Visual Abstract
Abstract
Recently, PET systems with a long axial field of view have become the current state of the art. Total-body PET scanners enable unique possibilities for scientific research and clinical diagnostics, but this new technology also raises numerous challenges. A key advantage of total-body imaging is that having all the organs in the field of view allows studying biologic interaction of all organs simultaneously. One of the new, promising imaging techniques is total-body quantitative perfusion imaging. Currently, 15O-labeled water provides a feasible option for quantitation of tissue perfusion at the total-body level. This review summarizes the status of the methodology and the analysis and provides examples of preliminary findings on applications of quantitative parametric perfusion images for research and clinical work. We also describe the opportunities and challenges arising from moving from single-organ studies to modeling of a multisystem approach with total-body PET, and we discuss future directions for total-body imaging.
Until today, PET scanners have had a limited scanning area, with approximately 90% of the body outside the field of view (FOV), yielding no signal (1). Consequently, it has not been possible to simultaneously study interactions between organs and whole-body effects of interventions. Recently, PET/CT systems with a long axial FOV, such as the Biograph Vision Quadra (Siemens Healthineers) and the uEXPLORER (United Imaging), have become the state of the art. These new camera systems provide wider (>1 m) anatomic coverage and a significant increase in counting statistics and system sensitivity. The total-body PET scanners enable unique possibilities for scientific research but also for clinical diagnostics (2–7), but this new technology also raises numerous challenges (8–11).
A key advantage of total-body imaging is that having all the organs in the FOV allows studying biologic interaction of all organs simultaneously using a whole-body dynamic scan. For example, when performing cardiac scans, one can measure network-level interactions between the brain, heart, cardiovascular system, liver, pancreas, bowel, and muscles noninvasively. Higher sensitivity allows detecting fast-acting changes. Additionally, in most studies arterial or venous catheterization and blood sampling for tracer input calculation can be avoided because the input function can be obtained from the left ventricular cavity or the aorta.
In this paper, we describe one promising total-body PET imaging target: quantitative perfusion imaging from the perspective of our multidisciplinary total-body PET research team. We discuss the current state of total-body imaging and the advantages of the total-body approach for different imaging targets and clinical questions. We then describe the technical challenges and solutions of quantitative parametric perfusion imaging. Finally, with case examples we discuss the opportunities and challenges arising from moving from single-organ studies to modeling of a multisystem biology approach with total-body PET.
TOTAL-BODY PERFUSION WITH 15O-WATER PET
Although there are reports demonstrating the feasibility of measuring blood flow with different perfusion tracers in specific organs such as 82Rb (12–16), 15O-water has many unique properties that make it particularly suitable for studying perfusion at the whole-body level. The other cardiac tracers have specific limitations that need to be overcome if used outside the heart as the target organ. Tracer extraction may not be ideal and may be nonlinear to flow, and kinetics may depend on metabolism, which needs to be considered.
15O-water is a unique tracer because of its biologically inert characteristics. Its kinetics do not reflect chemical binding or metabolic processes. 15O-water is also freely diffusible across the capillary membrane when passing through the capillary bed and is exchangeable with all tissue water molecules. The chemically inert characteristics of 15O-water also mean that the kinetics of this tracer are solely dependent on the perfusion in the body organs, whereas the areas under the curve are essentially equal to those in the arterial input function in any organs. Modeling of 15O-water PET data is based on the 1-tissue-compartment model (17), related to the Fick principle describing the exchange of inert gas between blood and tissues, in addition to the assumption of instantaneous exchange of 15O-water with all tissue water molecules. PET imaging with 15O-water provides quantitative accuracy for a wide range of transiently changing radioactivity, which is reflected by tissue perfusion values with a dynamic range of 0.04 mL/min/g (e.g., the skeletal muscle at rest) to 6.0 mL/min/g (e.g., the myocardium during stress).
