Search

Parallel magnetic resonance imaging achieves reduction in scan time by collecting a partial set of signals using an array of receiving coils each with a local sensitivity pattern. An image is then reconstructed from the partial dataset using the additional information of coil sensitivity. GRAPPA (generalized auto calibrating partially parallel acquisitions) is one of the most successful reconstruction techniques in which the missing k-space lines are interpolated from the acquired data in the whole coil array using a convolution kernel estimated from a fully sampled data patch in the center of k-space. The interpolation kernel is usually small but fixed in size for all coils. Here, we show that a variable kernel with a size dependent on the coil sensitivity can lead to better image quality. The kernel size is estimated from the ratio of the coil sensitivities obtained from a reference scan or from the same dataset. Conventional GRAPPA kernel estimation and image reconstruction is modified to employ the variable-size kernel for improved reconstruction. The new technique shows improved image quality compared to GRAPPA.

We propose a technique that allows the improvement of lateral resolution in ultrasound imaging using a deconvolution-based strategy. We first derive a formulation for the problem in terms of an arbitrary spatially-variant beam pattern and show that it is possible to optimally estimate the values of the image by solving a large linear inverse problem. This linear system depends on the shape and extent of the point spread function as well as the desired resolution of the resultant image. We show that this linear system is sparse and therefore sparse matrix techniques for storage and algebra are used to make the computational cost reasonable. The strategies used to solve this problem are proposed based on truncated singular value decomposition or regularized conjugate gradient method that allows an equivalent regularization to imposing a quadratic inequality constraint. This allows the condition number of the problem to be kept sufficiently low, thus ensuring a robust solution. For a given ultrasound line with specific transmit and receive focusing characteristics, this problem is solved for the whole image and we show that it is possible to implement the solution in a look-up table form similar to what is used in image reconstruction in current ultrasound systems. This accounts for the variations of the point spread function at different spatial positions.

Here, we propose a new technique that allows the improvement of SNR without lowering the frame rate. The basic version of the proposed technique works by exciting multiple non-overlapping apertures at the same time. Assuming N apertures are to be acquired simultaneously, N acquisitions from such arrangement are made with different aperture coding in each acquisition. In particular, the excitation to each aperture is multiplied by a different predetermined factor for each acquisition. By properly selecting these factors, the problem of estimating the different lines become equivalent to solving a well-conditioned linear system of size N. Also, we show that such factors can be chosen to make the system matrix orthogonal, thus enabling simple solution to be obtained. We describe the implementation of such procedure and demonstrate that it is possible to achieve in practice even for systems in which the sign of excitation can only be changed. For more advanced system where coded excitations can be used, more accurate implementation taking into account element sensitivity variations within each probe for closer adherence to the desired characteristics. The proposed system results in noise reduction that is equivalent to averaging N acquisitions similar to phase encoding in magnetic resonance imaging. At the same time, the system maintains the same lateral resolution and frame rate as conventional acquisition strategies. We also discuss the use of sub-optimal overlapped apertures and describe the deterioration of SNR gain as a result of using independent rather than orthogonal apertures.

We propose a new nonparametric technique for clutter rejection. We consider the Doppler data sampled using a sufficiently large dynamic range to allow for the clutter rejection to be implemented on the digital side. The Doppler signal is modeled as the summation of the true velocity signal, a clutter component, and a random noise component. To simplify the analysis, the first two components are assumed as deterministic yet unknown signals. The Doppler data are collected from the sample volume of interest as well as from several sample volumes in its neighborhood. Given that the shape of the clutter component will be similar in all these signals and given its relatively higher magnitude, it is possible to separate this component using principal component analysis (PCA). In particular, the clutter component appears as the first eigenvector (principal component) in PCA. Given this principal component, the projection of the Doppler signal of interest onto this component is removed and the remaining signal is subsequently used to derive the Doppler spectrogram. We describe an efficient implementation methodology that allows the added computational complexity of the new system to be reasonable.

We propose two new methods that allow the determination of the phase delays corresponding to phase aberration efficiently. We derive a new optimization methodology to compute the best compensation phase delays in successive steps. In particular, we start with an array consisting of one element with a specific excitation pattern. Then, another element is added and the dynamic receive delays are iteratively computed such that the obtained echoes are optimal in strength. A third element is added and the process is repeated. This process continues until all elements in the aperture are added. Hence, instead of solving the conventional N-dimensional problem of adjusting the delays of N elements together to achieve optimal characteristics, we transform the problem into the one of solving N-1 consecutive one-dimensional optimization problems. Given the fact that the set of available delay values is finite, the one-dimensional problem is shown to be a classical combinatorial optimization problem. The other technique based on Fourier transform tries to align signals based on information from a single frequency selected as the center frequency of the probe. This method is simple, computationally efficient and lends itself to real-time implementation. The proposed methods were implemented to correct real data from a resolution phantom and the results particularly indicate the potential of the second method.

