Quantitative Total-Body Imaging of Blood Flow with High Temporal Resolution Early Dynamic ¹⁸F-Fluorodeoxyglucose PET Kinetic Modeling

Abstract

Quantitative total-body PET imaging of blood flow can be performed with freely diffusible flow radiotracers such as ¹⁵O-water and ¹¹C-butanol, but their short half-lives necessitate close access to a cyclotron. Past efforts to measure blood flow with the widely available radiotracer ¹⁸F-fluorodeoxyglucose (FDG) were limited to tissues with high ¹⁸F-FDG extraction fraction. In this study, we developed an early-dynamic ¹⁸F-FDG PET method with high temporal resolution kinetic modeling to assess total-body blood flow based on deriving the vascular transit time of ¹⁸F-FDG and conducted a pilot comparison study against a ¹¹C-butanol reference.

Methods:

The first two minutes of dynamic PET scans were reconstructed at high temporal resolution (60×1 s, 30×2 s) to resolve the rapid passage of the radiotracer through blood vessels. In contrast to existing methods that use blood-to-tissue transport rate (${\mathrm{K}}_1$) as a surrogate of blood flow, our method directly estimates blood flow using a distributed kinetic model (adiabatic approximation to the tissue homogeneity model; AATH). To validate our ¹⁸F-FDG measurements of blood flow against a flow radiotracer, we analyzed total-body dynamic PET images of six human participants scanned with both ¹⁸F-FDG and ¹¹C-butanol. An additional thirty-four total-body dynamic ¹⁸F-FDG PET scans of healthy participants were analyzed for comparison against literature blood flow ranges. Regional blood flow was estimated across the body and total-body parametric imaging of blood flow was conducted for visual assessment. AATH and standard compartment model fitting was compared by the Akaike Information Criterion at different temporal resolutions.

Results:

¹⁸F-FDG blood flow was in quantitative agreement with flow measured from ¹¹C-butanol across same-subject regional measurements (Pearson R=0.955, p<0.001; linear regression y=0.973x–0.012), which was visually corroborated by total-body blood flow parametric imaging. Our method resolved a wide range of blood flow values across the body in broad agreement with literature ranges (e.g., healthy cohort average: 0.51±0.12 ml/min/cm³ in the cerebral cortex and 2.03±0.64 ml/min/cm³ in the lungs, respectively). High temporal resolution (1 to 2 s) was critical to enabling AATH modeling over standard compartment modeling.

Conclusions:

Total-body blood flow imaging was feasible using early-dynamic ¹⁸F-FDG PET with high-temporal resolution kinetic modeling. Combined with standard ¹⁸F-FDG PET methods, this method may enable efficient single-tracer flow-metabolism imaging, with numerous research and clinical applications in oncology, cardiovascular disease, pain medicine, and neuroscience.

Introduction

Imaging blood flow has garnered considerable interest over the past 50 years as its dysfunction is characteristic in many diseases.^1–3 PET imaging with a blood flow-specific radiotracer such as ¹¹C-butanol or ¹⁵O-water is widely considered the gold standard for blood flow imaging.^4–6 These flow radiotracers are freely diffusible across capillary membranes^4–6 and accordingly the measured PET signal is closely proportional to blood flow. Blood flow can then be quantified by a simple one-tissue compartment model due to the complete or near-complete extraction of these freely diffusible flow radiotracers.^4–6 Importantly, these flow radiotracers are highly extracted in tissue across the entire body and allow total-body imaging of blood flow.^4,5,7

However, the short half-lives of the radioisotopes in flow radiotracers create practical challenges that hinder their broader accessibility. ¹⁵O-water has a half-life of 2.04 minutes, which necessitates an onsite cyclotron and a dose delivery system.^{8 11}C-butanol comparably has a longer half-life of 20.40 minutes, but still requires nearby production, thus limiting access to urban or research PET centers. Other flow radiotracers, such as ⁸²RbCl and ¹³N-ammonia, similarly have short radioisotope half-lives and high costs in addition to having non-linear uptake with flow.⁹ A blood flow imaging method using a widely available radiotracer such as ¹⁸F-fluorodeoxyglucose (FDG) may mitigate these challenges and open attractive opportunities for imaging of blood flow and glucose metabolism with a single-tracer dynamic scan.

