Search

Background and purposeThe purpose of this work is to present the clinical experience from the first-in-human trial of real-time tumor targeting via MLC tracking for stereotactic ablative body radiotherapy (SABR) of lung lesions.Methods and materialsSeventeen patients with stage 1 non-small cell lung cancer (NSCLC) or lung metastases were included in a study of electromagnetic transponder-guided MLC tracking for SABR (NCT02514512). Patients had electromagnetic transponders inserted near the tumor. An MLC tracking SABR plan was generated with planning target volume (PTV) expanded 5 mm from the end-exhale gross tumor volume (GTV). A clinically approved comparator plan was generated with PTV expanded 5 mm from a 4DCT-derived internal target volume (ITV). Treatment was delivered using a standard linear accelerator to continuously adapt the MLC based on transponder motion. Treated volumes and reconstructed delivered dose were compared between MLC tracking and comparator ITV-based treatment.ResultsAll seventeen patients were successfully treated with MLC tracking (70 successful fractions). MLC tracking treatment delivery time averaged 8 minutes. The time from the start of CBCT to the end of treatment averaged 22 minutes. The MLC tracking PTV for 16/17 patients was smaller than the ITV-based PTV (range -1.6% to 44% reduction, or -0.6 to 18 cc). Reductions in mean lung dose (27 cGy) and V20Gy (50 cc) were statistically significant (p ConclusionThe first treatments with lung MLC tracking have been successfully performed in seventeen SABR patients. MLC tracking for lung SABR is feasible, efficient and delivers high-precision target dose and lower normal tissue dose.

Dynamic positron emission tomography (PET) imaging can be used to quantify changes in synaptic concentrations of endogenous neurotransmitters during cognitive tasks or pharmacological interventions. Existing pharmacokinetic models, such as the linear parametric neurotransmitter PET (lp-ntPET) method, can be used to model the measured dynamic data and characterize small transient changes in neurotransmitter levels. Application of these models to the voxel level is challenging, however, due to the high levels of noise in dynamic data, leading to a high number of false-positive responses (i.e., low specificity). In this article, we investigated the suitability of machine learning algorithms (MLAs), including support vector machine (SVM) classifiers, shallow feedforward neural networks, convolutional neural networks (CNNs) and long short-term memory (LSTM) networks, to detect and classify transient changes in voxel-wise time-activity curves. We also investigated whether the reconstruction framework (post versus direct reconstruction) had any impact on the performance of the MLAs. We used computer simulations to generate dynamic PET data, representing a [11C]raclopride study, with known activation responses, across a wide range of noise levels. Different simulated data sets were used to train and test the MLAs across a range of noise levels and activation response magnitudes. Results showed the MLAs offered a large improvement in specificity without a corresponding decrease in sensitivity across all noise levels tested compared to direct application of the lp-ntPET model. They also offered a modest benefit over the currently accepted method (statistical $F$ -test combined with cluster size analysis), for both 2-D+time data when incorporated within direct or post-reconstruction frameworks, and 4-D GATE data.

Rigid motion correction remains an important challenge in PET. Convolutional Neural Networks (CNNs) have recently matched or outperformed traditional solutions for some medical imaging problems but are largely unexplored for PET motion correction. A limitation of applying supervised neural networks is the requirement for large amounts of labelled data-and this is either unavailable or impractical for the large rigid-body motion parameter space. Therefore, the development of novel and efficient techniques to robustly train CNN models for this application is crucial. Here we present a potential PET “data engine” that efficiently generates realistic motion-corrupted exemplars. We investigated the feasibility of the engine using simulated data of an [18F]FDG brain PET exam and examined factors that could potentially affect the accuracy of the engine's exemplars, such as the reconstruction algorithms and attenuation correction

Objective: Dynamic PET imaging is extensively used in brain imaging to estimate parametric maps. Inter-frame motion can substantially disrupt the voxel-wise time-activity curves (TACs), leading to erroneous maps during kinetic modelling. Therefore, it is important to characterize the robustness of kinetic parameters under various motion and kinetic model related factors. Methods: Fully 4D brain simulations ([15O]H2O and [18F]FDG dynamic datasets) were performed using a variety of clinically observed motion patterns. Increasing levels of head motion were investigated as well as varying temporal frames of motion initiation. Kinetic parameter estimation was performed using both post-reconstruction kinetic analysis and direct 4D image reconstruction to assess bias from inter-frame emission blurring and emission/attenuation mismatch. Results: Kinetic parameter bias heavily depends on the time point of motion initiation. Motion initiated towards the end of the scan results in the most biased parameters. For the [18F]FDG data, k4 is the more sensitive parameter to positional changes, while K1 and blood volume were proven to be relatively robust to motion. Direct 4D image reconstruction appeared more sensitive to changes in TACs due to motion, with parameter bias spatially propagating and depending on the level of motion. Conclusion: Kinetic parameter bias highly depends upon the time frame at which motion occurred, with late frame motion-induced TAC discontinuities resulting in the least accurate parameters. This is of importance during prolonged data acquisition as is often the case in neuro-receptor imaging studies. In the absence of a motion correction, use of TOF information within 4D image reconstruction could limit the error propagation.

Current pharmacokinetic models, such as the linear parametric neurotransmitter PET (lp-ntPET) model have been developed to detect and quantify transient changes in receptor occupancy caused by variations in the concentration of endogenous neurotransmitters. However, it often performs poorly when applied at the voxel level due to high statistical noise. In this paper, we propose a new method to detect transient changes in neurotransmitter concentration in dynamic PET data using deep learning. Activation onset time and response magnitude of neurotransmitter were directly estimated using a convolution neural network (CNN) and compared to the lpntPET model. Computer simulations, as well as realistic GATE simulations were used to generate dynamic PET data, representing a [11C]raclopride study, with a known range of activation onset times and response magnitudes, across a wide range of noise levels. Results showed that the proposed neural network had better quantitative performance in estimating activation onset time and response magnitude than the conventional lp-ntPET method, especially where noise is high.

A comprehensive understanding of how the brain responds to a changing environment requires techniques capable of recording functional outputs at the whole-brain level in response to external stimuli. Positron emission tomography (PET) is an exquisitely sensitive technique for imaging brain function but the need for anaesthesia to avoid motion artefacts precludes concurrent behavioural response studies. Here, we report a technique that combines motion-compensated PET with a robotically-controlled animal enclosure to enable simultaneous brain imaging and behavioural recordings in unrestrained small animals. The technique was used to measure in vivo displacement of [11C]raclopride from dopamine D2 receptors (D2R) concurrently with changes in the behaviour of awake, freely moving rats following administration of unlabelled raclopride or amphetamine. The timing and magnitude of [11C]raclopride displacement from D2R were reliably estimated and, in the case of amphetamine, these changes coincided with a marked increase in stereotyped behaviours and hyper-locomotion. The technique, therefore, allows simultaneous measurement of changes in brain function and behavioural responses to external stimuli in conscious unrestrained animals, giving rise to important applications in behavioural neuroscience.