We propose a technique for suppression of both intra-slice and inter-slice types of motion artifacts simultaneously. Starting from the general assumption of rigid body motion, we consider the case when the acquisition of the k-space is in the form of bands of finite number of lines arranged in a rectilinear fashion to cover the k-space area of interest. We also assume that an averaging factor of at least 2 is desired. Instead of acquiring a full k-space of each image and then average the result, we propose a new acquisition strategy based on acquiring the k-space in consecutive bands having 50% overlap going from one end of the phase encoding direction to the other end. In case of no motion, this overlap can be used as the second acquisition (NEX=2). When motion is encountered, both types motion are reduced to the same form under this acquisition strategy. In particular, detection and correction of motion between consecutive bands result in suppression of both motion types. In this work, this is achieved by utilizing the overlap area to estimate the motion, which is then taken into consideration in further reconstruction (or even acquisition if real-time control is available on the MR system). We demonstrate the accuracy and computational efficiency of this motion estimation approach. Once the motion is estimated, we propose a simple strategy to reconstruct artifact-free images from the acquired data that take into account the possible voids in the acquired k-space as a result of rotational motion between blades.

The application of independent components analysis (ICA) to functional magnetic resonance imaging data has been proven useful to decompose the signal in terms of its basic sources. The main advantage is that ICA requires no prior assumption about the neuronal activity or the noise structure, which are usually unknown in fMRI. This enables the detection of true activation components free of random and physiological noise. Hence, this technique is superior to other techniques such as subspace modeling or canonical correlation analysis, which have underlying assumptions about the signal components. Nevertheless, this technique suffers from a fundamental limitation of not providing a consistent ordering of the signal components as a result of the whitening step involved in ICA. This mandates human intervention to pick out the relevant activation components from the outcome of ICA, which poses a significant obstacle to the practicality of this technique. In this work, a simple yet robust technique is proposed for ranking the resultant independent components. This technique adds a second step to ICA based on canonical correlation analysis and the prior information about the activation paradigm. This enables the proposed technique to provide a consistent and reproducible ordering of independent components. The proposed technique was applied to real event-related functional magnetic resonance imaging data and the results confirm the practicality and robustness of the proposed method.

The advent of event-related functional magnetic resonance imaging (fMRI) has resulted in many exciting studies that have exploited its unique capability. However, the utility of event-related fMRI is still limited by several technical difficulties. One significant limitation in event-related fMRI is the low signal-to-noise ratio (SNR). In this work, a new non-parametric technique for noise suppression in event related fMRI data is proposed based on spectrum subtraction. The new technique is based on generalized spectral subtraction that allows correlated noise components to be treated robustly. Moreover, it adaptively estimates a nonparametric model for random and physiological components of noise from the acquired data in a simple and computationally efficient manner. This allows the new method to overcome the limitations of previous methods while maintaining a robust performance given its fewer assumptions and suggests its value as a useful preprocessing step for fMRI data analysis.

This paper introduces an efficient algorithm for segmentation of fetal ultrasound images using the multiresolution analysis technique. The proposed algorithm decomposes the input image into a multiresolution space using the packet two-dimensional wavelet transform. The system builds features vector for each pixel that contains information about the gray level, moments and other texture information. These vectors are used as inputs for the fuzzy c-means clustering method, which results in a segmented image whose regions are distinct from each other according to texture characteristic content. An Adaptive Center Weighted Median filter is used to enhance fetal ultrasound images before wavelet decomposition. Experiments indicate that this method can be applied with promising results. Preliminary experiments indicate good results in image segmentation while further studies are needed to investigate the potential of wavelet analysis and fuzzy c-means clustering methods as a tool for detecting fetus organs in digital ultrasound images.

The objective of this study is to introduce a simple mathematical model for the brain image under both resting and activated states to facilitate both the understanding of the underlying neurophysiology and the realization of data sets. Two data sets were composed to simulate fMRI data. First set consists of a small spot that simulates the activated region superimposed on real baseline data. To simulate the signal enhancement, the hemodynamic response vector multiplies all pixels in the activated spot. Therefore, the resultant spots were added sequentially to the baseline images to create the first data set. The second set was formed by using the proposed model. The model took into account both random and physiological noise that are found in fMRI data. The random noise was assumed to vary from one frame to another while the physiological pattern was assumed of similar pattern throughout the brain with smooth spatial variations. A threshold cross-correlation technique was used on both data sets to compare the resultant activation maps. A falsehood measure was proposed and used as to test the accuracy of the activation detection. Finally, the results between the two data sets are compared to demonstrate the accuracy of the model.

