Exponential dosing to standardize myocardial perfusion image quality with rubidium-82 PET

Background 82Rb PET is commonly performed using the same injected activity in all patients, resulting in lower image quality in larger patients. This study compared 82Rb dosing with exponential vs proportional functions of body weight on the standardization of myocardial perfusion image (MPI) quality. Methods Two sequential cohorts of N = 60 patients were matched by patient weight. Rest and dipyridamole stress 82Rb PET was performed using 0.1 MBq·kg−2 exponential and 9 MBq·kg−1 proportional dosing. MPI scans were compared qualitatively with visual image quality scoring (IQS) and quantitatively using the myocardium-to-blood contrast-to-noise ratio (CNR) and blood background signal-to-noise ratio (SNR) as a function of body weight. Results Average (min–max) patient body weight was 81 ± 18 kg (46–137 kg). Proportional dosing resulted in decreasing CNR, SNR, and visual IQS with increasing body weight (P < 0.05). Exponential dosing eliminated the weight-dependent decreases in these image quality metrics that were observed in the proportional dosing group. Conclusion 82Rb PET dosing as an exponential (squared) function of body weight produced consistent stress perfusion image quality over a wide range of patient weights. Dramatically lower doses can be used in lighter patients, with the equivalent population dose shifted toward the heavier patients to standardize diagnostic image quality. Supplementary Information The online version contains supplementary material available at 10.1007/s12350-023-03303-6.


INTRODUCTION
][7][8] Despite these advantages of 82 Rb PET, image quality can still be affected by the patient's body habitus as an increase in the body dimension leads to higher fractions of attenuated and scattered photons resulting in fewer recorded counts and increased image noise. 9electing an appropriate imaging protocol including administered activity appropriate for each patient's body habitus is very important to standardize diagnostic image quality.Current SPECT imaging guidelines from the American Society of Nuclear Cardiology (ASNC) suggest ''…an effort to tailor the administered activity to the patient's habitus and imaging equipment should be made… [however] strong evidence supporting one particular weight-based dosing scheme does not exist.'' 10,11Similarly for PET, the current ASNC perfusion imaging guidelines suggest that ''Large patients may benefit from higher doses'' but no specific recommendations are provided to ensure consistent image quality for 82 Rb MPI. 12 Image smoothing can help to reduce noise and improve image quality, but at the expense of lower spatial resolution. 9][15] Historically, 82 Rb PET imaging has been performed using a single constant dose for all patients 16 due in part to limitations of early generator systems which were calibrated for dose delivery at a single activity value 17 but this is known to result in lower count-density and corresponding lower image quality in larger patients.We have shown previously that this variation of image quality can be mitigated to some degree by the administration of activity in proportion to body weight (15)  using a new generation 82 Rb elution system. 18Contrary to 18 FDG PET imaging however, longer scan times can not be used to improve 82 Rb image quality in these patients due to the ultra-short half-life of 75 seconds.
The European Association of Nuclear Medicine (EANM) guidelines for PET MPI currently recommends weight-based tracer dosing for 82 Rb imaging in 3Dmode at 10 MBqÁkg -1 (with a minimum dose of 740 MBq and maximum of 1480 MBq), 19 whereas the ASNC PET MPI guidelines still accept the use of a single constant dose of 82 Rb ranging from 740 to 1110 MBq depending on the PET-CT device sensitivity. 10The common lower limit of 740 MBq may not allow adequate dose reduction in very small patients, whereas the upper limit of 1110 to 1480 MBq may not allow adequate image quality in the largest patients.
Our center has, for several years, used weight-based dosing as a proportional function of patient weight (9-10 MBqÁkg -1 ) to reduce variations of image quality depending on body habitus, and to reduce detector saturation during the tracer first-pass for accurate blood flow quantification. 1,20Despite this approach, larger patients still appear to suffer from reduced 82 Rb PET image quality which is not aligned with the recommended principles of patient-centered imaging. 21,22herefore, the aim of this study was to investigate whether 82 Rb dosing as an exponential (squared) function of weight may help to standardize PET MPI quality across a wide range of patient body sizes, following a similar protocol validated previously for whole-body 18 FDG PET. 23,24