Computational methods, such as the linear parametric neurotransmitter PET (lp-ntPET) method, have been developed to characterize the transient changes in radiotracer kinetics in the target tissue during endogenous neurotransmitter release. In this paper, we describe and evaluate a parametric reconstruction algorithm that uses an expectation maximization framework, along with the lp-ntPET model, to estimate the endogenous neurotransmitter response to stimuli directly from the measured PET data. Computer simulations showed that the proposed direct reconstruction method offers improved accuracy and precision for the estimated timing parameters of the neurotransmitter response at the voxel level ( ${t}_{d}=1\pm 2 $ min, for activation onset bias and standard deviation) compared with conventional post reconstruction modeling ( ${t}_{d}=4\pm 7 $ min). In addition, we applied the proposed direct parameter estimation methodology to a [11C]raclopride displacement study of an awake rat and generated parametric maps illustrating the magnitude of ligand displacement from striatum. Although the estimated parametric maps of activation magnitude obtained from both direct and post reconstruction methodologies suffered from false positive activations, the proposed direct reconstruction framework offered more reliable parametric maps when the activation onset parameter was constrained.

The quality of reconstructed dynamic PET images, as well as the statistical reliability of the estimated pharmacokinetic parameters is often compromised by high levels of statistical noise, particularly at the voxel level. Many denoising strategies have been proposed, both in the temporal and spatial domain, which substantially improve the signal to noise ratio of the reconstructed dynamic images. However, although most filtering approaches are fairly successful in reducing the spatio-temporal inter-voxel variability, they may also average out or completely eradicate the critically important temporal signature of a transient neurotransmitter activation response that may be present in a non-steady state dynamic PET study. In this work, we explore an approach towards temporal denoising of non-steady state dynamic PET images using an artificial neural network, which was trained to identify the temporal profile of a time-activity curve, while preserving any potential activation response. We evaluated the performance of a feed-forward perceptron neural network to improve the signal to noise ratio of dynamic [11C]raclopride activation studies and compared it with the widely used highly constrained back projection (HYPR) filter. Results on both simulated Geant4 Application for Tomographic Emission data of a realistic rat brain phantom and experimental animal data of a freely moving animal study showed that the proposed neural network can efficiently improve the noise characteristics of dynamic data in the temporal domain, while it can lead to a more reliable estimation of voxel-wise activation response in target region. In addition, improvements in signal-to-noise ratio achieved by denoising the dynamic data using the proposed neural network led to improved accuracy and precision of the estimated model parameters of the lp-ntPET model, compared to the HYPR filter. The performance of the proposed denoising approach strongly depends on the amount of noise in the dynamic PET data, with higher noise leading to substantially higher variability in the estimated parameters of the activation response. Overall, the feed-forward network led to a similar performance as the HYPR filter in terms of spatial denoising, but led to notable improvements in terms of temporal denoising, which in turn improved the estimation activation parameters.

Image space decomposition based on tetrahedral voxels are interesting candidates for use in emission tomography. Tetrahedral voxels provide many of the advantages of point clouds with irregular spacing, such as being intrinsically multi-resolution, yet they also serve as a volumetric partition of the image space and so are comparable to more standard cubic voxels. Additionally, non-rigid displacement fields can be applied to the tetrahedral mesh in a straight-forward manner. So far studies incorporating tetrahedral decomposition of the image space have concentrated on pre-calculated, node-based, system matrix elements which reduces the flexibility of the tetrahedral approach and the capacity to accurately define regions of interest. Here, a list-mode on-the-fly calculation of the system matrix elements is described using a tetrahedral decomposition of the image space and volumetric elements-voxels. The algorithm is demonstrated in the context of awake animal PET which may require both rigid and non-rigid motion compensation, as well as quantification within small regions of the brain. This approach allows accurate, event based, motion compensation including non-rigid deformations.

We consider the application of approximate Bayesian Computation (ABC) in the context of medical imaging data. We consider the parameter estimation of compartmental models in PET imaging analysis, and provide a simple ABC algorithm for its estimation. We demonstrate the utility of the proposed estimation methods on a neurotransmitter response model, and compare our approach to existing methods.

Accurate characterisation of the scanner's point spread function across the entire field of view (FOV) is crucial in order to account for spatially dependent factors that degrade the resolution of the reconstructed images. The HRRT users' community resolution modelling reconstruction software includes a shift-invariant resolution kernel, which leads to transaxially non-uniform resolution in the reconstructed images. Unlike previous work to date in this field, this work is the first to model the spatially variant resolution across the entire FOV of the HRRT, which is the highest resolution human brain PET scanner in the world. In this paper we developed a spatially variant image-based resolution modelling reconstruction dedicated to the HRRT, using an experimentally measured shift-variant resolution kernel. Previously, the system response was measured and characterised in detail across the entire FOV of the HRRT, using a printed point source array. The newly developed resolution modelling reconstruction was applied on measured phantom, as well as clinical data and was compared against the HRRT users' community resolution modelling reconstruction, which is currently in use. Results demonstrated improvements both in contrast and resolution recovery, particularly for regions close to the edges of the FOV, with almost uniform resolution recovery across the entire transverse FOV. In addition, because the newly measured resolution kernel is slightly broader with wider tails, compared to the deliberately conservative kernel employed in the HRRT users' community software, the reconstructed images appear to have not only improved contrast recovery (up to 20% for small regions), but also better noise characteristics.

In this paper we investigated clustering as a potential aid in reducing false positive rates when modelling neurotransmitter activation using the lp-ntPET model. The aim was to investigate whether clustering time-activity curves (TACs) before kinetic modelling improves specificity for detecting activation states. We used two popular unsupervised clustering algorithms, k-means and Gaussian Mixture Models (GMM), which were applied prior to voxel-wise kinetic modelling. We generated statistically independent sets of TACs corresponding to [$^{11}$C]raclopride kinetics and we investigated the impact of: (a) the level of noise in the TACs, (b) the activation magnitude and (c) the ratio between the number of TACs with and without activation on the classification accuracy of each clustering approach. The classification performance was assessed by calculating the sensitivity and specificity for each clustering approach and was compared against conventional single-voxel modeling. Results showed that applying clustering before voxelwise kinetic modeling improves classification of TACs between active and non active states, at least for moderate to less noisy TACs. The single voxel modelling results in better specificity at higher noise levels and clustering assures better results at noise levels from 1k counts to 1M counts.

This paper describes the use of (18)F-FDG and micro-PET/CT imaging to determine in vivo glucose metabolism kinetics in mice (and is transferable to rats). Impaired uptake and metabolism of glucose in multiple organ systems due to insulin resistance is a hallmark of type 2 diabetes. The ability of this technique to extract an image-derived input function from the vena cava using an iterative deconvolution method eliminates the requirement of the collection of arterial blood samples. Fitting of tissue and vena cava time activity curves to a two-tissue, three compartment model permits the estimation of kinetic micro-parameters related to the (18)F-FDG uptake from the plasma to the intracellular space, the rate of transport from intracellular space to plasma and the rate of (18)F-FDG phosphorylation. This methodology allows for multiple measures of glucose uptake and metabolism kinetics in the context of longitudinal studies and also provides insights into the efficacy of therapeutic interventions.

