The problem of blood flow measurement in X-ray angiography using measurements of the leading edge of the contrast bolus as it traverses the vascular bed is considered. A new technique for velocity measurement is presented based upon the ratio of the temporal derivative to the spatial derivative of the contrast bolus in thedirection of flow. With the addition of a small correction factor, the value obtained is shown to reflect the transport velocity, or the velocity at which the contrast is transported down the vessel of interest. Because of the streaming that occurs due to laminary flow conditions, the measured transport velocity is found to be somewhere between the average and the peak(central)fluid velocities for measurements taken during the traversal of the bolus leading edge. The spatial and temporal variation of the transport velocity are found to be consistent with the bolus motion expected in the presence of laminar flow. From X-ray images of contrast passage through simple tubes, we find that the derivative method measures the transport velocity during passage of the bolus leading edge. In most cases of laminar blood flow, the leading edge transport velocity can be 20-40% higher than the average blood velocity.

The velocity of blood in the anterior cerebral arteries of 22 patients was measured as a function of vessel diameter using digital subtraction angiography. A series of lateral images of the cerebral vessels were acquired as the vessels opacified. We placed Several regions of interest on the images over each vessel segment. The time of arrival for each ROI was determined by calculating the time when the vessel at each ROI was halfway opacified. The blood velocity for each vessel segment was calculated from a plot of time of arrival vs. distance along the vessel segment. We measured the diameter of each fully opacified vessel from the film using a jewelers loop and then corrected for magnification. We found that the velocity of blood decreased from 30-40 cm/s in 1.5-2 mm diameter vessels to 5-10 cm/s in 0.3-0.7 mm diameter vessels.

Imaging of small vessels is of critical importance in evaluation of strokes, tumors, aneurysms and vascular malformations. While x-ray angiography is effective at producing the required small vessel image information, the potential morbidity of the procedure (.1 to 3%), cumulative risk of ionizing radiation and the required one day hospital stay have encouraged researchers to find a noninvasive procedure. Therefore, any area of overlap between MRA and x-ray angiography techniques would be of significant benefit to patients.

Hardware used to perform MRA has nearly been optimized, thus leaving post-processing as the next route to improving small vessel images. We are investigating how these filtering techniques can be used as a vessel enhancement tool, improving the vessel contrast-to-noise ratio present how this promising processing technique can be used to obtain an impreoved contrast to noise ratio.

With magnetic resonance angiography pushing the limits of what gains can be acquired by exploiting the physical parameters of the system being imaged, an examination of the reconstruction techniques is necessary for maximizing the vessel detail obtainable. A potentially limiting aspect of MRA image reconstruction concerns the asymmetric Fourier spectrum that is obtained. Current reconstruction algorithms make approximate corrections for this asymmetry by assuming that the phase in the image is slowly varying. Where blood is flowing with a velocity distribution along a gradient the phase behavior of the image may be rapidly varying. We examine the phase behavior in 3D time-of-flight images and how adequately various reconstruction algorithms account for these non-ideal signal characteristics. Images acquired with varying amounts of spectral asymmetry are reconstructed using homodyne demodulation, projection onto convex sets, and the chirp z-transform. Regions of signal loss from the asymmetry for each reconstruction algorithm are examined by comparison to full echo data. Reconstruction algorithms are compared by evaluating vessel contrast-to-noise ratio.

Magnetic resonance images acquired with phased array RF coils exhibit background intensity variations due to the non-uniform sensitivity of the coils. This sensitivity variation can be greater than the true variations in the tissue signal intensities, producing less than optimal images of the subject. Several basic methods have been proposed for removing the coil-dependent intensity variations including modeling of the coil sensitivity, measuring the sensitivity on phantoms, and image filtering techniques. In addition, traditional signal processing techniques such as function fitting and rank leveling may also be useful. We will discuss several intensitycorrection techniques including those previously proposed as well as several used in traditional signal processing.

3D Fast Spin Echo (3DFSE) is a versatile MR imaging sequence that is capable of generating images of high resolution in all 3 spatial dimensions in reasonable periods of time. From the 3D image data set, it is possible to obtain reformatted planar images in any orientation, therefore, it may be possible to replace multiple 2D imaging sequences at different orientations (i.e., axial and coronal acquisitions) and ultimately shorten imaging exam times. The 3DFSE technique has found many applications including the inner auditory canals, the lumbar spine, black blood angiography of intracranial vessels, musculoskeletal (i.e., knees and ankles) structures, and high resolution autopsy studies.

A brief description of the basic 3DFSE pulse sequence will be given. Two specific applications and the corresponding sequence developments will be discussed: (1) The Inner Ear. A localized version of 3DFSE sequence has been developed. This sequence has the advantages of shorter scan times, elimination of the slab boundary artifact, and smaller fields-of-view. (2) Black Blood Angiography. We have demonstrated that 3DFSE is capable of generating high-resolution images that rival what is seen with conventional 3DTOF MOTSA techniques.

The goal of this research is to develop an endcap birdcage radio frequency (RF) head resonator for the purpose of increasing the signl-to-noise-ratio (SNR) and contrast-to-noise ratio (CNR) in Magnetic Resonance Angiography (MRA) images. This will allow for better visualization of the smaller vessels in the brain. The birdcage resonator has become a standard coil for magnetic resonance imaging because of its large homogeneous imaging volume, its high characteristic Q-factor, and relatively high SNR. Its simple symmetric design also lends itself easily to quadrature excitation and detection of spins of interest. However, because of the characteristic profile of the standard birdcage, less signal is obtained towards its open ends. In this work, special consideration was given to increase the field homogeneity of the coil towards the top of the head, improve the Q-factor and achieve a better SNR than available with standard (commercial) birdcage coil with the use of a coil endcap. The prototype coil measures 24 cm x 20.5 cm, resulting in a better filling factor than the standard birdcage head coil. The conductive endcap has increased field homogeneity and helped reduce resistive losses. Using this prototype coil, MRA images have been obtained, demonstrating the desired coil characteristics.

The interest in measuring temperature results from the idea that temperature contrasts may provide additional information in studying breast lesions and improve the specificity of diagnosis.

We build a phantom in which we could control the temperature with two liquid baths at different temperatures, thus creating a linear temperature gradient across the phantom. The liquid used for heating and cooling is a perfluorocarbon that does not have any hydrogen protons and therefore cannot cause artifacts as it does not produce any signal. The phantom was filled with an agar gel doped with copper sulfate. Two thermistors embedded in the agar measured the temperature in the phantom. Various pulse sequences based on measuring temperature via the diffusion coefficient, proton resonance frequency shift, and longitudinal relaxation time were compared for precision and temporal resolution.

The same heating/cooling system that we used with the phantom was also used to induce temperature changes in the breast. In this case the temperature controlled liquid flows into a standard "heating" blanket that is then placed inside the breast coil. We hypothesize that the difference in vascular response of normal vs. diseased breast tissue due to external temperature changes may be an indicator of malignancy.

* Not presented at this symposium

Kinetic parameters of dynamic SPECT teboroxime data sets may provide a more sensitive measure of coronary artery disease than conventional (Tl, MIBI) static myocardial perfusion studies. A time-consuming and subjective step of the data analysis is drawing Regions of Interest (ROIs) to delineate blood pool and myocardial tissue regions. The time-activity curves for the regions are then used to estimate local kinetic parameters. In this work, the appropriate time-activity curves are found semi-automatically, in a manner similar to that used for calculating circumferential profiles in conventional static cardiac studies. The approach is shown to be comparable to manually selected ROIs when noisy simulated data (the MCAT phantom) is used. Furthermore, the potential pitfall of extremely high liver uptake, common in dynamic teboroxime studies, is addressed by an iterative scheme in which constraints are imposed to ensure reasonable ROIs are selected. The constraints are based on the shape of the time-activity curve of the circumferential profile members and on an assumption that the short axis slices are approximately circular. The constraints help to reduce the effects of noise and liver uptake when automatically finding ROIs.

