Marc A Kropholler

Development of a Tracer Kinetic Plasma Input Model for (R)-[11C]PK11195 Brain Studies

(R)-[11C]PK11195 ([1-(2-chlorophenyl)-N-methyl-N-(1-methylpropyl]-3-isoquinoline carboxamide) is a ligand for the peripheral benzodiazepine receptor, which, in the brain, is mainly expressed on activated microglia. Using both clinical studies and Monte Carlo simulations, the aim of this study was to determine which tracer kinetic plasma input model best describes (R)-[11C]PK11195 kinetics. Dynamic positron emission tomography (PET) scans were performed on 13 subjects while radioactivity in arterial blood was monitored online. Discrete blood samples were taken to generate a metabolite corrected plasma input function. One-tissue, two-tissue irreversible, and two-tissue reversible compartment models, with and without fixing K1/k2 ratio, k4 or blood volume to whole cortex values, were fitted to the data. The effects of fixing parameters to incorrect values were investigated by varying them over a physiologic range and determining accuracy and reproducibility of binding potential and volume of distribution using Monte Carlo simulations. Clinical data showed that a two-tissue reversible compartment model was optimal for analyzing (R)-[11C]PK11195 PET brain studies. Simulations showed that fixing the K1/k2 ratio of this model provided the optimal trade-off between accuracy and reproducibility. It was concluded that a two-tissue reversible compartment model with K1/k2 fixed to whole cortex value is optimal for analyzing (R)-[11C]PK11195 PET brain studies.

Optimization algorithms and weighting factors for analysis of dynamic PET studies

Positron emission tomography (PET) pharmacokinetic analysis involves fitting of measured PET data to a PET pharmacokinetic model. The fitted parameters may, however, suffer from bias or be unrealistic, especially in the case of noisy data. There are many optimization algorithms, each having different characteristics. The purpose of the present study was to evaluate (1) the performance of different optimization algorithms and (2) the effects of using incorrect weighting factors during optimization in terms of both accuracy and reproducibility of fitted PET pharmacokinetic parameters. In this study, the performance of commonly used optimization algorithms (i.e. interior-reflective Newton methods) and a simulated annealing (SA) method was evaluated. This SA algorithm, known as basin hopping, was modified for the present application. In addition, optimization was performed using various weighting factors. Algorithms and effects of using incorrect weighting factors were studied using both simulated and clinical time-activity curves (TACs). Input data, taken from [(15)O]H(2)O, [(11)C]flumazenil and [(11)C](R)-PK11195 studies, were used to simulate time-activity curves at various variance levels (0-15% COV). Clinical evaluation was based on studies with the same three tracers. SA was able to produce accurate results without the need for selecting appropriate starting values for (kinetic) parameters, in contrast to the interior-reflective Newton method. The latter gave biased results unless it was modified to allow for a range of starting values for the different parameters. For patient studies, where large variability is expected, both SA and the extended Newton method provided accurate results. Simulations and clinical assessment showed similar results for the evaluation of different weighting models in that small to intermediate mismatches between data variance and weighting factors did not significantly affect the outcome of the fits. Large errors were observed only when the mismatch between weighting model and data variance was large. It is concluded that selection of specific optimization algorithms and weighting factors can have a large effect on the accuracy and precision of PET pharmacokinetic analysis. Apart from carefully selecting appropriate algorithms and variance models, further improvement in accuracy might be obtained by using noise reducing strategies, such as wavelet filtering, provided that these methods do not introduce significant bias.

Evaluation of reference tissue models for the analysis of [11C](R)-PK11195 studies.

[11C](R)-PK11195 is a marker of activated microglia, which can be used to measure inflammation in neurologic disorders. The purpose of the present study was to define the optimal reference tissue model based on a comparison with a validated plasma input model and using clinical studies and Monte Carlo simulations. Accuracy and reproducibility of reference tissue models were evaluated using Monte Carlo simulations. The effects of noise and variation in specific binding, nonspecific binding and blood volume were evaluated. Dynamic positron emission tomography scans were performed on 13 subjects, and radioactivity in arterial blood was monitored online. In addition, blood samples were taken to generate a metabolite corrected plasma input function. Both a (validated) two-tissue reversible compartment model with K1/k2 fixed to whole cortex and various reference tissue models were fitted to the data. Finally, a simplified reference tissue model (SRTM) corrected for nonspecific binding using plasma input data (SRTMpl_corr) was investigated. Correlations between reference tissue models (including SRTMpl_corr) and the plasma input model were calculated. Monte Carlo simulations indicated that low-specific binding results in decreased accuracy and reproducibility. In this respect, the SRTM and SRTMpl_corr performed relatively well. Varying blood volume had no effect on performance. In the clinical evaluation, SRTMpl_corr and SRTM had the highest correlations with the plasma input model (R2 = 0.82 and 0.78, respectively). SRTMpl_corr is optimal when an arterial plasma input curve is available. Simplified reference tissue model is the best alternative when no plasma input is available.

