Variance in parametric images: direct estimation from parametric projections

Recent work has shown that it is possible to apply linear kinetic models to dynamic projection data in PET in order to calculate parameter projections. These can subsequently be back-projected to form parametric images--maps of parameters of physiological interest. Critical to the application of these maps, to test for significant changes between normal and pathophysiology, is an assessment of the statistical uncertainty. In this context, parametric images also include simple integral images from, e.g., [O-15]-water used to calculate statistical parametric maps (SPMs). This paper revisits the concept of parameter projections and presents a more general formulation of the parameter projection derivation as well as a method to estimate parameter variance in projection space, showing which analysis methods (models) can be used. Using simulated pharmacokinetic image data we show that a method based on an analysis in projection space inherently calculates the mathematically rigorous pixel variance. This results in an estimation which is as accurate as either estimating variance in image space during model fitting, or estimation by comparison across sets of parametric images--as might be done between individuals in a group pharmacokinetic PET study. The method based on projections has, however, a higher computational efficiency, and is also shown to be more precise, as reflected in smooth variance distribution images when compared to the other methods.     

Maximum a posteriori reconstruction of the Patlak parametric image from sinograms in dynamic PET

Parametric imaging using the Patlak graphical method has been widely used to analyze dynamic PET data. Conventionally a Patlak parametric image is generated by reconstructing a sequence of dynamic images first and then performing Patlak graphical analysis on the time-activity curves pixel-by-pixel. However, because it is rather difficult to model the noise distribution in reconstructed images, the spatially variant noise correlation is simply ignored in the Patlak analysis, which leads to sub-optimal results. In this paper we present a Bayesian method for reconstructing Patlak parametric images directly from raw sinogram data by incorporating the Patlak plot model into the image reconstruction procedure. A preconditioned conjugate gradient algorithm is used to find the maximum a posteriori solution. The proposed direct method is statistically more efficient than the conventional indirect approach because the Poisson noise distribution in PET data can be accurately modeled in the direct reconstruction. The computation cost of the direct method is similar to reconstruction time of two dynamic frames. Therefore, when more than two dynamic frames are used in the Patlak analysis, the direct method is faster than the conventional indirect approach. We conduct computer simulations to validate the proposed direct method. Comparisons with the conventional indirect approach show that the proposed method results in a more accurate estimate of the parametric image. The proposed method has been applied to dynamic fully 3D PET data from a microPET scanner.     

Deblurring of breathing motion artifacts in thoracic PET images by deconvolution methods

In FDG-PET imaging of thoracic tumors, blurring due to breathing motion often significantly degrades the quality of the observed image, which then obscures the tumor boundary. We demonstrate a deblurring technique that combines patient-specific motion estimates of tissue trajectories with image deconvolution techniques, thereby partially eliminating breathing-motion induced artifacts. Two data sets were used to evaluate the methodology including mobile phantoms and clinical images. The clinical images consist of PET/CT co-registered images of patients diagnosed with lung cancer. A breathing motion model was used to locally estimate the location-dependent tissue location probability function (TLP) due to breathing. The deconvolution process is carried by an expectation-maximization (EM) iterative algorithm using the motion-based TLP. Several methods were used to improve the robustness of the deblurring process by mitigating noise amplification and compensating for motion estimate uncertainties. The mobile phantom study with controlled settings demonstrated significant reduction in underestimation error of concentration in high activity case without significant superiority between the different applied methods. In case of medium activity concentration (moderate noise levels), less improvement was reported (10%-15% reduction in underestimation error relative to 15%-20% reduction in high concentration). Residual denoising using wavelets offered the best performance for this case. In the clinical data case, the image spatial resolution was significantly improved, especially in the direction of greatest motion (cranio-caudal). The EM algorithm converged within 15 and 5 iterations in the large and small tumor cases, respectively. A compromise between a figure-of-merit and entropy minimization was suggested as a stopping criterion. Regularization techniques such as wavelets and Bayesian methods provided further refinement by suppressing noise amplification. Our initial results show that the proposed method provides a feasible framework for improving PET thoracic images, without the need for gated/4-D PET imaging, when 4-D CT is available to estimate tumor motion.     

