Search

Techniques to estimate dynamic parameters from the data acquired during a single rotation of a standard (single, dual and triple head) SPECT camera are being investigated. The performance of the constrained linear least squares (C-LS) method is being assessed. This method uses a simple assumption that the activity within each pixel at a given projection angle is less than or equal to the activity seen at the prior angle. In simulations, we have used an analytical heart model with single and bi-exponential activity decrease, with and without background activity and statistical noise in the data. Studies using a single, a 90 [degrees] dual and a 120 [degrees] triple head camera were modeled with a total acquisition time of 20min, performing rotation over 180 [degrees] . These simulations were followed by experimental scans of our dynamic phantom used with and without attenuation and subsequently a preliminary study involving a human subject was performed. Dynamic images from simulations and phantom experiments reconstruct well with an accuracy of reconstructed half-lives for most reconstructions within a 20% error range. Some streaking artifacts which are evident in images corresponding to early reconstruction times subside upon incorporating further constraints and regularizations.

We have designed and built a dynamic heart-in-thorax phantom to be used as the primary tool for experimental verification of a quantitative dynamic functional imaging method we are developing for standard rotating single photon emission computed tomography (SPECT) cameras. The phantom consists of two independent parts (i) a dynamic heart model which allows up to two 'defects' to be mounted inside it and, (ii) a non-uniform thorax model with lungs and spinal cord. The principle behind the design of the phantom is the fact that the washout of a tracer by dilution is governed by a linear first order equation, the same type of equation as is used to model time-activity distribution in myocardial viability studies. Tests of the dynamic performance of the phantom using the planar scanning mode have confirmed the validity of these assumptions. Also the preliminary results obtained from dynamic data acquired using a simplified version of the phantom in SPECT mode show that the values of characteristic times could be experimentally determined and that these values agreed well with the values preset on the phantom. We consider that the phantom is ready for use in the development of the dynamic SPECT method.

A new image analysis method, called contiguous volume analysis, has been developed to automatically extract 3D information from emission images. The method considers volumes of activity and display,s data about them in a format which allows quantitative image comparison. This method of numerical analysis enables us to show, for example, whether or not information has been gained, lost or changed through the use of different filters and different attenuation and scatter correction, and reconstruction algorithms. Since the analysis method is consistent with a visual inspection of the data, intuitive insights into the meaning of the data are possible, allowing a better understanding of the effects of the different image processing techniques on the images. The method can be used to find patterns of activity in sets of images, and may be used to quantify noise, allowing an objective determination of which volumes in an image are meaningful.

A simple configuration for a transmission source for single photon emission computed tomography (SPECT) is proposed, which utilizes a series of collimated line sources parallel to the axis of rotation of the camera. The detector is equipped with a standard parallel hole collimator. We have demonstrated that this type of source configuration can be used to generate sufficient data for the reconstruction of the attenuation map when using 8-10 line sources spaced by 3.5-4.5cm for a 30x40cm detector placed 65cm from the sources. Transmission data for a nonuniform thorax phantom was simulated, then binned and reconstructed using filtered backprojection (FBP) and iterative methods. It was found that optimum maps were obtained with data binned into 2-3 bins and FBP reconstruction. The activity in the source was investigated for uniform and exponential activity distributions. The effect of gaps and overlaps of the neighboring fan beams was also studied. A prototype of the line source system has been built and the experimental verification of the simulation results has started.

We have developed a method for analytically calculating photon distributions in SPECT projections. This method models primary photon distributions as well as first and second order Compton scattering and Rayleigh scattering. It uses no free fitting parameters and so the projections produced are completely determined by the characteristics of the SPECT camera system, the energy of the isotope, an estimate of the source distribution and an attenuation map of the scattering object. The method was previously validated by comparison with Monte Carlo simulations and we are now verifying its accuracy with respect to phantom experiments. We have performed experiments using a Siemens MS3 SPECT camera system for a 99mTc point source (2mm in diameter) within a homogeneous water bath, a small spherical source (1cm in diameter) within a homogeneous water cylinder, the small spherical source within a non-homogeneous medium consisting of air and water, and the small spherical source together with a larger spherical source (2.8cm in diameter) in a non-homogeneous air and water medium. Our technique reproduces well the distribution of photons in the experimentally acquired projections for all of these cases.

Dynamic SPECT (dSPECT) is an image reconstruction method capable of determining three dimensional regional estimates of physiological kinetic rates while still maintaining the simple acquisition protocol of a single, slow camera rotation. Through the use of inequality constraints in a least squares optimization routine, we are able to determine both the rate of tracer uptake and subsequent tracer washout over the acquisition time of the SPECT scan. Simulations modelling different camera acquisition geometries and with variable dual exponential kinetic parameters and model complexity (eg, statistical noise, object attenuation) were performed. It was found that kinetic parameters were accurate to within 50% of the true values for the case of fast uptake and slow washout for a single head acquisition and improved to within 95% when acquired with a triple, head camera. Experimental phantom tests corraborated these results and are shown to have an accuracy of about 90%. A preliminary patient study of renal dynamics has also been performed and compared with conventional planar dynamic imaging. Results from this study indicate that the dSPECT reconstruction method is able to determine kinetic information with accuracy comparable to that of conventional planar dynamic imaging methods while providing additional three dimensional time behaviour information.

Simultaneous dual-isotope SPECT scans are clinically desirable because they require less acquisition time than sequential scans, and provide perfect spatial co-registration of the acquired images. However, they often suffer from extensive contamination caused by cross-talk, the detection of photons from one isotope in the energy window of the second. This contamination can make it difficult to distinguish between the detected distributions of the two isotopes, particularly when the emitted photons are of similar energies. The contamination can be reduced by using asymmetric windows but doing so does not completely remove the effect of cross-talk and there is a corresponding loss in sensitivity due to the reduction in size and/or the change in location of the energy window. We propose to distinguish the source distributions of the two isotopes by fitting calculated projections of potential sites of activity to the acquired projection data. The initial estimates of the potential sites are assumed to be identified using a reconstruction of the uncorrected data. Once the activities of the sources have been determined, corrected projections, if desired, can be generated allowing for the reconstruction of cross-talk free images. We have performed an initial feasibility study of this approach using small [sup.123]I and [sup.99m]Tc sources in homogeneous and inhomogeneous media filled with [sup.99m]Tc. The results show good agreement between the true source activity and our estimated activities.