Study design
This was an interrupted time series cohort comparison study performed as part of the clinical quality improvement (CQI) program in the Cardiac Imaging department at the University of Ottawa Heart Institute, therefore the requirement for informed patient consent was waived by the Ottawa Health Science Network Research Ethics Board.An exponential dosing protocol was designed to increase the 82 Rb activity as a squared function of body weight, while maintaining the same injected activity as the previous proportional dosing function for patients with our historical population average weight of 90 kg, as illustrated in Figure 1A.
PET image quality is determined by count statistics which follow a Poisson distribution.As a first-order approximation, with statistical iterative reconstruction methods the local image variance is proportional to the mean activity concentration (or the total number of radioactive decays recorded), and therefore the local image signal-to-noise ratio (SNR) should be proportional to the square-root of the local activity concentration (or total injected activity) and imaging time, i.e., SNR ¼ ffiffiffiffi ffi At p Â k, where the parameter k is a constant specific to the PET scanner, image reconstruction protocol and target organ.
The standard relationship above was extended by de Groot to include patient weight effects observed empirically in 18 FDG PET studies of the liver 20 according to Eq. 1.
For 82 Rb PET, the scan time (t) is essentially fixed, therefore SNR is determined solely by the injected activity (A).In the case of constant injected activity, SNR LIVER has been shown to decrease as an exponential function of weight (b = -1) as illustrated in Figure 1B. 23With proportional dosing (A µ Weight) image SNR still decreases with patient weight, but with a lesser dependence (i.e., b = -0.5).Finally, if activity is administered as a squared function of weight (A = e 9 Weight 2 ) and scan time is fixed, then SNR is expected to remain constant (b = 0) across different patient weights as derived in Eq. 2 and illustrated in Figure 1B.
where the dosing parameter e is site-dependent and can be adjusted to obtain the desired SNR Constant value in the target organ using a particular scanner and image reconstruction protocol.In this study a value of e = 0.1 MBqÁkg -2 was selected to maintain the same injected activity (810 MBq) in our historical average patient weight of 90 kg.

Patient population
A control group of 50 consecutive patients was identified initially who underwent clinically indicated 82 Rb MPI imaging with proportional dosing (9 MBqÁkg -1 ) during a 2-week period in November 2020.Following a short transition period, an additional 50 consecutive patients who underwent clinically indicated 82 Rb myocardial perfusion imaging (MPI) with the exponential dosing protocol (0.1 MBqÁkg -2 ) were identified during a 1-week period in January 2021.The distribution of patient weights was compared between cohorts in 10 kg intervals as shown in Figure 2. In those intervals with unequal numbers, subsequent consecutive patients in each cohort (N = 10) were added to obtain a final matched weight distribution consisting of N = 60 patients in both groups.Both proportional and exponential cohort scans were acquired on a Biograph Vision600 PET-CT scanner (Siemens Healthcare, Hoffman Estates, IL) following our standard clinical protocols. 25Briefly, a single low-dose CT scan was performed at normal endexpiration for attenuation correction of the rest and stress PET scans.Dynamic PET imaging was performed at rest and again during dipyridamole stress (0.14 mgÁkg -1 Ámin -1 9 4 min).For both scans, a 30seconds square-wave injection of Rubidium Rb 82 Chloride injection (RUBY-FILL TM , Jubilant Radiopharma, QC) was administered followed by a 20 mL saline-push. 25Ungated static images were reconstructed from 2 to 8 minutes, ECG-gated images (8 bins per cycle) from 1 1/2 to 8 minutes following tracer injection to maximize count statistics following the blood clearance phase.The vendor iterative OSEM reconstruction method was used including time-of-flight with 5 subsets, 4 iterations, 128 matrix size with 4 9 4 9 3 mm 3 voxels and 6 mm Gaussian post-filtering.

