides such as 64Cu, or 124I are available in PET, these have inferior imaging characteristics in terms of their positron emission probability and high effective dose [2, 3], which diminish some of the advantages of PET imaging. A further advantage of SPECT is that its radiopharmaceuticals can be labelled with different radionuclides, offering the possibility to image multiple physiological processes at the same time. Studies with CZT SPECT systems are showing the ability to perform simultaneous 99mTc/123I imaging [4], which, for example, could be applied in simultaneous perfusion/innervation studies in the heart, and perfusion/DaT (dopamine transporter) availability in the brain. But it is the widespread availability of SPECTCT and the wide range of accompanying radiopharmaceuticals which provides the greatest advantage of quantitative SPECT-CT over PET-CT, with much wider global market penetration due to the lower cost of imaging equipment, and the straightforward radiopharmaceutical production. Although at an early stage, the applications of quantitative SPECT-CT are becoming clearer. The driving force in the increasing interest in quantitative SPECT-CT comes from the growth in radionuclide therapy and theranostics, and the associated growth in personalised treatment planning. The unique advantage of many nuclear medicine therapies employing gamma-emitting therapeutic radionuclides is that they allow us to image and quantify the radiopharmaceutical bio-distribution after administration by SPECT-CT imaging. Because of this, internal absorbed dose calculations of radionuclide therapies using current commonly used therapeutic radionuclides such as 177Lu and 131I [5, 6] present the opportunity to verify treatment delivery and/or personalise the treatment over the course of multi-cycle administrations. Furthermore, imaging surrogates used with quantitative SPECT-CT can be used for diagnosis and patient selection prior to radionuclide therapy, in addition to assessment of disease progression [7, 8] or treatment response. Applications in the quantification of bone tracers for orthopaedic [9, 10] and cancer applications [11] are developing, while opportunities in cardiac [12] and neurological [13] imaging also exist. Given these early, but rapidly developing applications, it is vital that integration of such technology into clinical practice is performed correctly. This is necessary in terms of correct image interpretation and to facilitate the pooling of data to improve scientific knowledge. Furthermore, for successful adoption, it is necessary that techniques are robust, consistent and correctly understood by the user for their application. There must also be clear applications of the technology. For example, quantitative SPECT-CT may enable a better understanding of why a certain subgroup of patients does not benefit from radionuclide therapy despite adequately expressing the target. And in the development of new radionuclide therapies with a narrower therapeutic index, quantitative SPECT-CT has the potential to support 981European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 clinical development in identifying the optimal activity to be administered to each patient. The incremental value of using quantitative SPECT-CT for clinical care needs, however, still needs to be demonstrated. A definition of how to appropriately perform quantitative SPECT-CT is therefore essential to provide a stable platform for the understanding and application of this technology. These guidelines have been written with the objective of defining standards for quantitative SPECT-CT. Reviewing the current status of quantitative imaging in SPECT-CT and highlighting limitations and areas in need of development, we demonstrate where quantitative imaging is potentially most beneficial. In these guidelines, we describe the procedural considerations that must be met in both achieving and implementing quantitative measurements for SPECT imaging and provide clinical applications for which its implementation has already been successful. Imaging protocols forquantitative SPECT‑CT Overview The scope of this section is to provide technical recommendations on image acquisition and reconstruction to achieve reliable quantitative SPECT-CT images. They are intended to be used with gamma cameras with conventional planar detectors rotating around the patient during the SPECT acquisition. Other gamma camera designs such as those with solid state detectors or novel non-parallel collimation are not covered within these guidelines, although some of the concepts described are still relevant. The primary aim for quantitative SPECT-CT is to produce a tomographic image with voxel values representing activity concentration. The utilisation of these voxel values will then be dependent on the clinical application. For radionuclide dosimetry (for diagnostic and therapeutic applications), the absolute activity in Becquerels (Bq) within a delineated volume or organ is extracted from the images, and hence accuracy is paramount. For other diagnostic applications, voxel values may be converted to standardised uptake values (SUV). In this latter example, the SUV should translate to a clinically relevant biomarker and arguably the reproducibility of this metric is more relevant than its absolute accuracy. This is particularly pertinent to follow-up studies monitoring disease response to treatments. While activity concentration for dosimetry is almost always defined in a volume representing an organ or cancerous deposit, SUV can be measured using several metrics. Like in dosimetry, the mean value of SUV (SUVmean) in a feature can be defined. However, SUVmax is more commonly used as a better representation of the intensity of uptake in a feature, even with its limitation of being reliant on a single pixel value and therefore being more susceptible to image noise affecting both the bias and precision of the measure. SUVpeak is an alternative measure for quantitative SPECT, capturing the status of metabolically active features while mitigating image noise [14, 15]. Given that it can mitigate for differences in image quality and contrast recovery between imaging systems it is also well-suited to multi-centre studies [16]. All three SUV metrics are used in clinical practice depending on the objective of the measurement, but it is important to understand the advantages and the limitations of each SUV metric. There are many image-related factors, both controllable and uncontrollable, that must be considered when performing optimisation of quantitative SPECT-CT. Most of these factors will lead to negative bias (an underestimation of the actual activity concentration) in the quantitative values while others will lead to positive bias (an overestimation of activity concentration). Assuming that reconstruction corrections, such as measured attenuation and scatter, and a robust crosscalibration between the radionuclide calibrator and gamma camera are performed correctly, negative bias is largely due to partial volume effects from limitations in spatial resolution. Positive bias is mainly attributed to Poisson noise, arising from relatively low pixel values due to the constraints of scan time and patient administered activity, but may also be associated with reconstruction artefacts in some situations. Some technical factors that will influence these negative and positive bias factors are given in Table1. The optimisation process of quantitative SPECT-CT must evaluate the influence of the controllable variables on the accuracy and reproducibility of image-derived measurements. Furthermore, the parameters should be chosen such that they minimise, as far as possible, the influence of the non-controllable factors. It is essential that this is characterised using suitable phantom data, which will allow errors in the measurement to be determined. This section describes the considerations needed from both the image acquisition and reconstruction processes in achieving reliable quantitative data from SPECT-CT images. It should be highlighted that centres intending to perform quantitative SPECT-CT should have suitable equipment to assess system performance with a minimum of uniform phantom to assess calibration and an IEC image quality (IQ) Phantom, a Jaszczak Phantom or similar to characterise activity concentration recovery coefficients (ACR) for various object sizes. Scanner calibration andcharacterisation The differentiating factor between traditional SPECT-CT and quantitative SPECT-CT is the sensitivity calibration factor. Hence, the first step is to determine a reliable crosscalibration factor between a radionuclide calibrator and the 982 European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 gamma camera. The exact technique may differ depending on the system manufacturer and the radionuclide being used. A prerequisite for scanner calibration is a robust measurement of activity in a radionuclide calibrator. Activities must be traceable to national and international standards. International guidelines recommend accuracies of 5–10% for diagnostic, and 5% for therapeutic radionuclides [17, 18]. The use of activity measurements traceable to primary standards is not common practice in all countries [19], even though larger variabilities have been observed in international comparison exercises [20]. Therefore, radionuclide calibrators should be regularly maintained and calibrated against a primary standard for the radionuclide and measurement geometry of interest following available good practice guidelines [21, 22]. There are two distinct considerations that will influence the accuracy of quantitative data extracted from reconstructed SPECT-CT images. These are the activity calibration and characterisation of the system. It is important to appreciate the difference in these concepts. Calibration of a system is the act of including the activity calibration factor of the system to produce activity