Mars Shot for Nuclear Medicine, Molecular Imaging, and Molecularly Targeted Radiopharmaceutical Therapy

Data Science and Artificial Intelligence

Effective collection and sharing of image and patient data to drive machine learning approaches will strengthen clinical capabilities. There is significant interest in the use of advanced data sciences, including machine learning (especially deep learning) methods, in health care. Although their use provokes uncertain (and sometimes inflated) expectations, significant benefits are expected relative to benchmarked standards of care (1). Systematic efforts may subsequently enable better standardization of data acquisition, reduced radiation dose or scan times, and significantly improved diagnostic, predictive, prognostic and treatment response assessment capabilities. To these ends, the community needs to expand intersocietal collaborations and academic-industry partnerships; promote data sharing; encourage standardization and harmonization initiatives; link imaging studies with clinical reports, laboratory test results, genomic or proteomic assays, and clinical outcomes; and focus on generalizability and critical appraisal of the methods toward translation to routine clinical practice (in connection with regulatory bodies). The focus should not be purely on deep learning methods. For example, in oncology, there is significant potential for the use of radiomics and explicitly defined texture and shape measures, some of which are reproducible and effective (2). There are other clinical application opportunities. Data-driven (markerless, software-based) motion detection and associated correction techniques have potential to significantly improve the quality and quantitative accuracy of NM images (an increasingly important topic given improved spatial resolution in scanners). We envision a future where motion correction (including respiratory, cardiac, and gross patient motion) is applied routinely in the clinic. In neurology, observer-independent procedures such as stereotactic normalization and automated voxel-based comparison of patient brain imaging data to healthy control populations are well established. There are now opportunities to develop automated image interpretation algorithms (3). Regarding the high discriminatory accuracy of many methods of nuclear neuroimaging, these tests may form an ideal basis for training of deep learning procedures and translation into clinical practice.

Novel Imaging Systems

Clinical applications of PET/MRI systems will become more mainstream; total-body PET/CT systems will define new limits of clinical imaging and pharmacokinetic study; and there will be more development and optimization of dedicated instruments for wider use. The biologic and biochemical development of specifically targeted radiotracers will be augmented by technologic developments in scanners.

Integrated PET/MRI, a developing field in the last decade, is still in its early stages of application and holds potential for improved clinical diagnostic performance in several applications (4,5). Dedicated, diverse resources and thoughtful, synergistic efforts can lead to significant advancements and opportunities for larger-scale deployment.

Total-body PET is now 510(k)-approved and available with a field of view large enough to image the entire body simultaneously (6). The technical advantages of total-body PET include the potential to increase sensitivity by a factor of about 40. Improved image quality, enhanced detection of smaller lesions, and imaging of radiotracers beyond their currently used time frames, such as ¹⁸F imaging 18 h after tracer administration, are creating a myriad of new clinical and research opportunities (7). This technology also has the potential for significant reductions in patient doses or substantially decreased imaging times, which should be investigated. Total-body acquisitions over multiple time points for full pharmacokinetic characterization of new tracers are another opportunity.

There is significant scope for development and optimization of other instruments with high sensitivity, spatial resolution, or temporal resolution for specific applications and study of prevalent diseases. Opportunities exist for optimized pediatric imaging and for design of wearable, organ-specific or patient-configurable PET or SPECT imaging systems. For example, smaller brain-specific imaging devices may become critical in view of the increasing emphasis on early detection of diseases such as dementia (including Alzheimer disease) (8). Overall, dedicated imaging systems could provide significant cost or performance benefits and enable applicability to wider populations.

Pharmacokinetic Quantitation

Advances in scanner, reconstruction, and artificial intelligence technology push physiologically and biologically based imaging biomarkers from the research lab to the clinic.

Current clinical protocols are optimized primarily for qualitative assessment (as opposed to fully quantitative assessment). Yet, radiopharmaceutical distribution is a dynamic process with substantial pharmacokinetic variations between tumors and normal organs and among patients. Assessment of such dynamic processes has the potential to improve clinical diagnostic performance (9). Dynamic imaging enables quantitative analysis (e.g., via kinetic modeling) and generation of quantitative parametric data or images. Thus far, quantitative evaluation of myocardial blood flow in coronary artery disease evaluation has made its way into the clinic. However, there is a broad opportunity for expanded application in oncology and neurology. Here, dynamic imaging can generate parametric images in addition to conventional SUVs from a single imaging session optimized to the clinical workflow (10). This task can be accomplished in single-bed or multibed imaging and in fixed versus organ-specific systems. Given ongoing technical developments, advanced image reconstruction methods, and application of artificial intelligence to low-count datasets, dynamic images of increasing quality and quantitative accuracy are being obtained. Aided by this progress, there is a real opportunity for exploring shorter scan times in which physiologically or biologically meaningful kinetic parameters are generated as effective imaging biomarkers of disease, opening up venues to add significant value in a clinical setting.