This paper describes the use of 18F-FDG and micro-PET/CT imaging to determine in vivo glucose metabolism kinetics in mice (and is transferable to rats). Impaired uptake and metabolism of glucose in multiple organ systems due to insulin resistance is a hallmark of type 2 diabetes. The ability of this technique to extract an image-derived input function from the vena cava using an iterative deconvolution method eliminates the requirement of the collection of arterial blood samples. Fitting of tissue and vena cava time activity curves to a two-tissue, three compartment model permits the estimation of kinetic micro-parameters related to the 18F-FDG uptake from the plasma to the intracellular space, the rate of transport from intracellular space to plasma and the rate of 18F-FDG phosphorylation. This methodology allows for multiple measures of glucose uptake and metabolism kinetics in the context of longitudinal studies and also provides insights into the efficacy of therapeutic interventions.

International audience; Awake and/or freely moving small animal single photon emission imaging allows the continuous study of molecules exhibiting slow kinetics without the need to restrain or anaesthetise the animals. Estimating motion free projections in freely moving small animal planar imaging can be considered as a limited angle tomography problem, except that we wish to estimate the 2D planar projections rather than the 3D volume, where the angular sampling in all three axes depends on the rotational motion of the animal. In this study, we hypothesise that the motion corrected planar projections estimated by reconstructing an estimate of the 3D volume using an iterative motion compensating reconstruction algorithm and integrating it along the projection path, will closely match the true, motion-less, planar distribution regardless of the object motion. We tested this hypothesis for the case of rigid motion using Monte-Carlo simulations and experimental phantom data based on a dual opposed detector system, where object motion was modelled with 6 degrees of freedom. In addition, we investigated the quantitative accuracy of the regional activity extracted from the geometric mean of opposing motion corrected planar projections. Results showed that it is feasible to estimate qualitatively accurate motion-corrected projections for a wide range of motions around all 3 axes. Errors in the geometric mean estimates of regional activity were relatively small and within 10% of expected true values. In addition, quantitative regional errors were dependent on the observed motion, as well as on the surrounding activity of overlapping organs. We conclude that both qualitatively and quantitatively accurate motion-free projections of the tracer distribution in a rigidly moving object can be estimated from dual opposed detectors using a correction approach within an iterative reconstruction framework and we expect this approach can be extended to the case of non-rigid motion.

In emission tomographic imaging, the stochastic origin ensembles algorithm provides unique information regarding the detected counts given the measured data. Precision in both voxel and region-wise parameters may be determined for a single data set based on the posterior distribution of the count density allowing uncertainty estimates to be allocated to quantitative measures. Uncertainty estimates are of particular importance in awake animal neurological and behavioral studies for which head motion, unique for each acquired data set, perturbs the measured data. Motion compensation can be conducted when rigid head pose is measured during the scan. However, errors in pose measurements used for compensation can degrade the data and hence quantitative outcomes. In this investigation motion compensation and detector resolution models were incorporated into the basic origin ensembles algorithm and an efficient approach to computation was developed. The approach was validated against maximum liklihood—expectation maximisation and tested using simulated data. The resultant algorithm was then used to analyse quantitative uncertainty in regional activity estimates arising from changes in pose measurement precision. Finally, the posterior covariance acquired from a single data set was used to describe correlations between regions of interest providing information about pose measurement precision that may be useful in system analysis and design. The investigation demonstrates the use of origin ensembles as a powerful framework for evaluating statistical uncertainty of voxel and regional estimates. While in this investigation rigid motion was considered in the context of awake animal PET, the extension to arbitrary motion may provide clinical utility where respiratory or cardiac motion perturb the measured data.

Parametric imaging in thoracic and abdominal PET can provide additional parameters more relevant to the pathophysiology of the system under study. However, dynamic data in the body are noisy due to the limiting counting statistics leading to suboptimal kinetic parameter estimates. Direct 4D image reconstruction algorithms can potentially improve kinetic parameter precision and accuracy in dynamic PET body imaging. However, construction of a common kinetic model is not always feasible and in contrast to post-reconstruction kinetic analysis, errors in poorly modelled regions may spatially propagate to regions which are well modelled. To reduce error propagation from erroneous model fits, we implement and evaluate a new approach to direct parameter estimation by incorporating a recently proposed kinetic modelling strategy within a direct 4D image reconstruction framework. The algorithm uses a secondary more general model to allow a less constrained model fit in regions where the kinetic model does not accurately describe the underlying kinetics. A portion of the residuals then is adaptively included back into the image whilst preserving the primary model characteristics in other well modelled regions using a penalty term that trades off the models. Using fully 4D simulations based on dynamic [O-15]H2O datasets, we demonstrate reduction in propagation-related bias for all kinetic parameters. Under noisy conditions, reductions in bias due to propagation are obtained at the cost of increased noise, which in turn results in increased bias and variance of the kinetic parameters. This trade-off reflects the challenge of separating the residuals arising from poor kinetic modelling fits from the residuals arising purely from noise. Nonetheless, the overall root mean square error is reduced in most regions and parameters. Using the adaptive 4D image reconstruction improved model fits can be obtained in poorly modelled regions, leading to reduced errors potentially propagating to regions of interest which the primary biologic model accurately describes. The proposed methodology, however, depends on the secondary model and choosing an optimal model on the residual space is critical in improving model fits.

Attenuation correction in positron emission tomography brain imaging of freely moving animals is a very challenging problem since the torso of the animal is often within the field of view and introduces a non negligible attenuating factor that can degrade the quantitative accuracy of the reconstructed images. In the context of unrestrained small animal imaging, estimation of the attenuation correction factors without the need for a transmission scan is highly desirable. An attractive approach that avoids the need for a transmission scan involves the generation of the hull of the animal?s head based on the reconstructed motion corrected emission images. However, this approach ignores the attenuation introduced by the animal?s torso. In this work, we propose a virtual scanner geometry which moves in synchrony with the animal?s head and discriminates between those events that traversed only the animal?s head (and therefore can be accurately compensated for attenuation) and those that might have also traversed the animal?s torso. For each recorded pose of the animal?s head a new virtual scanner geometry is defined and therefore a new system matrix must be calculated leading to a time-varying system matrix. This new approach was evaluated on phantom data acquired on the microPET Focus 220 scanner using a custom-made phantom and step-wise motion. Results showed that when the animal?s torso is within the FOV and not appropriately accounted for during attenuation correction it can lead to bias of up to 10% . Attenuation correction was more accurate when the virtual scanner was employed leading to improved quantitative estimates (bias?2%), without the need to account for the attenuation introduced by the extraneous compartment. Although the proposed method requires increased computational resources, it can provide a reliable approach towards quantitatively accurate attenuation correction for freely moving animal studies.

Spatial resolution on the High Resolution Research Tomograph (HRRT) is of high importance, due to the need for accurate quantification of small brain structures. Thus accurate characterization of the scanner’s resolution properties and subsequent inclusion of such information within image reconstruction, requires measuring its point spread function (PSF). In this study we measured in detail the spatial variation of the image space PSF and assessed the impact of the scanner’s depth of interaction (DOI) capability on the spatial resolution. Furthermore, we characterized the dependency of the PSF on progressively increasing count rate statistics. An array of 15 × 11 printed point sources was scanned twice, initially on its own to measure the PSF spatial dependency, and subsequently with an extension line containing $\sim $ 300 MBq of carbon-11, to assess the count rate dependency. PSF data were reconstructed with the scanner’s default OP-OSEM and invariant PSF OP-OSEM algorithms, followed by image space model fitting. The axial, radial and tangential components of the PSF were found to vary under radial and angular transformations, but being radially symmetric and almost axially independent. The FWHM improves by $\sim $ 1.3 mm using the PSF-OP-OSEM but is still radially variable with $\sim $ 1 mm degradation within the FOV boundaries. When DOI was not taken into account, an additional degradation up to 0.7 mm was seen in the radial FWHM using OP-OSEM. Using the invariant PSF OP-OSEM, this degradation was less notable ( $\sim $ 0.5 mm), despite the fact an invariant kernel is used. In terms of count rate dependency, a clear resolution degradation was seen for mean count rates above 5–10 kcps. However, this degradation was found to vary within the FOV. The spatially variant PSF at low count rates could be incorporated within a resolution modelling image reconstruction. However, including a count rate dependent PSF model is less straightforward.

