Ethical Considerations for Artificial Intelligence in Medical Imaging: Data Collection, Development, and Evaluation.

The development of artificial intelligence (AI) within nuclear imaging involves several ethically fraught components at different stages of the machine learning pipeline, including during data collection, model training and validation, and clinical use. Drawing on the traditional principles of medical and research ethics, and highlighting the need to ensure health justice, the AI task force of the Society of Nuclear Medicine and Molecular Imaging has identified 4 major ethical risks: privacy of data subjects, data quality and model efficacy, fairness toward marginalized populations, and transparency of clinical performance. We provide preliminary recommendations to developers of AI-driven medical devices for mitigating the impact of these risks on patients and populations.

Ethical Considerations for Artificial Intelligence in Medical Imaging: Deployment and Governance.

The deployment of artificial intelligence (AI) has the potential to make nuclear medicine and medical imaging faster, cheaper, and both more effective and more accessible. This is possible, however, only if clinicians and patients feel that these AI medical devices (AIMDs) are trustworthy. Highlighting the need to ensure health justice by fairly distributing benefits and burdens while respecting individual patients' rights, the AI Task Force of the Society of Nuclear Medicine and Molecular Imaging has identified 4 major ethical risks that arise during the deployment of AIMD: autonomy of patients and clinicians, transparency of clinical performance and limitations, fairness toward marginalized populations, and accountability of physicians and developers. We provide preliminary recommendations for governing these ethical risks to realize the promise of AIMD for patients and populations.

Synthetic PET via Domain Translation of 3-D MRI.

Historically, patient datasets have been used to develop and validate various reconstruction algorithms for PET/MRI and PET/CT. To enable such algorithm development, without the need for acquiring hundreds of patient exams, in this article we demonstrate a deep learning technique to generate synthetic but realistic whole-body PET sinograms from abundantly available whole-body MRI. Specifically, we use a dataset of 56 18F-FDG-PET/MRI exams to train a 3-D residual UNet to predict physiologic PET uptake from whole-body T1-weighted MRI. In training, we implemented a balanced loss function to generate realistic uptake across a large dynamic range and computed losses along tomographic lines of response to mimic the PET acquisition. The predicted PET images are forward projected to produce synthetic PET (sPET) time-of-flight (ToF) sinograms that can be used with vendor-provided PET reconstruction algorithms, including using CT-based attenuation correction (CTAC) and MR-based attenuation correction (MRAC). The resulting synthetic data recapitulates physiologic 18F-FDG uptake, e.g., high uptake localized to the brain and bladder, as well as uptake in liver, kidneys, heart, and muscle. To simulate abnormalities with high uptake, we also insert synthetic lesions. We demonstrate that this sPET data can be used interchangeably with real PET data for the PET quantification task of comparing CTAC and MRAC methods, achieving ≤ 7.6% error in mean-SUV compared to using real data. These results together show that the proposed sPET data pipeline can be reasonably used for development, evaluation, and validation of PET/MRI reconstruction methods.

Multicenter Evaluation of Frequency and Impact of Activity Infiltration in PET Imaging, Including Microscale Modeling of Skin-Absorbed Dose.