Imaging of Neuronal Function

Imaging neuronal function and dysfunction has been a domain of ¹⁸F-FDG PET for more than 4 decades; however, movement to new synaptic density measurements may be appropriate for certain indications. ¹⁸F-FDG PET can still be considered the Swiss army knife of brain imaging in NM, not only because the tracer is widely accessible and relatively affordable (and reimbursed for some indications in several countries, including the United States) but also because ¹⁸F-FDG PET imaging of the brain provides highly reliable and sensitive information on the extent and location of neuronal dysfunction in a large spectrum of disorders of the central nervous system, ranging from epilepsy to neurodegeneration (11–13). However, some limitations apply to the tracer. One is the lack of information on the underlying cause of changes in neuronal function (14). Other limitations stem from the observation that neuronal activity may not necessarily reflect neuronal health (15,16). Some of these limitations may not apply to imaging of synaptic density. Recently, several promising tracers have been introduced that target mainly the synaptic vesicle proteins, which may allow measurement of synaptic density in vivo in the human brain (17–19). These novel tools may bear great potential to improve understanding of differences between neuronal activity and synaptic density in the healthy and the diseased brain (e.g., neurodegeneration or epilepsy). Furthermore, they may allow for closely following the effects of specific therapies.

Imaging of Protein Pathology

The development of tools for MI of amyloid deposits has been one of the most visible and game-changing achievements in neuroimaging, NM, and even in neuroscience altogether in recent years (20). The opportunity to detect a pathogenomic molecular pathology in the brain in vivo without the need for a biopsy or invasive procedure has significantly advanced research in the field of neurodegeneration, particularly regarding Alzheimer disease. Numerous publications have considerably impacted the understanding of these disorders, regarding the onset and course of disease or the interaction between pathology and dysfunction (21). Inclusion of patients in modern Alzheimer disease–related treatment trials is already based predominantly on the results from amyloid imaging, and this method may play a crucial role as a gatekeeper as soon as (hopefully) the first disease-modifying drugs arise. However, amyloid deposits represent only one type of protein aggregation, and it is commonly assumed the neurodegenerative disorders in general are based on the deposition of different types of pathologic proteins in the brain. Thus, great efforts are currently invested in the development and evaluation of tracers for tau, synuclein, and several others that may come into focus in the future (22). This endeavor is clearly challenging and, to be successful, will require considerable future commitment and investment. Importantly, a first compound for imaging tau aggregates (¹⁸F-flortaucipir [Tauvid; Lilly]) has recently been approved by the Food and Drug Administration (FDA) (23).

Imaging of Inflammation

Neuroinflammatory processes are increasingly implicated in a multitude of disorders of the central nervous system and may play a crucial role in many of them. These include not only primarily immunogenic disorders such as multiple sclerosis but also brain tumors, neurodegenerative disorders, psychiatric disorders (e.g., schizophrenia or depression), brain trauma, and ischemia. Imaging neuroinflammation not only is of great interest to potentially clarify its influence on the development and course of disease but could represent a most attractive target for therapeutic interventions. Tracer development for imaging inflammation by targeting the translocator protein has been hampered for many years by low specific signal, the high interindividual variability in physiologic binding affinity (high, intermediate, and low binders), or limited availability due to ¹¹C-labeling (24). However, although novel and improved approaches to imaging translocator protein tracers are being developed, tracers targeting other mechanisms involved in neuroinflammation (e.g., expression of the cyclooxygenases in the brain) may open interesting opportunities to investigate the role of neuroinflammation and potentially introduce imaging of neuroinflammation into clinical application (25,26).