The short half-life of 15O (∼122 s) also provides numerous advantages for assessing perfusion. First, repeated PET imaging for detecting transient changes during a single examination is doable with reasonably short administration intervals. The radiation dose for the patient is considerably small (e.g., 0.39 mSv for administration of 350 MBq of 15O-water), enabling a wide range of applications to radiation-sensitive subject populations, for example, pediatric research subjects, repeat examinations in follow-up studies, normal-physiology experiments, and screening of high-risk subjects for studies.
The introduction of the perfusable tissue fraction improves accuracy in the quantitation of absolute myocardial perfusion by compensating for the partial-volume effects attributed to myocardial wall motion and the limited spatial resolution relative to the thickness of the myocardial wall (18). This approach may also be applied to other organs that have small or thin structures, such as the small intestine and colon. The perfusable tissue fraction concept is applicable in detecting and quantitating the fractional mass of tissue that is water-perfusable in the infarcted myocardium (19,20). Perfusable tissue fraction and its related index could also be good markers in predicting postoperative wall motion recovery after coronary artery revascularization (21,22). The presence of diffuse fibrosis in idiopathic cardiomyopathy could also be detected (23). The approach of using perfusable tissue fraction to assess viability in several organs such as the kidney and liver is promising.
QUANTITATIVE TOTAL-BODY PERFUSION IMAGING APPLICATIONS
Dynamic quantitative imaging of myocardial perfusion is feasible already with conventional PET scanners with a limited axial FOV and has been part of clinical routine for many years (24). The main clinical indication for myocardial perfusion imaging is the detection of obstructive coronary artery disease (CAD) in patients with chest pain or equivalent symptoms. Conventional, nonquantitative myocardial perfusion imaging relies on the detection of relative regional differences in myocardial perfusion, whereas quantitative imaging enables the absolute measurement of myocardial blood flow and flow reserve, which provides improved characterization of the extent and severity of multivessel CAD (16,25–27). Furthermore, quantitative PET myocardial perfusion imaging can detect reduced perfusion due to microvascular dysfunction, which often coexists with obstructive CAD but can also be present in the absence of epicardial CAD (28).
Obstructive CAD shares the same pathophysiologic atherosclerotic process that may be present in, for example, arteries supplying the brain (i.e., cerebrovascular disease) or lower limbs (i.e., peripheral artery disease), and a significant fraction of patients with atherosclerotic cardiovascular disease has more than one organ system involved (29). Similarly, systemic risk factors such as diabetes may promote microvascular dysfunction not only in the heart but also in other organs, such as the brain, eyes, and kidneys (30). Furthermore, cardiovascular diseases involve complex bidirectional interactions between the heart and other organs, including the immune system (31), nervous system (32), kidneys (33), and endocrine system (34). For example, cardiovascular diseases have effects on the nervous system whereas neurologic disorders may affect the cardiovascular system, sometimes called the heart–brain axis (35). Many cardiovascular therapies are also partly based on their effects outside the cardiovascular system, such as in the neurohormonal, endocrine, or immune system (32,36). These findings highlight the need to investigate the interplay of different organ systems, and total-body perfusion imaging provides one solution for this need.
Silent brain infarcts and cerebral hypoperfusion are associated with an increased risk of symptomatic stroke and dementia, and thus their detection in individuals with vascular disease in other parts of the body is important for preventive measures. Whole-body PET imaging could reveal silent brain infarcts or cerebral perfusion deficit in, for example, a subject with atherosclerosis imaged for other indications. The more specific nature of the deficit could be investigated with multimodal PET imaging addressing the cerebral metabolic rate of oxygen, oxygen extraction ratio, and cerebral blood volume together with MRI (37).