Early dynamic ¹⁸F-FDG PET has been used to measure blood flow in selected tissue such as tumors,¹⁰ liver,¹¹ and myocardium¹² where ¹⁸F-FDG is moderately to highly extracted. The first 2-to-3-minute dynamic ¹⁸F-FDG PET signal is principally weighted towards the initial tissue delivery of the radiotracer¹³ and the higher regional extraction fraction makes the analysis amenable to simplified modeling like that of freely diffusible flow radiotracers. However, these approaches are not generally applicable to other regions like the brain, which has lower ¹⁸F-FDG extraction fraction.^14,15

An intravenously injected tracer is delivered to local tissue vasculature at a rate equal to blood flow. Standard compartmental models neglect this transient process, but distributed kinetic models explicitly model the blood flow and transit time associated with the radiotracer traversing the blood vessels.^16,17 Although described several decades ago, distributed models had limited application in PET due to the poor temporal resolution and statistical quality of time-activity curves measured with conventional PET scanners.^18,19

Total-body PET has substantially greater sensitivity^20–22 over conventional PET systems and allows high-temporal resolution dynamic imaging^21,23 and kinetic modeling.^13,24,25 This may revitalize opportunities to apply distributed kinetic models for blood flow estimation with ¹⁸F-FDG in various tissues. In this study, we developed an early dynamic ¹⁸F-FDG PET method for total-body blood flow imaging with high temporal resolution kinetic modeling and conducted a pilot validation against a ¹¹C-butanol PET reference standard.

Methods

Total-body Dynamic PET

Two human cohorts were pooled in this study, each separately approved by the Institutional Review Board at the University of California, Davis. Written informed consent was obtained for all participants. All participants received total-body dynamic imaging on the uEXPLORER PET/CT system (United Imaging Healthcare, Shanghai, China) with the scan commencing immediately prior to bolus injection of the radiotracer.

The first cohort comprised six participants (4 female; mean age: 67±15 years) with chronic low-back myofascial pain who underwent total-body dynamic PET, receiving bolus injections of both ¹⁸F-FDG (98±9 MBq) and ¹¹C-butanol PET (268±6 MBq) at two scanning sessions within 14 days (clinicaltrials.gov identifier: NCT05876858). The median interval between scans was 9 days (range: 0 to 14). Two participants were scanned on the same day with ¹¹C-butanol scanning commencing first followed by at least a 3-hour interval before ¹⁸F-FDG PET to allow ¹¹C to decay to negligible levels. The second cohort comprised of 34 healthy participants (21 female; mean age: 51±13 years) with no self-reported history of cancer or myocardial infarction in the past 5 years.²⁶ Participants were scanned with total-body dynamic ¹⁸F-FDG PET (mean injected activity: 358±33 MBq, bolus injection) and was used for methodological development and validation against literature blood flow ranges. Two of the six participants from the first cohort and twenty of thirty-four participants in the second cohort self-identified as belonging to racial/ethnic minorities.²⁶

For all dynamic scans, the first two minutes were reconstructed at high temporal resolution (HTR; 60×1 s, 30×2 s) using reconstruction software provided by the vendor. This involved a time-of-flight ordered subset expectation-maximum algorithm-based reconstruction without point spread function modeling and with 4 iterations, 20 subsets, and standard corrections for attenuation, scatter, randoms, dead time, and decay.²² We used a matrix size of 150×150×486 and an isotropic voxel size of 4 mm.

Tracer Kinetic Modeling of Blood Flow from Dynamic FDG Data

Existing methods to measure blood flow with ¹⁸F-FDG have been limited to selected tissue with high extraction fraction such that the blood-to-tissue transport rate ${\mathrm{K}}_1$ approximates blood flow directly^10,11 or by non-linear calibration.¹² ${\mathrm{K}}_1$ is defined as the product of blood flow ($\mathrm{F}$) and extraction fraction ($\mathrm{E}$):

$$K_1=FE,$$ (1)

Equation (1) shows that ${\mathrm{K}}_1$ is a good approximation of blood flow only when $\mathrm{E}$ is close to 1.