SPECT images are degraded by the large number of scattered photons which are detected. These scattered photons typically result in a decrease in the contrast of the images. The problem of scatter in SPECT is particularly difficult in inhomogeneous media, such as the thorax, due to the strong object dependence of the scatter distribution.

An efficient model for predicting the distribution of scattered photons in inhomogeneous media has been developed. Accurate scatter correction can be achieved by incorporating the scatter model into an iterative reconstruction algorithm. However, the reconstruction time required makes this method impractical for routine clinical use. Therefore, the current aim is to speed up the implementation of the scatter model.

A couple of methods for accelerating the model-based scatter correction scheme are proposed. These include using an unmatched projector-backprojector (in which scatter is only modeled in the forward projector). Monte-Carlo simulated projection data are used to thoroughly investigate the convergence properties of the accelerated reconstruction algorithms. The results show that the accelerated methods result in a slight reduction in the accuracy of the reconstructed image but offer significant time savings.

Geometric Point Response Correction (GPRC) is incorporated in the modified ordered-subset ML-EM (maxium-likelihood expectation-maximization) algorithm for iterative reconstruction of 511 keV SPECT projection data. The modified algorithm has a relaxation factor which effectively controls the step size at each iteration. Monte Carlo projections and patient/phantom projection data acquired from a Picker PRISM 3000 three-detector SPECT systerm are reconstruted by using both the proposed algorithm with GPRC and the filtered backprojection algorithm. It is observed that the iterative algorithm with GPRC gives a higher resolution in the reconstruction than the filtered backprojection algorithm when the ultra-high energy collimators are used in data acquisition. However, the resolution improvement in the reconstruction is not as obvious for projection data acquired from low energy collimators. The introduced relaxation factor needs to be empirically chosen. Reconstruction time is clinically practical for this algorithm.

Dynamic cardiac SPECT has the potential for quantitatively measuring myocardial perfusion. Perfusion is quantified by estimating rate constants which describe the wash-in of teboroxime from the blood to the myocardial tissue. The estimates are performed with image derived time-activity curves obtained from regions of interest in the left ventricle blood pool and the myocardial tissue. Cardiac motion tends to cause contamination in the image derived regions which in turn produces bias in estimates of wash-in. This talk will present results from simulated and experimental studies carried out to study the effect of cardiac motion on measurements of perfusion. The simulations utilize a beating version of the MCAT heart and torso phantom. The experimental results are from ongoing canine studies which are being used to assess the ability of dynamic SPECT to quantify myocardial perfusion.

The Compton camera has been proposed as an alternative to the Anger camera in SPECT. The advantage of the Compton camera is its high geometric efficiency due to electronic collimation. The Compton camera collects projections that are integrals over cone surfaces. Although some progress has been made toward image reconstruction from cone projections, at present no filtered backprojection algorithm exists. This paper investigates a simpler 2D version of the imaging problem. An analytical formula is developed for 2D reconstruction from data acquired by a 1D Compton camera that consists of two linear detectors,one behind the other. Coincidence photon detection allows the localization of the 2D source distribution to two lines in the shape of a "V" with the vertex on the front detector. A set of "V" projection data can be divided into subsets whose elements can be viewed as line-integrals of the original image added with its mirrored shear transformation. If the detector has infinite extent, reconstruction of the original image is possible using data from only one suchsubset. Computer simulations were performed to verify the newly developed algorithm.

