First clinical experience of a ring‐configured cadmium zinc telluride camera: A comparative study versus conventional gamma camera systems

A novel semiconductor cadmium zinc telluride (CZT) gamma camera system using a block sequential regularized expectation maximization (BSREM) reconstruction algorithm is now clinically available. Here we investigate how a multi‐purpose ring‐configurated CZT system can be safely applied in clinics and describe the initial optimization process.

Imaging single-photon emitting radionuclides can be traced to the development of the sodium iodide (NaI(Tl)), the scintillation camera invented by Hal Anger in the 1950s (Anger, 1952).The design of a NaI(Tl) scintillation crystal, coupled with an array of photomultiplier tubes with a collimator, has formed the basis for clinical nuclear medicine imaging systems for decades.Two-dimensional (2D) planar imaging was standard practice for many years.Later, single-photon emission computed tomography (SPECT) was developed and enabled three-dimensional (3D) imaging.SPECT is now commonly coupled to computed tomography (CT) for attenuation correction (AC) and localization.
Hardware advances have included cameras with solid-state cadmium zinc telluride (CZT) detectors (Madsen, 2007;Peterson & Furenlid, 2011).Solid-state detectors offer higher energy resolution (Desmonts et al., 2020;Goshen et al., 2018;Keider et al., 2016;Le Rouzic & Zananiri, 2021;Takahashi et al., 2013).Recently, new multipurpose CZT cameras with 12 detectors placed in a ring configuration were introduced first by Spectrum Dynamics (Caesarea, Israel) in 2018 (Veriton) and then by GE Healthcare (Haifa, Israel) in 2021 (Starguide).Standard SPECT systems consist of two detector heads, and a scan is generally performed by several planar acquisitions at different angles around the patient and then reconstructed to a 3D volume.
Advances in reconstruction algorithms have also been achieved in parallel with the development of better hardware.One example is the block sequential regularized expectation maximization (BSREM) algorithm commercially known as Q.Clear (Green, 1990;Mustafovic & Thielemans, 2001;Ross, 2014).Compared to ordered subset expectation maximization (OSEM), BSREM can maintain a low noise level as the number of iterations increases.This reduces the need to 'stop early' as a strategy for noise control and allows for more iterations while keeping the noise level at clinically acceptable levels.
An increasing number of iterations enhances the probability of detecting small lesions.The GE Starguide system is equipped with the BSREM reconstruction algorithm.
In 2022, two GE Starguide SPECT-CT systems were installed at Skåne University Hospital.Here, we present our initial experiences and protocol optimization for common nuclear medicine examinations.