concentration measurements in Bq/mL. Characterisation will describe the performance of a given measurement technique, for example SUVmax, derived from images obtained from a specific combination of acquisition and reconstruction parameters. Both will be impacted by the acquisition and reconstruction parameters chosen as has been described in Table1. Clearly, a key component of quantitative SPECT-CT is the accuracy of the calibration factor relating reconstructed counts from the scanner to activity measured in a radionuclide calibrator whose measurements should be traceable to a primary standard. Errors in this calibration factor will lead to systematic bias in quantitative values derived from the images. It is important to appreciate that this is a two-stage process: one of calibration, and another of verification of quantitative accuracy following the calibration process. The calibration process itself will vary according to the manufacturer’s requirements, which may range from a petri-dish source planar measurement; a long-lived sealed point source planar measurement; or volumetric measurements from a large uniformly filled phantom acquired, reconstructed and corrected using the same parameters used for patient imaging. Regardless of the method, it is recommended to perform this calibration using the method and with the frequency suggested by your system manufacturer. If no recommendations are given, the calibration should be performed at least annually, or after any major changes to hardware or software using a volumetric approach given that it better represents imaging conditions [23, 24] Once calibration has been performed, it must be verified before clinical use. This should be performed using a large uniform volumetric phantom filled with the appropriate Table 1 Controllable and uncontrollable factors that influence the quantitative accuracy of activity concentration measurements obtained from SPECT-CT images Factor Controllable? Impact of quantitative accuracy Administered activity Sometimes For SUVmax and other regions based upon threshold of SPECT voxel values, positive bias can occur due to changes in noise level Note that geometric regions, manually delineated on whole organs or lesions where all encompassed SPECT voxel values are averaged, are less susceptible to variations in image noise levels It should be noted that for diagnostic applications, the administered activity is defined locally while, for therapeutic applications, it is likely to be either a fixed activity or defined by the planned therapy absorbed dose for the patient Acquisition time Yes Positive bias from image noise—see comments above regarding administered activity Collimator Yes Negative bias due to degrading spatial resolution Matrix size Yes Negative bias due to changes in spatial sampling Positive bias due to changes in noise SPECT orbit radius Yes Negative bias due to degrading spatial resolution for larger radii Number of updates (product of iterations and subsets) for iterative reconstruction Yes Negative bias due to under-converged image Positive bias due to image noise Post-reconstruction smoothing filter Yes Negative bias due to additional image blurring Positive bias due to control of image noise Dead-time No Negative bias due to dead-time effects at very high count-rate levels Lesion/organ size and shape No Negative bias due to degrading spatial resolution for small volumes Organ-to-background contrast No Variable due to changes in spill-over from surrounding activity Organ location No Negative bias due to increasing distance from the detector Patient movement No Negative bias due to increasing image blurring 983European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 radionuclide and acquired on the SPECT-CT with a clinical imaging protocol. If the initial calibration is performed using a volumetric method, the verification scan should not be performed immediately after the calibration using the same phantom as any errors in the activity measurement or phantom filling will simply transfer across to the verification scan. The verification should be performed with a freshly filled phantom to allow a test of the entire process of activity measurement, phantom filling and image acquisition. The importance of establishing a good technique of measuring activity in the radionuclide calibrator should not be neglected and should mirror the measurement approach used for patients. If patient injections are performed in specific syringe sizes and diluted to a given volume, this should be replicated when measuring the activity for the calibration and verification phantom scans. The outcome of verification is to ensure that the measured activity concentration agrees with the true activity concentration. Once a system has been calibrated, it is essential to characterise the relationship between object size and measurement accuracy. This relationship is commonly demonstrated in the form of activity concentration recovery (ACR) curves, which are commonly derived in PET-CT [25]. These measurements are performed using phantoms, such as the NEMA IEC IQ or Jaszczak phantoms, with a representative range of fillable volumes. Recovery curves are specific to the contrast and measurement of activity concentration that has been performede.g., SUVmax, SUVmean and hence care should be taken to ensure clinically relevant measurements are taken when characterising the system. Figure1 shows an example of an ACR curve. Acquisition Acquisition parameters can vary depending on the patient investigation being performed and, on the characteristics of the camera. Each centre should therefore optimise its own acquisition parameters. Some steps that should be followed in the acquisition process are: • Steps should be taken to limit the possibility of patient motion. It is important that the patient remains in the same position during both the CT and SPECT acquisition to ensure good image registration and accurate CT attenuation correction. • The optimal collimator will depend on the radionuclide being imaged. Relevant imaging guidelines should be followed when choosing an appropriate collimator. • Step and shoot or continuous acquisition mode of acquisition can be used. The latter can offer a 1–2min saving on scanning time over 60 rotation angles. • Detector auto-contouring is advised to minimise the distance between the detectors and patient to provide optimal spatial resolution. However, for some applications detectors can be kept at a fixed but close distance. • Acquisition should typically be performed with opposing detectors at 180° from one another, but for cardiac applications a 90° configuration may be used. • A pixel size smaller than half the full width at half maximum (FWHM) spatial resolution of the system for the radionuclide used is recommended to ensure appropriate spatial sampling. Commonly, a matrix size of 128 × 128 is used. It should be noted that decreasing the pixel size results in a noisier image. • The number of projections is recommended to be similar to the matrix size (e.g. 120–128 projections for a 128 × 128 matrix) to ensure appropriate angular sampling. Cardiac applications using a 90° detector configuration may use a reduced arc and number of projections, although distortion and inaccurate quantification will occur when there are insufficient projection data—typically distant from the heart. • The time per projection will depend on the amount of radioactivity in the patient. As noise in the projection data follows a Poisson distribution, and in reconstructed data is much worse [26], imaging time must be high enough to reduce image noise as much as possible. If multiple fields of view (FOVs) are acquired, the time per projection may have to be decreased for patient comfort. Fig. 1 Example of an activity concentration recovery (ACR) curve obtained from the six hot spheres of a NEMA IEC image quality phantom filled with 10:1 contrast to background. In this example, activity concentration was measured by the maximum voxel value, which is akin to SUVmax 984 European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 Reconstruction Iterative methods are recommended to reconstruct the acquired SPECT projections. Normally, the algorithm used will be that included in the software provided by the vendor of the gamma camera; however, third-party algorithms are also available. For quantitative purposes, the number of updates, defined by the product of the number of iterations and subsets, may be greater than for reconstructions with qualitative purposes [27]. Preferably, that number should be obtained from phantom measurements in which the convergence of activity concentration is studied, paying attention not to reach excessive noise levels or introduce image artefacts, such as Gibbs artefacts from the use of resolution modelling. Ideally, such optimisation should be performed using anthropomorphic phantoms which mimic the clinical situation, but such phantoms are not always available. If NEMA IEC IQ or Jaszczak phantoms are used as part of the optimisation process, it should be noted that SPECT recovery curves are far more dependent on sphere position than PET. This is likely due to the variable detector radius acquisition resulting from auto-contouring [28] and so it is recommended that multiple sphere configurations are evaluated. Figure2 shows an example of how the variability of ACR can be aligned across different sphere configurations when sufficient updates are performed. Corrections Attenuationcorrection Attenuation correction based on CT data should be used for quantitative SPECT-CT. The CT images should be converted to a map of attenuation coefficients for the appropriate radionuclide energy and incorporated into the iterative reconstruction process [29]. It must also be possible to measure the full path length of the gamma-ray from the Fig. 2 NEMA IEC image quality phantom filled with 99mTc at 10:1 contrast in three different