¹⁸F-FDG ${\mathrm{K}}_1$ can be measured with early dynamic imaging and a standard one-tissue compartment (S1TC) model as the phosphorylation and dephosphorylation of ¹⁸F-FDG is not identifiable during the first few minutes of the dynamic scan.^13,27 The impulse response function, $R^{S1TC}\left(t\right)$, of the S1TC is (Figure 1):

$$R^{S1TC}(t)=\left\{\begin{array}{cc}{v}_b& t=0,\\ \left(1-v_b\right)K_1{e}^{-k_{2}t}& t>0,\end{array}\right.$$ (2)

where $v_b$ is the blood volume, $K_1$ and $k_2$ the blood-to-tissue transport and clearance rates, respectively. Here, the value of $v_b$ at $t=0$ reflects the compartmental assumption that radiotracer instantaneously and uniformly mixes in regional blood vessels.

Figure 1.

The impulse response functions (a, b) and time-activity curve fits in the cortical grey matter (c, d) using the standard one-tissue compartment (S1TC) model (a, c) and the adiabatic approximation to the tissue homogeneity (AATH) model (b, d) at high temporal resolution (60×1 s, 30×2 s). The dashed red and green lines represent the intravascular and extravascular components of the fitted curve, respectively, and the black arrows (c) indicate areas where S1TC fitting was poor.

In reality, the radiotracer requires a non-zero transit time to traverse the length of the blood vessels at a rate equal to blood flow. This can be explicitly modeled in distributed parameter models.^16,17 Here, we used the adiabatic approximation to the tissue homogeneity (AATH) model,¹⁷ a distributed kinetic model with a closed-form time-domain solution that explicitly models blood flow and a mean vascular transit time. The impulse response function, $R^{AATH}\left(t\right)$ is (Figure 1):

$$R^{AATH}(t)=\left\{\begin{array}{cc}F& 0\le t\le T_c,\\ K_1{e}^{-k_2\left(t-T_c\right)}& t>T_c,\end{array}\right.$$ (3)

where $F$ is blood flow and $T_c$ is the mean vascular transit time for the radiotracer to pass through the length of the blood vessels. The blood volume is accordingly the product of the volumetric blood flow rate and the average time required to traverse the vascular volume ($v_b=F{T}_c$).

The AATH impulse response function describes a finite-time vascular phase ($0\le t\le T_c$) during which the radiotracer traverses the blood vessels at a rate equal to the blood flow. After this mean vascular transit time $\left(t>T_c\right)$, radiotracer exchanges between blood and tissue like a compartment model and thus the impulse response follows an exponential decay like the S1TC model. Accordingly, the impulse response of the AATH and S1TC mainly differ by the presence of a non-zero length vascular phase in the AATH model. We expect that the AATH and S1TC fittings may perform similarly at high extraction fractions as blood flow becomes tightly correlated with ${\mathrm{K}}_1$_. For a general arterial input, $C_a\left(t\right)$, the tissue time-activity curve, $Q\left(t\right)$, can be derived as:

$$Q\left(t\right)=C_a\left(t-t_d\right)\otimes R\left(t\right)$$ (4)

where $t_d$ is the time delay between radiotracer arrival at the measured arterial input location and local tissue vasculature. We used a basis function method for parameter estimation using parametric forms of each model as described previously^28,29 and detailed in the Supplementary Materials. The AATH model was applied on both ¹⁸F-FDG and ¹¹C-butanol.

Image Analysis

Total-body PET enabled non-invasive measurement of an image-derived input function for kinetic analysis. The ascending aorta was used for kinetic modeling of all tissue except the lungs for which a right ventricle input function was used.^24,30,31 Early ¹⁸F-FDG kinetics were quantified by analyzing regional time-activity curves obtained from tissue segmentations in 10 regions of interest (Supplementary Materials).

