[ 15 O]H 2 O PET: Potential or Essential for Molecular Imaging?

Imaging water pathways in the human body provides an excellent way of measuring accurately the blood ﬂow directed to different organs. This makes it a powerful diagnostic tool for a wide range of diseases that are related to perfusion and oxygenation. Although water PET has a long history, its true potential has not made it into regular clinical practice. The article highlights the potential of water PET in molecular imaging and suggests its prospective role in becoming an essential tool for the 21st century precision medicine in different domains ranging from preclinical to clinical research and practice. The recent technical advances in high-sensitivity PET imaging can play a key accelerating role in empowering this technique, though there are still several challenges to overcome.


Introduction
W ater is a key molecule that empowers life on our planet. Indeed, about 60% of the human body consists of water. It also is the main component of blood plasma and following the flow of water can provide information on perfusion. As blood flow may be altered in disease, imaging water may help to differentiate healthy from diseased tissue. Water can be imaged quantitatively using positron emission tomography (PET) together with tracer amounts of oxygen- 15  O uptake in tissue is linearly related with actual blood flow, whereas this relationship is non-linear for metabolically trapped tracers.
Oxygen-15 has a short half-life of 122 seconds, resulting in short scan protocols and a relatively low overall radiation burden. Consequently, a [ 15 O]H 2 O scan can easily be added to scan protocols with other established or novel tracers. This provides the possibility to correct uptake of those other tracers for flow effects. Quantification is reasonably easy using relatively straightforward kinetic analysis of dynamic PET data, with a typical scanning duration of 5 to 10 minutes. On the other hand, due to the short half-life of oxygen-15, an onsite cyclotron is needed for its production. Ideally, this should be a dedicated (small size) cyclotron in order to ensure optimal availability and minimal impact on the production of other PET tracers within the same facility.
Perfusion imaging using [ 15 (MDR) in Europe, combined with cGMP regulations, its use is possible. With the emergence of digital PET and long axial field-of-view (LAFOV) PET, also referred to as "Total Body PET," dynamic imaging of [ 15 O] H 2 O across large anatomical areas and organ systems within a single scan has become feasible. As LAFOV PET enables the measurement of the image derived arterial input function without having to insert an arterial cannula, use of [ 15 O]H 2 O is expected to continue to expand within both research and clinical applications providing comprehensive information for more accurate diagnosis, precision therapy evaluation and other applications.
The purpose of this review is to outline and dissect current opportunities and challenges for [ 15

History
The use of [ 15 O]H 2 O to measure blood flow was first reported by Ter-Pogossian et al. 1 Originally, the technique was developed for the brain and, as no tomographic techniques were available, detection sensitivity decreased with the depth in tissues and recorded values of flow represented mixtures of grey and white matter. 2,3 Following the introduction of quantitative PET, 4 9 and Raichle et al. 10 This so-called autoradiographic method was based on a method originally developed by Kety. 11 In these early methods, the volume of distribution of water had to be fixed, as the use of a single static scan allowed for estimation of only a single parameter. Currently, nearly all [ 15 O]H 2 O studies are based on dynamic scanning, and when necessary, accompanied with simultaneous sampling of the arterial plasma input function, providing more accurate quantification of both blood flow and volume of distribution of water. 12