Higher mental processes and their neural bases are tightly linked with the internal homeostatic control of the body (38), and the brain responds continuously to the visceral, immunologic, metabolic, and sensorimotor signals (39). Integration of these afferent signals is used for altering bodily functions and driving complex motivated behaviors. The central–peripheral axis and particularly cardiorespiratory functions have emerged as a potential imaging target for affective disorders. For example, anxiety is common in patients with cardiovascular diseases (40,41)—not just because of the fear of cardiac disease but because of altered integration of the central–peripheral emotion circuits. Conversely, anxiety is recognized as an independent risk factor for cardiovascular disease due to an emotion-induced increase in cardiorespiratory tone (42,43), and resting-state amygdala activity and concomitant anxiety are associated with increased risk for cardiovascular disease events and arterial inflammation (44). Nevertheless, systems-level understanding of the crosstalk of the central–peripheral circuits in anxiety and cardiovascular disease remains elusive, and the systemic crosstalk between the brain and the peripheral organs during acute emotions and prolonged perturbations of the emotion circuit remain poorly understood. Total-body PET imaging can potentially transform human systems neuroimaging by shifting the focus from dualistic studies on the brain and periphery to high-resolution activation imaging across the brain and peripheral organs, creating an innovative new multisystem-biology approach for studying the central and peripheral bases of higher mental processes and their dysfunctions.
Alteration in tissue perfusion is also characteristic of metabolic syndrome, and therefore the cellular- and tissue-level mechanisms contributing to the coincidence of cardiovascular disease and insulin resistance continue to be a target of intense research. One line of research has shown that in obesity, insulin resistance, and type 2 diabetes, the vasodilatory effect of insulin is impaired, with insulin resistance thought to associate with endothelial dysfunction. With the novel scanners it has now finally become possible to assess changes in tissue blood flow and the distribution of blood flow occurring in metabolic syndrome and to clarify the interactions between downregulated tissue metabolism and blood flow (45–48).
There are a number of other potential research applications for total-body perfusion PET imaging, but these details are beyond this review. Tumor and metastasis perfusion could help in detecting cancer tissues because perfusion is typically increased in cancer (49) and could be a useful imaging marker for predicting treatment responses (50,51). The heart–kidney connection appears to be important in various conditions and diseases. For example, in the cardiorenal syndrome, the altered function of the heart and the kidneys is linked to regulation of the sympathetic nervous system (52).
TECHNICAL CHALLENGES AND SOLUTIONS OF WHOLE-BODY 15O-WATER PET IMAGING
Tracer Dose
High sensitivity achieved by total-body PET/CT is of value in improving image quality or lowering injected dose. Image quality is known to be worsened when the 15O-water injection dose is low because of a lack of count statistics in the PET images. On the other hand, image quality is degraded if the injection dose is too high, attributed to the increased random rates causing increased noise and systematic errors. We have tested the performance of the Quadra scanner with various 15O-water doses (from 20 to 700 MBq) to understand the performance and limitations of the scanner and to define the image acquisition protocol that can provide reasonable image quality with maintained quantitative accuracy. Our preliminary results suggest that the largest injected dose for optimal image quality is 700 MBq, with no systematic errors on quantitative perfusion being present. On the basis of these results, we have been using an injected dose of 350–400 MBq in ongoing clinical projects.
Automated Production and Injection Control
If very high radioactivity is present in the infusion catheter and in the arm when the injected radioactivity reaches the left ventricle and other organs, there could be problems due to counting rate limitations because the infusion tube has a very weak attenuation. A dedicated 15O-water delivery system (Hidex) was therefore programmed to flush the tracer by saline and minimize the period when high radioactivity is in the catheter, allowing most radioactivity to reach the circulation very quickly. Because of strong attenuation, the total count rate reduces to a reasonable level when tracer is distributed to the body. Automated tracer delivery control is important for reproducibility.
Image Acquisition
List-mode acquisition in total-body studies yields large datasets whose analysis is computationally prohibitive. A typical list-mode raw data file after a 4-min 40-s scan with 350 MBq of 15O-water with a Siemens Vision Quadra is 25 Gb. Although raw data can be archived, in practice the predefined dynamic reconstruction to, for example, 24 frames (14 × 5 s, 3 × 10 s, 3 × 20 s, and 4 × 30 s) is performed with a reconstructed voxel size of 1.65 × 1.65 × 2.8 mm.