Total-body parametric images of early kinetics were generated by voxel-wise kinetic modeling on 4-mm isotropic reconstructions. The dynamic images and parametric images were smoothed by the kernel method, which is analogous to nonlocal means denoising.^32,33 Composite image priors were derived from multiple static PET images (FDG: 0–5, 5–20, 20–40, and 40–60 minutes; ¹¹C-butanol: 0–1, 1–2, and 2–3 minutes) and we used 49 nearest neighbors within a 9×9×9 voxel neighbourhood as in our previous work.^32,33

Evaluating Time-Activity Curve Fitting

We compared the quality of the AATH and S1TC model time-activity curve fits using the Akaike information criterion ($\mathrm{A}\mathrm{I}\mathrm{C}$):³⁴

$$AIC=M\mathrm{l}\mathrm{n}\frac{\sum _{m=1}^M{\left(Q\left(t_m\right)-\hat{Q}\left(t_m\right)\right)}^2}{M}+2n+\frac{2n(n+1)}{M-n-1}$$ (5)

where $Q\left(t\right)$ and $\hat{Q}(t)$ are the measured and fitted time-activity curves, respectively, $M$ is the number of frames, $t_m$ is the midpoint time of the $m$th frame, and $n$ is the number of model parameters. The AATH model comprised $n=5$ parameters $\left(t_d,T_c,F,K_1,k_2\right)$ while the S1TC had $n=4\left(t_d,v_b,K_1,k_2\right)$. We computed the difference in AIC (AATH minus S1TC) for each region of interest. A lower AIC indicated better fitting after accounting for the number of model parameters and the residual fitting error. Practical identifiability analysis was also performed as in previous work²⁷ to determine the reliability of AATH parameter estimates.

To evaluate the effect of temporal resolution on the suitability of the AATH model, we frame-averaged each measured regional time-activity curve in the healthy ¹⁸F-FDG PET cohort at 1, 2, 3, 5, and 10 s frame intervals. The resampled data was fitted with the AATH and S1TC models and AIC differences were compared for each region and frame interval.

Validating ¹⁸F-FDG Blood Flow Quantification

The mean and standard deviation of regional blood flow values estimated with the AATH model were computed for all participants. In participants with both ¹⁸F-FDG and ¹¹C-butanol PET, we performed correlation and Bland-Altman analysis³⁵ of regional blood flow estimates between radiotracers. Total-body blood flow parametric images were visually compared between radiotracers. For the healthy ¹⁸F-FDG PET cohort, we compared their average regional values against literature ranges (summarized in Supplementary Table 1) mainly derived from flow-tracer PET.

Results

Time-Activity Curve Fitting and Model Selection

An example high-temporal resolution ¹⁸F-FDG time-activity curve fitting in the cortical grey matter with the S1TC and AATH models is shown in Figure 1. The first-pass peak, which was accurately measured with high temporal resolution dynamic imaging, was better fitted with the AATH model compared to the S1TC model. Furthermore, the peak of the intravascular component (dashed red line) of the AATH fitted curve better aligned with the peak of the measured curve. The intravascular distribution of the S1TC-fitted curve was smaller than that of the AATH model fitting, and to compensate, the extravascular distribution of the S1TC-fitted curve grew larger than that of the AATH. In all regions of interest investigated, the AATH model was preferred on average over the S1TC model across 34 high-temporal resolution dynamic ¹⁸F-FDG scans of healthy participants (Figure 2a).

Figure 2.

Difference in the Akaike Information Criterion (AIC) between the adiabatic approximation to the tissue homogeneity (AATH) and standard one-tissue compartment (S1TC) models at (a) the original high-temporal resolution data (60×1 s, 30×2 s per frame) and (b) at different simulated frame intervals. A negative AIC indicates a preference towards the AATH model balancing fitting error and the number of model parameters. GM indicates grey matter.

