Neal H. Clinthorne

Grouped-coordinate ascent algorithms for penalized-likelihood transmission image reconstruction

Presents a new class of algorithms for penalized-likelihood reconstruction of attenuation maps from low-count transmission scans. We derive the algorithms by applying to the transmission log-likelihood a version of the convexity technique developed by De Pierro for emission tomography. The new class includes the single-coordinate ascent (SCA) algorithm and Langes convex algorithm for transmission tomography as special cases. The new grouped-coordinate ascent (GCA) algorithms in the class overcome several limitations associated with previous algorithms. (1) Fewer exponentiations are required than in the transmission maximum likelihood-expectation maximization (ML-EM) algorithm or in the SCA algorithm. (2) The algorithms intrinsically accommodate nonnegativity constraints, unlike many gradient-based methods. (3) The algorithms are easily parallelizable, unlike the SCA algorithm and perhaps line-search algorithms. We show that the GCA algorithms converge faster than the SCA algorithm, even on conventional workstations. An example from a low-count positron emission tomography (PET) transmission scan illustrates the method.

Penalized weighted least-squares image reconstruction for dual energy X-ray transmission tomography

Presents a dual-energy (DE) transmission computed tomography (CT) reconstruction method. It is statistically motivated and features nonnegativity constraints in the density domain. A penalized weighted least squares (PWLS) objective function has been chosen to handle the non-Poisson noise added by amorphous silicon (aSi:H) detectors. A Gauss-Seidel algorithm has been used to minimize the objective function. The behavior of the method in terms of bias/standard deviation tradeoff has been compared to that of a DE method that is based on filtered back projection (FBP). The advantages of the DE PWLS method are largest for high noise and/or low flux cases. Qualitative results suggest this as well. Also, the reconstructed images of an object with opaque regions are presented. Possible applications of the method are: attenuation correction for positron emission tomography (PET) images, various quantitative computed tomography (QCT) methods such as bone mineral densitometry (BMD), and the removal of metal streak artifacts.

Preconditioning methods for improved convergence rates in iterative reconstructions

Because of the characteristics of the tomographic inversion problem, iterative reconstruction techniques often suffer from poor convergence rates-especially at high spatial frequencies. By using preconditioning methods, the convergence properties of most iterative methods can be greatly enhanced without changing their ultimate solution. To increase reconstruction speed, spatially invariant preconditioning filters that can be designed using the tomographic system response and implemented using 2-D frequency-domain filtering techniques have been applied. In a sample application, reconstructions from noiseless, simulated projection data, were performed using preconditioned and conventional steepest-descent algorithms. The preconditioned methods demonstrated residuals that were up to a factor of 30 lower than the assisted algorithms at the same iteration. Applications of these methods to regularized reconstructions from projection data containing Poisson noise showed similar, although not as dramatic, behavior.

Model-based estimation for dynamic cardiac studies using ECT

The authors develop a strategy for joint estimation of physiological parameters and myocardial boundaries using ECT (emission computed tomography). They construct an observation model to relate parameters of interest to the projection data and to account for limited ECT system resolution and measurement noise. The authors then use a maximum likelihood (ML) estimator to jointly estimate all the parameters directly from the projection data without reconstruction of intermediate images. They also simulate myocardial perfusion studies based on a simplified heart model to evaluate the performance of the model-based joint ML estimator and compare this performance to the Cramer-Rao lower bound. Finally, the authors discuss model assumptions and potential uses of the joint estimation strategy.

List-mode maximum likelihood reconstruction of Compton scatter camera images in nuclear medicine

A Maximum Likelihood (ML) image reconstruction technique using list-mode data has been applied to Compton scattering camera imaging. List-mode methods are appealing in Compton camera image reconstruction because the total number of data elements in the list (the number of detected photons) is significantly smaller than the number of possible combinations of position and energy measurements, leading to a much smaller problem than that faced by traditional iterative reconstruction techniques. For a realistic size device, the number of possible detector bins can be as large as 10 billion per pixel of the image space, while the number of counted photons would typically be a very small fraction of that. The primary difficulty in applying the list-mode technique is in determining the parameters which describe the response of the imaging system. In this work, a simple method for determining the required system matrix coefficients is employed, in which a back-projection is performed in list-mode, and response coefficients determined for only tallied pixels. Projection data has been generated for a representative Compton camera system by Monte Carlo simulation for disk sources with hot and cold spots and energies of 141, 364, and 511 keV, and reconstructions performed.