Logistics and Workflow
Oxygen-15 with a half-life of 122 seconds is produced with a cyclotron. Due to its short half-life, it is crucial to optimize Clinical and preclinical scanners preferably are connected to an efficient dose delivery system CE marked software available Needs arterial cannulation for arterial input function if the heart or a large artery is not within the field-of-view High heart-to-background contrast Metabolically inert, freely diffusible (ideal) Long axial field-of-view PET allows many more opportunities logistics of production and scanning, and to have the cyclotron in close proximity of the scanner room.
There are several methods to produce oxygen-15 with a cyclotron, the most often used is taking oxygen-15 in the chemical form of the preproduced [ 15 Clark and Buckingham (1975), and in work of 1976 by the research groups of Welch. 13,14 The production was performed using a cyclotron that accelerates deuterons in combination with a low energy beam. To produce 15 O via deuteron acceleration, the target must be filled with a mixture of highly purified nitrogen gas and a trace of oxygen gas. Because of this relatively cheap starting gas mixture, this procedure is suitable for a pass-through gas target, enabling continuous production of [ 15 O]H 2 O. A detailed description of this process is provided by Clark et al. 15 The book "short-lived radioactive gases for clinical use" by Clark and Buckingham describes a complete overview of 15 O radionuclide development processes. 14 A recent publication has brought attention to the eventual microbiological risks of radiopharmaceuticals, as their selfsterilization potential is absent. This means that also the microbiological risk of the instant production of [ 15 O]H 2 O needs to be investigated in a systematic way in order to minimize any microbiological risks. 16 One of the solutions to address microbiological risks in combination with continuous [ 15 O]H 2 O production is to use an examination room based device that has a fully disposable and quickly replaceable sterile injection system, with inline 0.22 mm filters, valves and diffusion chamber next to the patient, such as the Hidex radioactive water generator system (Hidex Oy, Turku, Finland) which is also compatible with the presence of high magnetic fields. 17 Such a bedside system not only allows for rapid patient flow, as little time is needed to make the system ready for the next patient / injection, but also allows for simultaneous monitoring of both bolus shape and total delivered [ 15 O]H 2 O dose to the patient, in turn improving repeatability of the overall [ 15 O]H 2 O bolus shape. Such a system requires the direct communication of the injector and monitoring system with a cyclotron. Another system with a similar bedside philosophy, yet also incorporating the entire tracer production at the bedside, in contrast to the previous system, is the MedTrace system (MedTrace, Hørsholm, Denmark). 18 The introduction of medical cyclotrons designed for high yield 18 F-production, and thus proton only cyclotrons, had a significant boost on the production of oxygen-15. 19 [20][21][22] Preclinical Small animal imaging is a useful resource to understand health and disease, and treatment mechanisms in mammals. Preclinical PET offers the advantage of studying the same molecules in animals before using these in humans. This is also the case with [ 15 O]H 2 O, however, there are some practical limitations and difficulties to perform such scans. Within these, the very short half-life of 15 O makes very close proximity to a cyclotron a necessity. Although clinical installations may be designed with this in mind, preclinical imaging facilities often do not make use of [ 15 O]H 2 O PET as it is seen as too complex. 23 Nevertheless some scientists have warm interest in studying the biological response to [ 15 O]H 2 O. For example, the SAFIR collaboration in Zurich has dedicated a lot of resources to build a pioneering PET insert for preclinical PET -magnetic resonance image (PET-MRI) that aims to allow measuring relatively high injected dose (eg, 500 MBq) of [ 15 O]H 2 O with short frames (up to 5 s) and high resolution (ie, 2 mm) for at least 10 minutes after its injection in order to study the rat brain with higher detail than ever before. 24 Brain Applications [ 15 O]H 2 O brain PET in a preclinical setting has mostly been used to study cerebral perfusion in rats. In animal models of stroke, [ 15 O]H 2 O PET was successfully used to measure changes in perfusion after occlusion of the middle cerebral artery [25][26][27][28][29] or anterior cerebral artery. 30 For example, Martín et al. 25 showed hypoperfusion during middle cerebral artery occlusion followed by reperfusion immediately after the occlusion was released. Then, hypoperfusion could be observed at days 1 and 2, and hyperperfusion at days 4 and 7.
While [ 15 O]H 2 O PET has widely been used to study brain activation in humans, such studies in rats are limited. Wehrl et al. 31 41,42 As discussed earlier, there is a perceived lack of small animal-studies, which reason may be twofold. The relatively long positron range of 15 O (R mean = 3.0 mm), compared to other PET radionuclides such as 18

Clinical Applications
Clinical PET/CT imaging is a key resource to evaluate the (patho)physiological processes underlying some of the most important diseases in the human body. PET/CT offers the advantage of adding information about the physiology on top of the anatomical evaluation. Particularly with [15O] H 2 O, there is the possibility to truly quantify absolute and regional blood flow supply to different organs, as this tracer has a perfect linear relationship with the arterial circulation. This opens up an opportunity of evaluating diverse organs for a wide variety of pathological processes, as differences in the hemodynamic status of tissues have been described between healthy and unhealthy states.