Image quality analysis
Visual image quality was determined for the stress ECG-gated series independently by two experienced physicians (AT, RSB) blinded to the study cohorts and to each other's results.Image quality scores in the heart (IQS HEART ) were assessed using a 5 point-scale (poor, fair, good, very good, excellent) based on the visual interpretation of heart-to-blood contrast and background noise as shown in Supplemental Figure S1.Intermediate (1/2 point) scores were also allowed resulting in 9 discrete scoring levels.Reliability between operators was assessed using Bland-Altman analysis, and the averaged scores were used in the final analysis.
Quantitative stress image analysis was performed using Corridor-4DM software v2018 (INVIA Medical Solutions, Ann Arbor, MI).Myocardium signal was measured as the maximum LV activity (LV MAX ) to avoid the effects of tracer uptake defects due to regional coronary disease.The blood background signal and noise were measured as the left atrium cavity mean and standard deviation (Blood MEAN and Blood SD ) in a blood region drawn manually as shown in Figure 3. Contrast-tonoise in the heart (CNR HEART ) = (LV MAX -Blood MEAN )/ Blood SD and SNR BLOOD = Blood MEAN /Blood SD were calculated for both the ungated (static) and ECG-gated (enddiastolic) stress PET images.Measurements of LV SD were not available in the 4DM software therefore a myocardialspecific SNR was not computed.To ensure reliability of these semi-automated measurements, two operators performed the heart CNR and blood SNR analyses (AT, RDK), blinded to the study cohorts and to the results of the other operator.These values were averaged between operators and used in the analyses of weight-based and dosing-based effects.To enable direct comparison of 82 Rb to the 18 FDG exponential dosing results of de Groot et al. image quality was also measured in the liver. 23SNR was measured as the mean divided by the standard deviation (SD) of activity in a large volume of interest (VOI) drawn in an area of uniform uptake in the liver, i.e., SNR LIVER = Liver MEAN /Liver SD as shown in Figure 3.
To characterize the dependence of image quality on patient body weight, the visual IQS HEART , and quantitative CNR HEART , SNR BLOOD , and SNR LIVER values were plotted against patient weight, and the data fit to exponential power functions as shown in Eq. 3.
where the parameter a ¼ ffiffi e p Â k from Eq. 2, and the exponent b indicates whether image quality is increasing (b [ 0), decreasing (b \ 0) or is constant (b = 0) as a function of patient weight.

Statistical analysis
Measurements of IQS, SNR, and CNR were compared between operators using Bland-Altman analysis.The weight-dependence of image quality on body weight (b coefficients) were compared between the exponential and proportional dosing groups using 95% confidence intervals.Variances were compared using non-parametric Levene's tests.Mean values were compared using paired Student t-tests, and median values using Mann-Whitney U tests.P \ 0.05 was considered statistically significant.Statistical analysis was performed using Excel 2019 with Real Statistics 8.1.

