Regional impact of time-to-equilibrium on indirect Patlak whole-body parametric imaging: a multi-tissue class analysis at the entire body level

Abstract

Purpose

Patlak parametric imaging is widely employed for kinetic modeling due to its simplicity and robustness. The time-to-equilibrium (t*), which must be defined to estimate kinetic parameters, is currently set empirically and uniformly across the entire body. In this study, we evaluate the regional impact of varying t* values on kinetic parameter estimates using a multi-tissue segmentation approach at the whole-body level.

Methods

Data from 53 patients who underwent one-hour dynamic 18 F-FDG PET/CT scans were retrospectively analyzed. Parametric maps of the net influx rate (K_i) and blood distribution volume (dv) were calculated for four t* values (10, 20, 30, and 45 min) using in-house software (PET KinetiX). Voxel-wise K_i and dv values were extracted from 10 predefined tissue structures through automated segmentation. Using t* = 30 min as the widely accepted reference, relative mean errors and relative absolute mean errors of K_i and dv estimated at t*shifts = 10, 20 and 45 min were calculated for each tissue. Pearson correlation coefficients between K_i or dv reference values and those estimated at t* shifts = 10, 20, and 45 min were also computed.

Results

Compared to the reference t*30, K_i estimates ranged from − 21.4% (liver) to 7.3% (SAT) at t*10, and from − 13.8% (lungs) to 2.4% (brain) at t*20. Median absolute bias was 12.8% at t*10 (6.5% brain to > 25% liver) and 8.6% at t*20 (3.2% brain to > 15% lungs and liver). At t*45, K_i was consistently overestimated, with a median bias of 19.4% (2.7% brain to > 33% lungs and liver) and median absolute bias of 19.8% (5.5% brain to > 33% lungs and liver). For dv, biases ranged from − 25.2% (brain) to 8.6% (spleen) at t*10; − 13.7% (brain) to 5.7% (lungs) at t*20; − 15.5% (liver) to 8.8% (brain) at t*45. Median absolute biases were 14.0% at t*10 (9.8% heart to 25.2% brain), 9.4% at t*20 (7.7% heart to 14.1% brain), and 15% at t*45 (12.4% skeletal muscle to 18.5% brain). Regardless of t*, K_i values exhibited strong linear correlations (r > 0.7) across all organs, whereas dv correlations showed greater variability, falling below 0.7 in 80% of organs at t*45.

Conclusion

Kinetic parameter sensitivity to time-to-equilibrium (t*) varies across organs in Patlak whole-body parametric imaging, underscoring the necessity of adopting flexible or adaptive t* values at the whole-body level.

Introduction

PET imaging enables the quantification of pathophysiological processes at the molecular level. By applying mathematical models to dynamic PET data, advanced parametric images are generated, providing deeper insights into biological processes [1]. Since the late 1970s, these methods have primarily been used for research purposes due to limitations in PET gantry coverage, processing time, and other practical challenges. However, the recent commercial availability of PET systems with an extended field-of-view (LAFOV) greater than 1 m, compared to the traditional short axial field-of-view (SAFOV), represents a significant milestone in the clinical integration of these techniques [2, 3–4].

Among these techniques, Patlak parametric imaging is the most widely used kinetic method, appreciated for its simplicity and robustness [5, 6]. The Patlak model is based on two key assumptions: (i) the trapping of the radiotracer metabolites within the cells is irreversible, and (ii) once steady-state conditions are reached (the time to equilibrium after radiotracer injection, denoted as t*), the relationship between volumes of distribution and the normalized time-integral of the tracer concentration in the input fluid becomes linear. Under these assumptions, the slope of the regression line to the data segment starting at time t* is the net influx rate constant Ki (in mL/min/cm³), representing the irreversible uptake of the radiotracer. Currently, t* is empirically set to 30–35 min (t* >30 min) after radiotracer injection for body PET imaging [7, 8–9].

The one-hour acquisition time required for kinetic modeling poses a significant barrier to the widespread clinical adoption of these techniques. As a result, various strategies have been proposed to shorten dynamic scan times. These approaches include combining population-based input functions with abbreviated scan durations that begin at the equilibrium state [7, 9, 10, 11, 12, 13–14] or reducing the acquisition time to the equilibrium time (t*) following radiotracer injection [15]. Although whole-body kinetic modeling has been extensively revisited with the advent of LAFOV PET systems, new challenges have emerged—particularly regarding the determination of t*—regardless of the type of PET gantry used (SAFOV or LAFOV). Due to differences in detection sensitivity, temporal resolution, and image quality [16], extrapolating results to more widely available SAFOV PET systems remains difficult. Furthermore, most current studies focus on small regions of interest (ROIs) within the body. However, with the emergence of AI-based segmentation tools, automatic segmentation of all tissue structures at the body level has become possible [17, 18]. This enables voxel-wise extraction of the corresponding PET data, offering a unique opportunity to assess the impact of t* in a more comprehensive way at the whole-body level.

The objective of this SAFOV PET study was to assess the regional impact of time-to-equilibrium (t*) on indirect Patlak whole-body parametric imaging at the entire body level, by using cutting-edge multi-tissue segmentation methods.

