PubMed Health⌕ Search

Biomedical subjects

F J Beekman

Publications and source records attributed to F J Beekman.

At least 19 recordsLinked to original sources

Evaluation of accelerated iterative x-ray CT image reconstruction using floating point graphics hardware.

Statistical reconstruction methods offer possibilities to improve image quality as compared with analytical methods, but current reconstruction times prohibit routine application in clinical and micro-CT. In particular, for cone-beam x-ray CT, the use of graphics hardware has been proposed to accelerate the forward and back-projection operations, in order to reduce reconstruction times. In the past, wide application of this texture hardware mapping approach was hampered owing to limited intrinsic accuracy. Recently, however, floating point precision has become available in the latest generation commodity graphics cards. In this paper, we utilize this feature to construct a graphics hardware accelerated version of the ordered subset convex reconstruction algorithm. The aims of this paper are (i) to study the impact of using graphics hardware acceleration for statistical reconstruction on the reconstructed image accuracy and (ii) to measure the speed increase one can obtain by using graphics hardware acceleration. We compare the unaccelerated algorithm with the graphics hardware accelerated version, and for the latter we consider two different interpolation techniques. A simulation study of a micro-CT scanner with a mathematical phantom shows that at almost preserved reconstructed image accuracy, speed-ups of a factor 40 to 222 can be achieved, compared with the unaccelerated algorithm, and depending on the phantom and detector sizes. Reconstruction from physical phantom data reconfirms the usability of the accelerated algorithm for practical cases.

Algorithms↗

Parallel statistical image reconstruction for cone-beam x-ray CT on a shared memory computation platform.

Statistical reconstruction methods offer possibilities of improving image quality as compared to analytical methods, but current reconstruction times prohibit routine clinical applications. To reduce reconstruction times we have parallelized a statistical reconstruction algorithm for cone-beam x-ray CT, the ordered subset convex algorithm (OSC), and evaluated it on a shared memory computer. Two different parallelization strategies were developed: one that employs parallelism by computing the work for all projections within a subset in parallel, and one that divides the total volume into parts and processes the work for each sub-volume in parallel. Both methods are used to reconstruct a three-dimensional mathematical phantom on two different grid densities. The reconstructed images are binary identical to the result of the serial (non-parallelized) algorithm. The speed-up factor equals approximately 30 when using 32 to 40 processors, and scales almost linearly with the number of cpus for both methods. The huge reduction in computation time allows us to apply statistical reconstruction to clinically relevant studies for the first time.

Algorithms↗

Evaluation of the ordered subset convex algorithm for cone-beam CT.

Statistical methods for image reconstruction such as maximum likelihood expectation maximization (ML-EM) are more robust and flexible than analytical inversion methods and allow for accurate modelling of the photon transport and noise. Statistical reconstruction is prohibitively slow when applied to clinical x-ray cone-beam CT due to the large data sets and the high number of iterations required for reconstructing high resolution images. One way to reduce the reconstruction time is to use ordered subsets of projections during the iterations, which has been successfully applied to fan-beam x-ray CT. In this paper, we quantitatively analyse the use of ordered subsets in concert with the convex algorithm for cone-beam x-ray CT reconstruction, for the case of circular acquisition orbits. We focus on the reconstructed image accuracy of a 3D head phantom. Acceleration factors larger than 300 were obtained with errors smaller than 1%, with the preservation of signal-to-noise ratio. Pushing the acceleration factor towards 600 by using an increasing number of subsets increases the reconstruction error up to 5% and significantly increases noise. The results indicate that the use of ordered subsets can be extremely useful for cone-beam x-ray CT.

Algorithms↗

Experimental validation of a rapid Monte Carlo based micro-CT simulator.

We describe a newly developed, accelerated Monte Carlo simulator of a small animal micro-CT scanner. Transmission measurements using aluminium slabs are employed to estimate the spectrum of the x-ray source. The simulator incorporating this spectrum is validated with micro-CT scans of physical water phantoms of various diameters, some containing stainless steel and Teflon rods. Good agreement is found between simulated and real data: normalized error of simulated projections, as compared to the real ones, is typically smaller than 0.05. Also the reconstructions obtained from simulated and real data are found to be similar. Thereafter, effects of scatter are studied using a voxelized software phantom representing a rat body. It is shown that the scatter fraction can reach tens of per cents in specific areas of the body and therefore scatter can significantly affect quantitative accuracy in small animal CT imaging.