The problem of detecting and tracking SPAMM tags has been the focus of cardiac MRI in the past few years. Among the many methods proposed to do that, the harmonic phase method (called HARP) has a large potential to enable real-time tracking. We propose two improvements to the original HARP technique to address the problem of magnetic field inhomogeneity and improve tracking. We derive a theoretical model for the problem of field inhomogeneity and propose an extension of the simulated phase evolution rewinding technique (SPHERE) whereby consecutive images are acquired using alternating echo time. Then, we propose a simple and computationally efficient approach that allows the problems of disappearing tag lines to be addressed. In particular, the geodesic horizontal and vertical distances between neighboring corner points in the detected tag lines are computed for each image. Once the correspondence between tag lines is established, the tracking of any point on the wall can be obtained given its relationship to the four corner points surrounding its location. The proposed improvements were implemented and applied to numerical phantoms as well as experimental data obtained for the original HARP technique to demonstrate the potential of the new techniques.

We propose a computerized system to accurately point laser to the diseased areas within the retina based on predetermined treatment planning. The proposed system consists of a fundus camera using red-free illumination mode interfaced to a computer that allows real-time capturing of video input. The first image acquired is used as the reference image for treatment planning. A new segmentation technique was developed to accurately discern the image features using deformable models. A grid of seed contours over the whole image is initiated and allowed to deform by splitting and/or merging according to preset criteria until the whole vessel tree is extracted. This procedure extracts the whole area of small vessels but only the boundaries of the large vessels. Correlating the image with a one-dimensional Gaussian filter in two perpendicular directions is used to extract the core areas of such vessels. Faster segmentation can be obtained for subsequent images by automatic registration to compensate for eye movement and saccades. Comparing the two sets of landmark points using a least-squares error provide an optimal transformation between the two point sets. This allows for real-time location determination and tracking of treatment positions.© (2001) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

We propose a novel technique for ultrasound speckle reduction based on iterative solutions to the coherent diffusion equation with the speckled image considered as the initial heat distribution. According to the extent of speckle, the model changes progressively from isotropic diffusion through anisotropic coherent diffusion to mean curvature motion. This structure maximally low-pass filters those parts of the image corresponding to fully-formed speckle, while preserving information associated with resolved object structures. The distance measure used to assess the deviation between images is embedded within the diffusivity tensor and is utilized as an intrinsic stopping criterion that ends the diffusion process completely in all directions when the deviation between the original and the filtered image exceeds the speckle limit. This model is termed pseudo-biased diffusion due to this unique formulation. Hence, there is no need for specifying the number of iterations in advance as with previous methods. Moreover, the steady state solution does not converge to the trivial single gray level solution, but rather to an image that is close in structure to the original but with speckle noise substantially reduced. Efficient discretization schemes allow large time steps to be used in obtaining the solution to achieve real-time processing.

We propose two techniques for robust motion artifact suppression in ultrasound images based on what is called navigator echoes. The navigator echo approach has been introduced and widely used in the area of magnetic resonance imaging where motion is detected along one of the image directions based on simple registration of one-dimensional projection of the image assuming rigid body motion. In ultrasound imaging, the same concept can be applied to correct for simple shifts (e.g., with linear array probes) or rotations (e.g., with convex array probes) along the width and depth direction. The first technique assumes a rigid body motion model. Hence, simple one-dimensional template matching can be used to obtain the motion parameters as in magnetic resonance imaging. Alternatively, the second technique considers a more general spatially- variant motion model. The motion parameters of this model can be obtained through localized optimization of an information theoretic criterion. The motion model parameters are then employed in the reconstruction thus providing images with substantially reduced artifacts. The proposed method was implemented on an experimental ultrasound system in which each line is obtained from 2-4 acquisitions from interleaved frames. The motion models are compared and their practical implementation for clinical systems is evaluated.