| SPECT system
The Starguide SPECT-CT system is a 12-detector pixelated CZT camera with the detector arranged in a ring configuration.Each detector is 3.93 cm wide and consists of seven pixelated CZT chips, each with 16 × 16 individual pixels for a total axial field of view (FOV) of 27.5 cm.The detectors are equipped with tungsten parallel-hole collimators aligned with the pixel array.To collect data from the entire volume within the FOV, the detectors have a sweeping motion on each detector at each acquisition step combined with a gantry rotation of a few angular steps-typically four steps, to include the entire patient volume.The detectors are positioned near the body contour using body scanning with infrared transmitters and receiver diodes.The acquisition can be performed in focus mode, thus enabling data collection in a user-defined region of interest (ROI).The percentage of the sweep time the detector spends in the focus area is set by the operator and is normally 80%-90% to avoid truncation.
The sweep time can be set to 1 s, enabling dynamic SPECT imaging with a frame time of as low as 1 s.During dynamic imaging, there is no gantry movement.Figure 1 shows the Starguide system, the full sweeping range for one detector, and the sweeping range within the user-defined ROI.
Image reconstruction on the Starguide can be performed using OSEM or BSREM.BSREM has been used to reconstruct positron emission computed tomography (PET) images for a while now (Lindström et al., 2018;Ross, 2014;Trägårdh et al., 2019Trägårdh et al., , 2020) ) and has now been introduced with the Starguide system for SPECT reconstruction.The BSREM reconstructions enable two regularization methods to control the noise: the relative difference prior (RDP) and the median root prior (MRP) (Green, 1990;Mustafovic & Thielemans, 2001).
F I G U R E 1 The Starguide architecture.A picture of the Starguide system where the detectors are fully extended (a).Illustration of the full sweeping range for each detector and the steps of the total gantry rotation (b).The detectors' sweeping range for focus mode within the userdefined ROI (c).ROI, region of interest.
The RDP uses the relative value in the adjacent voxels to identify the noise, and MRP uses the median voxel values in adjacent voxels to identify the noise.For RDP, a gamma value is also set that controls the shape of the a priori generating images with sharper edges.The ability to regulate both the strength of the penalty function and the edgepreserving gamma parameters makes the RDP regularization suitable for bone and lung scans.In comparison, MRP generates smoother images making this the regularization of choice for cardiac scans.
The Starguide software (SmartConsole; GE Healthcare) also enables reconstructions using functional anatomical mutual enhancement (FAME).FAME is a post-processing regularization method that aligns and limits the edges of the uptake in the SPECT image using anatomical structures in the CT.The 3D data can also generate reprojected 2D planar, anterior, and posterior images (pseudoplanar).The reprojection can be CT based, using the attenuation map from the CT to simulate the anterior and posterior detector view.The software also enables the management of list mode data, which enables optimization of the scan time and energy window setting after the acquisition is performed.
In this study, we compared images acquired with a conventional gamma camera used in clinical practice with images acquired with The energy window used for 99m Tc on DNMCT 670, DNM 630 and Starguide was 140 keV ± 10% with a scatter window of 120 keV ± 5%.The energy window was 140 keV ± 7.5% for DNM 530c.

| Clinical images
Seventy-six patients were examined with the Starguide system and on one of the conventional gamma cameras for comparison (Table 1).
The study was approved by the Swedish Ethical Review Authority (#2020-07213), and written consent was obtained before imaging.
The images acquired on the conventional gamma camera were reconstructed according to clinical practice.The reconstruction parameters for the Starguide were optimized with factory-defined parameters as a starting point.The initial scan time was set at a level that guaranteed good image quality for all scan types aiming for higher contrast, lower noise level and fewer artefacts compared to images from the conventional system.The validation consisted of a visual assessment of all studies for general image quality, noise level and contrast by eight experienced nuclear medicine physicians (three for bone scan, two for cardiac scan and three for lung scan) using Hermes software for assessment (HERMES Medical Solutions).

| Bone scan
Eighteen patients were injected with 570 MBq 99m Tc-labelled bisphosphonate and scanned on the first camera system 3-4 h after the administration.Forty-four per cent of the patients were first imaged on the DNMCT 670, followed by imaging on Starguide.

| Conventional gamma camera
The clinical protocol included a whole-body bone scan with a LEHR collimator (anterior and posterior views, 15 cm/min speed, zoom 1 and pixel size of 2.21 mm).SPECT-CT (step and shoot mode with 19 s/step and 60 views over a 180°arc per detector and voxel size 4.42 mm) was done if the information in the planar image was insufficient; this decision was made after a physician or technologist assessed the planar scintigraphy.Reconstruction was performed with OSEM (two iterations and 10 subsets and post-processing with Butterworth filter with order 10, cut-off frequency 0.6 cycle/pixel).
CT for AC and anatomic localization was performed (GE BrightSpeed, tube voltage 120 kV, rotation time 0.6 s, pitch 1.375 tube current auto mA 20-90 with noise index 40.2,attenuation map created separately), covering the same part as the SPECT.All images were reconstructed with resolution recovery and scatter correction (lower energy window).

