PET and MR imaging: the odd couple or a match made in heaven?

Although PET/CT and stand-alone MR are independently useful imaging modalities, there are numerous unmet medical needs that may benefit from this new hybrid technology. In the sections below we highlight some of these, starting with clinical situations where the benefits of PET/MR systems are most apparent, and moving to applications where future development might have even broader impact. FDG-PET/CT has improved the diagnostic accuracy in the majority of pediatric malignancies ( 26 – 28 ). Furthermore, studies have suggested that FDG uptake is a compelling and early surrogate marker of treatment efficacy in various pediatric malignancies ( 29 , 30 ). Yet, to date, these studies are limited, and prospective trials although needed, may be driven by the increasing consideration of radiation risk to the pediatric population ( 28 , 31 , 32 ). This is because whenever a PET scan is needed in these patients, a CT scan is also required for attenuation correction or for anatomical correlation. However, the radiation dose being delivered to the patient is of special concern in this population for at least two reasons. First, the effective dose from CT is several times higher in newborns and children than in adults for the same acquisition settings ( 33 ). Second, pediatric patients have a higher lifetime risk of developing cancer relative to adults ( 34 ). MR in PET/MR devices could replace CT for attenuation correction in these patients, reducing the radiation dose by at least 50% compared to a PET/CT study. Furthermore, collecting the data simultaneously limits anesthetic times for the pediatric populations who benefit from both examinations. Similarly, radiation exposure is of concern in patient populations in need for multiple PET/CT scans such as lymphoma patients that require repeated PET/CT scans at the time of diagnosis, interim restaging during therapy and end of treatment staging. However, this comes with significant exposure to ionizing radiation, on the order of 23–26 mSv, with the PET component contributing just 5–7 mSv of this dose. Thus, there is interest in an alternative imaging modality, such as MRI, that provides accurate anatomic localization and functional imaging without the associated excess radiation exposure of CT scanning. In the head and neck area, MRI is superior to CT in terms of accurate staging of tumor extent, involvement of vital soft tissue structures, and nodal involvement ( 35 , 36 ). Because of the added soft tissue discrimination capability of MRI, PET/MRI will likely improve the assessment of tumor extent, involvement of bony structures and bone marrow. Post-treatment surveillance presents many challenges because tissue distortion, scarring and fibrosis from radiation and surgery can obscure early detection of recurrence by conventional follow up. At our institution ~20% of patients with malignancies of the head and neck require both FDG PET/CT and MRI in order to marry the soft tissue discrimination of MRI with the metabolic specificity of FDG PET. MRI is the method of choice for evaluating pelvic malignancies (i.e. gynecological, rectal ( Fig. 2 ), prostate cancers), because of its improved soft tissue discrimination compared to CT. For example, in cervical ( Fig. 3 ) and edometrial cancers, MRI plays a principal role in staging in terms of parametrial invasion (T-staging). However, loco-regional control (discrimination of lymph node metastases) is not fully achieved in the majority of cases. While ovarian cancers are rare, benign adnexal masses (functional cysts, endometriosis, infectious processes, non-malignant growths) are common, thus it is challenging to identify those lesions that will benefit from surgical resection while sparing patients from the morbidity of an unnecessary surgery. MRI has demonstrated improved potential based on morphology alone and T1 and T2 signal intensity characteristics. In all these cases, combining the MR data with PET will likely improve clinical staging accuracy. MRI is also preferred for staging prostate cancer because of its accuracy at diagnosing extracapsular extent and neural invasion and its ability to incorporate multiple specific biomarkers that have shown promise in diagnosing prostate cancer (e.g. dynamic contrast enhanced (DCE) MRI, diffusion weighted imaging (DWI), and magnetic resonance spectroscopy (MRS)). Yet, its sensitivity is only approximately 80% for the primary malignancy, and often becomes much lower when considering bony metastases outside of the pelvis as well as lymph node metastases. A recent study showed that the combination of MRI with 11 C-acetate PET/CT was superior to the individual methods alone for detecting localized prostate cancer ( 37 ). Furthermore, a different study suggested that combining 11 C-choline and apparent diffusion coefficient (ADC) measurements improved the tissue-to-background contrast of Gleason ≥3+4 disease ( 38 ). Thus morphological and potentially multiparametric MRI may be the modality of choice for providing anatomical and physiological correlation to the PET findings, improving the accuracy of the assesment of the primary tumor ( Fig. 4 ) and distant metastases ( 39 ). In breast cancer, MRI has proven very useful for