Brain Applications
In the brain, [ 15 O]H 2 O PET can provide valuable insights in cerebral hemodynamic changes under healthy (eg, aging, task-based activities) and diseased (eg, neurodegenerative diseases, psychiatric disorders) conditions. Cerebral metabolic rate of O 2 consumption (CMRO 2 ), cerebral blood flow (CBF), and the balance between those 2, that is, oxygen extraction fraction (OEF), are important indicators of possible disease related changes in brain function 44 Fig. 1.
In a pathophysiological context, many diseases in the brain are highly associated with changes in CBF, either being the main manifestation of the underlying pathophysiological process of a disease, such as the case of cerebrovascular disorders (CVD) or playing a secondary-but still importantrole in disease onset and progression. In a study from Scarmeas and colleagues, CBF was negatively associated with overall cognitive reserves and worse prognosis for Alzheimer's disease (AD). 47 In another study, Xu and colleagues assessed CBF using both [ 15 O]H 2 O and arterial spin labelling (ASL) imaging. They reported a significant age-related decrease in CBF in patients with mild cognitive impairment (MCI). AD patients showed an even more pronounced decrease in CBF when compared with their control counterparts, suggesting a significant effect of AD in the CBF. 48 These findings confirm the possibility of using CBF measurements as a biomarker for diagnosis of AD but this may also function as a biomarker to observe early signs of the disease. 49,50 Although AD is the most studied neurodegenerative disorder, other ones such as Parkinson's disease and Huntington's disease (PD and HD, respectively) also showed altered blood flow comparable to a reduced metabolic activity in cortical and subcortical areas of the brain. 51 55,56 Regarding depression disorder, Dunn and colleagues reported a lack of correlation between blood flow and glucose metabolism in unipolar depression patients, a behavior not observed in patients with bipolar disorder, even when both groups of patients were clinically stable and on fixed-dose medication for at least 3 months. 55 To assess whether changes in CBF could be observed after treatment, 1 study assessed the effects of deep brain stimulation (DBS) on CBF using [ 15  O PET imaging showed differences in CBF between healthy controls and both combat veteran groups when all groups were exposed to an aversive stimulatory event during the scan. Interestingly, combat veterans with PTSD had altered neural responses in the medial prefrontal cortex and amygdala when compared with veterans without PTSD, suggesting disease-specific activation and change in blood flow in these specific regions. 60 These results Figure 1 Single-tissue compartment model parametric brain images with [ 15 O]H 2 O PET from a healthy volunteer from left to right the images correspond to cerebral blood flow (CBF ml¢min À1 ¢g À1 ), volume of distribution (V T -mL¢cm À3 ) and arterial blood volume (V A -mL/g). From top to bottom transaxial, sagittal, and coronal slices. In general, CBF illustrates the influx into grey matter (mainly shown with red and orange) and V T exhibits the ratio between the tracer concentration in the target tissue (light blue). In the transaxial planes CBF, the caudate nucleus and posterior limb of internal capsule can be identified mainly in red or yellow, respectively. In the same plane of V A is depicted part of the cerebral arterial circle (Willis), anterior communicating artery in the joint of the anterior cerebral left and right arteries and left and right posterior cerebral arteries. In the sagittal plane of CBF and V T can be identified in red and yellow (CBF) and light blue (V T ), respectively, the thalamus. In the sagittal plane of V A can be distinguished the internal carotid artery in its cavernous portion. In the coronal plane can be observed in red and yellow from the CBF, the corpus callosum. Finally, in the coronal plane of V A can be recognized the right and left common carotid arteries and their bifurcations into external and internal carotid arteries.
show the current use of [ 15 O]H 2 O PET for measurement of CBF in psychiatric disorders and, although not fully explored, opens the possibility of observing the changes of blood flow during the course of a treatment. Additionally, [ 15 O]H 2 O PET may be able to map different disease patterns that may be interesting in the basic understanding of psychiatric disorders and may contribute to devise a more individualized and precise treatment protocol for these neurological disorders.
In summary, there is a wide range of possibilities for using [ 15 O]H 2 O PET imaging both for a better comprehension of the healthy brain, 61 but also as a possible imaging marker for brain disorders. In combination with other PET tracers such as [ 18  O is freely diffusible across capillary and cell membranes, has an extraction fraction of essentially 100%, and is metabolically inert, it is the ideal tracer for MBF measurements along the entire spectrum of coronary artery disease (CAD). However, as it is not (metabolically) retained in tissues, voxel-based kinetic modelling is time consuming and sensitive to noise. Nevertheless, since the basis function method of the single-tissue compartment model, incorporating right ventricle spill-over, was implemented, it has been possible to automatically generate parametric images of absolute MBF in only a short time frame. 65