The isotope 153Gd is a common line source used in collection transmission data; applications range from SPECT imaging to bone densitometry. Currently only the higher of two 153Gd energy peaks is being used in the estimation of attenuation maps. By using the second peak as well, it is proposed that the resulting attenuation maps will be more accurate. This is accomplished by decomposing each reconstructed pixel as a mixture of two materials. To acheive density map estimates for each material at each pixel, an iterative algorithm that minimizes a weighted least squares objective and locally updates the pixels is employed. The final attenuation map at each pixel is the sum of mass fraction of material one multiplied by its attentuation, and mass fraction of material two multiplied by its attentuation at the appropriate energy.

* Not presented at this symposium

An efficient close-pixel projector model is developed to significantly improve the image quality, when the image is reconstructed by an iterative algorithm. The image quality dependence on ratio of size of non-overlapped square pixels and the projection bin size is studied. Attenuation and collimator point respond function can be taken into account. Approximation of a continuous image by linear combination of over-lapped function is considered to eliminate the aliasing artifacts. This approximation allow to solve the image equation because of the pixel dependency. This projector model will provide more reliable and more accurate SPECT images and will make it possible to use truncated projections and to construct cone-beam data.

It has been reported that clinical SPECT can have degradation of image quality if the patient's liver is close to the heart. This raises the question of whether this is also a problem for STEP(Simultaneous Transmission/Emmission Protocol). In addressing this problem we need to see not only if there is a problem but what the cause is and what can be done to correct for this problem as well. This investigation is being done by Monte Carlo techniques comparing many possible combinations: with scatter vs. without scatter, with noise vs. without noise, with attenuation correction vs. without attenuation correction, and most importantly with a "hot" liver vs. without a "hot" liver. Comparing these different combinations will allow us to know what causes the problem and then evaluate different methods to correct this problem.

* Not presented at this symposium

The spatially varying geometric response in SPECT results in significant resolution loss in addition to shape distortions and reconstructed density non-uniformity. A system geometric response function, which is the spatial photon distribution function relates the discrete projection data and the continuous image, is used to model the photon blurring effect. Two projection models approximate the system geometric response function used as the "natural pixel" in the projection space image reconstruction method. Parallel beam and fan beam projections of the Hoffman brain phantom were used to evaluate the performance of the geometric response correction. Image reconstructions were obtained by using the singular value decomposition (SVD) method. The reconstructed images using the two proposed projectors showed improved resolution when compared against a unit-strip natural pixel model and the filtered backprojection algorithm. When the reconstructed image was displayed into finer grid points, the strip artifact seen in the unit-strip natural pixel model was removed and the continuity and resolution of the image were still reserved.

* Not presented at this symposium

Last year I presented the status of the Small Bore MR Reserach Facility in its early stages. This presentation will cover the facility's current capabilities and the research that is planned for the coming year.

The system is now functional and has undergone a console upgrade which has improved the system noise figures somewhat. In addition, hardware has been added to improve the SNR of narowband single nucleus applications. An expanded set of sequences is now avaliable to the user. These include imaging sequences SE, GRASS, FLASH, spoiled SPGR, STIR, FSE (limited capabilities), turbo-FLASH and Diffusion weighted imaging. Localized spectroscopy sequences include STEAM and PRESS. EPI and 3D-GRE are works in progress. Automated prescan is being tested and should be avaialble in the near future.

Current research includes a collaborative project with Dr. Steve White in Pharmacology/Toxicology where the development of white matter lesions following repetitive electroshock and kainic acid treatment is being investigated. Dr. Wolfram Samlowski in Hematology is looking into the effects of nitroxyl-metabolite blockers on T2 contrast of murine tumors. With Dr. Mark Noble of the Huntsman Cancer Institute 1H/31P in vivo spectroscopy of select intracranial tumor cell lines will be initiated. Finally, Dr. Grant Gullberg will be start a MR perfusion study, to evaluate spin-tagging techniques to elucidate in-vivo tissue partition coefficients.