Imaging of Neurotransmission

One of the extraordinary strengths of MI is the ability to visualize the function and integrity of transmitter systems in the brain through tracers specifically reflecting the density of neuroreceptors or transporters. Only NM methods provide this type of noninvasive insight into cerebral neurochemistry on the molecular level. This potential is still not substantially exploited accordingly with regard to clinical purposes. So far, imaging of the presynaptic dopaminergic system in the diagnosis of Parkinsonian syndromes is well-established (27,28). However, numerous disorders of the brain are fundamentally based on a disturbed balance of the neuroreceptor/transmitter system, including neuropsychiatric disorders such as depression and schizophrenia, as well as neurodegeneration, addiction, and personality disorders. It appears obvious that greater effort is needed to further establish applicable and accessible approaches to neuroreceptor imaging in clinical routine. The value of these methods needs to be tested so that we can deepen our understanding of the involved pathomechanisms and potentially improve early diagnosis, prediction, therapy selection, and response monitoring. A prominent example may be the complex glutamatergic system (29). Glutamate is the predominant excitatory transmitter in the brain. Glutamate receptors are hypothesized to play a key role in both fast and slow excitatory neurotransmission (e.g., learning and neuroplasticity) in the healthy brain and also to be involved in numerous pathologic processes such as epilepsy, addiction, brain trauma, and neurodegeneration. Glutamate receptor imaging may be of high value to potentially guide personalized therapy and is currently an open and exciting area for future exploration.

Inflammation

In migrating from ¹⁸F-FDG–based inflammation imaging to more target-specific inflammation processes in cardiovascular disease, specific opportunities for MI include identification of atherosclerotic plaque, myocardial infarction, and progressive heart failure. Novel therapeutic approaches aim to mitigate the inflammatory process and facilitate endogenous healing. Previously, capitalizing on the glucose avidity of activated inflammatory leukocytes, ¹⁸F-FDG has been used to image and characterize inflammation in vasculature and myocardium, providing prognostic information on functional outcome (30). However, myocardial ¹⁸F-FDG imaging suffers from several limitations as a nonspecific marker. Radiopharmaceuticals specifically targeted to inflammatory cells and substrates are a priority. Several alternative targets highly expressed by activated inflammatory cells have been proposed, including chemokine receptors, somatostatin receptors, mitochondrial translocator protein (31), and receptors on fibroblasts, among others. The intersection between inflammation and fibrosis has also gained interest as a diagnostic and therapeutic target (32,33). Novel, specifically targeted radiotracers have been evaluated in animal models of myocardial infarction and atherosclerosis, with some small-scale clinical evaluation performed (32). Using a common target for imaging and therapy is particularly attractive, might facilitate identification of appropriate patient populations and guidance of antiinflammatory therapeutic decisions, and could be exploited as a surrogate endpoint for clinical drug trials (34).

Flow Quantification

Improved software and tracers with flow-imaging properties that are closer to ideal will drive more robust flow quantification and greater clinical utility. Myocardial perfusion imaging will remain a dominant component of nuclear cardiology. Greater accuracy of clinical assessment and absolute quantification continue to be top priorities. Deep machine learning may optimize and streamline nuclear cardiology assessment using SPECT (35), but further refinement is needed for routine application. Accurate and reproducible flow quantification using SPECT is desirable, with some centers reporting the feasibility of flow measurements with timing efficiency of cadmium-zinc-telluride detectors and favorable comparison to PET (36). Further clarification will be necessary, but this is an area ripe for development.

PET flow quantification, although more mature than SPECT, also requires further refinements in its application in the clinical setting. PET flow quantification is becoming increasingly common in the clinic, and its diagnostic and incremental prognostic values are well established. However, nonlinearities of radiotracer uptake at high flow are a challenge, as are inconsistent results between software packages, differences in flow reserve measurements, and differences between radiotracers (37). Further careful science would broaden clinical application. In addition, broader clinical application of PET myocardial perfusion imaging may require development of new cost-effective PET perfusion tracers with superior biologic and imaging characteristics, including a high extraction fraction, a short positron range, longer physical half-lives (for commercial distribution), and a high correlation between myocardial uptake and coronary blood flow without redistribution (37). Clinical development of such agents should remain a top priority for the field and will promote dissemination and standardization of the methodology.

Cardiac Amyloidosis

Old SPECT and new PET radiopharmaceuticals may drive this new clinical application, but fundamental mechanistic questions exist, and therapeutic response assessment applications require validation.

After the successful deployment of ¹¹C-Pittsburgh compound B and analogs for imaging of β-amyloid, there has been rising interest in applying similar tracers to amyloidosis in the heart. The spatial resolution of PET is particularly beneficial for amyloidosis, as the distribution pattern in the left ventricle is relevant to accurate diagnosis, particularly for amyloid subtypes (38). New-generation amyloid tracers have been evaluated, and further testing may increase the translation of these techniques.