Algorithms↗

Accelerated simulation of cone beam X-ray scatter projections.

Monte Carlo (MC) methods can accurately simulate scatter in X-ray imaging. However, when low noise scatter projections have to be simulated these MC simulations tend to be very time consuming. Rapid computation of scatter estimates is essential for several applications. The aim of the work presented in this paper is to speed up the estimation of noise-free scatter projections while maintaining their accuracy. Since X-ray scatter projections are often rather smooth, an approach is chosen whereby a short MC simulation is combined with a data fitting program that is robust to projection truncation and noise. This method allows us to estimate the smooth scatter projection rapidly. The speed-up and accuracy achieved by using the fitting algorithm were validated for the projection simulation of a small animal X-ray CT system. The acceleration that can be obtained over standard MC simulations is typically two orders of magnitude, depending on the accuracy required. The proposed approach may be useful for rapid simulation of patient and animal studies and for correction of the image-degrading effects of scatter in tomography.

Algorithms↗

Simultaneous (99m)Tc/(201)Tl dual-isotope SPET with Monte Carlo-based down-scatter correction.

In simultaneous technetium-99m/thallium-201 dual-isotope (DI) single-photon emission tomography (SPET), down-scatter of (99m)Tc photons contaminates the (201)Tl image, which leads to a decrease in lesion contrast and loss of quantitative accuracy. Correction for down-scatter can be achieved by first reconstructing the (99m)Tc activity distribution. Subsequently, the (99m)Tc down-scatter in the (201)Tl photopeak window is simulated and used for correction during iterative reconstruction of the (201)Tl image. In this work, the down-scatter projections are calculated using a dedicated Monte Carlo simulator which is able to efficiently model the detection of lead X-rays from the collimator. An anthropomorphic torso phantom with a cardiac insert with and without cold lesions was used for evaluation of the proposed method. Excellent agreement in lesion contrast and quantitative accuracy was found between the down-scatter corrected DI-SPET (201)Tl image and the virgin (i.e. separately acquired) (201)Tl image, in particular when the effects of lead X-rays were included. Compensation for the noise added by down-scatter to the (201)Tl image can be achieved by using a 15% lower dose of (99m)Tc, a 15% increase in scan time and a 12% increase in (201)Tl dose. In conclusion, the Monte Carlo-based down-scatter correction recovers lesion contrast and quantitative accuracy in DI-SPET (201)Tl images almost perfectly. In addition, degradations due to the added noise of down-scatter in simultaneous DI-SPET can be prevented by slight adaptations to the data acquisition protocol.

Humans↗

Choice of collimator for cardiac SPET when resolution compensation is included in iterative reconstruction.

In clinical cardiac single-photon emission tomography (SPET) studies, collimators of different spatial resolution and geometric efficiency are available for imaging. In selecting the appropriate collimator for clinical use, there is a trade-off between spatial resolution, which can limit the contrast of the reconstructed image, and detection efficiency, which determines the noise in the image. Our objective was to assess which collimator is best suited for cardiac SPET when reconstruction is performed with and without compensation for distance-dependent resolution (CDR). The dynamic MCAT thorax phantom was used to simulate 180 degree technetium-99m cardiac data, acquired using either a general-purpose (GP) or high-resolution (HR) collimator. For GP and HR, the resolution at 15 cm was 11.5 mm and 9.5 mm respectively, and the corresponding relative efficiency was 1.0 and 0.52 respectively. Distance-dependent resolution, attenuation and noise were included in the projection data; scatter was not included. Ordered subsets expectation maximisation reconstruction (subset size 4) was performed with and without CDR. Results were evaluated by comparing the myocardial recovery coefficient and contrast between myocardium and ventricle relative to the original phantom, each plotted for different noise levels corresponding to increasing iteration number. The study demonstrated that, without CDR, HR gave the best results. However, for any given noise level with CDR, GP gave superior recovery and contrast. These findings were confirmed in a physical phantom study. Results suggest that improved reconstruction can be achieved using a GP collimator in combination with resolution compensation.

Algorithms↗

Rapid SPECT simulation of downscatter in non-uniform media.