T A B L E 1
The number of patients scanned on both the conventional gamma camera and on Starguide.were evaluated side-by-side to assess which scan time had unacceptable quality.This information was then used to set a lower limit for the scan time with the chosen beta and gamma values.Image quality was evaluated by visually assessing the contrast, noise and artefacts side-by-side in the images.The reconstruction parameters were then applied to all 18 patients and evaluated by three nuclear medicine physicians.All images were viewed with and without FAME.
The SPECT data was then used to generate the pseudo-planar images, and assessments were done to verify that the chosen parameters were clinically acceptable by comparing them to the conventional planar images.

| Cardiac scan
Thirty-two patients with a mean body mass index (BMI) of 27 kg/m 2 were examined with a 1-day stress-rest protocol.The 99m Tctetrofosmin (2.5 MBq/kg) was administrated in stress (exerciseinduced stress in 46% of the patients and pharmacologic stress in others).A higher amount of activity was administered in rest three times if rest was assessed as required according to clinical routine (14% of the patients performed a rest scan).Fifty-seven per cent of the patients were first imaged on the DNM530c, followed by imaging on Starguide.

| Dedicated cardiac camera
The acquisition times on DNM 530c were 8 min (stress) and 3 min 40 s (rest).Patients were imaged supine with their arms positioned over the head, and electrocardiogram-gated scans were acquired using eight bins.Reconstruction was performed with the Maximum Likelihood Estimation Method with Green's regularization, which is a BSREM algorithm, using 40 iterations with an alpha of 0.4 and a beta of 0.4 for stress/rest (Bocher et al., 2010).The gated images were reconstructed with 40 iterations (stress) and 50 iterations (rest) and alpha 0.51 (stress) and 0.51 (rest) and beta 0.3 (stress) and 0.2 (rest).
All images used a Butterworth post-processing filter with a cut-off frequency of 0.4 cycle/pixel and an order of 10 for gated images; the cut-off was 0.37 cycle/pixel, and the order was 7 for non-gated images.Non-gated imaging was also performed in the prone position to facilitate the assessment of attenuation artefacts because the system does not have CT.

| Conventional gamma camera
SPECT imaging was performed using step-and-shoot mode with ELEGP collimators and 10 s/step for ventilation and 5 s/step for perfusion.There were 60 views over 180°per detector, a zoom of 1.33 and a pixel size of 6.64 mm.The reconstruction and visual assessment for the conventional gamma camera were performed in OASIS (Segami) using OSEM with three iterations and eight subsets and collimator correction.

| Starguide
The scan time for the ventilation and perfusion measurements were

| Bone scan
The bone scan optimization did not improve visual image quality with more than 10 iterations using RDP; therefore, the reconstruction was set to 10 iterations with 10 subsets.The gamma was set to two because larger gamma values tend to amplify noise which can be hard to distinguish from real hotspots.The selected beta values were 0.08 and 0.4 for NAC and AC images, respectively (Figure 2).The same number of iterations and gamma values were used for NAC images.A scan time of less than 4 min/BP resulted in excessively poor image quality in some of the 18 patients; therefore, the scan time was set to 4 min/BP for the Starguide, giving a total scan time of 20-30 min (Figure 2).
The visual interpretation of images from the two systems, both planar and SPECT images, provided similar clinical assessment for all bone scan patients.Additional rib uptake was observed in some patients imaged with the Starguide system (Figure 3).

| Cardiac scan
The optimized beta value for the 32 patients was set to 0.07 (NAC), 0.008 (AC) and 0.002 (gated), using regularization with MRP (50 iterations and 10 subsets) for stress and rest images (Figure 4).The scan time was reduced after visual image assessment; for stress scans from 16 to 8 min, and for rest scans from 6 min 40 s to 3 min 20 s.
The validation between the Starguide and 530c showed a good concordance, using regression analysis and Bland-Altman plots for EF and EDV (Figure 5).where there was a difference in the septal uptake.In the DNM 530c images, a small septal defect was seen, while there was a large septal perfusion defect in the Starguide images.

| Lung scan
For ventilation and perfusion, because there was no improvement in image quality with more iterations, the number of iterations and subsets were set to 20 and 10, respectively.No large differences in image contrast were seen after adjusting the gamma value; it was set to 0.5.The selected beta value was 0.03 (ventilation) and 0.02 (perfusion).
The images reconstructed with BSREM on Starguide show a reduction in image noise and a loss in image contrast versus images from the conventional gamma camera (Figure 6).A comparison of the OSEM reconstructed images from the conventional gamma camera and the Starguide system showed a higher contrast with the Starguide system but also higher noise levels.For the images reconstructed with BSREM, the visual comparison of the 26 patients showed equivalent results for diagnosing pulmonary embolism for DNMCT670 or DNM 630 and Starguide.