Materials and methods

The present study was conducted in compliance with the declaration of Helsinki. According to the rules of our institution, all patients are systematically informed for data collection and research purposes. Also, all the included patients were specifically informed and did not object to this study, which was approved by our University Ethical Committee (ref: CER-PARIS-SACLAY 2024-76).

Population study

Between May 2023 and August 2024, PET imaging data from 53 consecutive patients who consented to undergo a one-hour dynamic FDG PET acquisition for standard 18 F-FDG PET-CT imaging were retrospectively reprocessed and analyzed.

PET/CT imaging acquisitions

All acquisitions were performed using the same integrated SAFOV PET/CT device (Biograph mCT Flow Motion, Siemens Healthineers, Erlangen, Germany), and fulfilled the international procedure guidelines for FDG PET imaging, including fasting for 6 h and controlled blood glucose levels at the time of the imaging procedure. Each patient underwent one-hour whole-body dynamic PET acquisition (multi-bed multipass scheme, the first pass of 7 min being centered on the thorax), starting with the intravenous injection of 3.5 MBq/kg of ¹⁸F-FDG. The dynamic PET data were reconstructed using the same 3D iterative algorithm (3D time-of-flight–ordered subset expectation maximization method, two iterations and 21 subsets with time-of-flight and point spread function modeling, and with random, dead time, scatter, decay, and attenuation corrections; matrix size = 400 × 400), with postfiltering (Gaussian filter: 3.0 mm).

Image processing and data extraction

All the reconstructed 4D-PET data were processed with PET KinetiX, a clinically-oriented software solution we developed, enabling very fast indirect PET kinetic parametric modeling at the whole-field of view level [19, 20]. For each patient, whole-body parametric maps of the net influx rate of ¹⁸F-FDG within the cells (K_i) and blood distribution volume of ¹⁸F-FDG (dv) were computed voxel-wise according to the Patlak method [6]. For this purpose:

• An image-derived input function was defined by drawing a small sphere of interest within the descending thoracic aorta of each 4D-PET dataset.

• Four time-to-equilibrium values (t*) were applied independently: t* = 30 min (t*30), corresponding to the common historical reference; and t*10, t*20 and t*45, corresponding to times-to-equilibrium of 10 (very early), 20 (early) and 45 min (late) respectively. A total of 440 parametric maps were thus generated (for each of the 55 patients, 2 maps – K_i and dv – were generated 4-times, according to the different t* values).

An in-house and fully automated processing pipeline was developed in Python environment (Python version 3.6; Python Software Foundation, www.python.org) to extract voxel-wise, at the entire body level, the K_i and dv values of the 440 parametric maps within the following 10 predefined tissue structures: brain, heart, lungs, liver, spleen, axial skeleton, appendicular skeleton, subcutaneous adipose tissue (SAT), visceral adipose tissue (VAT), and skeletal muscles. For this purpose, the PET/CT series were spatially resampled within the PET native spaces and segmented using TotalSegmentator version 2, an open-source tool for robust and automated segmentation of anatomical structures (https://github.com/wasserth/TotalSegmentator) [18]; for each tissue structure, K_i and dv values were finally extracted voxel-wise from the 440 parametric maps.

Statistical analyses

Considering t*30 the historical reference of time-to-equilibrium, the relative mean error (rME) and relative absolute mean error (rMAE) of K_i and dv estimated at t*10, t*20, and t*45 were computed per organ, according to the following equations:

1

2

where is the 3D parametric map of K_i or dv estimated at t*10, t*20, or t*45, and . is the 3D parametric map of K_i or dv estimated at t*30. We also calculated Pearson correlation coefficients between K_i or dv values estimated at t* = 30 min and those estimated at t* = 10, 20, and 45 min, respectively, for each organ. Statistical significance was set to 0.05. All the statistical analyses were performed in Python environment (Python version 3.6; Python Software Foundation.

Results

Patient demographics are provided in Table 1. Briefly, the 53 patients were 64% male and 36% females, aged 56 (range 19–82 years). Weights were 72 ± 14.3 kg and heights were 1.7 ± 0.07 m. ¹⁸F-FDG PET/CT was performed for oncological purpose in 68% of cases, for lung nodule characterization in 15% of cases, and for inflammatory – infection purpose in 17% of cases. Blood serum glucose level at the time of PET/CT was under 2.0 g/l for all included patients.

Table 1. Patients characteristics

Total patients	N = 53
Gender
Male, n (%)   Female, n (%)	34 (64%) 19 (36%)
Median Age, years (range)	56 (19–82)
Height, cm (mean +/- SD)	1.71 +/- 0.07
Weight, Kg (mean +/- SD)	72 +/− 14.3
Median glycaemia, mmol/L (range)	5.3 (3.9–9.0)
Diabetes (%)	7 (13%)
Indications
Lung nodule, n (%)	8 (15%)
Lung cancer, n (%)	1 (2%)
Pancreatic cancer, n (%)	3 (6%)
Colorectal cancer, n (%)	4 (7%)
Esophageal cancer, n (%)	1 (2%)
Gastric cancer, n (%)	2 (4%)
Liver cancer, n (%)	1 (2%)
Breast cancer, n (%)	5 (9%)
Pelvic cancer, n (%)	3 (6%)
Tonsillar cancer, n (%)	1 (2%)
Hemopathy, n (%)	7 (13%)
Suspected neoplasia, n (%)	8 (15%)
Unexplained fever, n (%)	5 (9%)
Infection, n (%)	3 (6%)
Granulomatosis, n (%)	1 (2%)
FDG-avid lesions in target organs
Lungs, n (%)	12 (22%)
Liver, n (%)	2 (3%)
SAT/VAT, n (%)	2 (3%)
Axial Skeleton, n (%)	3 (5%)
Appendicular Skeleton, n (%)	0
Heart, n (%)	0
Spleen, n (%)	0
Skeletal muscle, n (%)	0
Brain, n (%)	0