Miscellaneous: Tissue, Limb, and Organ Perfusion
Peripheral vascular disease (PVD) is an entity caused by disfunction in the circulatory system, which results from damage, occlusion and/or inflammation of arteries and/or veins, excluding brain and heart vasculature. Most relevant diseases encompassed under the PVD umbrella include peripheral arterial disease (PAD), chronic venous insufficiency (CVI), and deep vein thrombosis (DVT). 73 PAD is an atherosclerotic disease that, when affecting the lower extremities, results in skeletal muscle ischemia, intermittent claudication, and, in more severe stages of disease, limb amputation and death. Several techniques have been used for the evaluation and detection of PVD, including ankleÀbrachial indices, duplex ultrasound, MRI, CT angiography, single photon emission computed tomography (SPECT), and PET. The ankle-brachial index is a widely applied diagnostic tool for the detection of PAD that uses the blood pressure differential between upper and lower extremities to detect a functionally significant arterial obstruction, but this technique can be problematic in the setting of microvascular disease and medial calcification. In vivo nuclear imaging approaches provide high sensitivity and, when using biologically targeted radiotracers, potentially offer novel methods for the investigation of PAD, with integration of perfusion and assessment of tissue oxygenation, metabolism, or biologic processes such as angiogenesis. 74  O reststress PET study found significant differences in flow reserve within the calves of PAD patients when compared with healthy volunteers, and these differences correlated with thermodilution-derived flow reserve values. 77 Another study found significantly reduced exercise-induced muscle blood flow in the distal legs of PVD patients who were referred for lower-limb amputation, suggesting that [ 15 O]H 2 O PET imaging may be a valuable tool for determining the level of subsequent amputations. 78 Kalliokoski et al. demonstrated that PET assessment of skeletal muscle blood flow and oxygen uptake in lowerextremities may be a useful tool for evaluating patient responses to exercise training programs. 79 Apart from assessing (skeletal) muscle, [ 15 O]H 2 O PET imaging has also been applied in the field of tendon studies in athletes during rest and during exercise. 80 Another field of interest is quantification of regional liver and splenic blood flow, where dynamic [ 15 O]H 2 O PET shows promise for clinical use. 81,82 Different factors make accurate estimation of liver blood flow difficult, such as the dual blood flow supply (hepatic artery and portal vein), inaccessibility of the portal vein for direct flow measurements, and changes in extraction kinetics in the pathological liver. 83 Despite these difficulties, standardized and routine quantification of liver blood flow could translate into advances in personalized therapies in patients with distinct hepatic insults. Also, splenic blood flow could be important in the analysis of regional hemodynamic of the spleen in patients with hypersplenism, portal hypertension, after traumatic spleen rupture, and in various other conditions. Brown adipose tissue (BAT) has emerged as a potential target to combat obesity and diabetes, but novel strategies to activate BAT are needed. PET with [ 15 O]H 2 O has been evaluated for direct measurements of adipose tissue perfusion, 84 resulting in a positive correlation with glucose uptake in obese and nonobese healthy subjects, as well as in patients with diabetes, including medical drug response monitoring. 85 Adenosine administration caused a maximal perfusion effect in human supraclavicular BAT, indicating increased oxidative metabolism. 86