sphere configurations reconstructed with 3 iterations and 6 subsets (upper set of images) and 20 iterations and 6 subsets (lower set of images). The corresponding recovery curves are given below for maximum voxel ACR. Note that all phantom images shown have corrections for attenuation, scatter and collimator response included and a 10mm Gaussian filter applied 985European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 point of emission to the patient boundary on the attenuation map. If the attenuation map is truncated, either by way of large patient habitus or inadequate CT field of view, then the attenuation correction will not be accurate and quantitative errors will occur. Hence, care should be taken when positioning the patient to minimise the likelihood of truncation. Where appropriate, and when available, it is recommended to use metal artefact correction algorithms [30] to produce more suitable attenuation maps [31]. Scatter correction To correct for scattered gamma-rays present within the photopeak window, multiple energy window scatter correction methods are typically applied, although model-based scatter correction can also be used if available [32]. Dual-energy window methods can be applied if there are no emissions above the photopeak of the radionuclide used; otherwise, a triple-energy window method should be implemented. Smoothing of the scatter window image may also be beneficial to reduce propagation of image noise from the correction to the reconstructed image. It is important to validate scatter correction techniques using appropriate phantoms containing areas of no activity, surrounded by uniform activity, to demonstrate that the algorithms do not over-correct the final images. Resolution modelling Resolution modelling is available in most modern reconstruction software and partially compensates for the limited spatial resolution due to the collimator and the detector by incorporating a depth-dependent collimator response model in the projection operation of the iterative reconstruction method [33]. It is recommended that resolution modelling techniques are used if available to improve quantification. Decay correction Understanding how and when decay correction is applied is important in quantitative SPECT-CT. Given the relatively long physical half-life of most SPECT radionuclides, its application to ensure differences are accounted for in the acquisition of the first and last projection are relatively minor. One exception being 81mKr SPECT-CT where differences in acquired projection counts will be large and application of such corrections essential. In multiple SPECT field of view studies where the study may take up to 1h, decay correction should also be performed to ensure consistency of relative pixel values across all acquired projections. A second area where decay correction is important is in the measurement of SUV metrics. With SUV normalising uptake to injected activity, it is important that SPECT data acquired several hours or even days later accounts for the decay (especially for radionuclides such as 177Lu). Ensuring clocks on scanners and in injection rooms are consistent is an important element of ensuring this is done appropriately. Partial volume (sphere andnon‑sphere) Resolution modelling will rarely fully compensate for the limited spatial resolution of the gamma camera. Hence, to quantify the activity in the volumes of interest (VOI), postreconstruction corrections can be applied. These recovery coefficients are commonly obtained from phantoms with spherical inserts of known volume that are filled with known activity concentrations [23]. Note that the applicability of this method has limitations for non-spherical volumes. For some anatomical regions (e.g. brain), methods based on anatomical information such as the geometric-transfer matrix method can be applied [34]. While partial volume corrections are not generally available with commercial scanner software and therefore not clinically available, ‘homemade’ site applied corrections can be useful for dosimetry, and some research applications. Note that for clinical use, any site written software should be in alignment with appropriate medical device legislation [35]. Dead‑time For acquisitions performed soon after administering therapeutic amounts of radiopharmaceuticals, e.g. treatment of neuroblastoma with 131I-mIBG or 177Lutetium peptide therapy of neuroendocrine tumours, dead-time correction may need to be applied [36, 37]. The non-linearity of the count rate will not be the same for every projection angle. Therefore, the dead-time correction should preferably be applied on each projection acquired [38]. However, for simplicity, an average correction could be applied based on the count rate averaged over all projections. Reconstruction post‑filtering The application of a post-reconstruction smoothing filter will inevitably degrade the spatial resolution and exaggerate partial volume effects. It is therefore often assumed