Boxplots comparing rME and rMAE of K_i estimated at equilibrium times of 10 min, 20 min, and 45 min post-injection (t*10, t*20 and t*45) versus the reference t*30 are provided in Figs. 1A and 2A, and the corresponding ranges are provided in Table 2. At t*10 and t*20, K_i values were underestimated in the majority of tissue structures (Fig. 1A), with median average bias of − 6.1% (range: − 21.4% for the liver to 7.3% for the SAT) at t*10 and median average bias of − 5.1% (range: − 13.8% for the lung to 2.4% for the brain) at t*20 (Table 2). The median absolute average bias (Fig. 2A and Table 2) was 12.8% at t*10 (min to max: 6.5% in the brain to over 25% in the liver) and 8.6% at t*20 (min to max: 3.2% in the brain to over 15% in the lungs and liver). On the opposite, K_i values were overestimated in all the structures at t*45 (Fig. 1A), with median average bias of 19.4% (min to max: 2.7% in the brain to over 33% in the lungs and liver, Table 2). The median absolute average bias (Fig. 2A and Table 2) was 19.8% (min to max: 5.5% in the brain to over 33% in the lungs and liver), more than 1.5 time higher compared to those observed at t*10 and t*20. To note, the lungs, liver, and spleen were the most affected tissue structures by the t* values. Voxel-wise relative error maps of K_i at the entire body level for a representative subject are provided in Fig. 3A, highlighting the regional % differences of K_i, maximized in these tissue structures (underestimated at t*10 and t*20, and overestimated at t*45).

[See PDF for image]

Fig. 1

Boxplots comparing rME of Ki (A) and dv (B) estimated at equilibrium times of 10 min, 20 min, and 45 min post-injection (t*10, t*20 and t*45) versus the reference t*30. The x-axis represents the organs, while the y-axis shows the percentage bias relative to the reference t* = 30 min

Table 2. Relative error biases of kinetic parameters at t* = 10, 20, and 45, using t*= 30 as the reference