| DISCUSSION
For the examinations compared here, the transition from a conventional gamma camera to the Starguide system was successful without any major issues.Bone, cardiac, and lung scans are now performed on the Starguide in clinical practice at Skåne University Hospital.Differences in the assessment could be seen for some patients, but overall, the systems were assessed to be interchangeable.
Published studies with a similar camera system (Veriton from Spectrum Dynamics) show improved sensitivity on phantom studies versus a conventional gamma camera (Desmonts et al., 2020;Goshen et al., 2018;Keider et al., 2016;Le Rouzic & Zananiri, 2021;Takahashi et al., 2013).For the Starguide, the total detector sensitivity is, at best, six times that of a conventional two-headed gamma camera system when all 12 detectors perform the acquisition in a non-sweep mode over the same volume.In clinical routine, the acquisition is performed in sweep mode to cover the entire patient volume, which gives a sensitivity equivalent to that of a conventional gamma camera system.The sensitivity can be increased when using focus mode, for example, over the heart.To avoid truncation artefacts, the focus sweep time is never set to 100%, and the cylindrical focus volume is generally larger than the detectors' width.
These two factors reduce the sensitivity compared to the theoretical value of six times higher than that of a conventional gamma camera, which is only achieved if all the detectors are stationary and aimed at the same volume for the whole measurement.
Regularization with RDP was chosen for bone scans because the penalizing term depends on the relative difference between the pixel values so that excessive smoothing can be avoided, which is preferable when detecting small objects, as in bone scans.Increasing the beta value decreases the contrast in the image; a reduced value is preferred to enhance contrast and avoid smoothing.For RDP, the gamma value can also be set to enhance the signal and enable the probability of detecting small lesions.The chosen gamma value was 2; this parameter is not fully optimized and needs further evaluation to see if it can improve the image quality by generating sharper edges.The scan time of a planar whole-body bone scan on a conventional gamma camera is about 15 min.Additional SPECT in one FOV takes another 10-20 min.The Starguide system enables whole-body 3D images with a total acquisition time of a maximum of 30 min for 4 min/BP.It has also been shown for dual head CZT that scan time can be reduced with preserved image quality (Koulikov et al., 2015;Yamane et al., 2019).When comparing the two systems, there is no distinctive time gain, but the Starguide yielded a wholebody SPECT.3D images usually yield more data, enabling more accurate localization and improved contrast (Schillaci et al., 2004).
The decision to perform additional SPECT-CT after the whole-body scan is not required with the Starguide, which optimizes the clinical workflow.
The MRP regularization uses the median pixel value to penalize the noise, reducing non-monotonic noise irregularities; MRP regularization was chosen for a cardiac scan where larger defect Optimizing the lung scintigraphy resulted in different beta values in the ventilation and perfusion reconstruction with BSREM RDP regularization.This is naturally due to the higher number of counts in the perfusion images (the administrated activity for the perfusion scan was four times higher, and the acquisition time only differed by a factor of 2).A higher number of counts permits a lower beta value, leading to a sharper image.
OSEM reconstruction for lung scintigraphy was performed with the same reconstruction parameters in Oasis and Starguide.The small differences between the systems, with some higher contrast with the Starguide, could be due to differences in the collimator corrections.A disadvantage of using Starguide for lung scans is that some patients (approximately 10%) with lungs larger than 25 cm need to be scanned with two-bed positions, which doubles the examination time.In this comparison, the visual improvement in image quality versus conventional gamma camera was not observed.However, the image quality could be maintained while reducing scan times, which can positively affect the clinical environment and patient comfort.