A deconvolution approach for PET-based dose reconstruction in proton radiotherapy

Positron emitters are activated by proton beams in proton radiotherapy, and positron emission tomography (PET) images can thus be used for dose verification. Since a PET image is not directly proportional to the delivered radiation dose distribution, predicted PET images are compared to measured PET images and an agreement of both indicates a successful irradiation. Such predictions are given on the basis of Monte Carlo calculations or a filtering approach which uses a convolution of the planned dose with specific filter functions to estimate the PET activity. In this paper, we describe and evaluate a dose reconstruction method based on PET images which reverses the just mentioned convolution approach using appropriate deconvolution methods. Deconvolution is an ill-posed inverse problem, and suitable regularization techniques are required in order to guarantee a stable solution. The basic convolution approach is developed for homogeneous media and additional procedures are necessary to generalize the PET estimation to inhomogeneous media. This generalization formalism is used in our dose deconvolution approach as well. Various simulations demonstrate that the dose reconstruction method is able to reverse the PET estimation method both in homogeneous and inhomogeneous media. Measured PET images are however degraded by noise and artifacts and the dose reconstructions become more difficult and the results suffer from artifacts as well. Recently used in-room PET scanners allow a decreased delay time between irradiation and imaging, and thus the influence of short-lived positron emitters on the PET images increases considerably. We extended our dose reconstruction method to process PET images which contain several positron emitters and simulated results are shown.     

Optimization of methods for quantification of rCBF using high-resolution [¹⁵O]H₂O PET images

This study aimed to derive accurate estimates of regional cerebral blood flow (rCBF) from noisy dynamic [1?O]H?O PET images acquired on the high-resolution research tomograph, while retaining as much as possible the high spatial resolution of this brain scanner (2-3 mm) in parametric maps of rCBF. The PET autoradiographic method and generalized linear least-squares (GLLS), with fixed or extended to include spatially variable estimates of the dispersion of the measured input function, were compared to nonlinear least-squares (NLLS) for rCBF estimation. Six healthy volunteers underwent two [1?O]H?O PET scans with continuous arterial blood sampling. rCBF estimates were obtained from three image reconstruction methods (one analytic and two iterative, of which one includes a resolution model) to which a range of post-reconstruction filters (3D Gaussian: 2, 4 and 6 mm FWHM) were applied. The optimal injected activity was estimated to be around 11 MBq kg?1 (800 MBq) by extrapolation of patient-specific noise equivalent count rates. Whole-brain rCBF values were found to be relatively insensitive to the method of reconstruction and rCBF quantification. The grey and white matter rCBF for analytic reconstruction and NLLS were 0.44 ± 0.03 and 0.15 ± 0.03 mL min?1 cm?3, respectively, in agreement with literature values. Similar values were obtained from the other methods. For generation of parametric images using GLLS or the autoradiographic method, a filter of ≥ 4 mm was required in order to suppress noise in the PET images which otherwise produced large biases in the rCBF estimates.     

Direct 4D parametric imaging for linearized models of reversibly binding PET tracers using generalized AB-EM reconstruction