RESULTS
Patient demographics are shown in Table 1.The proportional and exponential dosing cohorts had similar clinical characteristics, including patient weights (80.9 ± 18.2 kg and 81.0 ± 17.7 kg; P = 0.96) as expected based on the prospective cohort matching (Figure 2).The median injected activity was 12% lower using exponential vs proportional dosing (P = 0.04), as the median weight in our experimental cohort (80 kg) was slightly lower than the historical value of 90 kg used to design the exponential dosing protocol.The min-max range was substantially wider (211-1850 vs 433-1362 MBq) as expected using exponential vs proportional dosing.
With proportional dosing the measured activity values in the LV myocardium and blood were relatively constant, whereas with exponential dosing they both increased linearly with patient body weight (Figure 4A,  B). Background noise (Blood SD ) in both cohorts increased linearly with body weight and was unchanged between dosing protocols (Figure 4C).
For the measurements of cardiac IQS, CNR, and SNR, the inter-operator agreement was excellent with mean differences B 5% (details in Supplemental Table S1).The average values of IQS, CNR, and SNR are shown for both dosing cohorts in Table 2.In the exponential dosing cohort, there was an average decrease of -8.5% across all image quality metrics, consistent with the lower average injected activity as noted earlier.More importantly, there was 40% decreased variability of both the static and gated CNR HEART values in the exponential dosing cohort (P \ 0.001) demonstrating significantly improved consistency of image quality compared to proportional dosing.
Improved consistency was confirmed with the visual image quality scores (Figure 5) in the exponential dosing cohort, which showed no significant dependence on body weight (b = 0.11; P = 0.38).This was in contrast to the proportional dosing group which showed a significant decrease in image quality (b = -0.48;P \ 0.001) that was very similar to the value predicted by Eq. 1 and shown in Figure 1B (b = -0.5).Interestingly, the crossing point of equivalent IQS HEART values in both cohorts was close to 90 kg, further demonstrating validity of the noise model and dosing methods as described in the study design.Higher body weight was observed in the patients with lower IQS in the proportional dosing cohort (P \ 0.001) but with not exponential dosing (P = 0.82) where the distribution of weights was uniform across different visual IQS values (Supplemental Figure S2).The changes in visual image quality between dosing methods can be seen in the patient examples shown in Figure 6 and Supplemental Figure S3.
The quantitative CNR HEART values shown in Figure 7 demonstrated even more pronounced effects compared to the visual IQS HEART scores.Both the ECGgated and static images had better consistency of image quality in the exponential vs proportional dosing group (Figure 7A, B).Proportional dosing resulted in significantly decreased CNR HEART with increasing weight (b = -0.99 and -0.76, both P \ 0.001), whereas there was no significant weight effect in the exponential dosing cohort (b = 0.29 and 0.08, both P [ 0.05).The corresponding effects of dosing protocol on SNR HEART and SNR LIVER were also very similar, as shown in the Supplemental Figures S4 and S5.
The b coefficients summarizing the weight-dependence of all the image quality metrics are shown in Table 3.In the proportional dosing cohort, the average coefficient was (b = -0.56)confirming the negative effect of patient weight on image quality that was predicted in Figure 1B.In the exponential dosing cohort, the average coefficient was (b = 0.19) suggesting a possible small effect to actually increase quality in the gated and static images of the larger patients.This suggests that an exponential dosing coefficient slightly less than the squared function that we evaluated (exponent \ 2) may have been sufficient to remove the weight-dependence of image quality.On the other hand,  the squared function did produce very consistent results between visual IQS and quantitative CNR HEART which were both based on the combined evaluation of myocardium to blood contrast and background noise.

DISCUSSION
To our knowledge, this is the first report of a patient-centered approach using exponential dosing to standardize image quality for 82 Rb PET perfusion imaging.In the control group, when 82 Rb activity was administered in proportion to patient weight (9 MBqÁkg -1 ) image quality was observed to decrease significantly with increasing body weight (b values \ 0).For each 10 kg increase in patient weight, the ECG-gated CNR decreased by approximately 10%.This is equivalent to 50% reduction in CNR when the patient weight is doubled from 50 kg (110 lbs) to 100 kg (220 lbs), similar to the reduction shown in the patient examples of Figure 6A and B. Conversely, in the experimental group (Figure 6C and D) using exponential dosing (0.1 MBqÁkg -2 ) the image quality was more consistent (b values & 0) with less than 10% variation on average across a wide range of patient weights ranging from approximately 50 to 120 kg.The biggest changes in activity occurred at the extremes of patient weight, essentially redistributing the population dose from the smaller to the larger patients as needed to standardize image quality.