Recent studies have demonstrated the benefits of a resolution model within iterative reconstruction algorithms in an attempt to account for effects that degrade the spatial resolution of the reconstructed images. However, these algorithms suffer from slower convergence rates, compared to algorithms where no resolution model is used, due to the additional need to solve an image deconvolution problem. In this paper, a recently proposed algorithm, which decouples the tomographic and image deconvolution problems within an image-based expectation maximization (EM) framework, was evaluated. This separation is convenient, because more computational effort can be placed on the image deconvolution problem and therefore accelerate convergence. Since the computational cost of solving the image deconvolution problem is relatively small, multiple image-based EM iterations do not significantly increase the overall reconstruction time. The proposed algorithm was evaluated using 2D simulations, as well as measured 3D data acquired on the high-resolution research tomograph. Results showed that bias reduction can be accelerated by interleaving multiple iterations of the image-based EM algorithm solving the resolution model problem, with a single EM iteration solving the tomographic problem. Significant improvements were observed particularly for voxels that were located on the boundaries between regions of high contrast within the object being imaged and for small regions of interest, where resolution recovery is usually more challenging. Minor differences were observed using the proposed nested algorithm, compared to the single iteration normally performed, when an optimal number of iterations are performed for each algorithm. However, using the proposed nested approach convergence is significantly accelerated enabling reconstruction using far fewer tomographic iterations (up to 70% fewer iterations for small regions). Nevertheless, the optimal number of nested image-based EM iterations is hard to be defined and it should be selected according to the given application.

Regional quantification is often considered the end-point of many studies in Positron Emission Tomography (PET). However, in neurological small animal imaging, features can be sufficiently small that device resolution and voxel granularity inhibit the definition of segmentation boundaries that closely conform to organs or regions of interest. Even if well defined, estimation of uncertainty over small regions is problematic, particularly given that in some cases — such as in awake animal imaging — studies are unique and non-repeatable. The Origin Ensembles algorithm is an approach to image reconstruction based on Bayesian inference which utilises Monte-Carlo driven sampling to determine the posterior distribution of the emission counts given the measured data. The algorithm explores the full posterior and so can be used to estimate measures of uncertainty in quantitative parameters drawn from the data. However, being inference based, Origin Ensembles is sensitive to the dimension of the parameter space over which estimates are made. In this investigation the dimension of the image space is reduced using a regional segmentation based on an initial MLEM reconstruction of detected data. Origin Ensembles is used to reconstruct data directly to the regions of interest over which quantification is desired providing estimation of the full posterior distribution for each region. Origin Ensembles sampling allows the uncertainty in quantitative values to be determined both in terms of parameter variance and the covariance between recovered regional intensities. Direct regional reconstruction was employed in this study using simulated data and was shown to enhance the precision and accuracy of recovered values.

Motion corrected images from awake and freely moving animals often exhibit reduced resolution when compared to their stationary counterparts. This could be attributed to the combination of brief periods of fast animal motion and insufficient motion sampling speed. In this paper we hypothesise that we can measure the motion dependent point spread function of a given study and mitigate the motion blurring artifacts in the reconstructed images, in a similar way that a measured system response point spread function can improve resolution due to geometric effects (e.g. parallax errors). We investigated this hypothesis on a set of experimentally measured phantom data, which underwent a series of distinctively different motion patterns, ranging from slow to fast. Preliminary results showed that motion corrected images have reduced resolution compared to the stationary image and noticeable motion blurring artefacts, particularly for fast speed/acceleration settings. In addition, images deconvolved after reconstruction with the measured motion dependent PSF appear to be sharper compared to their unprocessed counterparts, yet without completely eliminating the motion blurring artefacts. Work is in progress to refine the methodology, by decomposing the geometric and motion components of the PSF, as well as including the deconvolution within the reconstruction algorithm.

Iterative expectation maximization (EM) techniques have been extensively used to solve maximum likelihood (ML) problems in positron emission tomography (PET) image reconstruction. Although EM methods offer a robust approach to solving ML problems, they usually suffer from slow convergence rates. The ordered subsets EM (OSEM) algorithm provides significant improvements in the convergence rate, but it can cycle between estimates converging towards the ML solution of each subset. In contrast, gradient-based methods, such as the recently proposed non-monotonic maximum likelihood (NMML) and the more established preconditioned conjugate gradient (PCG), offer a globally convergent, yet equally fast, alternative to OSEM. Reported results showed that NMML provides faster convergence compared to OSEM; however, it has never been compared to other fast gradient-based methods, like PCG. Therefore, in this work we evaluate the performance of two gradient-based methods (NMML and PCG) and investigate their potential as an alternative to the fast and widely used OSEM. All algorithms were evaluated using 2D simulations, as well as a single [(11)C]DASB clinical brain dataset. Results on simulated 2D data show that both PCG and NMML achieve orders of magnitude faster convergence to the ML solution compared to MLEM and exhibit comparable performance to OSEM. Equally fast performance is observed between OSEM and PCG for clinical 3D data, but NMML seems to perform poorly. However, with the addition of a preconditioner term to the gradient direction, the convergence behaviour of NMML can be substantially improved. Although PCG is a fast convergent algorithm, the use of a (bent) line search increases the complexity of the implementation, as well as the computational time involved per iteration. Contrary to previous reports, NMML offers no clear advantage over OSEM or PCG, for noisy PET data. Therefore, we conclude that there is little evidence to replace OSEM as the algorithm of choice for many applications, especially given that in practice convergence is often not desired for algorithms seeking ML estimates.

In order to remove the influence of anaesthetic agents on neurological function and to expand the range of imaging tasks available it is preferable to image small animals in an awake state. Accurate activity estimates in awake small animal Positron Emission Tomography rely on the measurement of animal head pose over the time-course of the scan. Pose measurements are then incorporated into the emission data during image reconstruction, compensating for animal motion. Uncertainty in pose measurement can impact reconstructed image quality by effectively degrading scanner resolution, and hence that of the reconstructed image. In small animal imaging, regions of interest can be small in comparison to image-space voxelisation so that the precision of estimates taken from a single reconstructed image can be difficult to gauge. Stochastic Origin Ensembles provides a means of estimating a more complete statistical description of the emission data than other methods of image reconstruction. In this investigation, rigid motion compensation is incorporated into the Stochastic Origin Ensembles algorithm and explored using simulated data. Realistic motion is modeled within a GATE simulation and measurement uncertainty is incorporated into the pose data using both simulated perturbations as well as experimental trials using a motion tracking system. Sampling from the posterior distribution is conducted using the Stochastic Origin Ensembles algorithm and compared to image reconstruction using the Maximum Likelihood-Expectation Maximisation algorithm. Using Stochastic Origin Ensembles both regional and single voxel parameters were investigated. The impact of varying levels of pose measurement uncertainty on image-space parameters was demonstrated and assessed.