ARTICLE IN PRESS
Potential or essential for molecular imaging

Oncology Applications
One of the first applications of using [ 15 O]H 2 O PET was to visualize cerebral tumors by measuring the regional cerebral blood flow together with oxygen utilisation. 87 Furthermore, the first application of [ 15 O]H 2 O PET outside the brain in oncology was the measurement of regional blood flow (in combination with oxygen utilization and blood volume) in patients with breast carcinoma. 88 Angiogenesis, the process in which new blood vessels originate from existing vasculature, is essential for tumor growth, progression, and development of metastases, 89,90 leading to increased blood flow in the tumor. Imaging of tumor blood flow (TBF) for tumor characterization and treatment response monitoring has therefore been studied in various cancers, such as non-small cell lung cancer, 91,92 colorectal cancer, 92,93 breast cancer, 94,95 head and neck cancer, 96 prostate cancer, 97 and brain cancer. 98 The field of cancer drug development would benefit from quantification of the vascular characteristics in tumors to assess the effectiveness of antiangiogenic agents, particularly the combination of [ 15 O]H 2 O with another PET tracer. 99 For example, antiangiogenic treatment was expected to normalize perfusion and therefore improve delivery of chemotherapy. However, a research group from Amsterdam found that reduction in radiolabeled docetaxel uptake was due to a reduction in perfusion after bevacizumab as shown by a combined [ 15 O]H 2 O and 11 C-Docetaxel PET scan. 99 The gold standard for (minimally invasive or even in some cases completely noninvasive) quantitative measurements of TBF is [ 15 102 an image derived input function (IDIF) for noninvasive TBF quantification can only be obtained for studies where the heart is in the (restricted) FOV, except for a recent generation of scanners with dedicated bed motion protocols allowing scanning of the heart area during a critical phase after the injection and also covering the target anatomy, such as multiparametric PET technology. 103 This limitation is resolved by the recent introduction of LAFOV PET, [104][105][106] where the heart together with all main organs and regions of interest can be captured in a single view, ensuring that an IDIF is also possible for structures further away from the heart. The use of an input function measured from the blood pool has been validated successfully in the past for cardiac studies 65 and also for the brain by Iida et al. who combined 2 PET scanners to simultaneously image the brain and the heart in a single scanning session. 107 Consequently, LAFOV dynamic scans allow acquisition of TBF information for multiple lesions within the FOV, which is important in case of intertumor heterogeneity. Given known associations of tumoral heterogeneity and resistance to targeted precision therapy, capturing all lesions simultaneously is important for response monitoring, 108 as overall patient response depends on the response of the poorest lesion. 102

Multitracer Imaging
The short radioactive half-life of oxygen-15 (122 s) together with the corresponding relatively low radiation exposure allows for repeated [ 15 O]H 2 O PET acquisitions at 10 min intervals. This, in turn, provides the possibility to directly measure short-term therapy effects in a single imaging session. Alternatively, a second scan can be performed using a different tracer for further characterization of, for example, tumor biology. 40 The use of more than 1 tracer in 1 scanning session has not been uncommon in clinical research. For example, at Hammersmith Hospital a usual scanning protocol involved imaging human participants with [

Validation and Radiation Dose
A fundamental advantage of [ 15 O]H 2 O used for PET imaging is the ultra-low radiation dose of only 1.2 mSv¢MBq À1 that is delivered to a person. 111 This corresponds to a dose as low as approximately 500 mSv for standard PET scanners, which could be meaningfully lowered even further with highly efficient state-of-the-art (ie, LAFOV) PET scanners, as already demonstrated with [18F]FDG. 112 Furthermore, the development of CT scanners with appropriate filters and acquisition parameters can allow to achieve an ultra-low CT dose (ie, less than 100 mSv), 113 and this would make it possible to use [ 15 O]H 2 O PET/CT as an imaging technique to scan even healthy young volunteers for obtaining "health-related knowledge," as of today, pathophysiological knowledge is "biased" as nuclear imaging techniques are performed only in patients with strict clinical indications due to radiation burden concerns. The fact that ultralow dose combined PET and CT can approach the natural background radiation levels could potentially permit the utilization of PET for measuring the perfusion of the embryo in utero. 114 This could offer new insights regarding the perfusion in the materno-placentalfetal system, and could allow the detection of significant abnormalities in the development of the embryo's organs at very early stages swiftly guiding clinicians to either intervention or termination of pregnancy. 114