local staging and treatment monitoring and it has greater sensitivity even than conventional imaging methods (i.e. X-ray mammography, sonography). On the other hand, its specificity is variable and could be improved when combined with spatially fused FDG-PET images ( 40 ). Several factors, however, limit the efficacy of spatially registering images acquired from standalone MRI and PET systems and near-perfect spatial co-registration requires simultaneous acquisition. Furthermore, axillary lymph node status, which is the most powerful prognostic indicator in breast cancer patients, can be better assessed using PET/MR using FDG or other radiolabelled targeted agents and MRI lymph node specific agents. The liver is a common site for distant metastases from many subtypes of cancers. MRI has been shown to be useful and superior in its ability to discriminate small lesions (e.g. <10 mm) ( 41 ). FDG-PET remains problematic due to the heterogeneous uptake in the normal liver, its low sensitivity to lesions smaller than 10 mm, and the concomitant decrease in sensitivity in patients with underlying liver disease including cirrhosis and non-alcoholic fatty liver disease ( 42 ). PET/MRI will likely play a large role in determining the true sensitivity of FDG-PET as compared to other liver specific reticuloendothelial, or hepatocyte specific agents for determining true extent of disease. Bone marrow involvement is one of the most important prognostic factors in patients with lymphoma. In a meta-analysis conducted by Wu et al., PET/CT was demonstrated to be superior to MRI or PET alone in the staging of lymphoma ( 43 ). However, this study came under scrutiny and it was suggested that prospective studies for comparing the distinct (and complementary) value of each imaging modality in specific settings are required ( 44 ). Multiple myeloma ( Fig. 5 ) is the most frequent primary neoplasm of the skeletal system. Whole-body MRI has been proposed for detecting infiltrative focal bone marrow lesions and demonstrated to have higher sensitivity than skeletal survey for this task. As compared to PET/CT, the radiation exposure would be minimized in the case of PET/MRI. Furthermore, as novel agents having demonstrated sensitivity in multiple myeloma or novel therapies are developed (e.g. Bortezomib), PET/MRI may be the perfect tool to test these strategies. A growing understanding of the underlying molecular biology of cancer has led to the development of novel therapies targeting various molecular pathways active in cancer. Unlike the conventional cytotoxic chemotherapeutic agents, many of the molecularly targeted agents are cytostatic, causing inhibition of tumor growth rather than tumor regression. In this context, conventional endpoints such as tumor volume reduction may be delayed as compared to other metabolic or physiologic parameters. Combined PET/MR studies may provide important biomarkers to predict and monitor targeted treatment response and to document pharmacodynamic response. The emerging importance of angiogenesis as a cancer therapy target makes assays of vascularity important to clinical research and future clinical practice related to targeted cancer therapy. DCE-MRI allows the assessment of tumor vascularity and detects changes associated with angiogenesis targeted therapy ( 45 – 47 ). For example, in glioblastoma patients treated with vascular endothelial growth factor inhibitors, evidence of tumor vascular normalization was demonstrated using DCE-MRI ( 48 ). However, using only DCE-MRI, the true anti-tumor effects of these agents cannot be completely understood and combining PET parameters (e.g. estimates of tumor glucose metabolism, cellular proliferation, aminoacid transport, etc.) and MRI methods may provide a better approach to this investigation. Furthermore, alternative probing of the microvascular system with magnetic nanoparticles (MNP) has shown promise at interrogating anti-vascular effects in preclinical models ( 49 ). Combined PET/MR measurements could help quantify precisely how tumor vascular properties (assessed by functional MR methods), proliferation and anti-tumor effects (assessed with PET) occur and interact. First, a richer data set is obtained using both imaging modalities ( Fig. 6 ). Second, the quantification of PET could be improved using the simultaneously acquired MR information (e.g. MR-assisted PET motion and partial volume effects correction, MR-based radiotracer arterial input function estimation). Third, the meaning of PET findings could be better understood using MR information. All these tools would enable a more precise understanding of tumor biology and therapeutic response, both for trials of new treatment protocols and perhaps even on an individual basis. The wide range of responses among patients suggests that further studies of individual responses to therapy, correlating structural and functional imaging, metabolic imaging and clinical findings (e.g., survival) will be helpful in understanding the mechanism of action of novel therapeutic agents. In the clinical assessment of patients with cardiovascular disease, PET allows the quantification of blood flow and it is considered the gold standard method for assessing myocardial tissue viability. MR can inform about