Evaluation of reference regions for (R)-[11C]PK11195 studies in Alzheimer's disease and Mild Cognitive impairment

Inflammation in Alzheimers disease (AD) may be assessed using (R)-[11C]PK11195 and positron emission tomography. Data can be analyzed using the simplified reference tissue model, provided a suitable reference region is available. This study evaluates various reference regions for analyzing (R)-[11C]PK11195 scans in patients with mild cognitive impairment (MCI) and probable AD. Healthy subjects (n = 10, 30 ± 10 years and n = 10, 70 ± 6 years) and patients with MCI (n = 10, 74 ± 6 years) and probable AD (n = 9, 71 ± 6 years) were included. Subjects underwent a dynamic three-dimensional (R)-[11C]PK11195 scan including arterial sampling. Gray matter, white matter, total cerebellum and cerebrum, and cluster analysis were evaluated as reference regions. Both plasma input binding potentials of these reference regions (BPPLASMA) and corresponding reference region input binding potentials of a target region (BPSRTM) were evaluated. Simulations were performed to assess cluster analysis performance at 5% to 15% coefficient of variation noise levels. Reasonable correlations for BPPLASMA (R2 = 0.52 to 0.94) and BPSRTM (R2 = 0.59 to 0.76) were observed between results using anatomic regions and cluster analysis. For cerebellum white matter, cerebrum white matter, and total cerebrum a considerable number of unrealistic BPSRTM values were observed. Cluster analysis did not extract a valid reference region in 10% of the scans. Simulations showed that potentially cluster analysis suffers from negative bias in BPPLASMA. Most anatomic regions outperformed cluster analysis in terms of absence of both scan rejection and bias. Total cerebellum is the optimal reference region in this patient category.

Evaluation of Methods for Generating Parametric (R)-[11C]PK11195 Binding Images

Activated microglia can be visualised using (R)-[11C]PK11195 (1-[2-chlorophenyl]-N-methyl-N-[1-methyl-propyl]-3-isoquinoline carboxamide) and positron emission tomography (PET). In previous studies, various methods have been used to quantify (R)-[11C]PK11195 binding. The purpose of this study was to determine which parametric method would be best suited for quantifying (R)-[11C]PK11195 binding at the voxel level. Dynamic (R)-[11C]PK11195 scans with arterial blood sampling were performed in 20 healthy and 9 Alzheimers disease subjects. Parametric images of both volume of distribution (Vd) and binding potential (BP) were obtained using Logan graphical analysis with plasma input. In addition, BP images were generated using two versions of the basis function implementation of the simplified reference tissue model, two versions of Ichise linearisations, and Logan graphical analysis with reference tissue input. Results of the parametric methods were compared with results of full compartmental analysis using nonlinear regression. Simulations were performed to assess accuracy and precision of each method. It was concluded that Logan graphical analysis with arterial input function is an accurate method for generating parametric images of Vd. Basis function methods, one of the Ichise linearisations and Logan graphical analysis with reference tissue input provided reasonably accurate and precise estimates of BP. In pathological conditions with reduced flow rates or large variations in blood volume, the basis function method is preferred because it produces less bias and is more precise.