Comparison to Guidelines and Previous Studies
The current ASNC guidelines advise either a constant dose for all patients or a proportional weightbased dose of 82 Rb for PET perfusion studies, 12   which have the limitation of producing lower quality images in obese patients.In the field of oncology PET, de Groot et al. found that 18 FDG activity administered as a squared function of patient weight provided wholebody PET images of consistent quality, i.e., liver SNR no longer varied with patient weight. 23This exponential relation between 18 FDG dose and body weight was also verified by Koopman 24 for general implementation and independently by Musarudin et al. 26 to provide constant liver image quality on a BGO PET-CT scanner.As a result of these studies, exponential or 'quadratic' dosing is now recommended for 18 FDG PET-CT imaging in the most recent EANM procedure guidelines for tumor imaging. 15In the present study, the effects of exponential vs proportional 82 Rb PET dosing on liver SNR were consistent with these previous studies of whole-body 18 FDG PET. 13,23,24,26,27The de Groot model of image quality shown in Eq. 1 predicts that SNR LIVER will decrease inversely as the square of patient weight (b = -0.5)which is consistent with the mean value of -0.48 observed in our control cohort (Table 3).This weight-dependence was effectively eliminated in the exponential dosing cohort with an average b \ 0.01, reproducing the results demonstrated previously using 18 FDG PET.
The effects of proportional dosing to produce constant LV MAX activity values in the heart (Figure 4) are partially consistent with results presented in the recent 82 Rb PET study by van Dijk et al. who reported that the number of recorded 'net' coincidences (prompts-randoms) was constant over a wide range of patient weights. 28However, unlike this previous study which found no differences in body weight among the different categories of visual image quality with proportional dosing, the present study demonstrated statistically significant decreases in image quality (assessed visually and quantitatively) as a function of body weight, consistent with the model that was developed and validated previously for 18 FDG wholebody PET. 20,24The pattern of decreasing image quality (despite constant tissue activity and 'net' coincidence counts) is likely due to the degrading effects of tissue attenuation on image quality.Our results suggest that the increasing noise effects of PET attenuation are approximately linear with patient weight, and these can be corrected with the exponential dosing protocol, to produce organ activity values that increase linearly with weight.It is surprising to us that van Dijk et al. 28 did not find a significant weight-effect of image quality using their proportional dosing protocol, however there are some methodological factors in their study which may have contributed: 1. Indirect evaluation of the weight distribution of patients across different image quality scores, 2. PMT-based PET scanner with lower sensitivity and resolution, 3. Visual evaluation of static images only where noise effects are less apparent vs ECG-gated, 4. Use of a 82 Rb generator system designed for single (constant) dose imaging. 29n contrast to our findings of improved standardization using exponential dosing with rubidium PET, a previous study with technetium SPECT perfusion imaging found that image noise in the LV myocardium could be standardized using the product of injected activity and scan-time adjusted as a proportional function of patient weight. 30While image quality using both these modalities is affected by the Poisson distribution of counting statistics, the noise effects and correction methods for the physical effects of scatter and  Our results have important implications for pediatric imaging studies such as Kawasaki Disease where PET imaging has been used to guide clinical management. 31In children, the effective dose constant (radiation risk) is typically higher per unit activity injected (e.g., 4.9 vs 1.1 mSvÁGBq -1 in a 5-year-old vs adult patient) reflecting the higher organ activity concentrations and smaller distances between organs. 32Our results suggest that the injected activity (and radiation effective dose) can be substantially reduced in the smallest patients while still maintaining diagnostic image quality.

Clinical implementation
The exponential dosing protocol for 82 Rb was easy to implement clinically by the PET technologists as a simple calculation, i.e., activity = weight (kg) 9 weight (kg)/10.For example, an 85 kg patient would be prescribed the 82 Rb dose of 85 9 8.5 = 722.5 MBq (19.5 mCi).Patients of 149 kg would be given the maximum dose of 2220 MBq (60 mCi) listed in the U.S. package insert 25 or 3700 MBq (100 mCi) for a 193 kg (425 lbs) patient as listed in the Canadian monograph. 33he activity available from the 82 Rb generator decreases over time according to the half-life of the parent 82 Sr, from 3700 MBq on day 0 to 700 MBq on day 60.Therefore, to implement exponential 82 Rb dosing in practice, patient scheduling needs to be adjusted accordingly, with maximum patient weights up to 193 kg on day 0 and up to 84 kg on day 60.
The present study results may be adapted to other PET perfusion imaging protocols, taking into account the differences in tracer retention fraction, isotope halflife, scan-time, and PET scanner sensitivity. 82Rb has approximately 30% tracer retention in the heart at a peak stress blood flow value of 3 mLÁmin -1 Ág -1 , whereas other PET tracers such as 13 N-ammonia or 18 F-flurpiridaz have approximately 60% retention at peak stress, resulting in higher myocardial activity and image quality for the same injected dose. 34These longer half-life tracers typically require lower injected activity and scantime that can be optimized for the desired image quality.These changes in imaging protocol should only affect the selected value of e in Eq. 2, whereas the weightdependence of cardiac PET image quality (b) is expected to remain the same regardless of these tracer and protocol changes.The present study value of e = 0.1 MBqÁkg -1 was selected to maintain the same 82 Rb image quality as our previous clinical standard dosing protocol (9 MBqÁkg -1 ) for our historical average patient weight of 90 kg.This value is higher than those reported previously (0.023 to 0.053 MBqÁkg -2 ) to standardize 18 FDG PET image quality, likely due to the ultra-short half-life of 82 Rb resulting in much lower count-rate and image quality recorded per unit activity (MBq) injected.Exponential dosing for 13 N-ammonia would likely use a value of e closer to those used in prior 18 FDG studies, as the typical scan times are close to the isotope half-life of 10 min.

