Compression and reconstruction of sorted PET listmode data.

BACKGROUND: In nuclear medicine data can be stored in histogram or listmode format. The most popular histogram format is the planar projection format. Due to the increase in detector blocks, the improved energy resolution and the trends towards time of flight, dynamic and gated imaging, it can be more appropriate to store the data in listmode format. The size of the storage in this format increases linearly with the number of properties (positions, energy, time info) while the histogram format increases exponentially. However, the datasize of listmode data also increases linearly with the number of coincidences. Due to the high number of counts in 3D PET this will lead to very large datasets. Therefore a good compression algorithm for listmode data is very important. METHODS: A sorting and compression method is proposed to reduce the amount of space needed to store the listmode dataset. One event is represented by one number without any information loss compared to the original listmode file. The next step is to sort all events into an array of increasing numbers. These data are compressed by the gzip routine. One of the advantages of 3D PET listmode reconstructions is that they result in a more uniform resolution across the field of view (FOV), which is not always true for other reconstruction algorithms. This improved resolution is shown for the listmode data of a gamma camera operating in PET mode. RESULTS: First the effect of positional accuracy in the listmode dataset is evaluated by comparing resolution in the reconstructions. It is shown that the highest accuracy is not necessary and a significant reduction in the size of the dataset can be obtained prior to lossless compression. A further reduction can be obtained by using the proposed sorting and compression techniques. It is shown that the storage space decreases linearly with the logarithm of the number of coincidences. The compression obtained by different acquisition matrices was compared. Finally it is shown that the 3D listmode reconstruction of sorted listmode data is faster because of improved cache behaviour. The method can be applied to any kind of listmode data. The compression factors will improve when the ratio of measured events to possible events increases.

Geometrical importance sampling and pulse height tallies in GATE.

Monte Carlo simulations are widely used to study the behavior and detection of gamma photons in medical imaging devices. Such simulations are computationally expensive. This is why geometrical importance sampling, a variance reduction technique, was recently incorporated into the GEANT4 Monte Carlo code. In order to use this technique for single photon emission computed tomography (SPECT) imaging, it needed to be made compatible with pulse height tallies. These tallies correspond to the number of detected pulses in distinct energy bins, covering an energy spectrum relevant to SPECT. Since each pulse is the combination of different detector hits, the tally bin is not known until the end of an event. In an analog simulation (without variance reduction) this poses no problems as each detected hit can be stored and the pulse can be calculated at the end of each event. Geometrical importance sampling combined with Russian Roulette however introduces branches into the particle history, which results in a much more complicated pulse calculation. This work describes how pulse height tallies are adjusted to geometrical importance sampling and Russian Roulette within GATE, a medical imaging and simulation application based on GEANT4. The validation of this technique is done through SPECT simulations comparing the analog result with the new method.

Detection of spike and wave discharges in the cortical EEG of genetic absence epilepsy rats from Strasbourg.

Genetic absence epilepsy rats from Strasbourg (GAERS) are a strain of Wistar rats in which all animals present spontaneous occurrence of spike and wave discharges (SWD) in the cortical electroencephalogram (EEG). In this paper, we present a method for the detection of SWD, based on the key observation that SWD are quasi-periodic signals. A spectral-comb based analysis method is used to extract the fundamental frequency and the percentage of energy explained by the harmonic spectral components is subsequently used as a detection parameter. It is shown that a maximum sensitivity and specificity of up to 96 per cent can be achieved. We also compared the performance of this method with the methods presented in the literature and conclude that the surplus value of the novel detection method lies in the higher specificity that can be obtained in the analysis of long-term EEG fragments, which are contaminated by artefacts and contain large portions of slow-wave sleep.

Chemical-shift imaging utilizing the positional shifts along the readout gradient direction.

In this work, we describe a method that uses the linear phase acquired during the readout period due to chemical shift to generate individual magnetic resonance (MR) images of chemically shifted species. The method utilizes sets of Fourier (or k-space) data acquired with different directions of the readout gradient and a postprocessing algorithm to generate chemical shift images. The methodology is developed for both Cartesian data acquisition and for radial data acquisition. The method is presented here for two chemically shifted species but it can be extended to more species. In this work, we present the theory, show the results in phantoms and in human images, and discuss the artifacts and signal-to-noise ratio of the images obtained with the technique.

Resolution recovery for list-mode reconstruction in SPECT.