Effect of Temporal Resolution on Model Selection

Figure 2b illustrates the difference in AIC between the AATH and S1TC models at different temporal resolutions and tissue regions for our 34 healthy ¹⁸F-FDG cohort. By the AIC, the AATH model had improved fitting over the S1TC model at 1 to 2 s frame intervals, though the magnitude of AIC differences was less at 2 s and a few more individual cases preferred the S1TC model. At 3 s frame interval, the AATH and S1TC models were similarly preferred, but beyond 3 s, the S1TC model was clearly preferred by the AIC.

Validation of ¹⁸F-FDG PET Blood Flow Against ¹¹C-Butanol PET

Correlation and Bland-Altman analysis between ¹¹C-butanol and ¹⁸F-FDG blood flow across all 6 participants, each with 10 tissue regions, are shown in Figure 3. ¹⁸F-FDG blood flow estimated with our proposed method had strong quantitative agreement with the ¹¹C-butanol reference measurement with a Pearson correlation coefficient of 0.955 (p<0.001) and a linear regression slope and intercept of 0.973 and −0.012, respectively. The mean difference (¹⁸F-FDG minus ¹¹C-butanol) was −0.031 ml/min/cm³, indicating that our ¹⁸F-FDG blood flow measures marginally underestimated that of ¹¹C-butanol on average. The Bland-Altman 95% limits of agreement were −0.445 to 0.383 ml/min/cm³ with the larger differences mainly driven by higher blood flow tissues. One participant had severe intra-frame respiratory motion during the ¹¹C-butanol scan, which prevented accurate lung blood flow quantification and substantial overestimation (>1.0 ml/min/cm³) with our ¹⁸F-FDG method. Further analysis stratified by regions with similar blood flow are shown in Supplementary Figures 1 and 2. Standard ¹⁸F-FDG ${\mathrm{K}}_1$ did not strongly agree with ¹¹C-butanol blood flow in general (Figure 3b).

Figure 3.

Correlation (left) and Bland-Altman (right) plots comparing ¹¹C-butanol blood flow against ¹⁸F-fluorodeoxyglucose (FDG) (a) blood flow with the adiabatic approximation to the tissue homogeneity (AATH) model and (b) standard 1-tissue compartment model ${\mathrm{K}}_1$ in the same participants. MD indicates mean difference; SD, standard deviation.

Total-body Parametric Imaging of Blood Flow with ¹⁸F-FDG

Total-body parametric images of blood flow generated with ¹⁸F-FDG and ¹¹C-butanol in the same participant are shown in Figure 4. Parametric images appeared similar both visually and in quantitative range across the body. A notable difference observed between the two blood flow maps was the absence of sagittal and transverse sinus in the ¹¹C-butanol parametric image. This is likely due to the high extraction fraction of ¹¹C-butanol in the brain resulting in its lower venous concentration. One participant had substantial differences in cerebral blood flow between ¹¹C-butanol and ¹⁸F-FDG (Supplementary Figure 3), which may be from a combination of physiological and methodological factors.

Figure 4.

Total-body parametric imaging of blood flow with the proposed early dynamic ¹⁸F-fluorodeoxyglucose (FDG) method compared against a ¹¹C-butanol flow-tracer PET reference in the same participant. The white arrows indicate the sagittal and transverse sinuses in the brain.

Regional ¹⁸F-FDG Early Kinetics in Healthy Participants

The distribution of blood flow estimates with our ¹⁸F-FDG method in 34 healthy participants is plotted in Figure 5. On average, all tissues were within the expected range except the subcortical grey matter and lungs, which were slightly below and above the upper range of average blood flow values reported in literature (Figure 5 and Supplementary Table 1), respectively.⁴ The identifiability of regional blood flow estimates with our proposed method was overall excellent (absolute mean error < 5%, standard deviation < 15%) across tissue regions except the skeletal muscle (mean overestimation of 6.4%; Supplementary Table 2).

Figure 5.

Regional blood flow in 34 healthy participants estimated with the proposed early dynamic ¹⁸F-fluorodeoxyglucose (FDG) method. Plots are separated by the range of blood flow values. Our average estimates mostly fall within the range of average blood flow values reported in literature (Supplementary Table 1) as indicated by the green boxes.