The temporal characterisation of endogenous neurotransmitter release during a cognitive task or drug intervention is an important capability for studying the role of neuro-transmitters in normal and aberrant brain function, including disease. Advanced kinetic models, such as the linear parametric neurotransmitter PET (lp-ntPET) have been developed to appropriately model the transient changes in the model parameters, such as the radiotracer efflux from the target tissue, during endogenous neurotransmitter release. Incorporation of the kinetic model within the tomographic reconstruction algorithm may lead to improved parameter estimates, both in terms of precision and accuracy, compared to the conventional two-step post reconstruction approach. In this study, we evaluate a direct reconstruction approach that uses an expectation maximisation framework to transfer the 4D spatiotemporal maximum likelihood problem into an image-based weighted least squares problem. This framework allows the use of well established kinetic models, such as the lp-ntPET model, to estimate the endogenous neurotransmitter response directly from the dynamic PET data. Dynamic GATE simulations using a realistic digital rat brain phantom showed that the proposed direct reconstruction method can provide higher temporal accuracy and precision for the estimated neurotransmitter response at the voxel level, compared to the conventional post reconstruction modelling. In addition, we applied this methodology to a [11C]raclopride displacement study in an awake and freely moving rat and generated voxel-wise parametric maps illustrating ligand displacement from striatum.

Quantitative measurements in dynamic PET imaging are usually limited by the poor counting statistics particularly in short dynamic frames and by the low spatial resolution of the detection system, resulting in partial volume effects (PVEs). In this work, we present a fast and easy to implement method for the restoration of dynamic PET images that have suffered from both PVE and noise degradation. It is based on a weighted least squares iterative deconvolution approach of the dynamic PET image with spatial and temporal regularization. Using simulated dynamic [(11)C] Raclopride PET data with controlled biological variations in the striata between scans, we showed that the restoration method provides images which exhibit less noise and better contrast between emitting structures than the original images. In addition, the method is able to recover the true time activity curve in the striata region with an error below 3% while it was underestimated by more than 20% without correction. As a result, the method improves the accuracy and reduces the variability of the kinetic parameter estimates calculated from the corrected images. More importantly it increases the accuracy (from less than 66% to more than 95%) of measured biological variations as well as their statistical detectivity.

Reliable interpretation of results from pre-clinical Emission Tomography studies is often hampered by the requirement that the animal be anesthetised, affecting certain neuro-transmission systems and cerebral blood flow. Animal tracking and motion compensation techniques have been exploited to account for that rigid motion associated with head movement - allowing brain imaging in small animals. However, rigid motion cannot be assumed for extra cranial activity. Tetrahedral mesh approaches have been applied in cardiac reconstruction to account for non-rigid motion in clinical environments. In this investigation a list-mode approach to calculation of the elements of the system matrix using a tetrahedral image space is developed and evaluated by applying rigid motion compensation to both simulated and experimental data. Experimental data for which only rigid motion was applied were used to demonstrate the application of variable voxel-size over the image space. A simple small animal model comprising rigid (head) and non-rigid (body) motion was developed and the resulting voxelised sources were simulated using GATE. Motion compensation based on rigid transforms allowed targeted voxel sampling and demonstrates the impact of the non-rigid motion of activity estimates. The tetrahedral mesh should allow future investigations to extend correction to also include non-rigid estimates of small animal motion.

Accurate motion compensated image reconstruction of freely moving small animals requires the exact calculation of the time-weighted sensitivity correction factors. Back-projection of all possible lines of response for every recorded pose is a computationally intensive task, which requires impractically long reconstruction times. In this work we investigated an approach to accelerate this task, by randomly sampling the lines of response and the poses that are used to calculate the time-averaged sensitivity image. Two phantom datasets, acquired on the microPET Focus220 scanner, were used to quantify errors introduced in the randomly sampled sensitivity images and propagated to the final reconstructed images. In addition, the qualitative performance of the proposed methodology was assessed by reconstructing a freely moving rat acquisition. Results showed that randomisation can severely amplify the noise in the reconstructed images, especially when few LORs are sampled. However, such errors can be suppressed by post-filtering the randomised sensitivity images prior to reconstruction (e.g. 2 mm FHWM). Such an approach can substantially reduce the computational time involved during the estimation of the time-averaged sensitivity image for motion compensated image reconstruction.

The kinetics of single photon emitting macromolecules (e.g. antibodies) are typically slower than small molecules, necessitating long or repeat acquisitions. We previously proposed the use of motion tracking and limited angle tomographic reconstruction to characterise tracer kinetics over extended periods in awake rodents. In this study, we explored this approach by imaging a contrast phantom, moved with 6 degrees of freedom, using a prototype preclinical SPECT scanner with parallel-hole collimation and a fixed detector located at 0 and 90 degrees. The position of the phantom was tracked and data were acquired in list mode. Each event was motion-corrected and reconstructed using LM-MLEM. Planar projections were created by summing the reconstructed volume along the x-axis. Line profiles of the contrast phantom were compared for a planar reference projection and projections generated from the motion-free, motion-corrupted and motion-corrected reconstructions. Projections created from the motion-corrected agreed well with the planar reference projection of the stationary object and exhibited similar contrast. Whilst this initial study was limited to rigid motion, it demonstrates the feasibility of motion-corrected planar projections of a moving object.

Single photon emission computed tomography (SPECT) brain imaging of freely moving small animals would allow a wide range of important neurological processes and behaviors to be studied, which are normally inhibited by anesthetic drugs or precluded due to the animal being restrained. While rigid body motion of the head can be tracked and accounted for in the reconstruction, activity in the torso may confound brain measurements, especially since motion of the torso is more complex (i.e., nonrigid) and not well correlated with that of the head. The authors investigated the impact of mispositioned events and attenuation due to the torso on the accuracy of motion corrected brain images of freely moving mice.Monte Carlo simulations of a realistic voxelized mouse phantom and a dual compartment phantom were performed. Each phantom comprised a target and an extraneous compartment which were able to move independently of each other. Motion correction was performed based on the known motion of the target compartment only. Two SPECT camera geometries were investigated: a rotating single head detector and a stationary full ring detector. The effects of motion, detector geometry, and energy of the emitted photons (hence, attenuation) on bias and noise in reconstructed brain regions were evaluated.The authors observed two main sources of bias: (a) motion-related inconsistencies in the projection data and (b) the mismatch between attenuation and emission. Both effects are caused by the assumption that the orientation of the torso is difficult to track and model, and therefore cannot be conveniently corrected for. The motion induced bias in some regions was up to 12% when no attenuation effects were considered, while it reached 40% when also combined with attenuation related inconsistencies. The detector geometry (i.e., rotating vs full ring) has a big impact on the accuracy of the reconstructed images, with the full ring detector being more advantageous.Motion-induced inconsistencies in the projection data and attenuation/emission mismatch are the two main causes of bias in reconstructed brain images when there is complex motion. It appears that these two factors have a synergistic effect on the qualitative and quantitative accuracy of the reconstructed images.