There has been significant recent interest in understanding both the frequency of nuclear medicine injection infiltration and the potential for negative impact, including skin injury. However, no large-scale study has yet correlated visualized injection site activity with actual activity measurement of an infiltrate. Additionally, current skin dosimetry approaches lack sufficient detail to account for critical factors that impact the dose to the radiosensitive epidermis. Methods: From 10 imaging sites, 1,000 PET/CT patient studies were retrospectively collected. At each site, consecutive patients with the injection site in the field of view were used. The radiopharmaceutical, injected activity, time of injection and imaging, injection site, and injection method were recorded. Net injection site activity was calculated from volumes of interest. Monte Carlo image-based absorbed dose calculations were performed using the actual geometry from a patient with a minor infiltration. The simulation model used an activity distribution in the skin microanatomy based on known properties of subcutaneous fat, dermis, and epidermis. Simulations using several subcutaneous fat-to-dermis concentration ratios were performed. Absorbed dose to the epidermis, dermis, and fat were calculated along with relative γ- and ß-contributions, and these findings were extrapolated to a hypothetical worst-case (470 MBq) full-injection infiltration. Results: Only 6 of 1,000 patients had activity at the injection site in excess of 370 kBq (10 µCi), with no activities greater than 1.7 MBq (45 µCi). In 460 of 1,000 patients, activity at the injection site was clearly visualized. However, quantitative assessment of activities averaged only 34 kBq (0.9 µCi), representing 0.008% of the injected activity. Calculations for the extrapolated 470-MBq infiltration resulted in a hypothetical absorbed dose to the epidermis of below 1 Gy, a factor of 2 lower than what is required for deterministic skin reactions. Analysis of the dose distribution demonstrates that the dermis acts as a ß-shield for the radiation-sensitive epidermis. Dermal shielding is highly effective for low-energy 18F positrons but less so with the higher-energy positrons of 68Ga. Conclusion: When quantitative activity measurement criteria are used rather than visual, the frequency of PET infiltration appears substantially below frequencies previously published. Shallow doses to the epidermis from infiltration events are also likely substantially lower than previously reported because of absorption of ß-particles in the dermis.

An International Study of Factors Affecting Variability of Dosimetry Calculations, Part 2: Overall Variabilities in Absorbed Dose.

Dosimetry for personalized radiopharmaceutical therapy has gained considerable attention. Many methods, tools, and workflows have been developed to estimate absorbed dose (AD). However, standardization is still required to reduce variability of AD estimates across centers. One effort for standardization is the Society of Nuclear Medicine and Molecular Imaging 177Lu Dosimetry Challenge, which comprised 5 tasks (T1-T5) designed to assess dose estimate variability associated with the imaging protocol (T1 vs. T2 vs. T3), segmentation (T1 vs. T4), time integration (T4 vs. T5), and dose calculation (T5) steps of the dosimetry workflow. The aim of this work was to assess the overall variability in AD calculations for the different tasks. Methods: Anonymized datasets consisting of serial planar and quantitative SPECT/CT scans, organ and lesion contours, and time-integrated activity maps of 2 patients treated with 177Lu-DOTATATE were made available globally for participants to perform dosimetry calculations and submit their results in standardized submission spreadsheets. The data were carefully curated for formal mistakes and methodologic errors. General descriptive statistics for ADs were calculated, and statistical analysis was performed to compare the results of different tasks. Variability in ADs was measured using the quartile coefficient of dispersion. Results: ADs to organs estimated from planar imaging protocols (T2) were lower by about 60% than those from pure SPECT/CT (T1), and the differences were statistically significant. Importantly, the average differences in dose estimates when at least 1 SPECT/CT acquisition was available (T1, T3, T4, T5) were within ±10%, and the differences with respect to T1 were not statistically significant for most organs and lesions. When serial SPECT/CT images were used, the quartile coefficients of dispersion of ADs for organs and lesions were on average less than 20% and 26%, respectively, for T1; 20% and 18%, respectively, for T4 (segmentations provided); and 10% and 5%, respectively, for T5 (segmentation and time-integrated activity images provided). Conclusion: Variability in ADs was reduced as segmentation and time-integration data were provided to participants. Our results suggest that SPECT/CT-based imaging protocols generate more consistent and less variable results than planar imaging methods. Effort at standardizing segmentation and fitting should be made, as this may substantially reduce variability in ADs.

Artificial Intelligence in Nuclear Medicine: Opportunities, Challenges, and Responsibilities Toward a Trustworthy Ecosystem.