The purpose of the study was to evaluate the resolution recovery in the list-mode iterative reconstruction algorithm (LMIRA) for SPECT. In this study we compare the performance of the proposed method with other iterative resolution recovery methods for different noise levels. We developed an iterative reconstruction method which uses list-mode data instead of binned data. The new algorithm makes use of a more accurate model of the collimator structure. We compared the SPECT list-mode reconstruction with MLEM, OSEM and RBI, all including resolution recovery. For the evaluation we used Gaussian shaped sources with different FWHM at three different locations and three noise levels. For these distributions we calculated the reconstructed images for a different number of iterations. The absolute error for the reconstructed images was used to evaluate the performance. The performance of all four methods is comparable for the sources located in the centre of the field of view. For the sources located out of the centre, the error of the list-mode method is significantly lower than that of the other methods. Splitting the system model into a separate object-dependent and detector-dependent module gives us a flexible reconstruction method. With this we can very easily adapt the resolution recovery to different collimator types.

The validation of the finite difference method and reciprocity for solving the inverse problem in EEG dipole source analysis.

The performance of the finite difference reciprocity method (FDRM) to solve the inverse problem in EEG dipole source analysis is investigated in the analytically solvable three-shell spherical head model for a large set of test dipoles. The location error for a grid with 2 mm and 3 mm node spacing is in general, not larger than twice the internode distance, hence 4 mm and 6 mm, respectively. Increasing the number of scalp electrodes from 27 to 44 only marginally improves the location error. The orientation error is always smaller than 4 degrees for all the test dipoles considered. We have also compared the sensitivity to noise using FDRM in EEG dipole source analysis with the sensitivity to noise using the analytical expression for the forward problem. FDRM is not more sensitive to noise than the method using the analytical expression.

Dipole location errors in electroencephalogram source analysis due to volume conductor model errors.

An examination is made of dipole location errors in electroencephalogram (EEG) source analysis, due to not incorporating the ventricular system (VS), omitting a hole in the skull and underestimating skull conductivity. The simulations are performed for a large number of test dipoles in 3D using the finite difference method. The maximum dipole location error encountered, utilising 27 and 53 electrodes is 7.6 mm and 6.1 mm, respectively when omitting the VS, 5.6 mm and 5.2 mm, respectively when neglecting the hole in the skull, and 33.4 mm and 28.0 mm, respectively when underestimating skull conductivity. The largest location errors due to neglecting the VS can be found in the vicinity of the VS. The largest location errors due to omitting a hole can be found in the vicinity of the hole. At these positions the fitted dipoles are found close to the hole. When skull conductivity is underestimated, the dipole is fitted close to the skull-brain border in a radial direction for all test dipoles. It was found that the location errors due to underestimating skull conductivity are typically higher than those found due to neglecting the VS or neglecting a hole in the skull.

PACS and multimodality in medical imaging.

A PACS (Picture Archiving and Communication System) is a system that is able to store, exchange, display and manipulate images and associated diagnoses from any modality within a hospital in a timely and cost-effective way. Several developments, such as the DICOM standard, fast and convenient networking, and new storage solutions for large amounts of data, make the setup of such a PACS system possible. As the information acquired with various imaging modalities is then available and often complementary, it is desirable for the clinician to have a point-by-point spatial co-registration of images from different modalities in order to enable a synergistic use of the multimodality imaging of a patient for increased diagnostic accuracy. Various types of algorithms are available for the matching of medical images from the same or from different modalities. Co-registration algorithms based on voxel properties consist of a similarity or dissimilarity measure and an iterative or non-iterative method minimizing the dissimilarity or maximizing the similarity between the two images by a transformation of one image relative to the other.

Influence of measurement noise and electrode mislocalisation on EEG dipole-source localisation.

Measurement noise in the electro-encephalogram (EEG) and inaccurate information about the locations of the EEG electrodes on the head induce localisation errors in the results of EEG dipole source analysis. These errors are studied by performing dipole source localisation for simulated electrode potentials in a spherical head model, for a range of different dipole locations and for two different numbers (27 and 148) of electrodes. Dipole source localisation is performed by iteratively minimising the residual energy (RE), using the simplex algorithm. The ratio of the dipole localisation error (cm) to the noise level (%) of Gaussian measurement noise amounts to 0.15 cm/% and 0.047 cm/% for the 27 and 148 electrode configurations, respectively, for a radial dipole with 40% eccentricity The localisation error due to noise can be reduced by taking into account multiple time instants of the measured potentials. In the case of random displacements of the EEG electrodes, the ratio of dipole localisation errors to electrode location errors amounts to 0.78 cm-1 cm and 0.27 cm-1 cm for the 27 and 148 electrode configurations, respectively. It is concluded that it is important to reduce the measurement noise, and particularly the electrode mislocalisation, as the influence of the latter is not reduced by taking into account multiple time instants.

EEG dipole source localization using artificial neural networks.