Study limitations
The effects of exponential versus proportional dosing were evaluated only on stress perfusion image quality, however similar results are expected for perfusion imaging at rest.Only weight-based dosing was investigated in the present study, whereas other measures of patient body habitus such as body mass index, body surface area, chest circumference, etc. could be considered as the patient-specific factor used to prescribe the injected activity.Many of these factors were investigated in the original 18 FDG study by de Groot which found that patient body weight was the best predictor of changes in image quality, 23 therefore we followed the same approach and observed similar dosing protocol-dependent results for 82 Rb PET.
Most of the patients evaluated in this study were in the range of 50 to 120 kg, however many patients at highest risk for CAD may be heavier than 120 kg.The maximum activity of 3700 MBq (100 mCi) available from the 82 Rb generator 33 enables exponential dosing in patients up to & 190 kg (420 lbs), but further studies are needed to confirm effectiveness in this obese population, and to evaluate the trend toward improved image quality in the largest patients.The small reduction of injected activity in the exponential-vs proportional-dosing cohort was a by-product of our average cohort weight \ 90 kg.Conversely, for patient populations [ 90 kg the average injected activity is expected to increase if the same exponential dosing factor is used, i.e., e = 0.1 MBq/kg 2 .
SNR in the LV myocardium could not be measured using the same method as the liver, i.e., SNR LV = LV MEAN /LV SD as the values of LV SD were not available in the Corridor-4DM analysis software, but could be the subject of future investigations.The values of LV SD would also be affected by variations in tracer uptake due to CAD, therefore any future studies of SNR LV would be recommended in subjects without CAD to ensure homogeneous tracer uptake.
We did not investigate the effects of exponential dosing on the quantification of myocardial blood flow (MBF).In a previous study, we have shown that PET detector saturation due to dead-time effects can bias the measurements of MBF when the bolus first-pass countrate exceeds the scanner's dynamic range. 18In centers performing MBF quantification, the injected activity must be kept below some maximum value which maintains accuracy of the bolus first-pass dynamic images, and this may limit the implementation of exponential dosing in larger patients.Saturation bias is PET scanner-specific and can be characterized easily as a function of the dynamic prompt coincidence countrate. 16,35Unfortunately, these values are not saved currently in the reconstructed image DICOM headers by the PET vendor used in this study; this may limit the ability to perform routine quality assurance of MBF accuracy in clinical practice when using the exponential dosing protocol.In these patients there remains a tradeoff between standardization of perfusion image quality versus accurate quantification MBF.The study of Moody et al. suggested that BMI-based dosing may be used to lower the incidence of PET saturation compared to proportional weight-based dosing. 36Patient BMI (kg/ m 2 ) is also proportional to weight therefore an exponential function of BMI may help to minimize saturation effects and maintain MBF accuracy while also standardizing 82 Rb PET image quality.

NEW KNOWLEDGE GAINED
Administration of 82 Rb activity as a fixed constant dose or in proportion to weight, as recommended in current guidelines, still results in stress PET perfusion image quality that decreases with patient weight.Exponential dosing as a squared function of patient weight (0.1 MBqÁkg -2 ) was found to standardize ECG-gated image quality across a wide range of weights, consistent with the goals of high-quality and patient-centered imaging.The proposed protocol can distribute the population dose from the smaller toward the larger patients as needed to maintain image quality, without increasing the average dose.