Quantification of (R)-[11C]PK11195 binding in rheumatoid arthritis

PurposeRheumatoid arthritis (RA) involves migration of macrophages into inflamed areas. (R)-[11C]PK11195 binds to peripheral benzodiazepine receptors, expressed on macrophages, and may be used to quantify inflammation using positron emission tomography (PET). This study evaluated methods for the quantification of (R)-[11C]PK11195 binding in the knee joints of RA patients.MethodsData from six patients with RA were analysed. Dynamic PET scans were acquired in 3-D mode following (R)-[11C]PK11195 injection. During scanning arterial radioactivity concentrations were measured to determine the plasma (R)-[11C]PK11195 concentrations. Data were analysed using irreversible and reversible one-tissue and two-tissue compartment models and input functions with various types of metabolite correction. Model preferences according to the Akaike information criterion (AIC) and correlations between measures were evaluated. Correlations between distribution volume (Vd) and standardized uptake values (SUV) were evaluated.ResultsAIC indicated optimal performance for a one-tissue reversible compartment model including blood volume. High correlations were observed between Vd obtained using different input functions (R2=0.80–1.00) and between Vd obtained with one- and two-tissue reversible compartment models (R2=0.75–0.94). A high correlation was observed between optimal Vd and SUV after injection (R2=0.73).Conclusion(R)-[11C]PK11195 kinetics in the knee were best described by a reversible single-tissue compartment model including blood volume. Applying metabolite corrections did not increase sensitivity. Due to the high correlation with Vd, SUV is a practical alternative for clinical use.

SPM analysis of parametric (R)-[11C]PK11195 binding images: Plasma input versus reference tissue parametric methods

(R)-[11C]PK11195 has been used for quantifying cerebral microglial activation in vivo. In previous studies, both plasma input and reference tissue methods have been used, usually in combination with a region of interest (ROI) approach. Definition of ROIs, however, can be labourious and prone to interobserver variation. In addition, results are only obtained for predefined areas and (unexpected) signals in undefined areas may be missed. On the other hand, standard pharmacokinetic models are too sensitive to noise to calculate (R)-[11C]PK11195 binding on a voxel-by-voxel basis. Linearised versions of both plasma input and reference tissue models have been described, and these are more suitable for parametric imaging. The purpose of this study was to compare the performance of these plasma input and reference tissue parametric methods on the outcome of statistical parametric mapping (SPM) analysis of (R)-[11C]PK11195 binding. Dynamic (R)-[11C]PK11195 PET scans with arterial blood sampling were performed in 7 younger and 11 elderly healthy subjects. Parametric images of volume of distribution (Vd) and binding potential (BP) were generated using linearised versions of plasma input (Logan) and reference tissue (Reference Parametric Mapping) models. Images were compared at the group level using SPM with a two-sample t-test per voxel, both with and without proportional scaling. Parametric BP images without scaling provided the most sensitive framework for determining differences in (R)-[11C]PK11195 binding between younger and elderly subjects. Vd images could only demonstrate differences in (R)-[11C]PK11195 binding when analysed with proportional scaling due to intersubject variation in K1/k2 (blood-brain barrier transport and non-specific binding).

Simulated annealing in pharmacokinetic modeling of PET neuroreceptor studies: accuracy and precision compared with other optimization algorithms

Fitting of PET pharmacokinetic parameters may suffer from bias or unrealistic outcomes, especially for noisy data. There are many readily available optimization algorithms, but each has different characteristics when fitting PET pharmacokinetic models. The purpose of this study is to evaluate the performance of four different types of optimization algorithms, including a modified simulated annealing method, in terms of precision and accuracy of PET pharmacokinetic parameters. The simulated annealing algorithm (SA), called Basin-hoping, was modified for present application. Input data, taken from [/sup 11/C]-PK11195 neuroreceptor ligand studies, was used to simulate time activity curves at various noise levels. Also the influence of incorrect weighting factors on algorithm performance was studied. Surprisingly, effects of using incorrect but reasonable weighting factors on bias and precision were negligible. Except when extreme and unrealistic weighting factors were used, an increase in bias and decrease in precision was observed. In general the modified SA provided smallest weighted squared residual error and was able to find the global minimum without the need for a proper start parameter selection. However, occasionally better fits were found with the interior-reflective Newton method, but only when implemented using a range of start parameters, centered on the expected value.

Increased microglial activation in MCI and AD: An (R)-111CIPK11195 PET study

Purpose: Recent evidence suggests that inflammatory mechanisms, like microglial activation, are involved in the pathogenesis of Alzheimers disease (AD). Microglial activation can be quantified in vivo using (R)-[C]PK11195 and PET. The purpose of the present study was to investigate extent and distribution of (R)-[C]PK11195 binding in AD and mild cognitive impairment (MCI) patients. A second aimwas to establish whether the presence of activated microglia in MCI is associated with progression to AD.