Extending PET imaging time with 15O-water does not improve the count statistics, particularly in organs that have high flow. Instead, injected dose and administration duration are the parameters that maximize PET image quality, as long as the counting rate performance of the PET device is not exceeded. Because of the wide range of tissue flow values, high counting rate performance and high sensitivity are essential requirements for simultaneous quantitation of total-body tissue perfusion.
Currently, the reconstruction of dynamic images with a matrix size of 440 × 440 × 380 and 24 frames takes 35 min. The single-image file size is 383 kb and the total series is 3.33 Gb, which can be handled with current image analysis software such as Carimas (Turku PET Centre). Quantitation of perfusion in high-flow tissues such as the heart is feasible also with shorter acquisition times (53), but for tissues with a lower flow, a longer scan duration such as explained above is warranted. In addition, the sensitivity profile of the scanner is uneven, depending on the scanning mode. To achieve maximum sensitivity, the brain should not be positioned at the edge of the FOV because sensitivity decreases toward the edges. Figure 1 shows the simulated time–activity curves of tissues with variable flow levels.
Simulated tissue time–activity curves (black lines) and arterial input function (red line) of dynamic 15O-water scans. AIF = aorta input function curve; f = flow.
Mathematic Model for Quantitation
A single-tissue-compartment model that includes the water-perfusable tissue fraction and the arterial blood component is an out-of-the-box solution for modeling most organs with 15O-water PET. However, a unified fitting solution for kinetic modeling of all organs requires additional work. For example, in the lungs and liver, two input functions are needed. The pulmonary artery carries a large volume of blood (54) whereas the blood supply to the bronchioles through the bronchial artery is much smaller, although oxygen and substrates are carried to the bronchioles under normal physiologic conditions. Lung tumors have been considered perfused by the pulmonary artery system (55). In the liver, the portal vein carries a higher volume of blood than the hepatic artery. One potential solution for modeling of liver perfusion involves estimation of portal vein input function from the measured hepatic arterial input function and liver time–activity curves based on mathematic modeling (56). Although there are several works on lung (57) and liver (56,58) perfusion, the detection of input function from the bronchial artery and the portal vein has been difficult in previous PET scanners. This might be more feasible with total-body PET scanners, which provide better spatial resolution and higher accuracy in measuring a wide range of perfusion levels. Further careful investigation is needed to investigate the potential impacts in case the fundamental assumptions that we applied, such as the instantaneous equilibrium, exchangeability of 15O-water to tissue water, homogeneity in the flow distribution, and no arterial-to-venous shunt, may be limited in some organs. Organ-specific optimization of the model with careful validation is warranted to provide physiologically verified reliable and disease-sensitive tissue perfusion values.
Patient Motion Prevention and Compensation
Patient motion is a challenge for total-body PET systems because body and limb movements have multiple degrees of freedom. Since the entire body is within the imaging FOV at all times, any changes in the patient’s position can result in blurring or issues with attenuation correction and region-of-interest delineation. Comfortable positioning of the patient, limb supports, and head fixation are important means for prospective motion prevention.
One solution to cyclic motion detection is deviceless gating. Advanced reconstruction methods can also address this problem, although reconstruction times may be significantly extended. Motion compensation and tracking in PET systems with a short axial FOV have been concentrated to tracking and correction for cardiac and respiratory motion, as well as their combination. However, motion patterns in PET systems with a long axial FOV are more complex, consisting of complex body motions; subtle motions such as swallowing, nodding, or head bobbing; and peristaltic, cardiac, and respiratory motions. This requires more refined methods in addition to traditionally applied motion correction methods and algorithms, whereas diffeomorphic registration and deep learning–based methods have been applied successfully so far (59,60).