Due to high noise levels in the voxel kinetics, development of reliable parametric imaging algorithms remains one of most active areas in dynamic brain PET imaging, which in the vast majority of cases involves receptor/transporter studies with reversibly binding tracers. As such, the focus of this work has been to develop a novel direct 4D parametric image reconstruction scheme for such tracers. Based on a relative equilibrium (RE) graphical analysis formulation (Zhou et al 2009b Neuroimage 44 661-70), we developed a closed-form 4D EM algorithm to directly reconstruct distribution volume (DV) parametric images within a plasma input model, as well as DV ratio (DVR) images within a reference tissue model scheme (wherein an initial reconstruction was used to estimate the reference tissue time-activity curves). A particular challenge with the direct 4D EM formulation is that the intercept parameters in graphical (linearized) analysis of reversible tracers (e.g. Logan or RE analysis) are commonly negative (unlike for irreversible tracers, e.g. using Patlak analysis). Subsequently, we focused our attention on the AB-EM algorithm, derived by Byrne (1998, Inverse Problems 14 1455-67) to allow inclusion of prior information about the lower (A) and upper (B) bounds for image values. We then generalized this algorithm to the 4D EM framework, thus allowing negative intercept parameters. Furthermore, our 4D AB-EM algorithm incorporated and emphasized the use of spatially varying lower bounds to achieve enhanced performance. As validation, the means of parameters estimated from 55 human (11)C-raclopride dynamic PET studies were used for extensive simulations using a mathematical brain phantom. Images were reconstructed using conventional indirect as well as proposed direct parametric imaging methods. Noise versus bias quantitative measurements were performed in various regions of the brain. Direct 4D EM reconstruction resulted in notable qualitative and quantitative accuracy improvements (over 35% noise reduction, with matched bias, in both plasma and reference-tissue input models). Similar improvements were also observed in the coefficient of variation of the estimated DV and DVR values even for relatively low uptake cortical regions, suggesting the enhanced ability for robust parameter estimation. The method was also tested on a 90 min (11)C-raclopride patient study performed on the high-resolution research tomograph wherein the proposed method was shown across a variety of regions to outperform the conventional method in the sense that for a given DVR value, improved noise levels were observed.     

3.5D dynamic PET image reconstruction incorporating kinetics-based clusters

Standard 3D dynamic positron emission tomographic (PET) imaging consists of independent image reconstructions of individual frames followed by application of appropriate kinetic model to the time activity curves at the voxel or region-of-interest (ROI). The emerging field of 4D PET reconstruction, by contrast, seeks to move beyond this scheme and incorporate information from multiple frames within the image reconstruction task. Here we propose a novel reconstruction framework aiming to enhance quantitative accuracy of parametric images via introduction of priors based on voxel kinetics, as generated via clustering of preliminary reconstructed dynamic images to define clustered neighborhoods of voxels with similar kinetics. This is then followed by straightforward maximum a posteriori (MAP) 3D PET reconstruction as applied to individual frames; and as such the method is labeled '3.5D' image reconstruction. The use of cluster-based priors has the advantage of further enhancing quantitative performance in dynamic PET imaging, because: (a) there are typically more voxels in clusters than in conventional local neighborhoods, and (b) neighboring voxels with distinct kinetics are less likely to be clustered together. Using realistic simulated (11)C-raclopride dynamic PET data, the quantitative performance of the proposed method was investigated. Parametric distribution-volume (DV) and DV ratio (DVR) images were estimated from dynamic image reconstructions using (a) maximum-likelihood expectation maximization (MLEM), and MAP reconstructions using (b) the quadratic prior (QP-MAP), (c) the Green prior (GP-MAP) and (d, e) two proposed cluster-based priors (CP-U-MAP and CP-W-MAP), followed by graphical modeling, and were qualitatively and quantitatively compared for 11 ROIs. Overall, the proposed dynamic PET reconstruction methodology resulted in substantial visual as well as quantitative accuracy improvements (in terms of noise versus bias performance) for parametric DV and DVR images. The method was also tested on a 90 min (11)C-raclopride patient study performed on the high-resolution research tomography. The proposed method was shown to outperform the conventional method in visual as well as quantitative accuracy improvements (in terms of noise versus regional DVR value performance).     