Rationale: Changes in the hippocampal concentrations of NAA(N-Acetyl-Aspartate), Cr (Creatin) and Cho (Choline) have been shown to reflect the hippocampal sclerosis (HS). The relative concentrations can be measured noninvasively using in vivo proton magnetic resonance spectroscopy (MRS). This study attempts to establish repeatable clinical standards of NAA/Cr and Cho/Cr measured with proton MRS to assist clinical diagnosis of HS.

Methods: All measurements were conducted at 1.5 T using the point resolved spectroscopy (PRESS) pulse sequence. In single voxel (SV) experiments (TR = 2000 ms, TE = 136 ms, 160 accumulations), a voxel of 0.56x1.0x1.5 cm3 was localized within the hippocampal formation. In the multi-voxel or spectroscopic imaging (SI) experiments (TR = 2000 ms, TE = 136 ms, 8 accumulations), a 12x2x1 voxel grid was localized in a 16x2x1.5 cm3 volume to cover both hippocampal formations. The final spatial resolution after zero filling is 0.25 x0.5x1.5 cm3. The data were processed and analyzed statistically using SUN work stations.

Results: The normal ratios of NAA/Cr (1.16+/-0.078, mean+/-SE) and Cho/Cr (1.21+/-0.09) were obtained from 16 volunteers (11 F, 5 M; age: 21-47, mean age: 35.2) in 21 measurements (19 SV and 2 SI). Ratios were lower than what has been previously reported due to the higher spatial resolution and lack of the partial volume artifact. Based on this result, ten patients were tested. Five have been identified to have altered metabolic ratios. Final pathological confirmation is pending.

Conclusion: Carefully performed MRS limited to the hippocampal formation may yield standardized results that can be used as a tool to detect HS.

Three active research projects are described: two involve nuclear medicine SPECT imaging and the other is an X-ray CT project. The common theme is that fully 3D techniques are required for the tomographic reconstruction.

The two SPECT projects involve rotating collimators, respectively cone-beam and multi-segment slant-hole collimators. In both cases the goal is increased photon sensitivity, while maintaining a simple circular motion of the detector gantry during the scan. Neither system achieves full detector utilization for imaging small organs in the patient, but both improve on conventional systems. The slant-hole approach presents a simpler image reconstruction problem.

The X-ray CT project also involves a cone-beam geometry. Using a large-area detector and a high-energy source, drums of potentially hazardous waste are scanned. The goal is real-time (video-rate) radiographic data-collection with efficient tomographic image reconstruction. Special techniques have been developed to avoid the relatively slow reconstruction times required by general cone-beam algorithms. Improved algorithms have been developed for a scanning configuration consisting of a rotation of the drum, followed by a few vertical translations of the x-ray source.

The use of intravascular stents for purposes of revascularization is often followed by re-stenosis of the vessel within 6 months. The re-stenosis is believed to be caused by proliferation of smooth muscle cells in the vessel around the stent site - referred to as intimal hyperplasia.

In 1989, Fischell and Fischell proposed irradiating the stent site with beta particles from P-32 to inhibit growth of these smooth cells. Recently, research activity has centered around using very expensive P-32 impregnated stents. A group in the Radiology Department has proposed the placement of an angioplasty balloon filled with P-32 in the stent area for a short period of time as a less expensive means to deal with the problem of intimal hyperplasia. The group plans to try this method on pigs. However, the question of exposure times and dose delivery needed to be first addressed.

Modeling of the radiation dose delivered to vessels using the latter method will be discussed, and compared to dosimetry measurements carried out in the laboratory under simulated conditions.

Purpose: To evaluate gadolinium enhanced MRA in the preoperative assessment of abdominal aortic aneurysms (AAA).

Materials and Methods: A retrospective study was performed on 47 consecutive patients who had preoperative gadolinium enhanced MRA and subsequent AAA repair. The original MRA interpretations were compared to the operative findings. Cases in which there was a discrepancy were re-reviewed on a work station.