Positron emission tomography (PET) with [(11)C]Raclopride is an important tool for studying dopamine D2 receptor expression in vivo. [(11)C]Raclopride PET binding experiments conducted using the Partial Saturation Approach (PSA) allow the estimation of receptor density (B(avail)) and the in vivo affinity appK(D). The PSA is a simple, single injection, single scan experimental protocol that does not require blood sampling, making it ideal for use in longitudinal studies. In this work, we generated a complete Monte Carlo simulated PET study involving two groups of scans, in between which a biological phenomenon was inferred (a 30% decrease of B(avail)), and used it in order to design an optimal data processing chain for the parameter estimation from PSA data. The impact of spatial smoothing, noise removal and image resolution recovery technique on the statistical detection was investigated in depth. We found that image resolution recovery using iterative deconvolution of the image with the system point spread function associated with temporal data denoising greatly improves the accuracy and the statistical reliability of detecting the imposed phenomenon. Before optimisation, the inferred B(avail) variation between the two groups was underestimated by 42% and detected in 66% of cases, while a false decrease of appK(D) by 13% was detected in more than 11% of cases. After optimisation, the calculated B(avail) variation was underestimated by only 3.7% and detected in 89% of cases, while a false slight increase of appK(D) by 3.7% was detected in only 2% of cases. We found during this investigation that it was essential to adjust a factor that accounts for difference in magnitude between the non-displaceable ligand concentrations measured in the target and in the reference regions, for different data processing pathways as this ratio was affected by different image resolutions.

Noninvasive functional imaging of awake, unrestrained small animals using motion-compensation removes the need for anesthetics and enables an animal's behavioral response to stimuli or administered drugs to be studied concurrently with imaging. While the feasibility of motion-compensated radiotracer imaging of awake rodents using marker-based optical motion tracking has been shown, markerless motion tracking would avoid the risk of marker detachment, streamline the experimental workflow, and potentially provide more accurate pose estimates over a greater range of motion. We have developed a stereoscopic tracking system which relies on native features on the head to estimate motion. Features are detected and matched across multiple camera views to accumulate a database of head landmarks and pose is estimated based on 3D-2D registration of the landmarks to features in each image. Pose estimates of a taxidermal rat head phantom undergoing realistic rat head motion via robot control had a root mean square error of 0.15 and 1.8 mm using markerless and marker-based motion tracking, respectively. Markerless motion tracking also led to an appreciable reduction in motion artifacts in motion-compensated positron emission tomography imaging of a live, unanesthetized rat. The results suggest that further improvements in live subjects are likely if nonrigid features are discriminated robustly and excluded from the pose estimation process.

List-mode image reconstruction with motion correction is computationally expensive, as it requires projection of hundreds of millions of rays through a 3D array. To decrease reconstruction time it is possible to use symmetric multiprocessing computers or graphics processing units. The former can have high financial costs, while the latter can require refactoring of algorithms. The Xeon Phi is a new co-processor card with a Many Integrated Core architecture that can run 4 multiple-instruction, multiple data threads per core with each thread having a 512-bit single instruction, multiple data vector register. Thus, it is possible to run in the region of 220 threads simultaneously. The aim of this study was to investigate whether the Xeon Phi co-processor card is a viable alternative to an x86 Linux server for accelerating List-mode PET image reconstruction for motion correction. An existing list-mode image reconstruction algorithm with motion correction was ported to run on the Xeon Phi coprocessor with the multi-threading implemented using pthreads. There were no differences between images reconstructed using the Phi co-processor card and images reconstructed using the same algorithm run on a Linux server. However, it was found that the reconstruction runtimes were 3 times greater for the Phi than the server. A new version of the image reconstruction algorithm was developed in C++ using OpenMP for mutli-threading and the Phi runtimes decreased to 1.67 times that of the host Linux server. Data transfer from the host to co-processor card was found to be a rate-limiting step; this needs to be carefully considered in order to maximize runtime speeds. When considering the purchase price of a Linux workstation with Xeon Phi co-processor card and top of the range Linux server, the former is a cost-effective computation resource for list-mode image reconstruction. A multi-Phi workstation could be a viable alternative to cluster computers at a lower cost for medical imaging applications.

Attenuation correction in positron emission tomography brain imaging of freely moving animals can be very challenging since the body of the animal is often within the field of view and introduces a non negligible atten- uating factor that can degrade the quantitative accuracy of the reconstructed images. An attractive approach that avoids the need for a transmission scan involves the generation of the convex hull of the animal’s head based on the reconstructed emission images. However, this approach ignores the potential attenuation introduced by the animal’s body. In this work, we propose a virtual scanner geometry, which moves in synchrony with the animal’s head and discriminates between those events that traverse only the animal’s head (and therefore can be accurately compensated for attenuation) and those that might have also traversed the animal’s body. For each pose a new virtual scanner geometry was defined and therefore a new system matrix was calculated leading to a time-varying system matrix. This new approach was evaluated on phantom data acquired on the microPET Focus 220 scanner using a custom-made rat phantom. Results showed that when the animal’s body is within the FOV and not accounted for during attenuation correction it can lead to bias of up to 10%. On the contrary, at- tenuation correction was more accurate when the virtual scanner was employed leading to improved quantitative estimates (bias

Estimation of nonlinear micro-parameters is a computationally demanding and fairly challenging process, since it involves the use of rather slow iterative nonlinear fitting algorithms and it often results in very noisy voxel-wise parametric maps. Direct reconstruction algorithms can provide parametric maps with reduced variance, but usually the overall reconstruction is impractically time consuming with common nonlinear fitting algorithms. In this work we employed a recently proposed direct parametric image reconstruction algorithm to estimate the parametric maps of all micro-parameters of a two-tissue compartment model, used to describe the kinetics of [ $$^{18}$$ F]FDG. The algorithm decouples the tomographic and the kinetic modelling problems, allowing the use of previously developed post-reconstruction methods, such as the generalised linear least squares (GLLS) algorithm. Results on both clinical and simulated data showed that the proposed direct reconstruction method provides considerable quantitative and qualitative improvements for all micro-parameters compared to the conventional post-reconstruction fitting method. Additionally, region-wise comparison of all parametric maps against the well-established filtered back projection followed by post-reconstruction non-linear fitting, as well as the direct Patlak method, showed substantial quantitative agreement in all regions. The proposed direct parametric reconstruction algorithm is a promising approach towards the estimation of all individual microparameters of any compartment model. In addition, due to the linearised nature of the GLLS algorithm, the fitting step can be very efficiently implemented and, therefore, it does not considerably affect the overall reconstruction time.