Joint model of motion and anatomy for PET image reconstruction

Anatomy-based positron emission tomography (PET) image enhancement techniques have been shown to have the potential for improving PET image quality. However, these techniques assume an accurate alignment between the anatomical and the functional images, which is not always valid when imaging the chest due to respiratory motion. In this article, we present a joint model of both motion and anatomical information by integrating a motion-incorporated PET imaging system model with an anatomy-based maximum a posteriori image reconstruction algorithm. The mismatched anatomical information due to motion can thus be effectively utilized through this joint model. A computer simulation and a phantom study were conducted to assess the efficacy of the joint model, whereby motion and anatomical information were either modeled separately or combined. The reconstructed images in each case were compared to corresponding reference images obtained using a quadratic image prior based maximum a posteriori reconstruction algorithm for quantitative accuracy. Results of these studies indicated that while modeling anatomical information or motion alone improved the PET image quantitation accuracy, a larger improvement in accuracy was achieved when using the joint model. In the computer simulation study and using similar image noise levels, the improvement in quantitation accuracy compared to the reference images was 5.3% and 19.8% when using anatomical or motion information alone, respectively, and 35.5% when using the joint model. In the phantom study, these results were 5.6%, 5.8%, and 19.8%, respectively. These results suggest that motion compensation is important in order to effectively utilize anatomical information in chest imaging using PET. The joint motion-anatomy model presented in this paper provides a promising solution to this problem.     

Equivalent Dipole Source Imaging of Brain Electric Activity by Means of Parametric Projection Filter

In the present study, spatial filters for inverse estimation of an equivalent dipole layer from the scalp-recorded potentials have been explored for their suitability in achieving high-resolution electroencephalogram (EEG) imaging. The performance of the parametric projection filter (PPF), which we propose to use for high-resolution EEG imaging, has been evaluated by computer simulations in the presence of a priori information on noise. An inhomogeneous three-concentric-sphere head model was used in the present simulation study to represent the head volume conductor. An equivalent dipole layer was used to model brain electric sources and estimated from the scalp potentials. Various noise conditions were simulated and the parametric projection filter was compared with standard regularization procedures such as the truncated singular value decomposition (TSVD) and the Tikhonov regularization (TKNV). The present simulation results suggest that the proposed method performs better than that of commonly used inverse regularization techniques, such as the general inverse using the TSVD and the TKNV, when the correlation between the original source distribution and the noise distribution is low, and performs similarly when the correlation is high. A method for determining the optimum regularization parameter, which can be applied to parametric inverse techniques, has also been developed. © 2001 Biomedical Engineering Society. PAC01: 8757Nk, 0230Zz, 8719Nn, 0260Dc     

A RVR-based Method for Bias Field Estimation in Brain Magnetic Resonance Images Segmentation

This paper presents a relevance vector regression(RVR) based on parametric approach to the bias field estimation in brain magnetic resonance(MR) image segmentation. Segmentation is a very important and challenging task in brain analysis,while the bias field existed in the images can significantly deteriorate the performance.Most of current parametric bias field correction techniques use a pre-set linear combination of low degree basis functions, the coefficients and the basis function types of which completely determine the field. The proposed RVR method can automatically determine the best combination for the bias field, resulting in a good segmentation in the presence of noise by combining with spatial constrained fuzzy C-means(SCFCM)segmentation. Experiments on simulated T1 images show the efficiency.     

Spatio-temporal diffusion of dynamic PET images

Positron emission tomography (PET) images are corrupted by noise. This is especially true in dynamic PET imaging where short frames are required to capture the peak of activity concentration after the radiotracer injection. High noise results in a possible bias in quantification, as the compartmental models used to estimate the kinetic parameters are sensitive to noise. This paper describes a new post-reconstruction filter to increase the signal-to-noise ratio in dynamic PET imaging. It consists in a spatio-temporal robust diffusion of the 4D image based on the time activity curve (TAC) in each voxel. It reduces the noise in homogeneous areas while preserving the distinct kinetics in regions of interest corresponding to different underlying physiological processes. Neither anatomical priors nor the kinetic model are required. We propose an automatic selection of the scale parameter involved in the diffusion process based on a robust statistical analysis of the distances between TACs. The method is evaluated using Monte Carlo simulations of brain activity distributions. We demonstrate the usefulness of the method and its superior performance over two other post-reconstruction spatial and temporal filters. Our simulations suggest that the proposed method can be used to significantly increase the signal-to-noise ratio in dynamic PET imaging.     