Whereas several tracking solutions have been developed to account for cardiac and respiratory motion, only proof-of-concept methods using optical tracking have been applied so far to track body motion in total-body PET imaging. Thus, there is a need to assess both tracking devices and PET data–based methods to derive motion signals as well as compensation methods applicable to PET systems with a long axial FOV. Because of increased sensitivity, dynamic PET reconstructions with even a subsecond duration are feasible, and novel motion correction approaches can be implemented by deriving motion signals from dynamic PET data alone. For example, changes and oscillations in the dynamic time–activity curves can be used to monitor changes due to respiratory and cardiac motion and to derive motion signals based on these oscillations (1). However, PET data–based methods might be more challenging to apply to tracers such as 15O-water because of fast tracer kinetics.
Tissue Segmentation
An effective and automated segmentation method is a prerequisite for analysis for the large amount of complex data a total-body system produces. Especially, dynamic images with temporal information and PET images with short–half-life traces, such as water, are challenging to segment manually, as this would lead to a significant workload and operator-dependent variability. Therefore, fully automated methods for delineation and analysis of organs, as well as smaller targets such as ischemic lesions, are required.
The state-of-the-art segmentation methods for 15O-water PET imaging consist of the semiautomatic cardiac analysis pipeline, CT-based total-body segmentation, and clustering of total-body PET images. Semiautomatic cardiac segmentation of dynamic PET images involves user-defined markers to segment the myocardium for modeling (61), but this approach does not generalize to the other parts of the image. Fully automated CT-based total-body segmentation algorithms, such as TotalSegmentator (62) and MIWBAS (63), use deep neural networks trained on labeled data. Despite their accuracy, these methods are not yet able to segment all different tissue types or subparts of an organ, such as the medulla and cortex of the kidney. Another drawback in CT-based segmentation applied to PET images is that the anatomic information of CT does not entirely match the PET functional information. It is also apparent that CT and PET images should be coregistered to eliminate potential misalignment, and PET data would require frame-by-frame realignment for correcting subject motion.
Use of unsupervised clustering for PET time–activity curves enables new PET-based total-body segmentation methods applying the physiologic information. The disadvantage is that functional data clustering does not always produce clinically meaningful results, but this problem can be resolved by combining PET-based clustering with CT-based segmentation. An example of total-body CT- and PET-based segmentation is shown in Figure 2.
Example of automated tissue segmentation of total-body PET/CT images done with TotalSegmentator (CT-based), MIWBAS (CT-based), and clustering (PET-based). Segmentations at middle coronal slice are visualized with original CT or PET image (sum over coronal slices to flatten them into 2 dimensions) in background for reference. TotalSegmentator and MIWBAS segment their predefined volumes of interest, whereas clustering segments all areas that end user defines as foreground without annotating resulting segments.
Determination of Arterial Input Function
A higher-quality input function is required for total-body PET examinations than for single-organ measurement, because of the higher range of perfusion values and more blood components in the tissues of some organs. Definition of the volume of interest to obtain the arterial input function is the first step in the postprocessing PET image analysis. The descending aortic artery is typically selected for detection of arterial input function because this location is near most organs. To that end, a single-line volume of interest with 1–2 pixels in each slice and 25–30 cm in length is defined manually. Patient movement can then be considered, and the volume-of-interest data can be corrected. To avoid labor-intensive work, a semiautomated technique is under development. This step must be well validated because of its importance for the quantitative analysis.
Correction for the delay and dispersion of the arterial input function in 15O-water PET studies is also critical. In total-body PET with aortic arterial input function, the magnitude of the correction for delay and dispersion is much smaller than in the classic examinations that used peripheral blood sampling. Of note is that the dispersion-added input function has the same formulation as the tissue radioactivity curve, where arterial input function and the tissue curve are close to each other at a high-flow region. Theoretically, the vascular volume (the third parameter in the model formulation for 15O-water) has the same interpretation as the dispersion time constant if one approximates the dispersion as the single exponential function (64). Delay could have similar effects. Obviously, a unique solution is required to determine the delay and dispersion in addition to the model parameter sets. The work of Wang et al. (65) also introduced an approach that defines the optimal model configuration for each voxel by means of the Akaike information criterion. A similar approach can be applied to 15O-water examinations to identify the model configuration.