A rotation-based Monte Carlo (MC) simulation method (RMC) has been developed, designed for rapid calculation of downscatter through non-uniform media in SPECT. A possible application is downscatter correction in dual isotope SPECT. With RMC, only a fraction of all projections of a SPECT study have to be MC simulated in a standard manner. The other projections can be estimated rapidly using the results of these standard MC calculations. For efficiency, approximations have to be made in RMC with regard to the final scatter angle of the detected photons. Further speed-up is obtained by combining RMC with convolution-based forced detection (CFD) instead of forced detection (FD), which is a more common variance reduction technique for MC. The RMC method was compared with standard MC for 99mTc downscatter in a 201Tl window (72 keV+/-10%) using a digital thorax phantom. The resulting scatter projections are in good agreement (maximum bias a few per cent of the largest value in the projection), but RMC with CFD is about three orders in magnitude faster than standard MC with FD and up to 25 times faster than standard MC with CFD. Using RMC combined with CFD, the generation of 64 almost noise-free downscatter projections (64 x 64) takes only a couple of minutes on a 500 MHz Pentium processor. Therefore, rotation-based Monte Carlo could serve as a practical tool for downscatter correction schemes in dual isotope SPECT.

Humans↗

Maximum likelihood estimation of cerebral blood flow in dynamic susceptibility contrast MRI.

For quantification of cerebral blood flow (CBF) using dynamic susceptibility contrast magnetic resonance imaging (DSC-MRI), knowledge of the tissue response function is necessary. To obtain this, the tissue contrast passage measurement must be corrected for the arterial input. This study proposes an iterative maximum likelihood expectation maximization (ML-EM) algorithm for this correction, which takes into account the noise in T2- or T2*-weighted image sequences. The ML-EM algorithm does not assume a priori knowledge of the shape of the response function; it automatically corrects for arrival time offsets and inherently yields positive response values. The results on synthetic image sequences are presented, for which the recovered flow values and the response functions are in good agreement with their expectation values. The method is illustrated by calculating the gray and white matter flow in a clinical example.

Cerebrovascular Circulation↗

The influence of backscatter material on 99mTc and 201Tl line source responses.

SPECT projections are contaminated by scatter, resulting in reduced image contrast and quantitative errors. When tissue is present behind the source, some of the detected photons backscatter via this tissue. Particularly in dual-isotope SPECT and in combined emission-transmission SPECT, backscatter constitutes a major part of the down-scatter contamination in lower-energy windows. In this paper, the effects of backscatter material were investigated. Planar images of 99mTc and 201Tl line sources between varying numbers of Perspex slabs were analysed using the photopeak windows and various scatter windows. In the 99mTc photopeak window no significant change in total counts due to backscatter material was measured. In the 201Tl photopeak window an increase of about 10% in total counts was observed. In the scatter windows an even more explicit influence of backscatter material was measured. For instance, at a forward depth of 10 cm, total counts of a 99mTc source detected in the 72 keV window eventually doubled with increasing backscatter material, compared with the situation without backscatter material. The backscatter contribution plateaued when more than 5-10 cm of scatter material was placed behind the source. In conclusion, backscatter should be taken into account, particularly in model-based down-scatter correction methods in dual-isotope SPECT and combined emission-transmission SPECT.

Gadolinium↗

Efficient SPECT scatter calculation in non-uniform media using correlated Monte Carlo simulation.

Accurate simulation of scatter in projection data of single photon emission computed tomography (SPECT) is computationally extremely demanding for activity distribution in non-uniform dense media. This paper suggests how the computation time and memory requirements can be significantly reduced. First the scatter projection of a uniform dense object (P(SDSE)) is calculated using a previously developed accurate and fast method which includes all orders of scatter (slab-derived scatter estimation), and then P(SDSE) is transformed towards the desired projection P which is based on the non-uniform object. The transform of P(SDSE) is based on two first-order Compton scatter Monte Carlo (MC) simulated projections. One is based on the uniform object (P(u)) and the other on the object with non-uniformities (P(nu)). P is estimated by P = P(SDSE) P(nu)/P(u). A tremendous decrease in noise in P is achieved by tracking photon paths for P(nu) identical to those which were tracked for the calculation of P(u) and by using analytical rather than stochastic modelling of the collimator. The method was validated by comparing the results with standard MC-simulated scatter projections (P) of 99mTc and 201Tl point sources in a digital thorax phantom. After correction, excellent agreement was obtained between P and P. The total computation time required to calculate an accurate scatter projection of an extended distribution in a thorax phantom on a PC is a only few tens of seconds per projection, which makes the method attractive for application in accurate scatter correction in clinical SPECT. Furthermore, the method removes the need of excessive computer memory involved with previously proposed 3D model-based scatter correction methods.

