Non-invasive kinetic modelling approaches for quantitative analysis of brain PET studies

Pharmacokinetic modelling with arterial sampling is the gold standard for analysing dynamic PET data of the brain. However, the invasive character of arterial sampling prevents its widespread clinical application. Several methods have been developed to avoid arterial sampling, in particular reference region methods. Unfortunately, for some tracers or diseases, no suitable reference region can be defined. For these cases, other potentially non-invasive approaches have been proposed: (1) a population based input function (PBIF), (2) an image derived input function (IDIF), or (3) simultaneous estimation of the input function (SIME). This systematic review aims to assess the correspondence of these non-invasive methods with the gold standard. Studies comparing non-invasive pharmacokinetic modelling methods with the current gold standard methods using an input function derived from arterial blood samples were retrieved from PubMed/MEDLINE (until December 2021). Correlation measurements were extracted from the studies. The search yielded 30 studies that correlated outcome parameters (V_T, DVR, or BP_ND for reversible tracers; K_i or CMR_glu for irreversible tracers) from a potentially non-invasive method with those obtained from modelling using an arterial input function. Some studies provided similar results for PBIF, IDIF, and SIME-based methods as for modelling with an arterial input function (R² = 0.59-1.00, R² = 0.71-1.00, R² = 0.56-0.96, respectively), if the non-invasive input curve was calibrated with arterial blood samples. Even when the non-invasive input curve was calibrated with venous blood samples or when no calibration was applied, moderate to good correlations were reported, especially for the IDIF and SIME (R² = 0.71-1.00 and R² = 0.36-0.96, respectively). Overall, this systematic review illustrates that non-invasive methods to generate an input function are still in their infancy. Yet, IDIF and SIME performed well, not only with arterial blood calibration, but also with venous or no blood calibration, especially for some tracers without plasma metabolites, which would potentially make these methods better suited for clinical application. However, these methods should still be properly validated for each individual tracer and application before implementation.

Short-time-window Patlak imaging using a population-based arterial input function and optimized Bayesian penalized likelihood reconstruction: a feasibility study

BACKGROUND

To explore the feasibility of short-time-window Ki imaging using a population-based arterial input function (IF) and optimized Bayesian penalized likelihood (BPL) reconstruction as a practical alternative to long-time-window Ki imaging with an individual patient-based IF. Myocardial Ki images were generated from 73 dynamic ¹⁸F-FDG-PET/CT scans of 30 patients with cardiac sarcoidosis. For each dynamic scan, the Ki images were obtained using the IF from each individual patient and a long time window (10-60 min). In addition, Ki images were obtained using the normalized averaged population-based IF and BPL algorithms with different beta values (350, 700, and 1000) with a short time window (40-60 min). The visual quality of each image was visually rated using a 4-point scale (0, not visible; 1, poor; 2, moderate; and 3, good), and the Ki parameters (Ki-max, Ki-mean, Ki-volume) of positive myocardial lesions were measured independently by two readers. Wilcoxon's rank sum test, McNemar's test, or linear regression analysis were performed to assess the differences or relationships between two quantitative variables.

RESULTS

Both readers similarly rated 51 scans as positive (scores = 1-3) and 22 scans as negative (score = 0) for all four Ki images. Among the three types of population-based IF Ki images, the proportion of images with scores of 3 was highest with a beta of 1000 (78.4 and 72.5%, respectively) and lowest with a beta of 350 (33.3 and 23.5%) for both readers (all p < 0.001). The coefficients of determination between the Ki parameters obtained with the individual patient-based IF and those obtained with the population-based IF were highest with a beta of 1000 for both readers (Ki-max, 0.91 and 0.92, respectively; Ki-mean, 0.91 and 0.92, respectively; Ki-volume, 0.75 and 0.60, respectively; and all p < 0.001).

CONCLUSIONS

Short-time-window Ki images with a population-based IF reconstructed using the BPL algorithm and a high beta value were closely correlated with long-time-window Ki images generated with an individual patient-based IF. Short-time-window Ki images using a population-based IF and BPL reconstruction might represent practical alternatives to long-time-window Ki images generated using an individual patient-based IF.

Population-based input function for TSPO quantification and kinetic modeling with [11C]-DPA-713

INTRODUCTION

Quantitative positron emission tomography (PET) studies of neurodegenerative diseases typically require the measurement of arterial input functions (AIF), an invasive and risky procedure. This study aims to assess the reproducibility of [11C]DPA-713 PET kinetic analysis using population-based input function (PBIF). The final goal is to possibly eliminate the need for AIF.

