Andriy Andreyev

Dual-isotope PET using positron-gamma emitters

Positron emission tomography (PET) is widely recognized as a highly effective functional imaging modality. Unfortunately, standard PET cannot be used for dual-isotope imaging (which would allow for simultaneous investigation of two different biological processes), because positron-electron annihilation products from different tracers are indistinguishable in terms of energy. Methods that have been proposed for dual-isotope PET rely on differences in half-lives of the participating isotopes; these approaches, however, require making assumptions concerning kinetic behavior of the tracers and may not lead to optimal results. In this paper we propose a novel approach for dual-isotope PET and investigate its performance using GATE simulations. Our method requires one of the two radioactive isotopes to be a pure positron emitter and the second isotope to emit an additional high-energy gamma in a cascade simultaneously with positron emission. Detection of this auxiliary prompt gamma in coincidence with the annihilation event allows us to identify the corresponding 511 keV photon pair as originating from the same isotope. Two list-mode datasets are created: a primary dataset that contains all detected 511 keV photon pairs from both isotopes, and a second, tagged (much smaller) dataset that contains only those PET events for which a coincident prompt gamma has also been detected. An image reconstructed from the tagged dataset reflects the distribution of the second positron-gamma radiotracer and serves as a prior for the reconstruction of the primary dataset. Our preliminary simulation study with partially overlapping (18)F/(22)Na and (18)F/(60)Cu radiotracer distributions showed that in these two cases the dual-isotope PET method allowed for separation of the two activity distributions and recovered total activities with relative errors of about 5%.

Fast image reconstruction for Compton camera using stochastic origin ensemble approach.

PURPOSE Compton camera has been proposed as a potential imaging tool in astronomy, industry, homeland security, and medical diagnostics. Due to the inherent geometrical complexity of Compton camera data, image reconstruction of distributed sources can be ineffective and/or time-consuming when using standard techniques such as filtered backprojection or maximum likelihood-expectation maximization (ML-EM). In this article, the authors demonstrate a fast reconstruction of Compton camera data using a novel stochastic origin ensembles (SOE) approach based on Markov chains. METHODS During image reconstruction, the origins of the measured events are randomly assigned to locations on conical surfaces, which are the Compton camera analogs of lines-of-responses in PET. Therefore, the image is defined as an ensemble of origin locations of all possible event origins. During the course of reconstruction, the origins of events are stochastically moved and the acceptance of the new event origin is determined by the predefined acceptance probability, which is proportional to the change in event density. For example, if the event density at the new location is higher than in the previous location, the new position is always accepted. After several iterations, the reconstructed distribution of origins converges to a quasistationary state which can be voxelized and displayed. RESULTS Comparison with the list-mode ML-EM reveals that the postfiltered SOE algorithm has similar performance in terms of image quality while clearly outperforming ML-EM in relation to reconstruction time. CONCLUSIONS In this study, the authors have implemented and tested a new image reconstruction algorithm for the Compton camera based on the stochastic origin ensembles with Markov chains. The algorithm uses list-mode data, is parallelizable, and can be used for any Compton camera geometry. SOE algorithm clearly outperforms list-mode ML-EM for simple Compton camera geometry in terms of reconstruction time. The difference in computational time will be much larger when full Compton camera system model, including resolution recovery, is implemented and realistic Compton camera geometries are used. It was also shown in this article that while correctly reconstructing the relative distribution of the activity in the object, the SOE algorithm tends to underestimate the intensity values and increase variance in the images; improvements to the SOE reconstruction algorithm will be considered in future work.

Reconstruction of dual isotope PET using expectation maximization (EM) algorithm

The dual-isotope (DI) imaging with positron emission tomography (PET) is particularly difficult because positron-electron annihilation always results in 511 keV photons, irrespective of the radioisotope. Therefore, differentiation of PET isotopes based on the energy discrimination is impossible. However, there are radioisotopes that in addition to a positron also emit at least one prompt high energy (HE) gamma (positron-gamma emitters). The detection of this gamma in coincidence with two annihilation photons tags this particular PET event as originating from the decay of the isotope with positron-gamma emissions. An example of such radioisotope is 60Cu which can be used in our approach for DI studies together with a pure positron emitter 18F. Two data sets are acquired in DI-PET imaging. The first, standard PET dataset results from detections of two coincident 511 keV photons, and the second dataset - from detection of two 511 keV photons in coincidence with a HE gamma. The challenge is then to reconstruct two separate activities using the information contained in these two datasets. In this work, we developed an Expectation Maximization (EM) algorithm for the estimation of distributions of both isotopes. The EM update equations were derived and evaluated using synthetic data. We used Monte Carlo package GATE to simulate the dual-isotope acquisition with a standard PET camera. Realistic efficiencies were assumed for the HE gamma detection varying from 3% to 9%. A set of seven partially overlapping 3D objects filled with 18F and 60Cu was located at the center of a water phantom of 30 cm diameter. A low-count (250k decays of each 18F and 60Cu) and a high-count (2.5M decays each) acquisitions were simulated. We found that the dual-isotope EM algorithm was able to successfully separate two PET tracers for all values of studied parameters. As expected the accuracy of this separation increased for higher HE gamma detection efficiency and higher total number of detected counts. A systematic analysis of the quantitative accuracy of these reconstructions is currently being performed.