that filtering should not be applied to images for quantitative applications. However, the application of a post-filter is intended to control the degree of noise in the image and hence the potential positive bias that may arise from noise. The application and choice of a post-reconstruction filter is dependent on the desired outcome from the image and nature of the measurement. Insituations where large regions are drawn, such as organ delineation, and where mean activity concentration measurements are derived from all voxels, then a post-filter is unlikely to be beneficial. However, 986 European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 insituations where images are drawn on small objects and maximum voxel values are extracted, i.e. where noise is a more significant influence, then a post-filter can be helpful in ensuring a greater level of consistency. Figure3 demonstrates how the application of a 10mm Gaussian postfilter can substantially improve the consistency of measurements of activity concentration. The filter has been applied to images reconstructed with the same number of iterative updates. It is shown that the post-filter increases the degree of negative bias in the smaller objects, but removes the positive bias, due to noise, present in the larger spheres of the unfiltered images. For clinical reporting, many nuclear medicine images are produced according to a visual preference of the clinician, and hence it may be appropriate to create a second reconstruction that is optimised for quantification. Quality control andharmonisation Quality control requirements Acceptance testing and quality control of SPECT-CT systems for SPECT and CT components should follow international guidelines [21, 39]. Given the importance of attenuation correction for quantitative accuracy, focus should also be given to the alignment between SPECT and CT, which must be checked periodically. The sensitivity calibration workflow can differ significantly between vendors. It is therefore recommended that the local physics team should determine the appropriate frequency in which to perform verification scans of sensitivity, based on manufacturer recommendations. The sensitivity calibration must be repeated with major changes in software or hardware. Fig. 3 NEMA IEC image quality phantom filled with 99mTc at 10:1 contrast in three different sphere configurations reconstructed with 120 updates with no post-filter (upper set of images) and a 10mm Gaussian post-filter (lower set of images). The corresponding recovery curves are given below for maximum voxel ACR 987European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 The need andmethods ofharmonisation Accurate and reproducible measurements of radioactivity are essential for quantitative SPECT-CT imaging, enabling the comparison of results from multi-centre studies [40]. This can be achieved by establishing traceable results [41] that can be related to a reference through a documented unbroken chain of calibrations, each contributing to the measurement uncertainty [42]. However, the uncertainties at all stages of the imaging acquisition and processing chain are not presently known or measured, partly due to the variations in available correction methods, reconstruction algorithms and the black-box nature of the available software. Full traceability is therefore not presently achievable, often stopping with the activity measurements in the radionuclide calibrator prior to use in phantom or patient studies. Harmonisation of 18F-FDG PET-CT has been addressed by the EANM Research Ltd EARL accreditation programme [43], and a similar approach is currently in development for SPECT-CT. The initiative in SPECT-CT was started following the recent interest in absolute quantification for this modality, mainly driven by the need of dosimetry following radionuclide therapy. Many investigators have explored the variability of activity quantification in interlaboratory and multi-centre studies for 99mTc [44–47], 123I [48, 49], 131I [49, 50], 133Ba [51], 177Lu [24, 52, 53], 223Ra [54], highlighting the need for harmonisation protocols. Accounting for differences in availability of local resources, a set of minimum requirements to harmonise SPECT-CT imaging across centres is recommended: • A radionuclide calibrator traceable to a national standards laboratory. • A suitable and accessible phantom set to calibrate the scanner and to assess partial volume effects through ACR curves following the recommendations described earlier. • Standard operating procedures (SOP) for traceable phantom preparation, image acquisition and reconstruction, assessment of partial volume effects, outlining volumes of interest and reporting of results [19, 55]. • A verification phantom study is recommended to assess the quantitative accuracy across centres. This can be a more realistic geometry, e.g. a circular or elliptical cylindrical phantom to match the clinical condition under consideration [24, 49]. All