Computer Simulation↗

Estimation of the depth-dependent component of the point spread function of SPECT.

The point spread function (PSF) of a gamma camera describes the photon count density distribution at the detector surface when a point source is imaged. Knowledge of the PSF is important for computer simulation and accurate image reconstruction of single photon emission computed tomography (SPECT) images. To reduce the number of measurements required for PSF characterization and the amount of computer memory to store PSF tables, and to enable generalization of the PSF to different collimator-to-source distances, the PSF may be modeled as the two-dimensional (2D) convolution of the depth-dependent component which is free of detector blurring (PSF(ideal)) and the distance-dependent detector response. Owing to limitations imposed by the radioactive strength of point sources, extended sources have to be used for measurements. Therefore, if PSF(ideal) is estimated from measured responses, corrections have to be made for both the detector blurring and for the extent of the source. In this paper, an approach based on maximum likelihood expectation-maximization (ML-EM) is used to estimate PSF(ideal). In addition, a practical measurement procedure which avoids problems associated with commonly used line-source measurements is proposed. To decrease noise and to prevent nonphysical solutions, shape constraints are applied during the estimation of PSF(ideal). The estimates are generalized to depths other than those which have been measured and are incorporated in a SPECT simulator. The method is validated for Tc-99m and T1-201 by means of measurements on physical phantoms. The corrected responses have the desired shapes and simulated responses closely resemble measured responses. The proposed methodology may, consequently, serve as a basis for accurate three-dimensional (3D) SPECT reconstruction.

Algorithms↗

Dual matrix ordered subsets reconstruction for accelerated 3D scatter compensation in single-photon emission tomography.

Three-dimensional (3D) iterative maximum likelihood expectation maximization (ML-EM) algorithms for single-photon emission tomography (SPET) are capable of correcting image-degrading effects of non-uniform attenuation, distance-dependent camera response and patient shape-dependent scatter. However, the resulting improvements in quantitation, resolution and signal-to-noise ratio (SNR) are obtained at the cost of a huge computational burden. This paper presents a new acceleration method for ML-EM: dual matrix ordered subsets (DM-OS). DM-OS combines two acceleration methods: (a) different matrices for projection and back-projection and (b) ordered subsets of projections. DM-OS was compared with ML-EM on simulated data and on physical thorax phantom data, for both 180 degrees and 360 degrees orbits. Contrast, normalized standard deviation and mean squared error were calculated for the digital phantom experiment. DM-OS resulted in similar image quality to ML-EM, even for speed-up factors of 200 compared to ML-EM in the case of 120 projections. The thorax phantom data could be reconstructed 50 times faster (60 projections) using DM-OS with preservation of image quality. ML-EM and DM-OS with scatter compensation showed significant improvement of SNR compared to ML-EM without scatter compensation. Furthermore, inclusion of complex image formation models in the computer code is simplified in the case of DM-OS. It is thus shown that DM-OS is a fast and relatively simple algorithm for 3D iterative scatter compensation, with similar results to conventional ML-EM, for both 180 degrees and 360 degrees acquired data.

Algorithms↗

Selection of task-dependent diffusion filters for the post-processing of SPECT images.

Iterative reconstruction from single photon emission computed tomography (SPECT) data requires regularization to avoid noise amplification and edge artefacts in the reconstructed image. This is often accomplished by stopping the iteration process at a relatively low number of iterations or by post-filtering the reconstructed image. The aim of this paper is to develop a method to automatically select an optimal combination of stopping iteration number and filters for a particular imaging situation. To this end different error measures between the distribution of a phantom and a corresponding filtered SPECT image are minimized for different iteration numbers. As a study example, simulated data representing a brain study are used. For post-reconstruction filtering, the performance of 3D linear diffusion (Gaussian filtering) and edge preserving 3D nonlinear diffusion (Catté scheme) is investigated. For reconstruction methods which model the image formation process accurately, error measures between the phantom and the filtered reconstruction are significantly reduced by performing a high number of iterations followed by optimal filtering compared with stopping the iterative process early. Furthermore, this error reduction can be obtained over a wide range of iteration numbers. Only a negligibly small additional reduction of the errors is obtained by including spatial variance in the filter kernel. Compared with Gaussian filtering, Catté diffusion can further reduce the error in some cases. For the examples considered, using accurate image formation models during iterative reconstruction is far more important than the choice of the filter.