| LIMITATIONS
This article focuses mainly on the implementation of Starguide into routine clinical use.The parameters that were chosen for optimization were the parameters that were believed to impact the image quality the most for the respective method.For the Starguide, the voxel size can be set to 2.46 or 4.92 mm, and it has a fixed collimator corresponding to a conventional LEHR collimator.Initially, 2.46 mm was used, but this was later changed to 4.92 mm since no difference in image quality was observed visually.When reconstructing images with noise regularization and resolution recovery, the data will be filtered and, therefore, not as affected by the pixel size but more by the count density.For the Starguide, all images are reconstructed with resolution recovery; therefore, the pixel size was not prioritized in the optimization in this paper.A limitation of this study is that a full-scale optimization has not been performed.According to the manufacturer's recommended protocols, the same energy window is used for the conventional NaI(Tl) gamma camera and the Starguide.A narrower energy window is possible for the CZT cameras, and these settings need to be further assessed for the Starguide to fully optimize the examinations.The impact of post-filtering needs to be evaluated.No quantitative measurements were done, only visual assessment, since the validation aims to mirror the clinical assessment.Only images acquired during stress were optimized for cardiac scans, and the same parameters were used for rest.

| CONCLUSIONS
Based on our findings, we conclude that the conventional gamma camera systems and Starguide for bone, cardiac, and lung scans are interchangeable, and Starguide has now been implemented into our clinical routine.For bone scans with an administered activity of 570 MBq, the acquisition time was set to 4 min/BP using BSREM reconstruction with RDP regularization, a beta of 0.4 and a gamma of 2. For cardiac scans using a 1-day protocol with the administered activity of 2.5 MBq/kg for rest and three times higher for stress, the reconstruction was performed using BSREM with MRP regularization beta of 0.07 (NAC) and 0.008 (AC).
Starguide.The conventional gamma camera systems used were Discovery NM/CT 670 (DNMCT 670; GE Healthcare), Discovery NM 630 (DNM 630; GE Healthcare) and Discovery NM 530c (DNM 530c; GE Healthcare).The DNMCT 670 and DNM 630 are dual head Anger cameras with a 3/8-inch NaI(Tl) crystal and a low energy high resolution (LEHR) (hole diameter 1.11 mm, septum thickness 0.2 mm and hole length 35 mm) or extended low energy general purpose (ELEGP) collimator (hole diameter 2.5 mm, septum thickness 0.4 mm and hole length 40 mm).The DNM 530c is a dedicated cardiac camera with a stationary gantry, and 19 CZT detectors with pinhole collimators focused on the heart.
Acquisitions of 16 min (stress) and 6 min 40 s (rest) in one BP were acquired with focus (90%) set over the heart followed by a CT (GE Optima CT 540, tube voltage 120 kV, time 0.6 s, pitch 1.375 tube current auto mA 20-90 with noise index 70 attenuation map created separately) for AC.Electrocardiogram-gated scans were acquired using eight bins.All images were reconstructed using the BSREM algorithm with MRP regularization.For three patients, the number of iterations varied from 10 to 60 in steps of 10 with 10 subsets.Since no visually improved image quality was obtained after 50 iterations and 10 subsets, this was used for further optimization.Image optimization was then performed in 12 patients for the chosen number of iterations assessing the beta values 0.004, 0.006 and 0.008 (AC) and 0.03, 0.05 and 0.07 (non attenuation corrected[NAC]).Reduced acquisition time of 80%, 60%, 50% and 40% was studied for the chosen beta values (AC and NAC for stress and rest) and evaluated side-by-side to assess which scan time had unacceptable quality.Gated images were optimized in 10 patients for beta values 0.001, 0.002, 0.004 and 0.01.The chosen parameters were then applied to all 32 patients to assess clinical performance compared to the conventional gamma camera system.Two experienced nuclear medicine physicians visually assessed the images simultaneously in QPS/QGS (Cedars-Sinai Medical Center in Hermes software) for contrast, noise, and artefacts side-by-side.Regression and Bland-Altman analyses were then used to evaluate the relationship between the ejection fraction (EF) and end-diastolic volume (EDV) for the chosen reconstruction parameters.2.5 | Lung scanCombined ventilation and perfusion SPECT (V/P SPECT) studies were performed on 26 patients on either DNMCT 670 or DNM 630 and the Starguide system.Each patient inhaled 99m Tc-labelled graphite particles (Technegas; approximately 25 MBq), followed by a ventilation study on one camera system (42% first on Starguide) and then on the other.The patient was injected with approximately 120 MBq 99m Tc-labelled albumin macroaggregates, depending on the count statistics registered on the conventional gamma camera, after the second ventilation imaging while remaining in the same position to collect perfusion data.The patient was then moved back to the first camera system.
10 min and 5 min, respectively, with a voxel size of 4.92 mm.All images were reconstructed using the BSREM algorithm with RDP regularization.For three patients, the number of iterations varied from 10 to 30 in steps of 10 with 10 subsets.OSEM reconstructions were performed on the Starguide using the same parameters as a conventional gamma camera (three iterations, eight subsets, collimator correction and 4.92-mm voxel size).The optimization was then performed on all 26 patients for the chosen number of iterations and beta values of 0.01, 0.02, 0.03 and 0.05 and gamma values of 0.5, 1, 1.5, 2 and 2.5 for V/P SPECT.Three experienced nuclear medicine physicians assessed all images by studying noise levels, contrast, and any presence of artefacts.The quality and clinical performance of the Starguide images were visually assessed using Hermes Hybrid Viewer since Starguide data cannot be reconstructed in OASIS.OASIS and Hermes software subtract the remaining ventilation activity from the perfusion data.The examinations from the two systems were compared to ensure the clinical interpretation of the images from the new camera system was the same as those interpretations made using the conventional gamma camera.