Organs	Ki						dv
	T*10		T*20		T*45		T*10		T*20		T*45
	rME	rMAE	rME	rMAE	rME	rMAE	rME	rMAE	rME	rMAE	rME	rMAE
Brain	5.2 [1.5; 10.2]	6.5 [3.0; 10.2]	2.4 [− 0.9; 4.8]	3.2 [1.7; 5.6]	2.7 [− 1.1; 7.0]	5.5 [2.1; 8.0]	− 25.2 [− 33.0; − 14.1]	25.2 [14.8; 33.0]	− 13.7 [− 20.3; − 4.5]	14.1 [6.5; 21.1]	8.8 [− 6.6; 25.9]	18.5 [7.5; 28.7]
Heart	− 7.5 [− 17.5; − 2.7]	9.1 [5.7; 18.1]	− 4.9 [− 14.5; − 0.9]	7.4 [3.3; 14.9]	14.5 [6.5; 47.4]	14.5 [6.5; 47.4]	7.6 [− 5.5; 18.0]	9.8 [7.1; 18.4]	3.6 [− 4.5; 10.8]	7.7 [4.0; 14.7]	− 13.8 [− 25.0; − 1.6]	16.2 [8.2; 26.6]
Lungs	− 17.3 [− 34.3; − 7.5]	19.6 [9.5; 34.5]	− 13.8 [− 22.8; − 2.0]	15.7 [6.9; 23.8]	33.6 [23.8; 53.9]	33.6 [23.8; 53.9]	8.1 [− 0.6; 23.6]	11.3 [5.8; 23.6]	5.7 [− 3.0; 13.5]	10.0 [4.5; 17.0]	− 13.4 [− 27.4; − 1.7]	16.0 [6.0; 27.6]
Liver	− 21.4 [− 34.9; − 6.1]	25.6 [13.3; 35.0]	− 12.8 [− 20.1; − 1.6]	15.2 [10.4; 20.6]	33.4 [16.9; 53.5]	33.4 [16.9; 53.5]	8.4 [− 1.7; 21.6]	10.9 [6.8; 21.6]	4.8 [− 1.0; 13.2]	9.1 [4.3; 14.5]	− 15.5 [− 26.9; − 6.0]	15.9 [8.5; 27.2]
Spleen	− 11.6 [− 24.3; − 0.8]	15.5 [7.3; 25.0]	− 8.1 [− 14.3; 0.9]	11.8 [5.0; 17.6]	24.4 [11.1; 37.3]	24.9 [12.1; 37.3]	8.6 [− 3.0; 24.6]	10.8 [7.2; 24.6]	5.0 [− 4.2; 14.2]	8.4 [4.6; 17.2]	− 14.1 [− 26.1; − 1.3]	16.2 [7.3; 27.2]
Axial skeleton	− 0.8 [− 8.3; 8.1]	8.3 [4.7; 13.6]	− 1.3 [− 7.2; 2.6]	5.3 [2.2; 10.3]	16.9 [8.6; 25.7]	16.9 [8.6; 25.7]	− 4.9 [− 14.6; 5.2]	12.2 [5.2; 17.7]	− 1.9 [− 9.0; 6.4]	8.3 [4.1; 14.0]	− 1.2 [− 15.7; 7.7]	14.0 [5.3; 22.1]
Appendicular skeleton	− 0.8 [− 8.4; 8.8]	8.5 [4.2; 14.1]	− 0.06 [− 6.5; 4.8]	5.7 [3.7; 9.0]	10.7 [5.8; 17.1]	11.1 [6.6; 17.1]	− 12.7 [− 22.7; 3.2]	16.2 [9.5; 24.9]	− 6.4 [− 13.6; 5.4]	10.7 [6.0; 16.1]	0.05 [− 7.2; 23.8]	14.21 [4.7; 25.7]
SAT	7.3 [− 3.6; 18.3]	11.8 [6.2; 23.2]	− 0.5 [− 8.7; 5.5]	6.7 [3.9; 11.9]	24.0 [13.2; 39.4]	24.0 [13.2; 39.4]	− 14.5 [− 23.8; − 0.4]	16.3 [6.8; 24.8]	− 6.5 [− 12.1; 2.6]	9.6 [4.3; 15.3]	− 3.2 [− 11.1; 14.1]	13.1 [6.5; 22.5]
VAT	− 12.6 (− 21.8; − 3.9)	13.5 (6.9; 24.5)	− 8.8 (− 15.7; − 3.1)	9,6 (4.1; 15,7)	16.2 [7.0; 26.5]	16.2 [7.0; 26.5]	− 8.1 (− 15.9; 6.8)	13.8 (6.9; 18.3)	− 6.1 [− 9.7; 5.5]	8.5 [6.1; 13.8]	− 1.0 [− 14.6; 10.3]	13.9 [6.1; 20.9]
Skeletal muscle	− 0.9 [− 9.6; 8.9]	9.5 [3.6; 14.2]	− 3.1 [− 8.6; 1.6]	5.9 [3.0; 11.5]	18.0 [10.4; 32.5]	18.0 [10.4; 32.5]	− 5.6 [− 16.9; 10.0]	13.3 [6.8; 17.6]	− 1.1 [− 8.9; 5.9]	8.0 [1.8; 14.3]	− 1.6 [− 12.8; 11.1]	12.4 [6.1; 21.8]
Median average (min; max)	− 6.1 (− 21.4; 7.3)	12.8 (6.5; 25.6)	− 5.1 (− 13.8; 2.4)	8.6 (3.2; 15.7)	19.4 (2.7; 33.6)	19.8 (5.5; 33.6)	− 3.9 (− 25.2; 8.6)	14.0 (9.8; 25.2)	− 1.7 (− 13.7; 5.7)	9.4 (7.7; 14.1)	− 5.5 (− 15.5; 8.8)	15.0 (12.4; 18.5)

Data are expressed as median [IQR]

Averages are expressed as median (min; max)

[See PDF for image]

Fig. 2

Boxplots comparing rMAE of K_i (A) and dv (B) estimated at equilibrium times of 10 min, 20 min, and 45 min post-injection (t*10, t*20 and t*45) versus the reference t*30. The x-axis represents the organs, while the y-axis shows the percentage bias relative to the reference t* = 30 min

[See PDF for image]

Fig. 3

Voxel-wise difference maps to the reference t*30 of K_i (A) and dv (B) at the entire body level for a representative subject. Images were smoothed with a 3D gaussian filter of 3 mm

Boxplots comparing rME and rMAE of dv estimated at t*10, t*20 and t*45 versus the reference t*30 are provided in Figs. 1B and 2B, and the corresponding ranges are provided in Table 2. At t*10, t*20 or t*45, dv values were either under- or either overestimated, depending on the tissue structure (Fig. 1B; Table 2). Median average biases were − 3.9% (min to max: − 25.2% in the brain to 8.6% in the spleen) at t*10; − 1.7% (min to max: − 13.7% in the brain to 5.7% in the lungs) at t*20; and − 5.5% (min to max: − 15.5% in the liver to 8.8% in the brain) at t*45. The median absolute average biases (Fig. 2B and Table 2) were 14.0% (min to max: 9.8% in the heart to 25.2% in the brain) at t*10; 9.4% (min to max: 7.7% in the heart to 14.1% in the brain) at t*20; and 15% (min to max: 12.4% in the skeletal muscle to 18.5% in the brain) at t*45. To note, brain was the most affected tissue structure whatever the t* values. Voxel-wise relative error maps of dv at the entire body level for a representative subject are provided in Fig. 3B. The lungs, liver, spleen showed overestimation at t*10 and t*20 and underestimation at t*45, mirroring K_i errors (Fig. 2A and Table 2).