ventricle function, structural changes, and using contrast agents, perfusion and tissue viability. It was suggested that the combination of these methods in a simultaneous PET/MR scanner ( Fig. 7 ) allows a more detailed risk assessment to be performed ( 50 ). Furthermore, the non-rigid body MR-assisted motion correction methods described above have the potential to significantly improve PET data quantification and reproducibility. Both PET and MRI are also used for molecular imaging cardiovascular applications. In the context of stem cell therapy monitoring, direct labeling either with FDG or gadolinium and MNP and reporter gene approaches allow the non-invasive imaging of stem cells and some of these methods have already been used in clinical trials ( 51 ). PET/MR could improve the short-term assessment of stem cell delivery and the long-term treatment efficacy. A number of groups have recently started to develop dual-labeled PET-MR probes. For example, MNP coupled to chelated 64 Cu have been proposed for targeting vascular inflammation ( 52 ) and tumor integrin α v β 3 expression ( 53 ). In our center we have focused on atherosclerotic plaque imaging as a potential first application for bi-modal probes ( 54 ). The burden from neuropsychiatric disorders (expressed as disability-adjusted life years lost) is higher than the burden from any other disease category in the developed world ( 55 ), yet despite the amazing advances in brain imaging over the last 30 years, imaging has had little to no impact on todays clinical practice ( 56 ). Nevertheless, both PET and MR imaging have had a profound impact on our understanding of neuropsychiatric diseases, from an improved understanding of neurotransmitter imbalances in schizophrenia to an evolving understanding of brain network perturbations in diseases such as depression and autism ( 57 ). Indeed, a comprehensive understanding of psychiatric diseases must encompass the integration of neurochemical, genetic, behavioral and circuit based models. Today the dominant tools for these investigations, at the brain level, are PET and MRI, thus their marriage becomes a natural one for the study of mental illnesses. While simultaneous PET/MR systems are thus likely to be profoundly useful for translational investigations, might this tool find its way into clinical practice? Perhaps the first will be in the study of patients with suspected Alzheimer’s disease (AD). The evolving understanding of AD as a disease encompassing a potentially long prodromal state ( 58 ) preceding definitive clinical manifestations, and the forthcoming arrival of disease modifying treatments will likely require both earlier, and more definitive, diagnosis. In this regard PET and MRI provide complementary information ( Fig. 8 ) in the assessment of AD patients ( 59 – 61 ), with PET’s ability to characterize amyloid (and soon tau) buildup regionally, and MR’s ability to see the associated neuronal degeneration and changes in circuit behavior ( 62 ). Imaging strategies in the future thus will likely extend beyond todays “rule out” with MRI alone (to exclude other organic causes of dementia like tumor or hydrocephalus) to comprehensive “rule in” studies, assessing amyloid and/or tau burden (through improved PET quantification facilitated by MR-assisted PET motion and partial volume effects correction) and their sequellae in terms of direct observation of ongoing neuronal degeneration and cortical dysfunction. Given the likely expense, and potential morbidity, associated with therapy altering treatments, the combined use of PET/MR may provide a cost effective way to assess who should, and should not, enter into such therapeutic regimens. Beyond AD the crystal ball is cloudier, but with no less potential for impact. Today, as novel treatments for disorders such as medically intractable depression are being explored ( 63 ), the need for pre-therapeutic diagnoses to match the precision of these treatments will present itself. Indeed Mayberg used both PET and MR for the pre and post-treatment evaluation of her patients, and it is likely that the combination of modalities will emerge alongside these new therapeutic approaches to provide the needed pharmacological and physiological information required to make informed treatment decisions. One tool that may emerge as a key asset in our exploration of therapeutic options for psychiatric diseases is the combination of PET and MR to study the dynamics of neurotransmission. PET investigations for the last 20 years have certainly provided the foundation for these studies, but the simultaneous collection of functional MR data should allow for important advances in these methods, both through the clearer definition of associated networks, and the concurrent collection of neurophysiologic parameters to refine (perhaps redefine) the traditional kinetic models used to analyze such PET data. Surgical assessment is the gold standard for the diagnosis of lymph node metastases. However, surgical lymphadenectomy confers an increased risk of immediate and delayed complications and noninvasive techniques that accurately identify lymph node metastases are needed. FDG-PET/CT has played a role in evaluating lymph node metastases of multiple oncologic etiologies but has