Localization of focal electrical activity in the brain using dipole source analysis of the electroencephalogram (EEG), is usually performed by iteratively determining the location and orientation of the dipole source, until optimal correspondence is reached between the dipole source and the measured potential distribution on the head. In this paper, we investigate the use of feed-forward layered artificial neural networks (ANNs) to replace the iterative localization procedure, in order to decrease the calculation time. The localization accuracy of the ANN approach is studied within spherical and realistic head models. Additionally, we investigate the robustness of both the iterative and the ANN approach by observing the influence on the localization error of both noise in the scalp potentials and scalp electrode mislocalizations. Finally, after choosing the ANN structure and size that provides a good trade off between low localization errors and short computation times, we compare the calculation times involved with both the iterative and ANN methods. An average localization error of about 3.5 mm is obtained for both spherical and realistic head models. Moreover, the ANN localization approach appears to be robust to noise and electrode mislocations. In comparison with the iterative localization, the ANN provides a major speed-up of dipole source localization. We conclude that an artificial neural network is a very suitable alternative for iterative dipole source localization in applications where large numbers of dipole localizations have to be performed, provided that an increase of the localization errors by a few millimetres is acceptable.

Reconstruction of MR images from data acquired on a general nonregular grid by pseudoinverse calculation.

A minimum-norm least-squares image-reconstruction method for the reconstruction of magnetic resonance images from non-Cartesian sampled data is proposed. The method is based on a general formalism for continuous-to-discrete mapping and pseudoinverse calculation. It does not involve any regridding or interpolation of the data and therefore the methodology differs fundamentally from existing regridding-based methods. Moreover, the method uses a continuous representation of objects in the image domain instead of a discretized representation. Simulations and experiments show the possibilities of the method in both radial and spiral imaging. Simulations revealed that minimum-norm least-squares image reconstruction can result in a drastic decrease of artifacts compared with regridding-based reconstruction. Besides, both in vivo and phantom experiments showed that minimum-norm least-squares image reconstruction leads to contrast improvement and increased signal-to-noise ratio compared with image reconstruction based on regridding. As an appendix, an analytical calculation of the raw data corresponding to the well-known Shepp and Logan software head phantom is presented.

MRI-SPET and SPET-SPET brain co-registration: evaluation of the performance of eight different algorithms.

The aim of this study was to assess the accuracy and computing time needed for MRI-SPET and SPET-SPET brain co-registration using eight different algorithms (Hermes software from Nuclear Diagnostics Ltd run on a SUN Ultra Sparc 2) to determine the clinically most suitable algorithm. MRI-SPET co-registration was evaluated using phantom studies. To approximate clinical dual-headed SPET studies, a Hoffman brain phantom was filled with 99Tcm. For MRI imaging (1.5 Tesla), the phantom was filled with water and doped with Gd-DTPA for contrast enhancement. For both modalities, phantom images were acquired and reconstructed using a routine clinical protocol. MRI and SPET images were matched by Downhill Simplex minimization of the sum of absolute Count Differences (CD), the sum of the Square Root of absolute count differences (SR), the Difference in Shape between the binary masks (SD), the number of Sign Changes in the subtracted image (SC), the Variance of intensities between corresponding pixels (VAR), the sum of absolute count differences between the 2D- and 3D-Gradient images (2DG-3DG) and, finally, the standard deviation of the Uniformity Index (UI), that is the intensity ratio between spatially corresponding voxels. Six degrees of freedom were allowed (three translation and three rotation parameters, three scaling parameters were constrained). The accuracy of the matching process with these different similarity measures was evaluated via the residual mismatch between external markers. We found that CD, SR, VAR nad UI give the most accurate registration compared with the other similarity measures. For the evaluation of SPET-SPET co-registration, five 99Tcm-ECD brain perfusion SPET scans were performed with a dual-headed gamma camera. These studies were then manually misaligned, and subsequently re-aligned using the methods outlined above. For this application, CD, SR and VAR were also found to give the most accurate registration. For all of these algorithms, the computing time required was clinically acceptable (i.e. less than 10 min).

Non-invasive diagnostic assessment of extensive vertebral artery tortuosity with enlargement of the intervertebral foramen.

Two cases with an incidental finding of a widened intervertebral foramen due to extensive tortuosity of the vertebral artery are presented. Computed tomography angiography and magnetic resonance imaging/magnetic resonance angiography (MRA) both proved to be useful as non-invasive techniques to provide the diagnosis. In-plane saturation is a disadvantage seen with the two-dimensional time-of-flight MRA technique.

Two motion-detection algorithms for projection-reconstruction magnetic resonance imaging: theory and experimental verification.

In this paper, projection-reconstruction (PR) magnetic resonance (MR) imaging is considered. We present two new quantitative techniques that allow to check when motion was present during a PR MR experiment. No a priori information about the motion is required since only the measured MR signals are used in the calculations. Moreover, the proposed methods can be implemented on a standard personal computer or workstation. It is experimentally shown that both methods are able to detect motion intervals with an accuracy of one repetition time.