In speckle motion tracking, blood velocity magnitude and direction are estimated from speckle pattern changes between successive images based on either 2D correlation of sum- absolute-difference (SAD) methods. Even though these techniques have been proven useful for flow mapping applications, they suffer from bias effects in estimating due to the presence of clutter induced from structural motion. In this work, we propose a technique for reducing the clutter effect, and hence enhancing the robustness of velocity estimation. The proposed technique relies on separating the speckle from the underlying specular structures. The basic idea is to employ a speckle reduction strategy based on nonlinear coherent diffusion filtering to obtain speckle free image of vessels from an original B- mode. Then, subtracting such image from the original image, an image for speckle is obtained. Nonlinear coherent diffusion filtering has been proven successful in removing Rayleigh distributed speckle pattern resulting mainly from blood scatterers in B-mode images while preserving structural information. This allows such scattering pattern to be utilized more accurately in calculating the velocity using an ultrasound research system and velocity estimates were obtained using 2D correlation.

Computerized ultrasound tissue characterization has become an objective means for diagnosis of diseases. It is difficult to differentiate diffuse liver diseases, namely cirrhotic and fatty liver from a normal one, by visual inspection from the ultrasound images. The visual criteria for differentiating diffused diseases is rather confusing and highly dependent upon the sonographer's experience. The need for computerized tissue characterization is thus justified to quantitatively assist the sonographer for accurate differentiation and to minimize the degree of risk from erroneous interpretation. In this paper we used the fuzzy similarity measure as an approximate reasoning technique to find the maximum degree of matching between an unknown case defined by a feature vector and a family of prototypes (knowledge base). The feature vector used for the matching process contains 8 quantitative parameters (textural, acoustical, and speckle parameters) extracted from the ultrasound image. The steps done to match an unknown case with the family of prototypes (cirr, fatty, normal) are: Choosing the membership functions for each parameter, then obtaining the fuzzification matrix for the unknown case and the family of prototypes, then by the linguistic evaluation of two fuzzy quantities we obtain the similarity matrix, then by a simple aggregation method and the fuzzy integrals we obtain the degree of similarity. Finally, we find that the similarity measure results are comparable to the neural network classification techniques and it can be used in medical diagnosis to determine the pathology of the liver and to monitor the extent of the disease.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.

Schistosomiasis is a major problem in Egypt, despite an active control program it is estimated to exist in about 1/3 of the population. Deposition of less functioning fibrous tissues in the liver is the major contributory factor to the hepatic pathology. Fibrous tissues consist of a complex array of connective matrix material and a variety of collagen isotopes. As a result of an increased stromal density (collagen content), the parenchyma became more ectogenic and less elastic (hard). In this study we investigated the effect of cardiac mechanical impulses from the heart and aorta on the kinetics of the liver parenchyma. Under conditions of controlled patient movements and suspended respiration, a 30 frame per second of 588 X 512 ultrasound images (cineloop, 32 pels per cm) are captured from an aTL ultrasound machine then digitized. The image acquisition is triggered by the R wave of the ECG of the patient. The motion that has a forced oscillation form in the liver parenchyma is quantified by tracking of small box (20 - 30 pels) in 16 directions for all the successive 30 frames. The tracking was done using block matching techniques (the max correlation between boxes in time, frequency domains, and the minimum SAD (sum absolute difference) between boxes). The motion is quantified for many regions at different positions within the liver parenchyma for 80 cases of variable degrees of schisto., cirrhotic livers, and for normal livers. The velocity of the tissue is calculated from the displacement (quantified motion), time between frames, and the scan time for the ultrasound scanner. We found that the motion in liver parenchyma is small in the order of very few millimeters, and the attenuation of the mechanical wave for one ECG cycle is higher in the schisto. and cirrhotic livers than in the normal ones. Finally quantification of motion in liver parenchyma due to cardiac impulses under controlled limb movement and respiration may be of value in the characterization of schisto. (elasticity based not scattering based). This value could be used together with the wide varieties of quantitative tissue characterization parameters for pathology differentiation and for differentiating subclasses of cirrhosis as well as the determination of the extent of bilharzial affection.

In this work, feature extraction algorithms are proposed to extract the tissue characterization parameters from liver images. Then the resulting parameter set is further processed to obtain the minimum number of parameters representing the most discriminating pattern space for classification. This preprocessing step was applied to over 120 pathology-investigated cases to obtain the learning data for designing the classifier. The extracted features are divided into independent training and test sets and are used to construct both statistical and neural classifiers. The optimal criteria for these classifiers are set to have minimum error, ease of implementation and learning, and the flexibility for future modifications. Various algorithms for implementing various classification techniques are presented and tested on the data. The best performance was obtained using a single layer tensor model functional link network. Also, the voting k-nearest neighbor classifier provided comparably good diagnostic rates.