Regional ¹⁸F-FDG extraction fraction values in the healthy cohort are summarized in Table 1. ¹⁸F-FDG extraction fraction varied greatly between tissues across the body. Accordingly, S1TC ¹⁸F-FDG ${\mathrm{K}}_1$ was in general agreement with ¹¹C-butanol blood flow only in tissues with high extraction fraction (Supplementary Figure 4).

Table 1.

¹⁸F-fluorodeoxyglucose (FDG) early kinetics across 34 healthy participants with the adiabatic approximation to the tissue homogeneity (AATH) model

Parameter \ Region	Blood Flow [ml/min/cm3]	K1 [ml/min/cm3]	Extraction Fraction	vb [ml/cm3]	Tc[s]
Cortical GM	0.507±0.122	0.136±0.018	0.278±0.046	0.036±0.006	4.4±0.9
White Matter	0.165±0.041	0.066±0.009	0.416±0.061	0.018±0.003	6.9±1.4
Subcortical GM	0.461±0.128	0.143±0.019	0.327±0.069	0.033±0.005	4.6±1.5
Brainstem	0.339±0.087	0.125±0.015	0.386±0.082	0.030±0.005	5.6±1.8
Cerebellum	0.447±0.104	0.145±0.016	0.336±0.052	0.037±0.005	5.1±1.0
Spleen	1.676±0.484	1.204±0.404	0.728±0.151	0.166±0.064	6.5±3.3
Renal Cortex	1.938±0.402	0.657±0.091	0.348±0.065	0.318±0.039	10.1±1.7
Skeletal Muscle	0.039±0.013	0.034±0.012	0.890±0.048	0.017±0.004	29.1±8.3
Bone Marrow	0.136±0.046	0.130±0.046	0.954±0.051	0.053±0.018	24.6±9.0
Lungs	2.031±0.639	0.072 (0.059–0.134)†	0.041 (0.033–0.067)†	0.143±0.031	4.4±0.8

Data are mean ± standard deviation except those marked with ^†(median [interquartile range]) for the lungs due to two measurements shifting the distribution (Supplementary Figure 5). K₁ indicates the blood-to-tissue transport rate; V_b, blood volume; T_c, mean vascular transit time; GM, grey matter

Discussion

We developed an early dynamic ¹⁸F-FDG PET method for total-body blood flow imaging with high-temporal resolution kinetic modeling and validated against a ¹¹C-butanol reference in a subset of participants scanned with both radiotracers. Conventional methods for ¹⁸F-FDG blood flow imaging have been limited to tissues with relatively high extraction fraction where blood-to-tissue transport rate ${\mathrm{K}}_1$ can approximate blood flow. Our proposed method instead uses a distributed kinetic model^16,17 that explicitly accounts for the blood flow delivery rate of radiotracer to blood vessels, which was resolved with high-temporal resolution dynamic imaging. ¹⁸F-FDG blood flow estimates were in quantitative agreement with ¹¹C-butanol in direct comparisons in the same subjects (Figure 3 and Figure 4). We further validated our method in 34 healthy participants showing regional blood flow values across the body were broadly within literature ranges (Figure 5). We report first data on ¹⁸F-FDG extraction fraction values across the body, which indeed varied between tissue types (4 to 95%; Table 1). To our knowledge, this is the first study to perform total-body blood flow imaging with ¹⁸F-FDG and compare against ¹¹C-butanol flow tracer PET in the same subjects.

Our data indeed showed that high temporal resolution of 1 to 2 s was required for the AATH model to be preferred over the S1TC model based on the AIC metric (Figure 2). The temporal resolution may need to be closer to 1 s for tissue such as the lung where the right ventricle input function often has a very fast, sharp bolus. Total-body PET now allows the requisite temporal resolution for blood flow imaging across the body using the widely available ¹⁸F-FDG radiotracer. Our method is generally applicable across the body in contrast to conventional ¹⁸F-FDG blood flow estimation methods that require high extraction fraction. Directly using ¹⁸F-FDG ${\mathrm{K}}_1$ as a surrogate of blood flow does not generalize across all tissue types (Figure 3b).