SPRINT II: a second generation single photon ring tomograph

SPRINT II is a stationary detector ring tomograph designed for brain imaging. Eleven two-dimensional sodium iodide camera modules that use maximum-likelihood position logic are arranged in a 50-cm-diameter ring with a scintillator packing fraction of 96%. A 34-cm-diameter rotating lead aperture ring containing either 10 or 12 slits is used for in-plane collimation, while the z-axis collimator is constructed of parallel lead foil rings. The field of view is 22 cm in diameter by 12 cm long. Sensitivity is 10 count/s/muCi for an on-axis (99m)Tc point source and 8500 count/s/muCi/cm(3) for 19.8-cm-diameter by 6.2-cm-long cylindrical source. Longitudinal resolution is 10 mm FWHM, and in-plane resolution varies from 8 mm FWHM on-axis to 5 mm FWHM at a radius of 9 cm. Performance results are presented.

C-SPRINT: a prototype Compton camera system for low energy gamma ray imaging

An electronically-collimated imaging system is being built using pixellated, low-noise, position-sensitive silicon as the first detector, and a sodium iodide scintillation detector ring as the second detector. The system consists of a single 3/spl times/3/spl times/0.1 cm/sup 3/ silicon pad detector module with 1 keV FWHM (noise-limited) energy resolution centered at the front face of a 50 cm diameter, 10 cm long NaI detector annulus. Custom acquisition and timing electronics have been manufactured to minimize system dead time. Monte Carlo modeling is used to predict system sensitivity and position resolution. Simulations using the existing setup show angular uncertainties of 4.1/spl deg/ and 2.1/spl deg/ FWHM for /sup 99m/Tc and /sup 131/I point sources, respectively (7.2 mm and 3.7 mm at 10 cm). Sensitivity can be improved by more than a factor of a hundred over the existing setup by stacking five 1 mm thick 9/spl times/9 cm/sup 2/ silicon arrays and redesigning the second detector geometry to accept a wider range of scattering angles. Lower bound calculations show that our electronically-collimated camera system challenges current mechanically-collimated systems for both /sup 99m/Tc and /sup 131/I despite the deleterious effects of Doppler broadening. Preliminary measurements show a timing resolution of 41 ns FWHM between the silicon detector and a single SPRINT module.

Regularized emission image reconstruction using imperfect side information

The authors report a preliminary investigation of a spatially variant penalized-likelihood method for tomographic image reconstruction based on a Gibbs penalty. The penalty weights are determined from structural side information, such as the locations of anatomical boundaries in high-resolution magnetic resonance images. Such side information will be imperfect in practice, and a simple simulation demonstrates the importance of accounting for the errors in boundary locations. The authors discuss methods for prescribing the penalty weights when the side information is noisy. Simulation results suggest that even imperfect side information is useful for guiding spatially variant regularization.<<ETX>>

Improved modeling of system response in list mode EM reconstruction of Compton scatter camera images

An improved List Mode EM method for reconstructing Compton scattering camera images has been developed. First, an approximate method for computation of the spatial variation in the detector sensitivity has been derived and validated by Monte Carlo computation. A technique for estimating the relative weight of system matrix coefficients for each gamma in the list has also been employed, as has a method for determining the relative probabilities of emission having some from pixels tallied in each list-mode back-projection. Finally, a technique has been developed for modeling the effects of Doppler broadening and finite detector energy resolution on the relative weights for pixels neighbor to those intersected by the back-projection, based on values for the FWHM of the spread in the cone angle computed by Monte Carlo. Memory issues typically associated with list mode reconstruction are circumvented by storing only a list of the pixels intersected by the back-projections, and computing the weights of the neighboring pixels at each iteration step. Simulated projection data has been generated for a representative Compton camera system (CSPRINT) for several source distributions and reconstructions performed. Reconstructions have also been performed for experimental data for distributed sources.

A Hybrid Maximum Likelihood Position Computer for Scintillation Cameras

Maximum likelihood (ML) estimators offer advantages of improved spatial resolution and linearity over traditional position estimates in position sensitive detectors. We have constructed a two board, multibus based hybrid position computer capable of performing the ML estimate at SPECT countrates. In addition, the board can implement any estimate linear in the photomultiplier tube outputs.