demonstrated mixed sensitivities and specificities. MRI using contrast agents administered interstitially or intravenously has been proposed as an alternative. Ultra-small superparamagnetic iron oxide (USPIO) (a.k.a. MNP) enhanced MRI is the method that has shown the most promise and has been used for N staging of patients with head and neck ( 64 ), breast ( 65 ), gastric ( 66 ) and prostate ( 67 ) cancer. Although USPIO agents are not yet approved for clinical use and further trials are needed to demonstrate their utility, recent provocative evidence has been shown with other FDA approved MRI contrast agents ( 68 ). Several studies investigated the relationship between MR and PET measurements for the assessment of metastatic lymph nodes. For example, statistically significant inverse correlation was observed between ADC values and standardized uptake values (SUVs) in metastatic lymph nodes of head and neck squamous cell carcinoma ( 69 ). In non-small cell lung cancer, short inversion time inversion recovery turbo spin-echo MRI proved more accurate than DWI-MRI and FDG-PET/CT ( 70 ). PET/CT is thought to have a higher sensitivity and specificity to detect retroperitoneal lymph node metastasis compared to current cross-sectional imaging modalities. Although these preliminary results suggest a complementary role for PET and MRI in the evaluation of lymph nodes, these studies would likely benefit from the perfect spatial co-registration and improved PET data quantification (e.g. partial volume effects correction) in a simultaneous PET/MR scanner. FDG is currently used with PET for primary staging, assessing treatment response, and follow-up in more than 90% of cancers ( 71 ). However, not all tumors show significant increase of metabolic activity on FDG-PET imaging secondary to various factors (e.g. variable expression of hexokinase and glucose transporters in hepatocellular carcinoma ( 72 )) and other tracers have entered clinical trials ( 73 ). In the context of drug development, PET allows the characterization of pharmacokinetics and pharmacodynamics of novel agents that can be radiolabeled with positron emitters. A PET/MR scanner may provide an ideal tool for testing and validating these agents. First, the reduced radiation exposure when compared to CT will facilitate the translation to human studies. Second, MRI provides the anatomical details needed for assessing the whole-body distribution of these tracers. Third, the potential for improved data quantification will allow researchers to go beyond semiquantitative methods (e.g. the SUV). Finally, advanced MR techniques will provide physiological information complementary to the PET data, including basic physiological and biophysical measurements, mapping of endogenous metabolites with MRS, etc. MR contrast agents induce relaxation of tissue water, and the extent of this relaxation enhancement, termed relaxivity, depends on a number of factors which influence the accessibility of water to the MR active agent (paramagnetic ion or superparamagnetic nanoparticle), as well as the overall concentration of the contrast agent. In a seminal paper, Louie and colleagues demonstrated that the relaxivity of a specifically designed contrast agent could be changed in the presence of the enzyme β-galactosidase ( 74 ). Numerous publications have followed that described “smart” agents responsive to other enzymes, pH, pO 2 and temperature, amongst other factors. However, a key limitation of these approaches is that the MR signal depends on both the relaxivity of the contrast agent and the local probe concentration – typically the product of these two factors. In vivo, the probe concentration is in general unknown, will change with time, and may vary in diseased versus normal tissue. Using a bimodal PET-MR probe, it has proven to be possible to use PET for estimating the overall concentration and the MR data to determine the molar relaxivity of the agent, allowing for determination of its specific biochemical or physiological “target” ( 75 ).

Despite the technical demands underlying the “matchmaking” between PET and MR, the differences between these two partners is the greatest source of strength in their potential marriage. Odd though the underlying technologies may be to each other, finding ways to bring these two pillars of medical imaging together can confer a high degree of synergy. And like any good marriage, the more intimate the connection between them, the stronger the resulting partnership. Many factors will decide the ultimate role of PET/MR systems within our overall health care system, not the least of which is the cost of such systems, and the degree to which the benefits accrued match the resources required to perform and interpret these studies in the clinic. Training the next generation of interpreters in the art and science of both PET and MR is another challenge, which will have to be met if this tool is to have widespread impact outside a small group of academic sites. Nevertheless, if the future of clinical practice is precision medicine, where therapeutic decision are designed around specific molecular pathological events at the earliest possible stage, then PET/MR systems may be the first of the next generation of molecular imaging tools for that future.