Region of interest motion compensation for PET image reconstruction

A motion-incorporated reconstruction (MIR) method for gated PET imaging has recently been developed by several authors to correct for respiratory motion artifacts in PET imaging. This method however relies on a motion map derived from images (4D PET or 4D CT) of the entire field of view (FOV). In this study we present a region of interest (ROI)-based extension to this method, whereby only the motion map of a user-defined ROI is required and motion incorporation during image reconstruction is solely performed within the ROI. A phantom study and an NCAT computer simulation study were performed to test the feasibility of this method. The phantom study showed that the ROI-based MIR produced results that are within 1.26% of those obtained by the full image-based MIR approach when using the same accurate motion information. The NCAT phantom study on the other hand, further verified that motion of features of interest in an image can be estimated more efficiently and potentially more accurately using the ROI-based approach. A reduction of motion estimation time from 450 s to 30 and 73 s was achieved for two different ROIs respectively. In addition, the ROI-based approach showed a reduction in registration error of 43% for one ROI, which effectively reduced quantification bias by 44% and 32% using mean and maximum voxel values, respectively.     

Accelerated median root prior reconstruction for pinhole single-photon emission tomography (SPET)

Pinhole collimation can be used to improve spatial resolution in SPET. However, the resolution improvement is achieved at the cost of reduced sensitivity, which leads to projection images with poor statistics. Images reconstructed from these projections using the maximum likelihood expectation maximization (ML-EM) algorithms, which have been used to reduce the artefacts generated by the filtered backprojection (FBP) based reconstruction, suffer from noise/bias trade-off: noise contaminates the images at high iteration numbers, whereas early abortion of the algorithm produces images that are excessively smooth and biased towards the initial estimate of the algorithm. To limit the noise accumulation we propose the use of the pinhole median root prior (PH-MRP) reconstruction algorithm. MRP is a Bayesian reconstruction method that has already been used in PET imaging and shown to possess good noise reduction and edge preservation properties. In this study the PH-MRP algorithm was accelerated with the ordered subsets (OS) procedure and compared to the FBP, OS-EM and conventional Bayesian reconstruction methods in terms of noise reduction, quantitative accuracy, edge preservation and visual quality. The results showed that the accelerated PH-MRP algorithm was very robust. It provided visually pleasing images with lower noise level than the FBP or OS-EM and with smaller bias and sharper edges than the conventional Bayesian methods.     

Monte Carlo simulations of clinical PET and SPECT scans: impact of the input data on the simulated images

Monte Carlo simulations of emission tomography have proven useful to assist detector design and optimize acquisition and processing protocols. The more realistic the simulations, the more straightforward the extrapolation of conclusions to clinical situations. In emission tomography, accurate numerical models of tomographs have been described and well validated under specific operating conditions (collimator, radionuclide, acquisition parameters, count rates, etc). When using these models under these operating conditions, the realism of simulations mostly depends on the activity distribution used as an input for the simulations. It has been proposed to derive the input activity distribution directly from reconstructed clinical images, so as to properly model the heterogeneity of the activity distribution between and within organs. However, reconstructed patient images include noise and have limited spatial resolution. In this study, we analyse the properties of the simulated images as a function of the properties of the reconstructed images used to define the input activity distributions in (18)F-FDG PET and (131)I SPECT simulations. The propagation through the simulation/reconstruction process of the noise and spatial resolution in the input activity distribution was studied using simulations. We found that the noise properties of the images reconstructed from the simulated data were almost independent of the noise in the input activity distribution. The spatial resolution in the images reconstructed from the simulations was slightly poorer than that in the input activity distribution. However, using high-noise but high-resolution patient images as an input activity distribution yielded reconstructed images that could not be distinguished from clinical images. These findings were confirmed by simulated highly realistic (131)I SPECT and (18)F-FDG PET images from patient data. In conclusion, we demonstrated that (131)I SPECT and (18)F-FDG PET images indistinguishable from real scans can be simulated using activity maps with spatial resolution higher than that used in routine clinical applications.     