Analysis Framework
Large datasets covering multiple organs and tissues propose a major challenge for modeling total-body PET data. A universal voxelwise model cannot be applied across the image because of the variations in tracer uptake kinetics across organs. Thus, radiotracer delay and dispersion parameters would need to be adjusted regionally, and a priori regions of interest need to be delineated for tissue-specific modeling. Neither manual delineation of tens of different regions of interest per subject nor warping of the total-body images accurately into a common stereotactic space is feasible.
For the reasons described above, we have been developing a novel modeling framework for fast, automated, and replicable analysis of total-body PET data. The pipeline uses openly available in-house software (http://www.turkupetcentre.net/petanalysis/) for whole-body and brain data analysis (66). The proposed method (Fig. 3) automates the PET data modeling by joint use of CT and PET images and time–activity curve information for the segmentation. CT-based segmentation (TotalSegmentator) of major organs and tissues is first run for generating tissue-type priors. Next, principal-component analysis–based segmentation on the voxelwise PET time–activity curves is generated for the PET data, whereas CT segmentation is used to restrict clustering within single organs. For brain, either CT- or PET-template–based normalization provides a way to map region-of-interest atlases, such as AAL3 (67) from MNI152 space to native PET space.
Developed framework for whole-body 15O-water PET data processing and kinetic modeling. Validation of this framework is currently in progress. TOTALPET database denotes final image archive of parametric images.
Consequently, a 1-tissue-compartment model is fitted for each segment both at the region-of-interest level and at the voxel level, accounting for the differences in radiotracer delay. This yields tissue-specific mean regional outcome values for statistical analysis (Fig. 4) and organ-specific parametric maps (Fig. 5). Our framework is flexible, allowing unique parametrization of each modeled organ, and thus it can also be readily adapted to model other radiotracers, such as [18F]FDG.
Mean 15O-water different PET time–activity curves within volume-of-interest regions (red) and mean compartment model fits (blue) corresponding to automatically segmented organs. Shaded areas illustrate SD of data and corresponding model fits (means and SD calculated over analyzed studies [n = 93]).
Quantitative parametric perfusion image with organ specific voxel level mathematic modeling that accounts for differences in radiotracer delay.
INITIAL EXPERIENCES WITH TOTAL-BODY QUANTITATIVE PERFUSION PET
Institutional Review Board
The institutional review board of the Welfare Services County of Southwest Finland has approved this study, and all study subjects signed a written informed consent form.
Dynamic Image Quality and Total-Body Parametric 15O-Water Images
Figure 6 shows the dynamic sequence of early frames after injecting 350 MBq of 15O-water. The 10-s bolus of 15O-water in 20 mL was given using the specific delivery system (Hidex) with a flush of 10 mL. The visual image quality was good even with 5-s frame durations, with clear visualization of body organs.
Dynamic 5-s images after 15O-water injection of one of first patients.
The next step was to test the quality of the data and the stability of performing voxel-level fitting of the whole-body dynamic 15O-water image set. The example in Figure 7 demonstrates that good-quality parametric images can be created using a single input function from the aorta and applying the same 1-tissue compartment model for each image voxel of the whole body (rotating videos are provided as Supplemental Video 1, available at http://jnm.snmjournals.org).
One of first cases of total-body parametric images with 15O-water shown by Carimas software. Images are processed to three 3-dimensional parametric components: tissue perfusion, perfusable tissue fraction, and arterial blood volume images. Delay and dispersion corrections were not applied, and same model was used in all organs. Therefore, these parametric images are not yet ideal; for example, K1 is influenced by arterial vessels largely attributed to delay and dispersion, and in heart, K2-weighted model is applied.