Purpose: Measuring and incorporating a scanner-specific point spread function (PSF) within image reconstruction has been shown to improve spatial resolution in PET. However, due to the short half-life of clinically used isotopes, other long-lived isotopes not used in clinical practice are used to perform the PSF measurements. As such, non-optimal PSF models that do not correspond to those needed for the data to be reconstructed are used within resolution modeling (RM) image reconstruction, usually underestimating the true PSF owing to the difference in positron range. In high resolution brain and preclinical imaging, this effect is of particular importance since the PSFs become more positron range limited and isotope-specific PSFs can help maximize the performance benefit from using resolution recovery image reconstruction algorithms. Methods: In this work, the authors used a printing technique to simultaneously measure multiple point sources on the High Resolution Research Tomograph (HRRT), and the authors demonstrated the feasibility of deriving isotope-dependent system matrices from fluorine-18 and carbon-11 point sources. Furthermore, the authors evaluated the impact of incorporating them within RM image reconstruction, using carbon-11 phantom and clinical datasets on the HRRT. Results: The results obtained using these two isotopes illustrate that even small differences in positron range can result in different PSF maps, leading to further improvements in contrast recovery when used in image reconstruction. The difference is more pronounced in the centre of the field-of-view where the full width at half maximum (FWHM) from the positron range has a larger contribution to the overall FWHM compared to the edge where the parallax error dominates the overall FWHM. Conclusions: Based on the proposed methodology, measured isotope-specific and spatially variant PSFs can be reliably derived and used for improved spatial resolution and variance performance in resolution recovery image reconstruction. The benefits are expected to be more substantial for more energetic positron emitting isotopes such as Oxygen-15 and Rubidium-82. (C) 2014 American Association of Physicists in Medicine.

Awake and/or freely moving small animal single photon emission imaging allows the pseudo-continuous study of 125I-labelled macromolecule kinetics, which could not otherwise be performed with systems involving restraint or anaesthesia. Estimating motion free projections in freely moving small animal planar imaging can be considered as a limited angle tomography problem (except that we wish to estimate the projections rather than the 3D volume), where the angular sampling depends on the observed motion. In this study, we hypothesise that the motion corrected planar projections estimated by reconstructing the 3D volume using an iterative reconstruction algorithm and integrating it along the projection path, will closdy match the true planar distribution, regardless of the observed motion. We tested this hypothesis using 3D simulations based on a dual opposed detector system, where motion was modelled with 6 degrees of freedom (i.e. rigid body). We also investigated the quantitative accuracy of regional activity extracted from the geometric mean of opposing motion corrected planar projections. Results showed that it is feasible to accurately estimate motion-corrected projections for a wide range of motions around all 3 axes. Errors were dependent on the observed motion, as well as the surrounding activity of overlapping organs, but geometric mean estimates of regional activity were within 10% of expected values. We conclude that quantitatively accurate motion-free projections of the tracer distribution in a freely moving animal can be estimated from dual opposed detectors using MLEM-based motion correction.

Brain dedicated scanners such as the High Resolution Research Tomograph (HRRT) can achieve spatial resolutions of 2–3 mm FWHM due to their small scintillators crystal size and a double crystal layer phoswich detector which permits some level of depth of interaction (DOI) to minimise parallax errors. Resolution can further be improved by resolution models (RM) within image reconstruction which may be spatially invariant (SIRM) or variant (SVRM). In previous work we demonstrated substantial suppression of parallax errors on the HRRT using spatial variant kernel maps within a resolution modelling (RM) image reconstruction. This approach could potentially be used as an alternative to DOI electronics. In this work we compare the resolution improvements achieved through iterative reconstruction algorithms incorporating resolution models with and without DOI. Greater resolution improvements were observed with RM reconstruction compared to DOI electronics however when both DOI and RM are employed additional benefits were still observed. In order to perform RM, assumptions which simplify the model are normally made, which might limit the resolution performance of RM algorithms. RM algorithms come with the penalty of introducing Gibb's artefacts which we observed to be more predominant when performing (SVRM) instead of more conservative approaches such as (SIRM).

Clinical PET imaging of the abdomen and thorax rely on standardised uptake values using static imaging but more robust quantification can be performed using dynamic protocols and estimating kinetic parameters from the dynamic PET data. However quantitative parameters both in static and dynamic imaging suffer from patient motion. In dynamic imaging the motion-induced data blurring and emission-attenuation mismatch results in erroneous time-activity curves potentially leading to erroneous parametric maps. Furthermore the effects of motion on direct parameter estimates using 4D image reconstruction algorithms are not known. In this work using a novel 5-D numerical body phantom we evaluate the impact of body motion on kinetic parameters in dynamic abdominal and thoracic PET imaging. Furthermore we compared parameter estimates obtained using both conventional post-reconstruction analysis as well as direct 4D image reconstruction.

Attenuation correction of small animal PET data is very important when quantitative images are of interest. Attenuation correction coefficients are conventionally obtained via a transmission or a computed tomography scan, which require anaesthetisation of the animal. However, in the context of awake and/or freely moving animals, where animal motion is compensated via appropriate motion tracking and correction techniques, anaesthetisation is no longer required. In this work we investigate the accuracy of a transmission-less attenuation correction approach based on the segmentation of the motion corrected emission image. Results on both phantom and real rat data acquired on the microPET Focus220 scanner, indicate good agreement between the segmentation based and conventional transmission based approach (∼ 2% difference). In addition, the segmentation based approach has the potential to eliminate noise propagation from the measured transmission data to the reconstructed attenuation corrected emission images.

Methods for adaptively fitting kinetic models to image data are explored. In conducting this fitting two goals are set: 1) that good fits are obtained for all image voxels; and 2) that where a primary biologically based model is appropriate, this model is used. The first goal is obtained through the inclusion of a more general secondary model. The second goal is achieved through the penalisation of estimates from this secondary model with the weighting of this penalisation optimised using generalised cross validation. The result is a modelling scheme where the secondary model is adaptively included for voxels that the primary model is unable to fit. The envisaged application is in 4D (space and time) reconstruction to prevent spatial propagation of errors from regions where a primary model is not appropriate. An evaluation of reconstructed clinical data using a dataset with challenging kinetics has been conducted. Application to the clinical dataset showed promising results with the modelling scheme able to fit kinetics associated with activity in the region of the injection site veins, the right ventricle, right atrium and the gall bladder. The method requires the careful selection of the secondary model and further work is needed to optimise this selection and evaluate the benefit for 4D reconstruction.

SPECT imaging of freely moving small animals would allow monitoring of many important physiological processes and behaviours, which are normally inhibited by anaesthetic drugs. However, in a freely moving animal it may be difficult, if not impossible, to distinguish between events that originate in the head from those that originate in the body and motion parameters cannot be assumed to be the same for the head and the body. This work employs a Monte CaIro simulation, based on a singleheaded SPECT camera, to assess the impact of inaccurate motion compensation of events originating in the body of a realistic mouse phantom, on the accuracy of motion corrected brain images. Results on non-attenuated data showed minimal motion-related bias

Abdominal and thoracic [18F]FDG PET/CT imaging is routinely used in clinical practice and drug development but parametric imaging based on full kinetic analysis is rarely used due to noise induced bias and variance in kinetic parameters and analysis is usually restricted to semi-quantitative indices. Direct parametric estimation using 4D image reconstruction can potentially provide parametric maps of reduced bias and variance but little or no evaluation has been done in the body. Previous studies have demonstrated that body [18F]FDG PET/CT imaging can benefit from these methods but analysis was restricted to macroparameters of interest. In work, we implement and apply a recently proposed direct 4-D algorithm using a 2 tissue compartment model on simulated and real oncology [18F]FDG PET/CT data to directly estimate microparameters of interest and assess potential improvements over the traditional post-reconstruction kinetic analysis. Results on the simulated data suggest clear reduction in bias for K1, k2 and blood volume and less for k3. Similarly, variance reduction was observed for K1, k2, k3 and Ki and less for blood volume. Evaluation on patient data demonstrated similar improvements using the direct 4-D method in agreement with simulations.