Stochastic image reconstruction method for Compton camera

Compton camera has been proposed as an imaging tool in astronomy, industry, homeland security and medical imaging where source location can be identified based on detection of Compton scattered and photoabsorbed photons. However, due to the geometrical complexity of the problem, image reconstruction from Compton camera data is difficult and/or ineffective when using techniques such as filtered backprojection (FBP) or maximum likelihood expectation maximization (MLEM). In this paper, we propose a novel stochastic origin ensembles (SOE) approach based on Markov chains (previously implemented and tested for PET) to the reconstruction of Compton camera data. The advantages of this method are that it is universal (i.e. works with any camera geometry), does not require any rebinning of acquired data, parallelizable, and does not require forward- and back-projection operations nor voxelization of the image space. During image reconstruction the origins of every measured events are the randomly assigned locations on the conical surface (which is the Compton camera analog of lines-of-response in PET). Therefore, the image is defined as an event ensemble holding the coordinates of all possible event origins. During the course of the reconstruction, origins of the events are randomly moved and the acceptance of the new event origin is determined by the acceptance probability, which is proportional to the change in event density. For example, if event density at the new location is higher than in the previous location, the new position is always accepted. As a result of many iterations, the reconstructed image converges to the quasi-stationary state which can be voxelized and displayed. Comparison to the list-mode MLEM shows that SOE algorithm has similar performance in terms of image quality while clearly outperforming it in terms of reconstruction time. Implementation of corrections for detector energy resolution can be done at almost no additional computational cost, which is a major advantage of SOE over other reconstruction methods.

Dual-radioisotope PET data acquisition and analysis

This study experimentally evaluates a novel approach for dual radioisotope PET imaging. The method relies on one radioisotope being a pure positron emitter and the other radioisotope emitting a prompt high energy gamma along with the positron. Using the differential count rates of dual and triple coincidences allows for quantitative reconstruction of the individual radioisotope activities. The objective of the present study was to perform an experimental proof-of-principle test of the method. We used two cMiCE detector modules (with 400 keV lower energy threshold) mounted directly across from one another and a third module (850 keV threshold) at 90 degrees to the first two modules. Coincidence logic was set to either monitor a coincidence between modules 1 and 2 and NOT 3, or a coincidence between all three modules. These were used to measure dual and triple coincidence count rates from 22Na and 68Ge point sources, scanned either separately or together. While 68Ge was considered to be a pure positron emitter, for 22Na, 90.4% of decays produce a positron and a 1.275 MeV prompt gamma. In addition to the count rate test, full tomographic data were collected and images reconstructed for both dual and triple coincidence data sets. Reconstructed images demonstrate the ability of the method to clearly distinguish between 22Na and 68Ge sources based on triple-coincidence criterium. For both dual and triple coincidence event modes, the coincidence rates for simultaneous acquisition of the 22Na and 68Ge sources were 10–19% higher than the sum of the coincident rates for the radioisotopes acquired individually. We speculate this is due to energy pulse-pile up and are continuing evaluations of this effect. We were able to demonstrate the basic validity of differentiating the individual activity levels of 22Na and 68Ge sources even when acquired at the same time. Quantitative accuracy can likely be improved using normalization methods, and evaluations of this approach are ongoing.

Reconstructed Image Spatial Resolution of Multiple Coincidences Compton Imager

We study the multiple coincidences Compton imager (MCCI) which is based on a simultaneous acquisition of several photons emitted in cascade from a single nuclear decay. Theoretically, this technique should provide a major improvement in localization of a single radioactive source as compared to a standard Compton camera. In this work, we investigated the performance and limitations of MCCI using Monte Carlo computer simulations. Spatial resolutions of the reconstructed point source have been studied as a function of the MCCI parameters, including geometrical dimensions and detector characteristics such as materials, energy and spatial resolutions.

Acceleration of blob-based iterative reconstruction algorithm using Tesla GPU

Blob-based iterative image reconstruction techniques provide high-quality denoised images in the photon starving emission tomography. However, the attractiveness of blob-based iterative algorithms is devalued by their high demands on the computation time. In this study we investigate the use of graphic processing units (GPU) to parallelize the ray-driven blob-based OSEM reconstruction algorithm for SPECT. We obtained a speed-up factor of 14.7 as compared to the blob reconstruction performed using CPU. Therefore, the reconstruction time of blob-based image on GPU was comparable to the reconstruction time of voxel-based image on the CPU. The algorithm can be further accelerated by more effective utilization of the GPU register space and shared GPU memory, which we plan to implement in the future.