The clinical utility of bone-seeking radiotracers in diagnosing transthyretin-related cardiac amyloidosis has recently been established (39). Questions remain on the utility of this technique in identifying cardiac involvement in other transthyretin-related mutations, on the generalizability to diverse populations, and on the ability of the technique to monitor disease progression over time. Finally, the mechanism by which these bone-seeking tracers binds to amyloid transthyretin amyloidosis more than light chain amyloid fibrillar deposits remains to be elucidated. With effective treatments now available for transthyretin-mediated amyloidosis, the optimal use of imaging to diagnose and guide therapies represents a major opportunity.

Imaging of Infection

Highly specific radiopharmaceuticals targeting infection imaging could replace current technologies in the myocardium and throughout the body and complement anatomic imaging methods.

Although ¹⁸F-FDG has been functional in defining localized enrichment of activated leukocytes, it cannot distinguish between sterile inflammation and infection. With the rising prevalence of implantable devices and in-hospital infections, reliable identification of bacterial infection by noninvasive imaging is desirable. The novel radiotracers specifically target proteins expressed solely by bacterial cells, such as the maltodextrin transporter, folates, siderophores, or proteins involved in cell wall synthesis. For example, labeled analogs of maltose sugars, including ¹⁸F-maltotriose, have been characterized in preclinical infection models and myocarditis (40), with clinical translation imminent. In addition to specificity, accurate and reproducible quantitation will be paramount for the successful rollout of candidate tracers, as the target patient population will inevitably be treated with antibiotics. Although cardiology-based infections have been a focus in the imaging of bacterial infections, applications in other fields for infection should develop in parallel. Opportunities for imaging viral diseases, such as coronavirus disease 2019–related myocarditis, also are emerging.

Clinical Outcome Studies

The growth of molecular cardiac imaging will depend on rigorous prospective clinical trial designs demonstrating clinical impact. As in other areas of NM, having well-designed, prospective clinical trials with relevant hard outcomes is essential to boost the quality of our evidence base to inform decision making. The large number of patients with cardiac disease in the United States lends itself to careful, well-designed clinical studies. Such studies are essential for public health decisions and for the appropriate integration of innovative imaging into clinical care pathways. Such studies should include quality-of-life assessments and hard clinical endpoints and, ideally, should include collecting cost data.

Use of the Full Potential of ¹⁸F-FDG PET

More robust application of quantitative ¹⁸F-FDG PET imaging, particularly in response to therapy, will help maximize its clinical potential. Currently, clinical oncologic PET focuses primarily on the use of ¹⁸F-FDG for the diagnosis, staging, and treatment response assessment of cancer. ¹⁸F-FDG PET, assessing cancer metabolism, can measure cancer activity and burden at multiple time points (41) and is commonly used to determine whether a therapy has been effective. A considerable opportunity for ¹⁸F-FDG PET includes early assessment of the efficacy of cancer therapies to help aid in intensification or deintensification of therapy protocols soon after cancer therapy has started, as opposed to after it has been completed. The success of expanding ¹⁸F-FDG’s utility in this area will be contingent on well-controlled prospective therapeutic clinical trials using ¹⁸F-FDG, with endpoints guided by quantitative metrics. Development and validation of automated software tools and standardized approaches to data collection and analysis, including newer metrics such as metabolic tumor volume and metabolic tumor burden, are likely necessary prerequisites to successful implementation.

New PET Imaging Agents for Cancer

Immune Modulation Therapy

Monoclonal antibodies and other small molecules disrupting immune checkpoints on T cells have drastically changed the landscape of therapy for many types of cancer. In addition, genetically engineered T cells and other types of cells are being applied to treat a variety of neoplasms. These therapies can be remarkably effective but are expensive and not devoid of toxicities. Determining whether these therapies are likely to work—and if they have worked—is an incompletely answered question for ¹⁸F-FDG PET, as it is challenged by imaging metabolic changes in both tumor and host response (78). There are great opportunities to use MI to better interrogate tumors for several variables: their metabolic changes in response to immunotherapy, their levels of checkpoint inhibitor targets in tumors, and their lymphocyte content (79–81). A first-in-humans trial of the anti-CD8 minibody ⁸⁹Zr-IAB22M2C has demonstrated successful targeting of CD8-positive T-cell–rich tissues, including infiltrates in tumors, and further trials are under way to determine whether imaging can inform immunotherapy decisions (82). There are other targets of possible interest as well. In addition, imaging of the presence and activity of genetically reengineered cells in vivo to predict efficacy represents a major opportunity for MI. Better imaging in immunotherapy represents a great opportunity to improve patient management and preserve economic resources.