Trustworthiness is a core tenet of medicine. The patient-physician relationship is evolving from a dyad to a broader ecosystem of health care. With the emergence of artificial intelligence (AI) in medicine, the elements of trust must be revisited. We envision a road map for the establishment of trustworthy AI ecosystems in nuclear medicine. In this report, AI is contextualized in the history of technologic revolutions. Opportunities for AI applications in nuclear medicine related to diagnosis, therapy, and workflow efficiency, as well as emerging challenges and critical responsibilities, are discussed. Establishing and maintaining leadership in AI require a concerted effort to promote the rational and safe deployment of this innovative technology by engaging patients, nuclear medicine physicians, scientists, technologists, and referring providers, among other stakeholders, while protecting our patients and society. This strategic plan was prepared by the AI task force of the Society of Nuclear Medicine and Molecular Imaging.

The RSNA QIBA Profile for Amyloid PET as an Imaging Biomarker for Cerebral Amyloid Quantification.

A standardized approach to acquiring amyloid PET images increases their value as disease and drug response biomarkers. Most 18F PET amyloid brain scans often are assessed only visually (per regulatory labels), with a binary decision indicating the presence or absence of Alzheimer disease amyloid pathology. Minimizing technical variance allows precise, quantitative SUV ratios (SUVRs) for early detection of ß-amyloid plaques and allows the effectiveness of antiamyloid treatments to be assessed with serial studies. Methods: The Quantitative Imaging Biomarkers Alliance amyloid PET biomarker committee developed and validated a profile to characterize and reduce the variability of SUVRs, increasing statistical power for these assessments. Results: On achieving conformance, sites can justify a claim that brain amyloid burden reflected by the SUVR is measurable to a within-subject coefficient of variation of no more than 1.94% when the same radiopharmaceutical, scanner, acquisition, and analysis protocols are used. Conclusion: This overview explains the claim, requirements, barriers, and potential future developments of the profile to achieve precision in clinical and research amyloid PET imaging.

Large Area, High Resolution Mapping of Approximate Rotational Symmetries in a Pd77.5Cu6Si16.5 Metallic Glass Thin Film.

Densely spaced four-dimensional scanning transmission electron microscopy (4D STEM) analyzed using correlation symmetry coefficients enables large area mapping of approximate rotational symmetries in amorphous materials. Here, we report the effects of Poisson noise, limited electron counts, probe coherence, reciprocal space sampling, and the probe-sample interaction volume on 4D STEM symmetry mapping experiments. These results lead to an experiment parameter envelope for high quality, high confidence 4D STEM symmetry mapping. We also establish a direct link between the symmetry coefficients and approximate rotational symmetries of nearest-neighbor atomic clusters using electron diffraction simulations from atomic models of a metallic glass. Experiments on a Pd77.5Cu6Si16.5 metallic glass thin film demonstrate the ability to image the types, sizes, volume fractions, and spatial correlations amongst local rotationally symmetry regions in the glass.

Nuclear Medicine and Artificial Intelligence: Best Practices for Evaluation (the RELAINCE Guidelines).

An important need exists for strategies to perform rigorous objective clinical-task-based evaluation of artificial intelligence (AI) algorithms for nuclear medicine. To address this need, we propose a 4-class framework to evaluate AI algorithms for promise, technical task-specific efficacy, clinical decision making, and postdeployment efficacy. We provide best practices to evaluate AI algorithms for each of these classes. Each class of evaluation yields a claim that provides a descriptive performance of the AI algorithm. Key best practices are tabulated as the RELAINCE (Recommendations for EvaLuation of AI for NuClear medicinE) guidelines. The report was prepared by the Society of Nuclear Medicine and Molecular Imaging AI Task Force Evaluation team, which consisted of nuclear-medicine physicians, physicists, computational imaging scientists, and representatives from industry and regulatory agencies.

Noise-Based Image Harmonization Significantly Increases Repeatability and Reproducibility of Radiomics Features in PET Images: A Phantom Study.