The correlation values between the reference t*30 and t*10, t*20 and t*45 are reported in Table 3 for K_i, and in Table 4 for dv. A strong linear relationship between Ki values across the different t∗ was consistently observed for all organs (r > 0.7; Table 3). In contrast, the strength of the linear relationships for dv values exhibited greater variability, depending on both the organ and the t∗ (Table 4). In particular, at t*45, correlation coefficients dropped below 0.7 in 80% of the organs.

Table 3. Correlation coefficient values of reference K_i (at t*30) versus K_i estimated at t*10, t*20 and t*45

Organs	Ki_t*30 vs. Ki_t*10	Ki_t*30 vs. Ki_t*20	Ki_t*30 vs. Ki_t*45
Brain	0.99	0.99	0.78
Heart	0.99	0.99	0.97
Lungs	0.96	0.98	0.96
Liver	0.86	0.92	0.92
Spleen	0.93	0.95	0.91
Axial skeleton	0.93	0.96	0.95
Appendicular skeleton	0.97	0.99	0.90
SAT	0.92	0.95	0.93
VAT	0.97	0.99	0.84
Skeletal muscle	0.92	0.95	0.87

All the correlations are statistically significant (p < 0.05)

Table 4. Correlation coefficient values of reference dv (at t*30) versus dv estimated at t*10, t*20 and t*45

Organs	dv_t*30 vs. dv_t*10	dv_t*30 vs. dv_t*20	dv_t*30 vs. dv_t*45
Brain	0.85	0.91	0.84
Heart	0.66	0.80	0.35
Lungs	0.85	0.89	0.76
Liver	0.65	0.79	0.63
Spleen	0.70	0.81	0.59
Axial skeleton	0.72	0.81	0.47
Appendicular skeleton	0.78	0.89	0.67
SAT	0.88	0.91	0.50
VAT	0.90	0.91	0.64
Skeletal muscle	0.70	0.78	0.48

All the correlations are statistically significant (p < 0.05)

Discussion

Patlak parametric imaging assumes an irreversible 2-tissue-compartment tracer kinetic model. At the local level, it requires a fixed time-to-equilibrium t* to estimate the linear fit slope—net influx rate K_{i −} of irreversible radiotracers kinetics. A t* of at least 30 min is currently admitted in practice. In the present study, the one-hour dynamic ¹⁸F-FDG PET/CT data of 53 consecutive patients were retrospectively processed with PET KinetiX – an academic research software we developed and previously validated [19, 20] – to generate whole-body Patlak parametric maps of K_i and dv at t*10, t*20, t*30 (the common reference) and t*45 minutes post-injection respectively. Compared to the empirical reference t*30, the K_i and dv values extracted voxel-wise from 10 tissue classes showed bias variabilities at the entire body level, depending both on the t* and tissue classes. K_i values were mostly underestimated before t*30, and overestimated after t*30. The lungs, liver, and spleen exhibited the highest biases. To note, K_i estimates in the brain, axial and appendicular skeleton, skeletal muscles, and heart remained highly stable at both t*10 and t*20, with biases of less than 10%. Additionally, only the K_i values in the brain were systematically overestimated, regardless of the t* shift applied to the reference t*30. This distinctive feature may be attributed to the blood-brain barrier, which has particular permeability-regulating properties [21, 22]. Regardless of the organ assessed, a strong linear relationship was consistently observed between reference K_i values and those estimated at t*10, t*20, and t*45. Conversely, dv values showed greater sensitivity to t* variations, with the impact differing across organs, while still maintaining a good linear relationship. Naturally, the high correlation observed between kinetic parameters estimated at different equilibrium times underscores the potential for multicenter parameter harmonization—an essential step in validating the diagnostic or predictive value of any biomarker in research practice. Nevertheless, the organ-specific variability in biases associated with a fixed, uniform t* at the whole-body level, as demonstrated in this study, raises significant concerns about the appropriateness of applying a single uniform t* in whole-body Patlak imaging.

Over the past decades, Patlak parametric imaging studies have consistently applied fixed, uniform t* values across the entire field of view, regardless of the underlying rationale [7, 9, 10, 15, 23, 24, 25, 26, 27–28]. However, the time required for the radiotracer to reach equilibrium throughout the body may be influenced by several regional factors. For example, the regional expression of glucose-6-phosphatase contributes to reversible FDG kinetics, especially in the liver, where this enzyme is abundantly expressed [29]. A generalized model incorporating a net efflux rate constant has recently been introduced to mitigate such bias [30]. Additionally, organ-specific blood volume fractions play a crucial role in FDG uptake [31]. This may help explain the greater sensitivity of K_i values to t* variations observed in the liver, spleen, and lungs in this study, which are characterized by relatively high blood volume fractions. However, this factor has not yet been incorporated into Patlak models, introducing potential biases in K_i estimates. Furthermore, while hexokinase activity is commonly studied in FDG PET imaging, tissue-dependent influx rates observed by Patronas and colleagues [32] suggest that other mechanisms, beyond hexokinase, could influence FDG uptake. In light of these considerations, the practice of applying a single arbitrary t* across all anatomical regions in whole-body PET kinetic modeling warrants reconsideration.