F
I G U R E 2 Optimization of bone scan parameters acquisition time and noise reduction beta value.Images were acquired for 4 min/ BP and reconstructed with RDP beta values of 0.1, 0.4 and 1, 10 iterations, 10 subsets, gamma 2, collimator correction, AC and SC (a).Images for another patient acquired with scan times of 8, 6, 4, 3 and 2 min/BP reconstructed with RDP beta value 0.4, 10 iterations, 10 subsets, gamma 2, resolution recovery (RR), AC and SC (b).AC, attenuation correction; BP, bed position; MIP, maximum intensity projection; RDP, relative difference prior; SC, scatter correction.The visual interpretation yielded similar results for 97% of the patients imaged on the Starguide and the DNM 530c.There were 9% of patients who had a difference in the extension and shape of the perfusion distribution between the Starguide and DNM 530c.Similar clinical conclusions were drawn from both image sets.In one case (3%), the interpretation of the images differed between the systems

F
I G U R E 3 Planar versus pseudo-planar images for bone scan.Pseudo-planar anterior and posterior images from Starguide were acquired with 4 min/BP and reconstructed with RDP 10 iterations, 10 subsets, beta 0.4, and gamma 2 scanned from the top of the head to the end of the femur.Reconstruction with RR, AC and SC and FAME (left) a transversal SPECT and SPECT/CT slice from Starguide (middle) and conventional gamma camera anterior and posterior planar images from DNM 670 acquired with the speed of 16 cm/min (right).AC, attenuation correction; BP, bed position; FAME, functional anatomical mutual enhancement; MIP, maximum intensity projection; RDP, relative difference prior; RR, resolution recovery; SC, scatter correction; SPECT, single-photon emission computed tomography.

F
I G U R E 4 Axes images and bullseye for the cardiac scan (stress) with different beta values and scan time for AC images.Short axis, vertical long axis, horizontal long axis images, and bullseye plots were acquired from Starguide for one patient using different beta values for AC images.The acquisition time was 16 min, and the reconstruction was performed with MRP, 50 iterations, 10 subsets and beta values of 0.004, 0.006 and 0.008 for AC.The images are scaled relative to their own maximum.Bullseye and vertical long axis images for another patient for AC images with beta value 0.006 and acquisition time of 16 min, 12.8 min, 9.6 min, 8 min and 6.4 min.AC, attenuation correction; MRP, median root prior.areas are studied.RDP regularization risks giving a non-uniform visual impression, which may increase the risk of false positive assessment.The difference in the appearance of tracer distribution between the DNM 530c system and the Starguide depends on the difference in the detector geometry.Rearrangement of the intestine, when the patient is moved from one camera to another, can also cause inconsistency between the two systems.The evaluation of scan time resulted in a reduction of the total scan time on the Starguide by 50% (stress and rest), which corresponds to the supine scan time on the GE 530c.This optimizes the clinical workflow since no imaging in prone position is required on the Starguide camera.Reducing scan times but maintaining count statistics can be advantageous for image quality because patient motion affects the image to a lesser degree.The administered activity can be reduced while maintaining image quality for myocardial perfusion imaging, which in turn optimizes the examinations on the Starguide(Carsuzaa et al., 2022).Additional scan time optimization would be possible for sites that administer higher amounts of activity(Oddstig et al., 2013).The EF and EDV show concordant results between 530c and Starguide which is to be expected.The systems we have compared both have CZT detectors and use BSREM reconstruction but with different regularizations.