For multicenter clinical studies, characterizing the robustness of image-derived radiomics features is essential. Features calculated on PET images have been shown to be very sensitive to image noise. The purpose of this work was to investigate the efficacy of a relatively simple harmonization strategy on feature robustness and agreement. A purpose-built texture pattern phantom was scanned on 10 different PET scanners in 7 institutions with various different image acquisition and reconstruction protocols. An image harmonization technique based on equalizing a contrast-to-noise ratio was employed to generate a "harmonized" alongside a "standard" dataset for a reproducibility study. In addition, a repeatability study was performed with images from a single PET scanner of variable image noise, varying the binning time of the reconstruction. Feature agreement was measured using the intraclass correlation coefficient (ICC). In the repeatability study, 81/93 features had a lower ICC on the images with the highest image noise as compared to the images with the lowest image noise. Using the harmonized dataset significantly improved the feature agreement for five of the six investigated feature classes over the standard dataset. For three feature classes, high feature agreement corresponded with higher sensitivity to the different patterns, suggesting a way to select suitable features for predictive models.

SIR-Spheres Activity Measurements Reveal Systematic Miscalibration.

The purpose of this work was to perform an independent and National Institute of Standards and Technology-traceable activity measurement of 90Y SIR-Spheres (Sirtex). γ-spectroscopic measurements of the 90Y internal pair production decay mode were made using a high-purity germanium detector. Methods: Measured annihilation radiation detection rates were corrected for radioactive decay during acquisition, dead time, source attenuation, and source geometry effects. Detection efficiency was determined by 2 independent and National Institute of Standards and Technology-traceable methods. Results: Measured SIR-Spheres vials (n = 5) contained more activity than specified by the manufacturer calibration; on average, the ratio of measured activity to calibrated was 1.233 ± 0.030. Activity measurements made using 2 distinct efficiency calibration methods agreed within 1%. Conclusion: The primary SIR-Spheres activity calibration appears to be a significant underestimate of true activity.

Quantification of uptake in pelvis F-18 FLT PET-CT images using a 3D localization and segmentation CNN.

PURPOSE: The purpose of this work was to develop and validate a deep convolutional neural network (CNN) approach for the automated pelvis segmentation in computed tomography (CT) scans to enable the quantification of active pelvic bone marrow by means of Fluorothymidine F-18 (FLT) tracer uptake measurement in positron emission tomography (PET) scans. This quantification is a critical step in calculating bone marrow dose for radiopharmaceutical therapy clinical applications as well as external beam radiation doses. METHODS: An approach for the combined localization and segmentation of the pelvis in CT volumes of varying sizes, ranging from full-body to pelvis CT scans, was developed that utilizes a novel CNN architecture in combination with a random sampling strategy. The method was validated on 34 planning CT scans and 106 full-body FLT PET-CT scans using a cross-validation strategy. Specifically, two different training and CNN application options were studied, quantitatively assessed, and statistically compared. RESULTS: The proposed method was able to successfully locate and segment the pelvis in all test cases. On all data sets, an average Dice coefficient of 0.9396 ± $\pm$ 0.0182 or better was achieved. The relative tracer uptake measurement error ranged between 0.065% and 0.204%. The proposed approach is time-efficient and shows a reduction in runtime of up to 95% compared to a standard U-Net-based approach without a localization component. CONCLUSIONS: The proposed method enables the efficient calculation of FLT uptake in the pelvis. Thus, it represents a valuable tool to facilitate bone marrow preserving adaptive radiation therapy and radiopharmaceutical dose calculation. Furthermore, the method can be adapted to process other bone structures as well as organs.

Nuclear Medicine and Artificial Intelligence: Best Practices for Algorithm Development.

The nuclear medicine field has seen a rapid expansion of academic and commercial interest in developing artificial intelligence (AI) algorithms. Users and developers can avoid some of the pitfalls of AI by recognizing and following best practices in AI algorithm development. In this article, recommendations on technical best practices for developing AI algorithms in nuclear medicine are provided, beginning with general recommendations and then continuing with descriptions of how one might practice these principles for specific topics within nuclear medicine. This report was produced by the AI Task Force of the Society of Nuclear Medicine and Molecular Imaging.