MATERIALS AND METHODS

Eighteen subjects including six healthy volunteers (HV) and twelve Parkinson disease (PD) subjects from two [11C]-DPA-713 PET studies were included. Each subject underwent 90 min of dynamic PET imaging. Five healthy volunteers underwent a test-retest scan within the same day to assess the repeatability of the kinetic parameters. Kinetic modeling was carried out using the Logan total volume of distribution (VT) model. For each data set, kinetic analysis was performed using a patient-specific AIF (PSAIF, ground-truth standard) and then repeated using the PBIF. PBIF was generated using the leave-one-out method for each subject from the remaining 17 subjects and after normalizing the PSAIFs by 3 techniques: (a) Weightsubject×DoseInjected, (b) area under AIF curve (AUC), and (c) Weightsubject×AUC. The variability in the VT measured with PSAIF, in the test-retest study, was determined for selected brain regions (white matter, cerebellum, thalamus, caudate, putamen, pallidum, brainstem, hippocampus, and amygdala) using the Bland-Altman analysis and for each of the 3 normalization techniques. Similarly, for all subjects, the variabilities due to the use of PBIF were assessed.

RESULTS

Bland-Altman analysis showed systematic bias between test and retest studies. The corresponding mean bias and 95% limits of agreement (LOA) for the studied brain regions were 30% and ± 70%. Comparing PBIF- and PSAIF-based VT estimate for all subjects and all brain regions, a significant difference between the results generated by the three normalization techniques existed for all brain structures except for the brainstem (P-value = 0.095). The mean % difference and 95% LOA is -10% and ±45% for Weightsubject×DoseInjected; +8% and ±50% for AUC; and +2% and ± 38% for Weightsubject×AUC. In all cases, normalizing by Weightsubject×AUC yielded the smallest % bias and variability (% bias = ±2%; LOA = ±38% for all brain regions). Estimating the reproducibility of PBIF-kinetics to PSAIF based on disease groups (HV/PD) and genotype (MAB/HAB), the average VT values for all regions obtained from PBIF is insignificantly higher than PSAIF (%difference = 4.53%, P-value = 0.73 for HAB; and %difference = 0.73%, P-value = 0.96 for MAB). PBIF also tends to overestimate the difference between PD and HV for HAB (% difference = 32.33% versus 13.28%) and underestimate it in MAB (%difference = 6.84% versus 20.92%).

CONCLUSIONS

PSAIF kinetic results are reproducible with PBIF, with variability in VT within that obtained for the test-retest studies. Therefore, VT assessed using PBIF-based kinetic modeling is clinically feasible and can be an alternative to PSAIF.

Assessment of population-based input functions for Patlak imaging of whole body dynamic 18F-FDG PET

BACKGROUND

Arterial blood sampling is the gold standard method to obtain the arterial input function (AIF) for quantification of whole body (WB) dynamic 18F-FDG PET imaging. However, this procedure is invasive and not typically available in clinical environments. As an alternative, we compared AIFs to population-based input functions (PBIFs) using two normalization methods: area under the curve (AUC) and extrapolated initial plasma concentration (CP*(0)). To scale the PBIFs, we tested two methods: (1) the AUC of the image-derived input function (IDIF) and (2) the estimated CP*(0). The aim of this study was to validate IDIF and PBIF for FDG oncological WB PET studies by comparing to the gold standard arterial blood sampling.

METHODS

The Feng 18F-FDG plasma concentration model was applied to estimate AIF parameters (n = 23). AIF normalization used either AUC(0-60 min) or CP*(0), estimated from an exponential fit. CP*(0) is also described as the ratio of the injected dose (ID) to initial distribution volume (iDV). iDV was modeled using the subject height and weight, with coefficients that were estimated in 23 subjects. In 12 oncological patients, we computed IDIF (from the aorta) and PBIFs with scaling by the AUC of the IDIF from 4 time windows (15-45, 30-60, 45-75, 60-90 min) (PBIFAUC) and estimated CP*(0) (PBIFiDV). The IDIF and PBIFs were compared with the gold standard AIF, using AUC values and Patlak Ki values.

RESULTS

The IDIF underestimated the AIF at early times and overestimated it at later times. Thus, based on the AUC and Ki comparison, 30-60 min was the most accurate time window for PBIFAUC; later time windows for scaling underestimated Ki (- 6 ± 8 to - 13 ± 9%). Correlations of AUC between AIF and IDIF, PBIFAUC(30-60), and PBIFiDV were 0.91, 0.94, and 0.90, respectively. The bias of Ki was - 9 ± 10%, - 1 ± 8%, and 3 ± 9%, respectively.

CONCLUSIONS

Both PBIF scaling methods provided good mean performance with moderate variation. Improved performance can be obtained by refining IDIF methods and by evaluating PBIFs with test-retest data.