F
I G U R E 5 EF and EDV comparison.The correlation plots (solid line correlation line and dashed line identity line) and Bland-Altman plots for EF (a) and EDV (b) obtained for 530c and Starguide.An overall bias of 1.3% for EF (95% limit of agreement: −11.82 to 14.36) and 2.10% for EDV (95% limit of agreement: −12.79 to 15.65).EDV, end-diastolic volume; EF, ejection fraction.
For lung scans with an administered activity of 25 MBq (ventilation) and 120 MBq (perfusion), the acquisition time was set to 5 min (ventilation) and 10 min (perfusion) using BSREM with RDP regularization with beta value 0.03 (ventilation) and 0.02 (perfusion) and gamma 0.5.The transition from the conventional gamma camera to the Starguide system was managed easily.AUTHOR CONTRIBUTIONSAll authors participate in the design of the study.Eva Persson, Henrik Mosén, Kristian Valind and Elin Trägårdh interpreted the images.Irma Cerić Andelius and David Minarik optimized the images and did the data analysis.Irma Cerić Andelius wrote the manuscript, and the other revised the manuscript.All authors read and approved the final manuscript.

F
I G U R E 6 Sagittal and transaxial slices for ventilation and perfusion with different reconstructions and with different types of software.Sagittal (top row) and transaxial (bottom row) images of ventilation (a, b, c) and perfusion (d, e, f) from the Starguide system (b, c, e, f) and a conventional gamma camera (a, d).Ventilation and perfusion images from the Starguide (b, e) are reconstructed with OSEM 3 iterations, 8 subsets and RR compared to reconstructions with RDP 10 iterations and 20 subsets using beta 0.03 and 0.02, respectively (c, f).The images from Oasis are reconstructed with three iterations, eight subsets and RR.Ventilation data have been subtracted from the perfusion images.OSEM, ordered subset expectation-maximization; RR, resolution recovery.
times were 8 min/bed position (BP); depending on the patient's length, the number of BP used per scan varied within the 5-7 BP range with voxel size of 4.92 mm.A dedicated sweep mode was used on the torso for the bone scan.This is a default parameter when studying bone where the sweeping area is set from the body contouring, and 60% of the acquisition time is spent on the spine area.A CT for AC and localization (Optima CT 540, tube voltage 120 kV, time 0.6 s, pitch 1.375 tube current auto mA 20-90 with noise index 70, attenuation map created separately) covering the same part as the SPECT was also performed.All images were reconstructed using the BSREM algorithm with RDP regularization.For each patient, planar and, if acquired, SPECT-CT data from the conventional gamma camera were compared to SPECT images from Starguide to evaluate any differences in image interpretation.Six 99m Tc-labelled albumin macroaggregates 120 MBq (perf) 30 (48)Note: Metrics include radiopharmaceutical used, injected activity, as well as time p.i. for the first and second scan for bone, cardiac and lung scans.Acquisition patients were chosen to optimize the number of iterations, 10-50 in steps of 10 with 10 subsets.Since no visually improved image quality was obtained after 10 iterations and 10 subsets, this was used for further optimization.For scan time, 8 min/BP (10 iterations and 10 subsets), beta 0.3, 0.4, 0.5 and 0.7 and gamma 2, 4 and 6 were assessed.Different scan times (1-8 min/BP) in steps of 1 min