the scans must be acquired and reconstructed with the same protocols used for the specific clinical condition. For dosimetry following radionuclide therapy and to enable a quantitative comparison between centres, it is recommended to calculate the uncertainties in the recovery coefficients following EANM guidelines [56]. The path toclinical use Quantitative SPECT-CT is an emerging imaging technology, but, as with any new technology, its success depends on whether routine clinical applications can be identified. Not every technical evolution is automatically translated into wide clinical acceptance, and depends on issues including impact, ease-of-use, cost, availability and an adequately trained workforce [57]. If quantitative SPECT-CT is here to stay, it should answer clinically relevant questions and impact patient treatment and outcome. Recent developments in nuclear medicine in association to new theranostic approaches [58] support the use of SPECTCT for some radionuclides, such as 177Lu and 131I to visualise the efficacy of treatments. Quantitative SPECT-CT can quantify how much activity is delivered to each tumour lesion and organ at risk, and consequently, the absorbed dose. This may be of high clinical importance for personalised medicine, even though confirmation in future studies is needed. The estimation of absolute activity concentration may also be attractive for several other purposes: (1) to deliver a reliable diagnosis, (2) for accurate therapy response monitoring, (3) for prognosis and to guide patient management decisions, (4) to improve the reproducibility of interpretations, (5) to allow comparison of data between centres (6) and to facilitate (semi)automatic analysis. Initial applications reported in the literature include the assessment of skeletal conditions (e.g. bone metabolism, detection of bone metastases, mandibular condyle asymmetry), coronary artery disease, amyloidosis and parkinsonism. For quantitative SPECT-CT, it will therefore be essential to define the limitations of the measures produced by this technique. As different applications have different accuracy/ precision requirements, understanding the technology’s limitations will help guide the focus towards areas with the highest likelihood of successful clinical implementation. Optimisation for a range of applications, radionuclides, geometries and activities will also be necessary, as will the transferability of results. The cost-effectiveness of the technique with regards to the humanistic and societal outcomes must also still be proven, which is beyond the scope of these guidelines. However, these guidelines should aid in the design of clinical trials with the appropriate methodology required to demonstrate the value of the technique. Clinical use cases Dosimetry Until recently, radionuclide therapies were dominated by the use of radioactive iodine, which has been used 988 European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3 87. Chan BH, Leung YY. SPECT bone scintigraphy for the assessment of condylar growth activity in mandibular asymmetry: is it accurate? Int J Oral Maxillofac Surg. 2018;47(4):470–9. 88. Kim J, etal. Maximum standardised uptake value of quantitative bone SPECT/CT in patients with medial compartment osteoarthritis of the knee. Clin Radiol. 2017;72(7):580–9. 89. Braun M, Kachovan M, Vija A, Pagenstert G, Wild D, Kretzschmar M. Quantitative 99mTc-DPD-SPECT/CT for the detection of prosthetic loosening in patients with hipand knee joint replacement—an interim analysis of a prospective study. J Nucl Med. 2019;60(supplement 1):92 LP – 92. 90. Beyer T, etal. What scans we will read: imaging instrumentation trends in clinical oncology. Cancer Imaging. 2020;20(1):38. 91. Liu J, Yang Y, Wernick MN, Pretorius PH, King MA. Deep learning with noise-to-noise training for denoising in SPECT myocardial perfusion imaging. Med Phys. 2021;48(1):156–68. 92. Ahn S, etal. Quantitative comparison of OSEM and penalized likelihood image reconstruction using relative difference penalties for clinical PET. Phys Med Biol. 2015;60(15):5733–51. 93. Ryden T, Marin I, van Essen M, Svensson J, Bernhardt P. Deep learning generation of intermediate projections and Monte Carlo based reconstruction improves 177Lu SPECT images reconstructed with sparse acquired projections. J Nucl Med. 2019;60(supplement 1):528–35. 94. Shao W, Rowe SP, Du Y. SPECTnet: a deep learning neural network for SPECT image reconstruction. Ann Transl Med. 2021;9(9):819–819. 95. Seifert R, Weber M, Kocakavuk E, Rischpler C, and Kersting D. AI and machine learning in nuclear medicine: future perspectives, Semin Nucl Med. 2020;51(2):170–7. 96. Arabi H, AkhavanAllaf A, Sanaat A, Shiri I, Zaidi H. The promise of artificial intelligence and deep learning in PET and SPECT imaging. Phys Medica. 2021;83:122–37. Publisher's note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations. 995European Journal of Nuclear Medicine and Molecular Imaging (2023) 50:980–995 1 3