Biophysical Phenomena↗

Accelerated iterative transmission CT reconstruction using an ordered subsets convex algorithm.

Iterative maximum likelihood (ML) transmission computed tomography algorithms have distinct advantages over Fourier-based reconstruction, but unfortunately require increased computation time. The convex algorithm [1] is a relatively fast iterative ML algorithm but it is nevertheless too slow for many applications. Therefore, an acceleration of this algorithm by using ordered subsets of projections is proposed [ordered subsets convex algorithm (OSC)]. OSC applies the convex algorithm sequentially to subsets of projections. OSC was compared with the convex algorithm using simulated and physical thorax phantom data. Reconstructions were performed for OSC using eight and 16 subsets (eight and four projections/subset, respectively). Global errors, image noise, contrast recovery, and likelihood increase were calculated. Results show that OSC is faster than the convex algorithm, the amount of acceleration being approximately proportional to the number of subsets in OSC, and it causes only a slight increase of noise and global errors in the reconstructions. Images and image profiles of the reconstructions were in good agreement. In conclusion, OSC and the convex algorithm result in similar image quality but OSC is more than an order of magnitude faster.

Algorithms↗

Half-fanbeam collimators combined with scanning point sources for simultaneous emission-transmission imaging.

UNLABELLED: One type of SPECT system often used for simultaneous emission-transmission tomography is equipped with parallel-hole collimators, moving line sources (MLS) and electronic windows that move in synchrony with the sources. Although downscatter from the emission distribution is reduced by the use of the electronic window, this still can represent a sizable fraction of the transmitted counts. These systems have relatively poor spatial resolution and use costly transmission sources. METHODS: Using a two-head SPECT system, with heads at right angles, two 153Gd line sources (5800 MBq each) were replaced by two 153Gd point sources of only 750 MBq each and positioned to move along the focal lines of two half-fanbeam collimators. A suitable acquisition protocol for a moving point source (MPS) system was selected by considering the results of a simulation study. With this protocol, physical phantom experiments were conducted. RESULTS: Simulations showed that by using two half-fanbeam collimators, a gantry rotation of 90 degrees, such as used for 180 degrees acquisition with parallel-beam collimators for cardiac imaging, was insufficient. A gantry rotation of 180 degrees resulted in attenuation maps where only an area to the posterior of a 400-mm wide thorax phantom was affected by truncation. The MPS system had a 14.7 times higher sensitivity for transmission counts than the MLS system. Despite the smaller sources in the MPS system, the number of acquired transmission counts was a factor 1.91 times higher compared with the MLS system, resulting in reduced noise. The relative downscatter contribution from 99mTc (140 keV) in the 153Gd moving electronic window (100 keV) was reduced by a factor of 1.81. Transmission images of a rod phantom with segments containing acrylic rods of different diameters showed an improvement of resolution in favor of the MPS system from about 11 mm to about 6 mm (five instead of two segments of rods were clearly visible). In addition, the noise level in the MPS thorax transmission images was significantly lower. CONCLUSION: The MPS system has important advantages when compared with the MLS system. The use of low-activity point sources is economically beneficial when compared with line sources and reduces radiation exposure to staff and patients.

Equipment Design↗

A clinical perspective of accelerated statistical reconstruction.

Although the potential benefits of maximum likelihood reconstruction have been recognised for many years, the technique has only recently found widespread popularity in clinical practice. Factors which have contributed to the wider acceptance include improved models for the emission process, better understanding of the properties of the algorithm and, not least, the practicality of application with the development of acceleration schemes and the improved speed of computers. The objective in this article is to present a framework for applying maximum likelihood reconstruction for a wide range of clinically based problems. The article draws particularly on the experience of the three authors in applying an acceleration scheme involving use of ordered subsets to a range of applications. The potential advantages of statistical reconstruction techniques include: (a) the ability to better model the emission and detection process, in order to make the reconstruction converge to a quantitative image, (b) the inclusion of a statistical noise model which results in better noise characteristics, and (c) the possibility to incorporate prior knowledge about the distribution being imaged. The great flexibility in adapting the reconstruction for a specific model results in these techniques having wide applicability to problems in clinical nuclear medicine.

Algorithms↗