Differential attenuation method for simultaneous estimation of activity and attenuation in multiemission single photon emission computed tomography

A penalized weighted least squares reconstruction algorithm is described that simultaneously estimates activity and attenuation distributions from emission sinogram data alone. This estimation technique is based on differential attenuation information and is applicable to any single photon emission computed tomography imaging isotope with emissions at two or more distinct energies, after accurate compensation for Compton scatter. A rotation-based forward projector is used to efficiently model photon attenuation at multiple emission energies, as well as distance-dependent spatial resolution. The algorithm was tested using simulated scatter-free 201T1 projection data from a single-slice numerical cardiac phantom with and without cold myocardial defects. Poisson noise was added to the projection data to mimic clinically realistic count densities. The activity estimates resulting from the proposed method had fewer artifacts and were substantially more accurate than images reconstructed with filtered backprojection without compensation for attenuation. Several techniques were employed to reduce the time required for the iterative routine to converge and to reduce the sensitivity of the solution to noise in the projection data. These included: (1) a preconditioning image variable transformation; (2) a coarse-to-fine grid initialization schedule; and (3) a convex hull image mask determined directly from the data. The combined effect of these techniques substantially reduced the compute time required for the reconstruction.     

Computer phantom study of brain PET glucose metabolism imaging using a rotating SPECT/PET camera

Positron emission tomography (PET) with [18F] fluoro-deoxy-glucose (FDG) provides information about glucose metabolism and is used to measure tissue glucose kinetics in the brain. The recent interest in hybrid SPECT/PET systems emerged as a practical approach to reduce the high cost of purchasing a dedicated ring-detector PET system. We have implemented interpolation methods for processing the projection data that could potentially reduce artifacts when reconstructing a dynamic imaging sequence in a PET study from a dual-head rotating SPECT/PET system. The computer simulations predict that parameter estimates from the dedicated PET system will be superior to results using the rotating camera system. However, the rotating camera system using projection interpolation may approach the accuracy of the dedicated PET system if the data noise is below 20%.     

AB-OSEM reconstruction for improved Patlak kinetic parameter estimation: a simulation study

The non-negativity constraint inherently present in OSEM reconstruction successfully reduces the standard deviation in cold regions but at the cost of introducing a positive bias, especially at low iteration numbers. For low-count data, as often encountered in short-duration frames in dynamic imaging protocols, it has been shown that it can be advantageous (in terms of bias in the reconstructed image) to remove the non-negativity constraint. In this work two competing algorithms that do not impose non-negativity in the reconstructed image are investigated: NEG-ML and AB-OSEM. It was found that the AB-OSEM reconstruction outperformed the NEG-ML reconstruction. The AB-OSEM algorithm was then further developed to allow a forward model that includes randoms and scatter background terms. In addition to static reconstruction the current analysis was extended to consider the important case of kinetic parameter estimation from dynamic PET data. Simulation studies (comparing OSEM, FBP and AB-OSEM) showed that the positive bias obtained with OSEM reconstruction can be avoided in both static and parametric imaging through use of a negative lower bound in AB-OSEM reconstruction (i.e. by lifting the implicit non-negativity constraint of OSEM). When quantification tasks are considered, the overall error in the estimates (composed of both bias and standard deviation) is often of primary concern. An important finding of this work is that in most cases the activity concentration and the kinetic parameters obtained from images reconstructed using AB-OSEM showed a lower overall root mean squared error compared to the popular choices of either OSEM or FBP reconstruction for both cold and warm regions. As such, AB-OSEM should be preferred instead of the standard OSEM and FBP reconstructions when kinetic parameter estimation is considered. Finally, this work shows example parametric images from the high-resolution research tomograph obtained using the different reconstruction methods.     