CONCLUSION 82
Rb PET perfusion image quality is degraded in larger patients when the injected activity is kept at a single constant value.This effect is still observed (but to a lesser degree) when the activity is increased in proportion to patient weight.Administration of 82 Rb activity as a squared function of patient weight was effective to reduce the weight-dependence of image quality for patients in the range of 50 to 120 kg.This dosing protocol is recommended to standardize MPI quality when feasible within the limits of 82 Rb generator activity levels.Further studies are needed to evaluate the interaction of exponential dosing and PET scanner dynamic range on the accuracy of MBF quantification, particularly in patients [ 120 kg where detector saturation effects are more pronounced.

Figure 1 .
Figure 1. 82Rb PET dosing protocols as a function of patient body weight.(A) Constant, proportional and exponential dosing curves intersect at a common injected activity (810 MBq) and average patient body weight (90 kg).(B) Predicted changes in signal-to-noise ratio (SNR) as a function of patient weight for 3 different dosing methods (scaled to 100% at 90 kg) based on previous 18 FDG PET studies by de Groot 23 and Koopman 24 .

Figure 2 .
Figure 2. Patient weight distributions in the exponential and proportional dosing cohorts were matched prospectively.

Figure 3 .
Figure 3. Regions-of-interest drawn in the heart (A) and liver (B) for measurement of CNR and SNR.LV MAX was taken within the three-dimensional region of the myocardial wall (white) identified automatically by the Corridor-4DM software.Blood mean and standard deviation were taken in a single-slice region drawn manually in the left atrial cavity (red) on a vertical long axis (VLA) image.Liver mean and standard deviation were taken in an ellipsoid volume-of-interest drawn manually near the diaphragm (yellow).

Figure 4 .
Figure 4. 82 Rb PET activity values on ECG-gated imaging with proportional and exponential dosing.LV MAX (A) values are constant with proportional dosing (orange) but increase linearly by weight with exponential dosing (blue).(C) Blood SD activity remains very similar between dosing protocols.

Figure 5 .
Figure 5. 82 Rb PET visual image quality score (IQS HEART ) was assessed on a 5-point scale (Excellent, Very Good, Good, Fair, Poor) which decreased by weight (A) in the proportional dosing group (orange) but was constant in the exponential dosing group (blue).There was no difference in the median ECG-gated image quality score (B) between dosing cohorts (P = 0.11).Lines of best-fit are IQS µ Weight b .

Figure 6 .
Figure 6. 82Rb PET static-ungated SA (top) and ECG-gated HLA & VLA (bottom) images acquired with proportional (A,B) and exponential (C,D) dosing.Proportional dosing resulted in visibly lower image quality in the large (B) vs small (A) patient (CNR = 39 vs 80).With exponential dosing the image quality was very similar between the large (D) and small (C) patient (CNR = 50 vs 55), and much improved vs the large patient with proportional dosing (B).

Figure 7 .
Figure 7. 82 Rb PET contrast-to-noise ratio (CNR HEART ) decreases with increasing patient body weight in the proportional dosing cohort but not in the exponential dosing cohort for both ECGgated (A) and ungated static (B) images.Box-plots of CNR HEART in (C) show there was a highly significant effect of exponential dosing to reduce the variability in image quality (CNR HEART ) among patients for both static and gated reconstructions (***P \ 0.001 lower cohort variance versus proportional dosing).Lines of best-fit are CNR µ Weight b .

Table 1 .
Patient demographics Values are mean ± standard deviation or N (%) No significant differences between dosing cohorts

Table 2 .
both of 82 Rb PET image quality measurements Values are mean ± standard deviation IQS, Image Quality Score, SNR, Signal-to-Noise Ratio, CNR, Contrast-to-Noise Ratio *P \ 0.001 lower variance versus proportional dosing cohort

Table 3 .
Weight-dependence of 82 Rb PET image quality