This study has several limitations. First, it was a retrospective monocentric design. However, the patient recruitment was applied randomly under daily practice conditions, covering a wide variety of diseases spectra and real-life data. We thus reasonably consider our results generalizable whatever the disease considered. Second, all the data were acquired on analogical SAFOV PET system, which provided intrinsic noisy 4D-PET data. Although better fitting procedure would be expected with LAFOV PET systems, this bias was inherently controlled at the intra-patient level. To note, the Patlak method is robust in the presence of noise, and has been historically recommended by the European Organization for Research and Treatment of Cancer for performing full kinetic analysis of dynamic 18 F-FDG PET studies [33]. And third, few patients exhibited FDG abnormal uptakes within several anatomical structures of interest. These few abnormal uptakes—tumor or inflammatory lesions (less than 5% of patients except lung nodules (22%)—may not necessarily share the same kinetic behavior as healthy tissues, and may have partially corrupted the kinetic parameters. Compared to the thousands to millions of voxels per tissue structure, it is highly probable that such uptakes have only marginally influenced the results. Furthermore, all these marginal abnormal uptakes were impacted in the same way by the different t* at the intra-patient level. The specific kinetic behavior of such lesions, particularly tumors, warrants evaluation in future dedicated studies.

Conclusion

Kinetic parameter sensitivity to time-to-equilibrium (t*) varies across organs in Patlak whole-body parametric imaging, underscoring the necessity of adopting flexible or adaptive t* values at the whole-body level.

Acknowledgements

During the preparation of this manuscript, the authors used ChatGPT 3.5 for language editing. After utilizing this tool, the authors carefully reviewed and revised the content as needed and take full responsibility for the final publication.

Author contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by Florent L. Besson and Abarnaa Sivapathasundaram. The first draft of the manuscript was written by Florent L. Besson and all authors commented on previous versions of the manuscript. All authors read and approved the final manuscript.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Data availability

The datasets generated during and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

All the patients gave their informed consent.

Competing interests

The authors have no relevant financial or non-financial interests to disclose.

Publisher’s note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

References

1. Pantel, AR; Viswanath, V; Muzi, M; Doot, RK; Mankoff, DA. Principles of tracer kinetic analysis in oncology, part I: principles and overview of methodology. J Nucl Med; 2022; 63, pp. 342-52.1:CAS:528:DC%2BB38XhtFGltLrL [DOI: https://dx.doi.org/10.2967/jnumed.121.263518] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35232879][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8978199]

2. Pantel, AR; Viswanath, V; Muzi, M; Doot, RK; Mankoff, DA. Principles of tracer kinetic analysis in oncology, part II: examples and future directions. J Nucl Med; 2022; 63, pp. 514-21.1:CAS:528:DC%2BB38XivVanurjP [DOI: https://dx.doi.org/10.2967/jnumed.121.263519] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35361713][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8973282]

3. Slart, RHJA; Tsoumpas, C; Glaudemans, AWJM; Noordzij, W; Willemsen, ATM; Borra, RJH et al. Long axial field of view PET scanners: a road map to implementation and new possibilities. Eur J Nucl Med Mol Imaging; 2021; 48, pp. 4236-45. [DOI: https://dx.doi.org/10.1007/s00259-021-05461-6] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34136956][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC8566640]

4. Dimitrakopoulou-Strauss, A; Pan, L; Sachpekidis, C. Kinetic modeling and parametric imaging with dynamic PET for oncological applications: general considerations, current clinical applications, and future perspectives. Eur J Nucl Med Mol Imaging; 2021; 48, pp. 21-39. [DOI: https://dx.doi.org/10.1007/s00259-020-04843-6] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/32430580]

5. Gjedde, A. Calculation of cerebral glucose phosphorylation from brain uptake of glucose analogs in vivo: a re-examination. Brain Res Rev; 1982; 4, pp. 237-74.1:CAS:528:DyaL38XkvVyitLY%3D [DOI: https://dx.doi.org/10.1016/0165-0173(82)90018-2]

6. Patlak, CS; Blasberg, RG; Fenstermacher, JD. Graphical evaluation of blood-to-brain transfer constants from multiple-time uptake data. J Cereb Blood Flow Metab; 1983; 3, pp. 1-7.1:STN:280:DyaL3s7gvFegtA%3D%3D [DOI: https://dx.doi.org/10.1038/jcbfm.1983.1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/6822610]

7. Sari, H; Eriksson, L; Mingels, C; Alberts, I; Casey, ME; Afshar-Oromieh, A et al. Feasibility of using abbreviated scan protocols with population-based input functions for accurate kinetic modeling of [18F]-FDG datasets from a long axial FOV PET scanner. Eur J Nucl Med Mol Imaging; 2023; 50, pp. 257-65.1:CAS:528:DC%2BB38XisF2hsb%2FK [DOI: https://dx.doi.org/10.1007/s00259-022-05983-7] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36192468]

8. Li S, Abdelhafez YG, Nardo L, Cherry SR, Badawi RD, Wang G. Total-body parametric imaging using relative patlak plot [Internet]. arXiv; 2024 [cited 2025 Mar 12]. Available from: https://arxiv.org/abs/2406.09720