Magnetic resonance imaging and the reduction of motion artifacts: review of the principles.

Magnetic resonance (MR) imaging is a non-invasive diagnostic tool which is widely used nowadays. In this paper, the basic principles of MR imaging are explained and it is shown how images can be reconstructed in case of standard 2D Fourier Transform (2DFT) imaging. Several aspects of MR signal encoding are described. Unfortunately, motion of the patient during a magnetic resonance experiment often causes severe artifacts in the images. For example, in 2DFT imaging blurring and ghosting are seen and the appearance of motion artifacts remains one of the major drawbacks in MR imaging. Several methods to reduce motion artifacts in MR imaging have been proposed in the past. An overview of the principles on which these methods are based is given in this paper. Both post-processing methods and techniques that rely on gating or the use of alternative acquisition schemes such as projection reconstruction are discussed.

Thionine-Feulgen-Congo red staining of cervical smears for the BioPEPR image-analysis system.

A staining method developed for use with BioPEPR (Biological Precision Encoding and Pattern Recognition), an automated image-analysis system for cervical smears, is described. The stain is a combination of the Feulgen procedure, with thionine-SO2 as the Schiff reagent, and Congo red, which is used as a counterstain. The stain resulted in smears suitable both for microscopic diagnosis and for BioPEPR measurements made on photonegatives at a single wavelength 545 nm. A high level of reproducibility and accuracy of nuclear and cytoplasmic area measurements was obtained. Nuclear integrated optical density could be well measured and was shown to be useful in discriminating between normal and abnormal cells. Using a combination of morphologic features, a high level of cell classification accuracy was reached. The possibility of using the stain for more detailed studies is discussed.

Thionine-Feulgen Congo Red -- a new staining technique for automated cytology.

A new staining method has been developed for an automated pre-screening device for cervical smears called BioPEPR. The method is a combination of the Feulgen technique with thionine as th Schiff reagent, and congo red. Thionine and congo red have absorption maxima at 585 nm and 500 nm respectively. The staining results in cells with an orange-red cytoplasm and brownish-blue nucleus. These smears can e easily screened using a normal microscope. The BioPEPR system scans photonegatives of the smears that are made at a magnification of ten using monochromatic light of 545 nm. The staining facilitates good discrimination between background and cytoplasm and between cytoplasm and nucleus, allowing morphologic parameters such as cytoplasmic area, nuclear area and nuclear integrated darkness to be well measured.

Field test results using the BioPEPR cervical smear prescreening system.

BioPEPR is a cathode ray tube scanning device that has been developed for the prescreening of cervical smears. Cervical samples are collected in suspension, syringed, counted and spread across a microscope slide. Feulgen-Thionine with Congo Red is used as the staining procedure. Fields of 6 X 8 mm are scanned at 1 mu resolution, using an intermediate photographic step. Morphologic parameters of the cells are measured, as called for by a hierarchical decision strategy. Analysis speed is currently about 4 min per smear; improvements of at least a factor of four are expected. A field test of BioPEPR is being undertaken in cooperation with a local population screening program. In the study two smears are made from each woman; the first is prepared according to the Papanicolaou technique while the second is prepared and stained according to the BioPEPR methods. In addition to smears from the population screening program a number of possibly abnormal smears are obtained from several gynecological clinics. To date approximately 3500 "second" smears have been analyzed by the BioPEPR system. The results show a missed positive rate on the order of a few percent (no missed positives were more severe than a slight dysplasia), and a false alarm rate of about 24%. A further analysis of the age distribution of the false alarm rate shows that a large proportion came from women above the age of 50; the false alarm rate is about 20% for women under the age of 50.

BioPEPR: a system for the automatic prescreening of cervical smears.

A feasibility study has indicated that a Prescion Encoding and Pattern Recognition (PEPR) cathode ray tube prescreening system for cervical smears can be both accurate and fast. Smears are prepared using a syringing technique and are stained with a Feulgen-type nuclear stain and a protein counter-stain. The use of film as an intermediate step between the cells and Bio PEPR allows the scanning of fields as large as 8 x 8 mm. The morphological features of the cells are measured as directed by a hierarchical decision strategy. Additional programs detect artifacts, overlaps, and leukocytes. For clean samples, false positive and false negative rates on the cell level have been obtained that will allow acceptable smear level rates (10% false positive, 1% false negative). These rates have been reached without compromising the required speed goals of 120 to 180 smears per hr. The efficiency of the system is dependent on the quality of the smears. Measurements on a set of 192 routinely prepared smears indicate acceptable false negative rates and a false positive rate of about 18%. A reduction of this rate is expected with small improvements in cell preparation and measuring software, leading to the overall system efficiency required for commercial feasibility.