E.V.R. Di Bella

On the dark rim artifact in dynamic contrast-enhanced MRI myocardial perfusion studies

A dark band or rim along parts of the subendocardial border of the left ventricle (LV) and the myocardium has been noticed in some dynamic contrast‐enhanced MR perfusion studies. The artifact is thought to be due to susceptibility effects from the gadolinium bolus, motion, or resolution, or a combination of these. Here motionless ex vivo hearts in which the cavity was filled with gadolinium are used to show that dark rim artifacts can be consistent with resolution effects alone. Magn Reson Med, 2005.

A comparison of rotation-based methods for iterative reconstruction algorithms

Rotation-based approaches to iterative reconstruction algorithms offer unique trade-offs between accuracy, computational complexity, and smoothing. We analyzed six different rotation methods for generating projections and reconstructions of simulated and clinical data. Nearest neighbor, upsampled nearest neighbor, bilinear interpolation, and bicubic interpolation were all used to determine the values at rotated grid points. A decomposition of the rotation transformation matrix into three components was also investigated. Linear and cubic interpolation were used in the three pass method. For all of the methods, mean normalized square errors were computed as was a measure of smoothness. Our results demonstrate the trade-offs associated with the different methods and suggest that the method of shears with cubic interpolation offers a computationally efficient approach with accuracy between that of bilinear and bicubic interpolation.

Factor analysis with a priori knowledge - application in dynamic cardiac SPECT

Two factor analysis of dynamic structures (FADS) methods for the extraction of time-activity curves (TACs) from cardiac dynamic SPECT data sequences were investigated. One method was based on a least squares (LS) approach which was subject to positivity constraints. The other method was the well known apex-seeking (AS) method. A post-processing step utilizing a priori information was employed to correct for the non-uniqueness of the FADS solution. These methods were used to extract 99mTc-teboroxime TACs from computer simulations and from experimental canine and patient studies. In computer simulations, the LS and AS methods, which are completely different algorithms, yielded very similar and accurate results after application of the correction for non-uniqueness. FADS-obtained blood curves correlated well with curves derived from region of interest (ROI) measurements in the experimental studies. The results indicate that the factor analysis techniques can be used for semi-automatic estimation of activity curves derived from cardiac dynamic SPECT images, and that they can be used for separation of physiologically different regions in dynamic cardiac SPECT studies.

Comparison of temporal filtering methods for dynamic contrast MRI myocardial perfusion studies

Dynamic contrast myocardial perfusion studies may benefit from methods that speed up the acquisition. Unaliasing by Fourier encoding the overlaps using the temporal dimension (UNFOLD), and a similar linear interpolation method have been shown to be effective at reducing the number of phase encodes needed for cardiac wall motion studies by using interleaved sampling and temporal filtering. Here such methods are evaluated in cardiac dynamic contrast studies, with particular regard to the effects of the choice of filter and the interframe motion. Four different filters were evaluated using a motion‐free canine study. Full k‐space was acquired and then downsampled to allow for a measure of truth. The different filters gave nearly equivalent images and quantitative flow estimates compared to full k‐space. The effect of respiratory motion on these schemes was graphically depicted, and the performance of the four temporal filters was evaluated in seven human subjects with respiratory motion present. The four filters provided images of similar quality. However, none of the filters were effective at eliminating motion artifacts. Motion registration methods or motion‐free acquisitions may be necessary to make these reduced FOV approaches clinically useful. Magn Reson Med 49:895–902, 2003.

Automated region selection for analysis of dynamic cardiac SPECT data

Dynamic cardiac SPECT using Tc-99m labeled teboroxime can provide kinetic parameters (washin, washout) indicative of myocardial blood flow. A time-consuming and subjective step of the data analysis is drawing regions of interest to delineate blood pool and myocardial tissue regions. The time-activity curves of the regions are then used to estimate local kinetic parameters. In this work, the appropriate regions are found automatically, in a manner similar to that used for calculating maximum count circumferential profiles in conventional static cardiac studies. The drawbacks to applying standard static circumferential profile methods are the high noise level and high liver uptake common in dynamic teboroxime studies. Searching along each ray for maxima to locate the myocardium does not typically provide useful information. Here we propose an iterative scheme in which constraints are imposed on the radii searched along each ray. The constraints are based on the shape of the time-activity curves of the circumferential profile members and on an assumption that the short axis slices are approximately circular. The constraints eliminate outliers and help to reduce the effects of noise and liver activity. Kinetic parameter estimates from the automatically generated regions were comparable to estimates from manually selected regions in dynamic canine teboroxime studies.