9. van Sluis, J; van Snick, JH; Glaudemans, AWJM; Slart, RHJA; Noordzij, W; Brouwers, AH et al. Ultrashort oncologic whole-body [18F]FDG Patlak imaging using LAFOV PET. J Nucl Med; 2024; 65, pp. 1652-7. [DOI: https://dx.doi.org/10.2967/jnumed.124.267784] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39353647]

10. Wu, Y; Feng, T; Shen, Y; Fu, F; Meng, N; Li, X et al. Total-body parametric imaging using the Patlak model: feasibility of reduced scan time. Med Phys; 2022; 49, pp. 4529-39. [DOI: https://dx.doi.org/10.1002/mp.15647] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35394071]

11. Dias, AH; Smith, AM; Shah, V; Pigg, D; Gormsen, LC; Munk, OL. Clinical validation of a population-based input function for 20-min dynamic whole-body 18F-FDG multiparametric PET imaging. EJNMMI Phys; 2022; 9, 60. [DOI: https://dx.doi.org/10.1186/s40658-022-00490-y] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36076097][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9458803]

12. Thuillier, P; Bourhis, D; Pavoine, M; Metges, J-P; Le Pennec, R; Schick, U et al. Population-based input function (PBIF) applied to dynamic whole-body 68Ga-DOTATOC-PET/CT acquisition. Front Nucl Med; 2022; 2, 941848. [DOI: https://dx.doi.org/10.3389/fnume.2022.941848] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39390995][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11464975]

13. Pavoine, M; Thuillier, P; Karakatsanis, N; Legoupil, D; Amrane, K; Floch, R et al. Clinical application of a population-based input function (PBIF) for a shortened dynamic whole-body FDG-PET/CT protocol in patients with metastatic melanoma treated by immunotherapy. EJNMMI Phys; 2023; 10, 79. [DOI: https://dx.doi.org/10.1186/s40658-023-00601-3] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38062278][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10703763]

14. Lan, W; Sari, H; Rominger, A; Fougère, CL; Schmidt, FP. Optimization and impact of sensitivity mode on abbreviated scan protocols with population-based input function for parametric imaging of [18F]-FDG for a long axial FOV PET scanner. Eur J Nucl Med Mol Imaging; 2024; 51, pp. 3346-59.1:STN:280:DC%2BB1cfgslakuw%3D%3D [DOI: https://dx.doi.org/10.1007/s00259-024-06745-3] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38763962][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11368996]

15. Liu, G; Shi, Y; Hou, X; Yu, H; Hu, Y; Zhang, Y et al. Dynamic total-body PET/CT imaging with reduced acquisition time shows acceptable performance in quantification of [18F]FDG tumor kinetic metrics. Eur J Nucl Med Mol Imaging; 2024; 51, pp. 1371-82.1:CAS:528:DC%2BB3sXisF2jtLbM [DOI: https://dx.doi.org/10.1007/s00259-023-06526-4] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/38078950]

16. Nadig, V; Herrmann, K; Mottaghy, FM; Schulz, V. Hybrid total-body pet scanners—current status and future perspectives. Eur J Nucl Med Mol Imaging; 2022; 49, pp. 445-59. [DOI: https://dx.doi.org/10.1007/s00259-021-05536-4] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34647154]

17. Shiyam Sundar, LK; Yu, J; Muzik, O; Kulterer, OC; Fueger, B; Kifjak, D et al. Fully automated, semantic segmentation of Whole-Body 18F-FDG PET/CT images based on Data-Centric artificial intelligence. J Nucl Med; 2022; 63, pp. 1941-8. [DOI: https://dx.doi.org/10.2967/jnumed.122.264063] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/35772962]

18. Wasserthal, J; Breit, H-C; Meyer, MT; Pradella, M; Hinck, D; Sauter, AW et al. TotalSegmentator: robust segmentation of 104 anatomic structures in CT images. Radiol Artif Intell; 2023; 5, e230024. [DOI: https://dx.doi.org/10.1148/ryai.230024] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37795137][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10546353]

19. Besson, FL; Faure, S. PET KinetiX—a software solution for PET parametric imaging at the whole field of view level. J Digit Imaging Inf Med; 2024; 37, pp. 842-50. [DOI: https://dx.doi.org/10.1007/s10278-023-00965-z]

20. Faure, S; Paillet, A; Comtat, C; Besson, FL. Fast 4D-PET parametric imaging computation at the whole field of view level: reliability under simulated conditions of PET KinetiX, a dedicated software solution. Eur J Nucl Med Mol Imaging; 2025; 52, pp. 4065-77. [DOI: https://dx.doi.org/10.1007/s00259-025-07285-0] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/40257613][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC12397132]

21. Harris, WJ; Asselin, M-C; Hinz, R; Parkes, LM; Allan, S; Schiessl, I et al. In vivo methods for imaging blood-brain barrier function and dysfunction. Eur J Nucl Med Mol Imaging; 2023; 50, pp. 1051-83. [DOI: https://dx.doi.org/10.1007/s00259-022-05997-1] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/36437425]