Clinical Cases and Preliminary Findings for Total-Body Perfusion Imaging
For nearly 20 y in our center, 15O-water PET has been routinely used to quantify perfusion in absolute terms in the hearts of patients with suspected CAD. Since fall 2022, we have performed all clinical cardiac scans with a total-body scanner aiming to investigate the association of reduced myocardial perfusion with perfusion in other tissues such as the brain and kidneys. Our hypothesis is that obstructive CAD is linked with changes in perfusion in other organs and that these changes may also have clinical consequences. Figure 8 shows an example of a patient who underwent a clinically indicated cardiac 15O-water PET scan because of suspected CAD. As the total-body perfusion was routinely performed, we were able to also study other organs of the body, including the brain. The patient demonstrated a clear perfusion deficit in the right parietooccipital area, and the brain infarction was subsequently confirmed with CT.
Brain infarction in parietooccipital watershed area detected by total-body perfusion PET: coronal PET slice (A), sagittal PET slice (B), transaxial PET slice (C), and transaxial corresponding CT image confirming brain infarction (D). Images are summed PET images of all frames of dynamic scan.
One current research question about the interaction of the brain–heart axis is the effect of anxiety on cardiac and cerebral integration. In this analysis, we benchmarked the automated segmentation and modeling tool (described above) for mapping interactions of the central and peripheral circuits in anxiety during adenosine-induced cardiac stress and at the resting baseline. Proof-of-concept results are shown in Figure 9. Self-reported bodily sensation maps indicate that adenosine evokes unpleasant bodily sensations (Fig. 9A) and that the experience of anxiety is associated with increased cerebral perfusion in key nodes of the limbic, paralimbic, and cortical emotion circuits (Fig. 9B). Interorgan regional correlations in perfusion are widespread during stress and become more focused during adenosine infusion, indicating stronger integration during pharmacologic stress. These results provide the first-ever (to our knowledge) total-body quantification of the central and peripheral circuits during acute anxiety, paving the way for novel network-level studies on the central–peripheral distributed mechanism of higher mental processes.
Proof-of concept results in patients with suspected CAD entering total-body perfusion PET. (A) Adenosine induces unpleasant bodily sensations. (B) Adenosine-induced anxiety is associated with increased perfusion in brain’s emotion circuits. (C and D) Regional correlations in perfusion are widespread during rest (C) but become increasingly focused during adenosine-induced stress (D). FDR = false discovery rate.
FUTURE DEVELOPMENT OF TOTAL-BODY PERFUSION IMAGING
Calculation of parametric images can be expedited using the basis function method (68), in which the convolution integral is tabulated as a function of rate constant K2, and the calculation could be parallelized. This is particularly important for the large datasets acquired with total-body PET. Use of a graphics processing unit can speed up the calculation even more, provided that the kinetic fitting programs are compatible with the unit specification. Several techniques established for cerebral blood flow imaging could also be extended to other organs in total-body data. Since the adequacy of the compartment model is not known in several organs, the graphic plot approach (69) is currently being evaluated to determine the K1 flow without assuming any compartment model configurations and to test the adequacy of the compartmental assumptions. The weighted integral approach (70,71) could also be used as a fast alternative to standard nonlinear least-squares methodology when calculating parametric images.
Efficient data processing and storage solutions are still a work in progress. Currently, our approach involves a centralized computer cluster that includes multiple computer nodes with fast central and graphics processing units and large memories. These cluster nodes may run multiple processers for functional image calculation for each organ, each patient, and different model configurations.
Artificial intelligence solutions will likely further enhance image quality. For example, applying deep learning methods could enhance workflow and speed up computationally intensive tasks related to data analysis, image registration, or reconstruction using models with a relatively short inference time. Moreover, there are multiple applications in which deep learning can be applied: automated analysis, data enhancement via denoising or resolution improvement, estimation of dispersion or delay corrections, estimation of tracer kinetics directly in kinetic modeling, and effective performance of otherwise complex tasks such as motion correction (72).