Factor analysis of dynamic structures in dynamic SPECT imaging using maximum entropy

Factor analysis of dynamic structures (FADS) is a technique used in the automatic extraction of time activity curves (TACs) from dynamic sequences. Although it has been reported that factor analysis with non-negativity constraints can in certain cases correlate with region of interest (ROI) measurements in SPECT and PET heart studies, the method does not ensure a unique solution. In this work it is shown that the FADS result is improved by using the Maximum Entropy Principle. Both the FADS technique and the new maximum entropy method were tested on simulated data and experimental /sup 99m/Tc-teboroxime canine cardiac data. The results showed that the FADS technique, using only non-negativity constraints, produced curves that did not always closely approximate the true curves. The new method, however, resulted in TACs that closely resembled the true curves. Thus, the inclusion of an entropy term is a useful method for obtaining more accurate extractions of TACs from dynamic SPECT data.

Parallelized formulation of the maximum likelihood-expectation maximization algorithm for fine-grain message-passing architectures

Recent architectural and technological advances have led to the feasibility of a new class of massively parallel processing systems based on a fine-grain, message-passing computational model. These machines provide a new alternative for the development of fast, cost-efficient Maximum Likelihood-Expectation Maximization (ML-EM) algorithmic formulations. As an important first step in determining the potential performance benefits to be gathered from such formulations, we have developed an ML-EM algorithm suitable for the high-communications, low-memory (HCLM) execution model supported by this new class of machines. Evaluation of this algorithm indicates a normalized least-square error comparable to, or better than, that obtained via a sequential ray-driven ML-EM formulation and an effective speedup in execution time (as determined via discrete-event simulation of the Pica multiprocessor system currently under development at the Georgia Institute of Technology) of well over two orders of magnitude compared to current ray-driven sequential ML-EM formulations on high-end workstations. Thus, the HCLM algorithmic formulation may provide ML-EM reconstructions within clinical time-frames.

Improved reconstruction of dynamic cardiac perfusion MRI with use of a reference frame

Lower spatial resolution can result when increasing the temporal resolution of dynamic imaging to track the uptake and washout of a bolus of paramagnetic contrast. This work investigates the maintenance or improvement of spatial resolution in ultra rapid dynamic cardiac MRI perfusion sequences when a higher resolution reference frame either from the same sequence, or from a steady-state cine sequence, is available. The generalized series and keyhole methods were used on dynamic data acquired with a bolus contrast injection. The image sequence of the heart region was extracted from the full field of view image sequence. The extracted images were registered for motion and also denoised before reconstruction by generalized series or keyhole methods was applied. Resolution improvement methods were used on low resolution data. Results were compared to the higher resolution data and showed improvement in resolution for the generalized series, with less artifacts than for the keyhole method.

SPECT imaging of teboroxime during myocardial blood flow changes

Kinetic parameters and static images from dynamic SPECT imaging of (99m)Tc-teboroxime have been shown to reflect blood flow in dogs and in humans at rest and during adenosine stress. When compartment modeling is used, steady-state physiological conditions are assumed. With standard adenosine stress protocols, imaging of teboroxime would likely involve significant changes in flow, even if performed only for five minutes. These flow changes may significantly bias the kinetic parameter estimates. On the other hand, when static imaging is performed, large flow changes during acquisition may improve contrast between normal and occluded regions. Computer simulations were performed to determine the effect of changing flows on kinetic parameter estimation and on static (average tissue uptake) images. Two canine studies were also performed in which adenosine was given with a standard protocol, and then imaging was repeated with adenosine infusion held constant. The simulations predicted biases on the order of 7% for kinetic washin parameter estimation and 18% for the washout parameter. Contrast for static studies was found to depend critically on the time-activity behavior of the distribution as well as on the stress protocol. The differences in washin contrast from the standard and continous adenosine dog studies was slightly larger than predicted from the simulations. Optimal imaging of teboroxime with adenosine using compartment modeling will require non-standard adenosine stress protocols, although sub-optimal imaging may still be useful clinically.

Theoretical analysis of blind identification of kinetic parameters in a three compartment model: nonuniqueness problem

Estimation of compartment model parameters can be done in dynamic medical imaging blindly - that is, without an input function. In this work we address the question of uniqueness of blind identification of kinetic parameters for a three compartment model such as is widely used with FDG. We show analytically that if only two regions of interest in the imaged organ are used, the solution is not unique. More regions are shown to lead to a unique solution.