Optimization of the Dose and Dose Schedule for MTRT

Prospective studies to understand the impact of activity and dosing intervals on tumor response and toxicity will drive informed, optimized dosing schedules. The dose of radiopharmaceuticals, both radioactive and mass dose, as well as the dose schedule, is currently largely empiric. Formal dose escalation studies, which are standard for oncologic drugs, are generally lacking. Since hematologic and other side effects of peptide-based MTRT have generally been mild, higher amounts of activity, shorter dosing intervals, or larger numbers of treatment cycles appear feasible and should be studied systematically. These studies must also investigate the long-term toxicity of MTRT. More safety data are critically necessary to advance application of MTRT to earlier tumor stages, as this toxicity may limit the efficacy of the treatment. Dose escalation using stem cell transplantation approaches is feasible and warrants further study as these methods improve.

Personalized Dosimetry

Powered by advances in both quantitative PET and SPECT/CT, personalized image-based dosimetry drives MTRT to greater effectiveness. With exciting new developments and the expanded use of MTRT, application of patient-specific (personalized) dosimetry may be critical to optimize efficacy while controlling normal-tissue toxicities (90). Substantial efforts are needed to systematically study the outcome of therapies based on actual delivered absorbed doses to tumors and normal tissue through prospective, controlled clinical trials. Efforts are also needed to usher in the opportunity to adjust the dose for optimized delivery of therapy to each patient (precision therapy) (91). To these ends, we envision routine dosimetry based on individualized internal organ geometries and rates of drug uptake and clearance. Such effort may include, but is not limited to, developing streamlined dosimetry methods based on patient-specific, image-based radiopharmaceutical distribution rather than previous-generation geometric models; optimizing the frequency and timing of SPECT/CT or planar imaging studies, as needed, to enable accurate personalized dosimetric calculations; and advancing microdosimetry methods for cellular or DNA-based dosimetry models to support an understanding of radiobiologic effectiveness at the cellular and subcellular levels. Such efforts will be critical in optimized development of α-particle–based therapies for which current dosimetry methods are inappropriate or limited.

Combination of Radionuclide Therapy with Other Forms of Cancer Therapy

MTRT will become a standard therapeutic approach used in combination with more standard cancer treatments at earlier tumor stages. So far, MTRT has been almost exclusively used as monotherapy in patients with very advanced disease and few or no other treatment options (92). Since the efficacy and safety of targeted radionuclide therapy has now been demonstrated unequivocally in this setting (93,94), there is enormous potential in using targeted radionuclide therapy at early tumor stages and in combining targeted radionuclide therapy with other forms of cancer therapy in order to increase the frequency and duration of tumor responses. In addition, efficacy and safety may be increased by combining 2 radiopharmaceuticals with different biodistributions or intratumoral distributions. Substantial and clinically impactful research opportunities exist in this space.

Mechanisms for Radiation-Induced Cytotoxicity

Improvements in understanding the radiobiology of MTRT will inform optimized dosing strategies to better predict response and enhance treatment effects. Although patient-specific dosimetry is likely to explain some of the variability of clinical responses to targeted radionuclide therapy, there are many clinical cases of exceptional responses without a particularly high tumor uptake of the radiopharmaceutical. Conversely, patients may not respond despite high tumor uptake. Possible explanations include heterogeneous expression of the target in different lesions (intermetastatic and intrametastatic heterogeneity) (95), inhomogeneous distribution of the radiopharmaceutical within a lesion because of differences in regional perfusion (96), presence or absence of DNA repair defects, and immunologic effects caused by radionuclide therapy. Systematic studies of these factors are scientifically highly interesting and will likely result in the identification of new rational combinations of radionuclide therapy with other targeted drugs. A much greater understanding of the radiobiology of MTRT will be essential, informed by the most modern scientific methods.

High-Quality Clinical Trials of MTRT

Carefully designed prospective trials comparing the efficacy of MTRT with that of conventional therapies will demonstrate clinical benefits, quality-of-life improvements, and medicoeconomic efficiencies. One of the limitations to the adoption of some NM methods has been a lack of high-quality prospective clinical data. For MTRT to be fully accepted, well-designed prospective clinical trials must be conducted in which the MTRT agents are directly compared with more standard therapies in a randomized fashion. This was done in the development of ²²³RaCl₂ (radium) and in the development of ¹⁷⁷Lu-DOTATATE. Such definitive trials are essential to moving the field ahead, to securing suitable reimbursement, and to achieving dissemination of the methods. Careful trials that show the value of dosimetry as compared with more empiric approaches will be essential. Tracking patient-reported quality of life outcomes and actual costs in delivering patient care will be important as well, to lift nuclear oncology practice to the evidence level of medical, radiation, or surgical oncology.