Challenges and Future Perspectives
The inert PET tracer [ 15 O]H 2 O represents the accepted gold standard for absolute quantification of tissue perfusion in myocardium, brain and other variety of pathological conditions including cancer. Multiple obstacles thus far have blocked the routine use of PET perfusion imaging, including dependence of a cyclotron, image processing, clinical standardization, regulatory approval, reimbursement, and feasible clinical workflows. 64,115 Fortunately, some of these obstacles have been overcome, especially with the introduction of mini cyclotrons opening the door for PET perfusion imaging to become standard clinical practice in more centers around the world. In the foreseeable future, it is possible that LAFOV PET perfusion imaging with [  O PET a valuable tool for better patient management especially when patients are scanned with LAFOV PET, where arterial cannulation would be avoided, allowing a "true" noninvasive approach. These scanners can measure brain and heart perfusion, organ crosstalk, and absolute tumor blood flow quantification in combination with glycolysis, which will provide important complementary information regarding the prognosis, treatment adequacy, and therapy response.
Furthermore, the application in the context of LAFOV can allow for extending brain activation studies across the linked organ axes / connected body systems, something that is simply not possible with current fMRI technologies.
Future developments in PET-MRI may allow for maximizing the synergistic benefits of PET and MRI in the context of PET reconstruction, including positron range reduction 116 and more advanced PET image reconstruction utilizing information from MRI. 117 Furthermore, simultaneous PET-MRI could potentially offer motion compensation, which is important for more accurate kinetic analysis. 118

Author Contributions
Riemer H. J. A. Slart: Conception and design of the work, data collection for the introduction, preclinical cardiovascular applications, clinical cardiovascular applications, clinical miscellaneous applications, challenges and future perspectives, drafting of the manuscript, critical revision of intellectual content, and final approval of the manuscript; T. Samara Martinez-Lucio: Design of the work, data collection for clinical brain applications, conclusion, drafting of the manuscript, critical revision of intellectual content, and final approval of the manuscript; Hendrikus H. Boersma: Data collection, drafting of the manuscript, critical revision of intellectual content, and final approval of the manuscript; Ronald H. Borra: Data collection for logistics & workflow, implementation, challenges and future perspectives, drafting of the manuscript, critical revision of intellectual content, and final approval of the manuscript; Bart Cornelissen: Data collection for the preclinical oncology applications, drafting of the manuscript, and final approval of the manuscript; Rudi A. J. O. Dierckx: Data collection for clinical brain applications, drafting of the manuscript, and final approval of the manuscript; Magdalena Dobrolinska: Data collection for clinical cardiovascular applications, drafting of the manuscript, and final approval of the manuscript; Janine Doorduin: Data collection for preclinical brain applications, drafting of the manuscript, and final approval of the manuscript; Paola A. Erba: Data collection for challenges and future perspectives, drafting of the manuscript, and final approval of the manuscript; Andor W. J. M. Glaudemans: Data collection for clinical oncology applications, drafting of the manuscript, and final approval of the manuscript; Bruno Lima Giacobbo: Data collection for clinical brain applications, drafting of the manuscript, and final approval of the manuscript; Gert Luurtsema: Data collection for logistics & workflow, drafting of the manuscript, and final approval of the manuscript; Walter Noordzij: Data collection for clinical cardiovascular applications, drafting of the manuscript, and final approval of the manuscript; Joyce van Sluis: Data collection for clinical oncology applications, drafting of the manuscript, and final approval of the manuscript; Charalampos Tsoumpas: Conception and design of the work, data collection for the introduction, multitracer imaging, implementation, challenges and future perspectives, drafting of the manuscript, critical revision of intellectual content, and final approval of the manuscript; Adriaan A. Lammertsma: Conception and design of the work, data collection for the introduction, history drafting of the manuscript, critical revision of intellectual content, and final approval of the manuscript.