The longer PET FOV also requires that total-body CT be performed for every subject on each examination, and this increases radiation dose. Even ultra-low-dose attenuation CT scans are currently available with less than a 0.5-mSv total radiation dose for the whole-body scan; alternative low-dose solutions are warranted, especially with pediatric studies. Among several approaches for CT-free attenuation correction, one is to create attenuation data using background γ-rays emitted from the lutetium oxyorthosilicate detector. A small amount of natural 176Lu in the lutetium oxyorthosilicate scintillation detector emits γ-rays of 307, 202, and 88 keV simultaneously (half-life, 38 × 109 y) (73). Using a total-body PET scanner, a μ-map can be calculated with systematic errors of 5%–7% in the same way as for the transmission scan (74). The systematic errors could be reduced by applying machine learning even for postemission attenuation μ-imaging (75). It is of interest to develop a technique without the need for the machine learning approach, because training and validation data may be difficult to acquire with radiation to patients.
A unique application is the myocardial perfusion study, in which respiration-related periodic motion may not be present because of temporal smoothing over the entire transmission acquisition period. This could be used as a reference to evaluate and improve CT-based attenuation correction during respiratory motion.
SUMMARY AND CONCLUSIONS
With new-generation extended-FOV PET/CT scanners, total-body quantitative perfusion imaging appears to have become a realistic possibility. 15O-water as a biologically inert and freely diffusible tracer is likely the tracer of choice for perfusion quantitation. Preliminary experience has already shown that automated image segmentation and analysis appear feasible and promising. Currently, several technical and analysis solutions are still under development and warrant further research. However, it already appears that biologic interaction of all body organs could be studied noninvasively. This ability will open new avenues for research and clinical work and is likely to transform our understanding of the abnormalities of various diseases such as in obesity, diabetes, and cardiovascular diseases.
DISCLOSURE
The study was supported by grants from the Finnish Cultural Foundation, the Jane and Aatos Erkko Foundation, the Finnish Foundation for Cardiovascular Research, Sigrid Juselius stiftelse and Academy of Finland, and Finnish State Research Funding. Juhani Knuuti received consultancy fees from GE Healthcare and Synektik and speaker fees from GE Healthcare, Bayer, Lundbeck, Boehringer-Ingelheim, Pfizer, Siemens Healthineers, and Merck, outside the submitted work. Antti Saraste received consultancy fees from Amgen, Astra Zeneca, Boehringer Ingelheim, and Pfizer and speaker fees from Abbott, Astra Zeneca, and Bayer. No other potential conflict of interest relevant to this article was reported.
Footnotes
Published online Nov. 1, 2023.
- © 2023 by the Society of Nuclear Medicine and Molecular Imaging.
REFERENCES
- Received for publication June 21, 2023.
- Revision received October 4, 2023.
In this issue
Jump to section
- Article
- Visual Abstract
- Abstract
- TOTAL-BODY PERFUSION WITH 15O-WATER PET
- QUANTITATIVE TOTAL-BODY PERFUSION IMAGING APPLICATIONS
- TECHNICAL CHALLENGES AND SOLUTIONS OF WHOLE-BODY 15O-WATER PET IMAGING
- INITIAL EXPERIENCES WITH TOTAL-BODY QUANTITATIVE PERFUSION PET
- FUTURE DEVELOPMENT OF TOTAL-BODY PERFUSION IMAGING
- SUMMARY AND CONCLUSIONS
- DISCLOSURE
- Footnotes
- REFERENCES
- Figures & Data
- Supplemental
- Info & Metrics
Related Articles
Cited By...
- Impact of norepinephrine versus phenylephrine on brain circulation, organ blood flow and tissue oxygenation in anaesthetised patients with brain tumours: study protocol for a randomised controlled trial
- Quantitative Total-Body Imaging of Blood Flow with High Temporal Resolution Early Dynamic 18F-Fluorodeoxyglucose PET Kinetic Modeling
- Mapping 18F-FDG Kinetics Together with Patient-Specific Bootstrap Assessment of Uncertainties: An Illustration with Data from a PET/CT Scanner with a Long Axial Field of View