Intravenously injected tracers are delivered to tissue vasculature by blood flow. Distributed models explicitly account for this process.^16,18,19 Historically, the AATH distributed model used in this study has been employed for blood flow imaging using inert contrast agents and high temporal resolution dynamic CT or MRI.^36,37 We now show that distributed modeling is applicable to a noninert metabolic radiotracer (¹⁸F-FDG) as well as a freely diffusible flow radiotracer (¹¹C-butanol) provided the requisite temporal resolution is used. This suggests that our method may be generally applicable to a wide range of tracers, enabling single-tracer multiparametric imaging of biologically and physiologically meaningful parameters, such as flow-metabolic imaging^1–3 with ¹⁸F-FDG or joint quantification of blood flow and amyloid burden with amyloid PET tracers.³⁸

Our study did not examine liver and myocardial blood flow. Kinetic modeling of the liver is complicated by its additional portal vein input,³⁹ which could not be accurately measured due to insufficient PET spatial resolution.³⁹ Further development is required to enable hepatic blood flow measurements with our method; however, existing methods^11,39 using ¹⁸F-FDG ${\mathrm{K}}_1$ as a surrogate of hepatic blood flow may be sufficient due to the high permeability of liver sinusoids.⁴⁰ Our initial analysis of the myocardium suggested that spillover from the right and left ventricles were substantial at high temporal resolution dynamic imaging of ¹⁸F-FDG first pass, resulting in substantial blood flow overestimation. We will investigate methods for correcting spillover⁴¹ and cardiac motion⁴² in the future for better quantification of myocardial blood flow with the proposed method.

This study had limitations. First, the sample size of participants scanned with both ¹⁸F-FDG and ¹¹C-butanol was small in this pilot study. Instead, the validity of our early dynamic ¹⁸F-FDG PET blood flow measurements was supported by comparisons of 34 additional healthy participants against literature blood flow ranges. Additional subjects scanned with both radiotracers will be analyzed in future studies. Second, participants were not recruited specifically for validation of ¹⁸F-FDG blood flow. One participant from the dual-tracer group was suspected to have physiological differences between ¹¹C-butanol and ¹⁸F-FDG scans. Future studies will better account for physiological confounds by measuring pCO₂, pO₂, and heart rate among others. Lastly, we did not study patients with major diagnosed blood flow defects such as those with peripheral, carotid, or coronary artery disease among others. Further validation is required under these disease conditions.

Conclusion

This study presented the development of an early dynamic ¹⁸F-FDG PET method with high-temporal resolution kinetic modeling for total-body blood flow imaging. Utilizing the ubiquitous ¹⁸F-FDG radiotracer for blood flow imaging may mitigate the need for a costly and practically-challenging flow-tracer PET scan. In combination with standard ¹⁸F-FDG PET methods for glucose metabolic imaging, our proposed method may allow efficient single-tracer imaging of blood flow and metabolism, resulting in lower radiation exposure to the patient, shorter scan times, and less infrastructural requirements and cost. Our method may be generally applicable to other radiotracers, broadening the possibility of single-tracer multiparametric imaging of biologically and physiologically meaningful parameters from a single dynamic PET scan.

Key Points.

Question:

Can high-temporal resolution early dynamic ¹⁸F-fluorodeoxyglucose (FDG) PET kinetic analysis be used for total-body blood flow imaging?

Pertinent Findings:

Blood flow estimates between ¹⁸F-FDG and gold standard ¹¹C-butanol PET in the same participants showed good agreement across the body. Regional blood flow measurements with the proposed early dynamic ¹⁸F-FDG PET method in 34 healthy participants were within well-established reference ranges in tissues across the body.

Implications for Patient Care:

Total-body blood flow imaging can be performed with the widely available ¹⁸F-FDG radiotracer, possibly mitigating the need for a dedicated flow radiotracer and expanding opportunities to efficiently study blood flow and glucose metabolism in combination with standard ¹⁸F-FDG metabolic imaging methods.