Feasibility study of dual isotope PET

Positron emission tomography (PET) is a highly effective functional imaging modality. Unfortunately, PET cannot be easily used for dual-isotope imaging (which would allow for simultaneous investigation of two different biological processes), because positron-electron annihilation products from different tracers are indistinguishable in terms of energy. Methods have been proposed for dual-isotope PET based on different half-lives of the participating isotopes, however, those approaches are based on many assumptions concerning kinetic behavior of the tracers and may not always lead to optimal results. In this manuscript we propose another approach for dualisotope PET and investigate its effectiveness using GATE simulations. Our method requires that one of the two radioactive isotopes is a pure positron emitter while the second isotope emits an additional high-energy photon in a cascade (simultaneously) with positron emission. Measurement of this auxiliary photon in coincidence with the annihilation event allows us to identify the corresponding 511 keV photon pair as originating from the same isotope. Two list-mode datasets are created: a primary dataset that contains all detected 511 keV photon pairs from both isotopes; and a second, auxiliary (much smaller) dataset that contains only those PET events for which coincident auxiliary gamma has also been detected. An image reconstructed from the auxiliary dataset reflects the distribution of the second radiotracer and serves as a prior for the reconstruction of the primary dataset. Our preliminary simulation study with partially overlapping 18F and 22Na radiotracer distributions suggest that this method allows for separation of two activities with relative errors less than 0.07.

EM reconstruction of dual isotope PET using staggered injections and prompt gamma positron emitters.

The aim of Dual Isotope Positron Emission Tomography (DIPET) is to create two independent images of two co-injected radiotracers. DIPET shortens the duration of the study, reduces patient discomfort and produces perfectly coregistered images. Its implementation is difficult because positron decay of any isotope creates only 511 keY photons. Recently, we have proposed a DIPET technique that uses a combination of radiotracer A which is a pure positron emitter (such as 18F or 11C) and radiotracer B in which positron decay is followed by the emission of a high-energy (HE) prompt gamma (such as 38K or 60Cu). Events that are detected as triple coincidences of HE gammas with the corresponding two 511 keY photons identify the line-of-responses (LORs) that originate from isotope B (~10% of all events). These LORs are used to separate the two intertwined distributions, using a dedicated image reconstruction algorithm based on Expectation Maximization (EM). However, this approach is suboptimal as it combines high-statistics data (radiotracers A and B), with low-statistics (radiotracer B) to create separate images of radiotracers A and B. We hypothesize that from the performance of DIPET could be improved if an independent estimate of radiotracer A distribution is available by using staggered injections method. Here we investigate this hypothesis by developing and testing an EM-based algorithm for image reconstruction of such data. Our algorithm is general, i.e. can reconstruct staggered data, positron and HE gamma data or a combination of both. Reconstruction of DIPET data obtained from Monte-Carlo simulations shows that the combined method provides the best results in terms of separation of images corresponding to radiotracers A and B. Our EM algorithm provides not only good image separation for the combined method but also when only staggered injections or only prompt gamma data are used.

Simulations of a micro-PET system based on liquid xenon

The imaging performance of a high-resolution preclinical micro-positron emission tomography (micro-PET) system employing liquid xenon (LXe) as the gamma-ray detection medium was simulated. The arrangement comprises a ring of detectors consisting of trapezoidal LXe time projection ionization chambers and two arrays of large area avalanche photodiodes for the measurement of ionization charge and scintillation light. A key feature of the LXePET system is the ability to identify individual photon interactions with high energy resolution and high spatial resolution in three dimensions and determine the correct interaction sequence using Compton reconstruction algorithms. The simulated LXePET imaging performance was evaluated by computing the noise equivalent count rate, the sensitivity and point spread function for a point source according to the NEMA-NU4 standard. The image quality was studied with a micro-Derenzo phantom. Results of these simulation studies included noise equivalent count rate peaking at 1326 kcps at 188 MBq (705 kcps at 184 MBq) for an energy window of 450-600 keV and a coincidence window of 1 ns for mouse (rat) phantoms. The absolute sensitivity at the center of the field of view was 12.6%. Radial, tangential and axial resolutions of (22)Na point sources reconstructed with a list-mode maximum likelihood expectation maximization algorithm were ≤0.8 mm (full-width at half-maximum) throughout the field of view. Hot-rod inserts of <0.8 mm diameter were resolvable in the transaxial image of a micro-Derenzo phantom. The simulations show that a LXe system would provide new capabilities for significantly enhancing PET images.