Repeatability of 68Ga-PSMA-HBED-CC PET/CT-Derived Total Molecular Tumor Volume.

Molecular tumor volume (MTV) is a parameter of interest in prostate cancer for assessing total disease burden on prostate-specific membrane antigen (PSMA) PET. Although software segmentation tools can delineate whole-body MTV, a necessary step toward meaningful monitoring of total tumor burden and treatment response through PET is establishing the repeatability of these metrics. The present study assessed the repeatability of total MTV and related metrics for 68Ga-PSMA-HBED-CC in prostate cancer. Methods: Eighteen patients from a prior repeatability study who underwent 2 test-retest PSMA PET/CT scans within a mean interval of 5 d were reanalyzed. Within-subject coefficient of variation and repeatability coefficients (RCs) were analyzed on a per-lesion and per-patient basis. For the per-lesion analysis, individual lesions were segmented for analysis by a single reader. For the per-patient analysis, subgroups of up to 10 lesions (single reader) and the total tumor volume per patient were segmented (independently by 2 readers). Image parameters were MTV, SUVmax, SUVpeak, SUVmean, total lesion PSMA, and the related metric PSMA quotient (which integrates lesion volume and PSMA avidity). Results: In total, 192 segmentations were analyzed for the per-lesion analysis and 1,662 segmentations for the per-patient analysis (combining the 2 readers and 2 scans). The RC of the MTV of single lesions was 77% (95% CI, 63%-96%). The RC improved to 33% after aggregation of up to 10 manually selected lesions into subgroups assessed per patient (95% CI, 25%-46%). The RC of the semiautomatic MTVtotal (the sum of all voxels in the whole-body total tumor segmentation per patient) was 35% (95% CI, 25%-50%), the Bland-Altman bias was -6.70 (95% CI, -14.32-0.93). Alternating readers between scans led to a comparable RC of 37% (95% CI, 28%-49%) for MTVtotal, meaning that the metric is robust between scanning sessions and between readers. Conclusion:68Ga-PSMA-HBED-CC PET-derived semiautomatic MTVtotal is repeatable and reader-independent, with a change of ±35% representing a true change in tumor volume. Volumetry of single manually selected lesions has considerably lower repeatability, and volumetry based on subgroups of these lesions, although showing acceptable repeatability, is less systematic. The semiautomatic analysis of MTVtotal used in this study offers an efficient and robust means of assessing response to therapy.

A Path to Qualification of PET/MRI Scanners for Multicenter Brain Imaging Studies: Evaluation of MRI-Based Attenuation Correction Methods Using a Patient Phantom.

PET/MRI scanners cannot be qualified in the manner adopted for hybrid PET/CT devices. The main hurdle with qualification in PET/MRI is that attenuation correction (AC) cannot be adequately measured in conventional PET phantoms because of the difficulty in converting the MR images of the physical structures (e.g., plastic) into electron density maps. Over the last decade, a plethora of novel MRI-based algorithms has been developed to more accurately derive the attenuation properties of the human head, including the skull. Although promising, none of these techniques has yet emerged as an optimal and universally adopted strategy for AC in PET/MRI. In this work, we propose a path for PET/MRI qualification for multicenter brain imaging studies. Specifically, our solution is to separate the head AC from the other factors that affect PET data quantification and use a patient as a phantom to assess the former. The emission data collected on the integrated PET/MRI scanner to be qualified should be reconstructed using both MRI- and CT-based AC methods, and whole-brain qualitative and quantitative (both voxelwise and regional) analyses should be performed. The MRI-based approach will be considered satisfactory if the PET quantification bias is within the acceptance criteria specified here. We have implemented this approach successfully across 2 PET/MRI scanner manufacturers at 2 sites.

Monte Carlo evaluation of hypothetical long axial field-of-view PET scanner using GE Discovery MI PET front-end architecture.