22. Chung, KJ; Abdelhafez, YG; Spencer, BA; Jones, T; Tran, Q; Nardo, L et al. Quantitative PET imaging and modeling of molecular blood-brain barrier permeability. Nat Commun; 2025; 16, 3076.1:CAS:528:DC%2BB2MXovFajs7c%3D [DOI: https://dx.doi.org/10.1038/s41467-025-58356-7] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/40159510][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11955546]

23. Cheebsumon, P; Velasquez, LM; Hoekstra, CJ; Hayes, W; Kloet, RW; Hoetjes, NJ et al. Measuring response to therapy using FDG PET: semi-quantitative and full kinetic analysis. Eur J Nucl Med Mol Imaging; 2011; 38, pp. 832-42.1:CAS:528:DC%2BC3MXkt1WrsLw%3D [DOI: https://dx.doi.org/10.1007/s00259-010-1705-9] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/21210109][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3070082]

24. Krak, NC; Hoeven, JJM; Hoekstra, OS; Twisk, JWR; Wall, E; Lammertsma, AA. Measuring [18F]FDG uptake in breast cancer during chemotherapy: comparison of analytical methods. Eur J Nucl Med Mol Imaging; 2003; 30, pp. 674-81.1:CAS:528:DC%2BD3sXjsF2ksb8%3D [DOI: https://dx.doi.org/10.1007/s00259-003-1127-z] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/12640556]

25. Coello, C; Fisk, M; Mohan, D; Wilson, FJ; Brown, AP; Polkey, MI et al. Quantitative analysis of dynamic 18F-FDG PET/CT for measurement of lung inflammation. EJNMMI Res; 2017; 7, 47. [DOI: https://dx.doi.org/10.1186/s13550-017-0291-2] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/28547129][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5445063]

26. Visser, EP; Philippens, MEP; Kienhorst, L; Kaanders, JHAM; Corstens, FHM; De Geus-Oei, L-F et al. Comparison of tumor volumes derived from glucose metabolic rate maps and SUV maps in Dynamic¹⁸ F-FDG PET. J Nucl Med; 2008; 49, pp. 892-8. [DOI: https://dx.doi.org/10.2967/jnumed.107.049585] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/18483085]

27. Liu, G; Yu, H; Shi, D; Hu, P; Hu, Y; Tan, H et al. Short-time total-body dynamic PET imaging performance in quantifying the kinetic metrics of 18F-FDG in healthy volunteers. Eur J Nucl Med Mol Imaging; 2022; 49, pp. 2493-503.1:CAS:528:DC%2BB3MXhvF2qt7vP [DOI: https://dx.doi.org/10.1007/s00259-021-05500-2] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/34417855]

28. Pan, L; Sachpekidis, C; Hassel, J; Christopoulos, P; Dimitrakopoulou-Strauss, A. Impact of different parametric Patlak imaging approaches and comparison with a 2-tissue compartment pharmacokinetic model with a long axial field-of-view (LAFOV) PET/CT in oncological patients. Eur J Nucl Med Mol Imaging; 2025; 52, pp. 623-37. [DOI: https://dx.doi.org/10.1007/s00259-024-06879-4] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/39256215]

29. Bender, D; Munk, OL; Feng, HQ; Keiding, S. Metabolites of (18)F-FDG and 3-O-(11)C-methylglucose in pig liver. J Nucl Med; 2001; 42, pp. 1673-8.1:CAS:528:DC%2BD38XktlOktL0%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/11696638]

30. Karakatsanis, NA; Zhou, Y; Lodge, MA; Casey, ME; Wahl, RL; Zaidi, H et al. Generalized whole-body Patlak parametric imaging for enhanced quantification in clinical PET. Phys Med Biol; 2015; 60, pp. 8643-73.1:CAS:528:DC%2BC28Xmslaht7Y%3D [DOI: https://dx.doi.org/10.1088/0031-9155/60/22/8643] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/26509251][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4710061]

31. Shen, S; DeNardo, SJ; Richman, CM; Yuan, A; Siantar, CH; O’Donnell, RT et al. Planning time for peripheral blood stem cell infusion after high-dose targeted radionuclide therapy using dosimetry. J Nucl Med; 2005; 46, pp. 1034-41.1:CAS:528:DC%2BD28XitFSqt74%3D [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/15937317]

32. Patronas, E-M; Balber, T; Miller, A; Geist, BK; Michligk, A; Vraka, C et al. A fingerprint of 2-[18F]FDG radiometabolites: how tissue-specific metabolism beyond 2-[18F]FDG-6-P could affect tracer accumulation. iScience; 2023; 26, 108137.1:CAS:528:DC%2BB3sXitFyqs7jE [DOI: https://dx.doi.org/10.1016/j.isci.2023.108137] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/37867937][PubMedCentral: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10585399]

33. Young, H; Baum, R; Cremerius, U; Herholz, K; Hoekstra, O; Lammertsma, AA et al. Measurement of clinical and subclinical tumour response using [18F]-fluorodeoxyglucose and positron emission tomography: review and 1999 EORTC recommendations. Eur J Cancer; 1999; 35, pp. 1773-82.1:STN:280:DC%2BD3c7jvV2ntA%3D%3D [DOI: https://dx.doi.org/10.1016/S0959-8049(99)00229-4] [PubMed: https://www.ncbi.nlm.nih.gov/pubmed/10673991]