Results: MRA correctly evaluated the extent of the AAA in 45 of 47 patients (96%). Ten of 15 accessory renal arteries were correctly identified (67%). Four of the 5 missed accessory renal arteries were identified on re-review. Renal artery stenosis was correctly diagnosed in 14 of 18 cases (78%). On re-review of the four cases of renal artery stenosis which were describedintraoperatively but not called on the MRA, three had detectable renal artery stenosis. MRA described three cases of renal artery stenosis which were felt to be normal intraoperatively. MRA correctly diagnosed four cases of mesenteric stenosis and 10 cases of occlusion. MRA correctly diagnosed six retroaortic left renal veins and two replaced right hepatic arteries.

Conclusion: Fast 3D spoiled gradient echo imaging during the dynamic intravenous administration of gadolinium is a safe and non-invasive method for the preoperative evaluation of abdominal aortic aneurysms. These studies need to be carefully evaluated on a work station, as there appears to be a learning curve to correctly identify the presence of accessory renal arteries and renal artery stenosis.

When reconstructing an image using iterative algorithms, projection and backprojection are performed at each iteration. The projector usually models the imaging physics, such as voxel weighting geometry, attenuation, system point response function, and scatter. The backprojection matrix is commonly chosen as the transpose of the projection matrix. The purpose of this study is to investigate the possibility of using backprojection matrices other than the transpose of the projection matrix, in order to significantly reduce the computation time. A sufficient condition for convergence is that the eigenvalues of the composed projection-backprojection matrix are positive. We have tested the following combinations: (1) a projector is ray-driven, attenuation weighted, while a backprojector is ray-driven without attenuation weighting; (2) a projector is rotation-based with a model of system point response function, while a backprojector is voxel-driven and does not model the system point response function; and (3) a cone-beam projector is ray-driven, line-length weighted, while a cone-beam backprojector is voxel-driven, bilinear interpolated. In all the studies, the composed matrices are strongly diagonally dominated with positive diagonal elements suggesting positive eigenvalues.

* Not presented at this symposium

After nearly 4 years of waiting, we have finally upgraded one of our MRI scanners to the GE echo planar system. The increase in gradient performance has been anticipated for some time, and we can now experimentally evaluate the expected improvements in imaging performance. Many of the improvements will have been discussed in other presentations, including more efficient high resolution volume imaging, more rapid volume angiography, and potentially more rapid temperature measurement studies. The improved gradient efficiency also makes possible the enhanced use of contrast agents in dynamic angiography and regional perfusion studies. At least a part of our work during the next few years will involve the evaluation of improved efficiency that can be obtained with the improved gradients. We also hope to continue the investigation of novel imaging techniques.

A new and exciting area of nuclear cardiology is the simultaneous imaging of perfusion and metabloism using SPECT. In a single imaging protocol 99mTc-labelled sestamibi and 18FDG are injected into the patient. The distribution of sestamibi is used to measure myocardial perfusion and the metabolism of 18FDG is used to determine myocardial viability. This protocol is used to determine if a patient with reduced flow but viable myocardial tissue is a good candidate for revascularization surgery. Dynamic cardiac SPECT may be a more sensitive way to measure simultaneously perfusion and metabolism with the injection of only one radiopharmaceutical. Dynamic three-dimensional reconstructions are obtained from serial tomographic projection sets of the 3D distribution with a temporal resolution as low as 5 seconds. From the dynamically reconstructed data, compartment modeling techniques are used to estimate kinetic parameters for the distribution of the radiopharmaceutical through the blood and tissue of regions in the heart. Three compartment models will be presented: a one compartment model that models the perfusion of 99mTc-teboroxime, a two compartment model that models the distribution and metabolism of 18FDG, and a three compartment model that models the distribution and metabolism of the fatty acid 123I-labelled iodophenylpentadecanoic acid (IPPA). Each model will be compared as to its feasibility to measure simultaneously myocardial tissue perfusion and metabolism.