PURPOSE: The development of total-body PET scanners is of growing interest in the PET community. Investigation into the imaging properties of a hypothetical extended axial field-of-view (AFOV) GE Healthcare SiPM-based Discovery MI (DMI) system architecture has not yet been performed. In this work, we assessed its potential as a whole-body scanner using Monte Carlo simulations. The aim of this work was to (1) develop and validate a Monte Carlo model of a four-ring scanner and (2) extend its AFOV up to 2 m to evaluate performance gain through NEMA-based evaluation. METHODS: The DMI four-ring geometry and its pulse digitization scheme were modeled within the GATE Monte Carlo platform using published literature. The GATE scanner model was validated by comparing results against published NEMA performance measurements. Following the validation of the four-ring model, the model was extended to simulate 8-, 20-, 30-, and 40-ring systems. Spatial resolution, sensitivity, NECR, and scatter fraction were characterized with modified NEMA NU-2 2018 standards; however, the image quality measurements were not acquired due to computational limitations. Spatial resolutions were simulated for all scanner ring configurations using point sources to examine the effects of parallax errors. NEMA count rates were estimated using a standard 70 cm scatter phantom and an extended version of scatter phantom of length 200 cm with (1-800) MBq of 18 F for all scanners. Sensitivity was evaluated using NEMA methods with a 70 cm standard and a 200 cm long line source. RESULTS: The average FWHM of the radial/tangential/axial spatial resolution reconstructed with filtered back-projection at 1 and 10 cm from the scanner center were 3.94/4.10/4.41 mm and 5.29/4.89/5.90 mm for the four-ring scanner. Sensitivity was determined to be 14.86 cps/kBq at the center of the FOV for the four-ring scanner using a 70 cm line source. Sensitivity enhancement up to 21-fold and 60-fold were observed for 1 and 2 m AFOV scanners compared to four-ring scanner using a 200 cm long line source. Spatial resolution simulations in a 2 m AFOV scanner suggest a maximum degradation of ∼23.8% in the axial resolution compared to the four-ring scanner. However, the transverse resolution was found to be relatively constant when increasing the axial acceptance angle up to ±70°. The peak NECR was 212.92 kcps at 22.70 kBq/ml with a scatter fraction of 38.9% for a four-ring scanner with a 70 cm scatter phantom. Comparison of peak NECR using the 200 cm long scatter phantom relative to the four-ring scanner resulted in a NECR gain of 15 for the 20-ring and 28 for the 40-ring geometry. Spatial resolution, sensitivity, and scatter fraction showed an agreement within ∼7% compared with published measured values. CONCLUSIONS: The four-ring DMI scanner simulation was successfully validated against published NEMA measurements. Sensitivity and NECR performance of extended 1 and 2 m AFOV scanners based upon the DMI architecture were subsequently simulated. Increases in sensitivity and count-rate performance are consistent with prior simulation studies utilizing extensions of the Siemens mCT architecture and published NEMA measurements with the uEXPLORER system.

An International Study of Factors Affecting Variability of Dosimetry Calculations, Part 1: Design and Early Results of the SNMMI Dosimetry Challenge.

In this work, we present details and initial results from a 177Lu dosimetry challenge that has been designed to collect data from the global nuclear medicine community aiming at identifying, understanding, and quantitatively characterizing the consequences of the various sources of variability in dosimetry. Methods: The challenge covers different approaches to performing dosimetry: planar, hybrid, and pure SPECT. It consists of 5 different and independent tasks to measure the variability of each step in the dosimetry workflow. Each task involves the calculation of absorbed doses to organs and tumors and was meant to be performed in sequential order. The order of the tasks is such that results from a previous one would not affect subsequent ones. Different sources of variability are removed as the participants advance through the challenge by giving them the data required to begin the calculations at different steps of the dosimetry workflow. Data from 2 patients after a therapeutic administration of 177Lu-DOTATATE were used for this study. The data are hosted in Deep Blue Data, a data repository service run by the University of Michigan. Participants submit results in standardized spreadsheets and with a short description summarizing their methods. Results: In total, 178 participants have signed up for the challenge, and 119 submissions have been received. Sixty percent of submissions have used voxelized dose methods, with 47% of those using commercial software. In initial analysis, the volume of organs showed a variability of up to 49.8% whereas for lesions this was up to 176%. Variability in time-integrated activity was up to 192%. Mean absorbed doses varied up to 57.7%. Segmentation is the step that required the longest time to complete, with a median of 43 min. The median total time to perform the full calculation was 89 min. Conclusion: To advance dosimetry and encourage its routine use in radiopharmaceutical therapy applications, it is critical that dosimetry results be reproducible across centers. Our initial results provide insights into the variability associated with performing dose calculations. It is expected that this dataset, including results from future stages, will result in efforts to standardize and harmonize methods and procedures.