Accurate characterisation of the point spread function across the entire field of view is crucial in order to account for spatially dependent factors that degrade the resolution of the reconstructed images. The HRRT users community resolution modelling reconstruction software includes a shift-invariant resolution kernel, which leads to transaxially non-uniform resolution recovery. In this work we have developed a spatially-variant image-based resolution modelling reconstruction dedicated to the HRRT scanner, using an experimentally measured shift-variant resolution kernel. Application of the newly developed resolution modelling reconstruction on both phantom and clinical data demonstrated improvements both in contrast and resolution recovery, particularly for those regions that are close to the edges of the field of view, with almost uniform resolution recovery across the entire transverse field of view. In addition, because the newly measured resolution kernel is slightly broader with wider tails, compared to the deliberately conservative kernel employed in the HRRT users community software, the reconstructed images appear to have not only improved contrast recovery, but also better noise characteristics, at the expense of minor development of ringing (i.e. Gibbs) artefacts.

Measuring and incorporating scanner specific point spread functions (PSFs) within image reconstruction has been shown to improve spatial resolution in reconstructed PET images. However due to the short half life of the clinically used isotopes other long lived isotopes not used in clinical practice are chosen to perform the PSF measurements consequently leading to over or under estimation of the true PSF width during reconstruction due to the difference in positron range. In high resolution brain and preclinical imaging this effect is of particular importance since the resolution becomes more positron range limited and isotope specific PSFs can help maximizing the performance benefit from using resolution recovery image reconstruction algorithms. In this work we use a printing technique to simultaneously measure multiple point sources and demonstrate the feasibility of deriving isotope dependent system matrices on the High Resolution Research Tomograph (HRRT) by measuring spatially variant and isotope specific PSFs using Fluorine 18 and Carbon 11. Initial results based on these 2 isotopes illustrate that even small differences in positron range can result in different PSF maps. The difference is more distinct in the centre of the field of view (FOV) where the full width at half maximum (FWHM) from the positron range has a larger contribution in the overall FWHM compared to the edge of the FOV where the parallax error dominates the overall FWHM. Further PSF measurements are underway to evaluate clinically used isotopes with larger positron ranges and significantly shorter half lives. These measurements could be used to create a database of isotope dependent system matrices to be used within image reconstruction. © 2012 IEEE.

This note presents a practical approach to a custom-made design of PET phantoms enabling the use of digital radioactive distributions with high quantitative accuracy and spatial resolution. The phantom design allows planar sources of any radioactivity distribution to be imaged in transaxial and axial (sagittal or coronal) planes. Although the design presented here is specially adapted to the high-resolution research tomograph (HRRT), the presented methods can be adapted to almost any PET scanner. Although the presented phantom design has many advantages, a number of practical issues had to be overcome such as positioning of the printed source, calibration, uniformity and reproducibility of printing. A well counter (WC) was used in the calibration procedure to find the nonlinear relationship between digital voxel intensities and the actual measured radioactive concentrations. Repeated printing together with WC measurements and computed radiography (CR) using phosphor imaging plates (IP) were used to evaluate the reproducibility and uniformity of such printing. Results show satisfactory printing uniformity and reproducibility; however, calibration is dependent on the printing mode and the physical state of the cartridge. As a demonstration of the utility of using printed phantoms, the image resolution and quantitative accuracy of reconstructed HRRT images are assessed. There is very good quantitative agreement in the calibration procedure between HRRT, CR and WC measurements. However, the high resolution of CR and its quantitative accuracy supported by WC measurements made it possible to show the degraded resolution of HRRT brain images caused by the partial-volume effect and the limits of iterative image reconstruction.

Bootstrap resampling has been successfully used in estimating statistical properties of PET images by generating a set of statistically equivalent datasets based on one or more original datasets. However, the bootstrap resampling is only valid when the original dataset well represents the underlying distribution. The purpose of this work is to assess the validity of nonparametric bootstrap resampling using a long acquisition of a planar brain phantom, ensuring a good representation of the underlying distribution of all possible events. The assessment is carried out in two stages corresponding to the two ‘worlds’: i) the real world—generation of K reference list-mode datasets with five statistical levels (0.01%, 0.1%, 1%, 10% and 20% of the original dataset) using resampling with replacement of the statistically very rich original dataset playing the role of the population; and ii) the bootstrap world—generation of equivalent K bootstrap replicates using five resampled dataset from stage i) for each of the five statistical levels. The distributions from the two stages or worlds are then compared using the metric of Jensen-Shannon (J-S) divergence to quantify the similarity of the two distributions from stages i) and ii). In order to apply the J-S divergence two different histogramming methods are used: i) with constant and ii) adaptable binning. The bootstrap distributions are found to be constantly different to the real world distributions regardless of the data size and binning method. However when statistics are very low (for single voxels and 0.01% datasets) the comparison fails as the distributions are limited by the non-negativity constraint.

Direct parametric image reconstruction has the potential to reduce variance in parameter estimates when applied to PET/CT data. One complication when estimating parametric maps in the body is the difficulty of finding one single model to describe all the different kinetics in the field of view (FOV). Contrary to the post-reconstruction kinetic analysis though, any errors (bias) from the discrepancy between the model and the observed kinetics in the direct 4D reconstruction can potentially propagate spatially from unimportant areas to areas of interest. In this work we investigate this effect on simulated 4-D datasets based on a digital body phantom. Different realistic cases were simulated including differential input functions in the FOV and organs with different kinetics. Micro-parameters (K 1 , k 2 ,Vd, bv) where estimated using a newly proposed spatiotemporal 4D image reconstruction algorithm as well as using post-reconstruction kinetic analysis on noiseless and noisy datasets simulating [15O] H 2 O kinetics in the body. Bias analysis both in noiseless and noisy data showed a bias from badly modelled areas spatially propagates to other regions of interest in the direct reconstruction. Critically though under noisy conditions even with the bias propagation, the direct reconstruction method still outperforms the conventional post-reconstruction methodology. Nevertheless there is a need to ensure that appropriate models are chosen to describe the kinetics in the entire FOV with approaches such as data-driven adaptive kinetic modelling worth exploring.

Spatial resolution is of high importance especially in preclinical and brain PET imaging and characterization of the scanner's resolution properties requires measuring its point spread function (PSF). In our previous work we measured the PSF in 2 PET/CT scanners using a printed point source array. Here we extend the work to accurately measure the spatial variation of the PSF on the High Resolution Research Tomograph (HRRT) and assess the impact of the scanner's depth of interaction (DOI) capability on the spatial resolution. Furthermore we characterize the dependency of the PSF to count rate. An array of 15×11 printed sources was scanned twice, initially on its own (10min scan) and subsequently with an extension line containing ∼300MBq of carbon-11 at the scan starts (15 × 10min contiguous frames). Data were reconstructed with OP-OSEM and invariant PSF OP-OSEM. The PSF was found to be radially dependent in all directions but importantly radially symmetric and almost axially independent. The FWHM improves by ∼1.3mm using the PSF-OP-OSEM but is still radially variable with almost 1 mm degradation within the boundaries within which the brain is typically located. When DOI was not taken into account, a degradation up to 0.7mm was seen in the radial FWHM. In terms of count rate dependency, a clear resolution deterioration was seen, for count rates above 5–10 kcps with such count rates observed during many clinical scans.