Tomographic digital subtraction angiography for lung perfusion estimation in rodents

In vivo measurements of perfusion present a challenge to existing small animal imaging techniques such as magnetic resonance microscopy, micro computed tomography, micro positron emission tomography, and microSPECT, due to combined requirements for high spatial and temporal resolution. We demonstrate the use of tomographic digital subtraction angiography (TDSA) for estimation of perfusion in small animals. TDSA augments conventional digital subtraction angiography (DSA) by providing three-dimensional spatial information using tomosynthesis algorithms. TDSA is based on the novel paradigm that the same time density curves can be reproduced in a number of consecutive injections of microL volumes of contrast at a series of different angles of rotation. The capabilities of TDSA are established in studies on lung perfusion in rats. Using an imaging system developed in-house, we acquired data for four-dimensional (4D) imaging with temporal resolution of 140 ms, in-plane spatial resolution of 100 microm, and slice thickness on the order of millimeters. Based on a structured experimental approach, we optimized TDSA imaging providing a good trade-off between slice thickness, the number of injections, contrast to noise, and immunity to artifacts. Both DSA and TDSA images were used to create parametric maps of perfusion. TDSA imaging has potential application in a number of areas where functional perfusion measurements in 4D can provide valuable insight into animal models of disease and response to therapeutics.     

Minimisation of parameter estimation errors in dynamic PET: choice of scanning schedules

We addressed the general problem of finding an optimal scan schedule in positron emission tomography (PET) dynamic studies which minimises the errors in estimating the transfer constants between a set of compartments. As an example, the influence of scan intervals in PET on the accuracy of estimation of the rate constants and vascular component in the deoxyglucose method was examined using an empirical noise model. The simulated noisy curves used in the analysis were compared with patient data to validate the noise model. A series of scan schedules were compared for accuracy of fit by evaluating the determinant of the variance-covariance matrix of the fitted parameters as an index of parameter accuracy. For realistic noise levels there is a monotonic improvement in the index of parameter accuracy with increasing sampling frequency, particularly over the initial minutes after the tracer injection. However, since faster schedules are more susceptible to errors introduced by time mismatches between plasma and tissue curves and impose greater computational and memory overhead, an initial scan duration of 30 s provide a practical trade-off for dynamic PET 18F-fluoro-deoxyglucose studies.     

Parametric image of myocardial blood flow generated from dynamic H2 15O PET using factor analysis and cluster analysis

Algorithm-based parametric imaging of myocardial blood flow (MBF), as measured by H₂ ¹⁵O PET, has been the goal of many research efforts. A method for generating parametric images of regional MBF by factor and cluster analysis on H₂ ¹⁵O dynamic myocardial PET was validated by its comparison with gold-standard MBF values determined invasively using radiolabelled microspheres. Right and left ventricular blood pool activities and their factor images were obtained by the application of factor analysis to dynamic frames. By subtraction of the factor images multiplied by their corresponding values on the factors from the original dynamic images for each frame, pure tissue dynamic images were obtained, from which arterial blood activities were excluded. Cluster analysis that averaged pixels having time-activity curves with the same shape was applied to pure tissue images to generate parametric MBF images. The usefulness of this method for quantifying regional MBF was evaluated using canine experiment data. H₂ ¹⁵O PET scans and microsphere studies were performed on seven dogs at rest and after pharmacological stress. The image qualities and the contrast of parametric images obtained using the proposed method were significantly improved over either the tissue factor images or the parametric images obtained using a conventional method. Regional MBFs obtained using the proposed method correlated well with those obtained by the region of interest method (r=0.94) and by the microsphere technique (r=0.90). A non-invasive method is presented for generating parametric images of MBF from H₂ ¹⁵O PET, using factor and cluster analysis.