Harmonization of PET image reconstruction parameters in simultaneous PET/MRI.

OBJECTIVE: Simultaneous PET/MRIs vary in their quantitative PET performance due to inherent differences in the physical systems and differences in the image reconstruction implementation. This variability in quantitative accuracy confounds the ability to meaningfully combine and compare data across scanners. In this work, we define image reconstruction parameters that lead to comparable contrast recovery curves across simultaneous PET/MRI systems. METHOD: The NEMA NU-2 image quality phantom was imaged on one GE Signa and on one Siemens mMR PET/MRI scanner. The phantom was imaged at 9.7:1 contrast with standard spheres (diameter 10, 13, 17, 22, 28, 37 mm) and with custom spheres (diameter: 8.5, 11.5, 15, 25, 32.5, 44 mm) using a standardized methodology. Analysis was performed on a 30 min listmode data acquisition and on 6 realizations of 5 min from the listmode data. Images were reconstructed with the manufacturer provided iterative image reconstruction algorithms with and without point spread function (PSF) modeling. For both scanners, a post-reconstruction Gaussian filter of 3-7 mm in steps of 1 mm was applied. Attenuation correction was provided from a scaled computed tomography (CT) image of the phantom registered to the MR-based attenuation images and verified to align on the non-attenuation corrected PET images. For each of these image reconstruction parameter sets, contrast recovery coefficients (CRCs) were determined for the SUVmean, SUVmax and SUVpeak for each sphere. A hybrid metric combining the root-mean-squared discrepancy (RMSD) and the absolute CRC values was used to simultaneously optimize for best match in CRC between the two scanners while simultaneously weighting toward higher resolution reconstructions. The image reconstruction parameter set was identified as the best candidate reconstruction for each vendor for harmonized PET image reconstruction. RESULTS: The range of clinically relevant image reconstruction parameters demonstrated widely different quantitative performance across cameras. The best match of CRC curves was obtained at the lowest RMSD values with: for CRCmean, 2 iterations-7 mm filter on the GE Signa and 4 iterations-6 mm filter on the Siemens mMR, for CRCmax, 4 iterations-6 mm filter on the GE Signa, 4 iterations-5 mm filter on the Siemens mMR and for CRCpeak, 4 iterations-7 mm filter with PSF on the GE Signa and 4 iterations-7 mm filter on the Siemens mMR. Over all reconstructions, the RMSD between CRCs was 1.8%, 3.6% and 2.9% for CRC mean, max and peak, respectively. The solution of 2 iterations-3 mm on the GE Signa and 4 iterations-3 mm on Siemens mMR, both with PSF, led to simultaneous harmonization and with high CRC and low RMSD for CRC mean, max and peak with RMSD values of 2.8%, 5.8% and 3.2%, respectively. CONCLUSIONS: For two commercially available PET/MRI scanners, user-selectable parameters that control iterative updates, image smoothing and PSF modeling provide a range of contrast recovery curves that allow harmonization in harmonization strategies of optimal match in CRC or high CRC values. This work demonstrates that nearly identical CRC curves can be obtained on different commercially available scanners by selecting appropriate image reconstruction parameters.