Results 1  10
of
34
A graph cut algorithm for generalized image deconvolution
 In ICCV
, 2005
"... The goal of deconvolution is to recover an image x from its convolution with a known blurring function. This is equivalent to inverting the linear system y = Hx. In this paper we consider the generalized problem where the system matrix H is an arbitrary nonnegative matrix. Linear inverse problems c ..."
Abstract

Cited by 19 (2 self)
 Add to MetaCart
(Show Context)
The goal of deconvolution is to recover an image x from its convolution with a known blurring function. This is equivalent to inverting the linear system y = Hx. In this paper we consider the generalized problem where the system matrix H is an arbitrary nonnegative matrix. Linear inverse problems can be solved by adding a regularization term to impose spatial smoothness. To avoid oversmoothing, the regularization term must preserve discontinuities; this results in a particularly challenging energy minimization problem. Where H is diagonal, as occurs in image denoising, the energy function can be solved by techniques such as graph cuts, which have proven to be very effective for problems in early vision. When H is nondiagonal, however, the data cost for a pixel to have a intensity depends on the hypothesized intensities of nearby pixels, so existing graph cut methods cannot be applied. This paper shows how to use graph cuts to obtain a discontinuitypreserving solution to a linear inverse system with an arbitrary nonnegative system matrix. We use a dynamically chosen approximation to the energy which can be minimized by graph cuts; minimizing this approximation also decreases the original energy. Experimental results are shown for MRI reconstruction from fourier data. 1. Generalized Image Deconvolution The goal of image deconvolution is to recover an image x from its convolution with a known blurring function h. This is equivalent to solving the linear inverse problem [1] y = Hx (1) for x given y, where H is the convolution matrix corresponding to h. In this paper we consider the generalized image deconvolution problem, where H is an arbitrary nonnegative matrix. This generalization is motivated by an important problem in medical imaging, namely the reconstruction of MRI images from fourier data. We will discuss this problem in more detail in section 5.1; for the moment, we
Improvements in magnetic resonance imaging using information redundancy
, 2005
"... This thesis describes a number of algorithms related to the acquisition, reconstruction and postprocessing of Magnetic Resonance data. The basic theme underlying each of these algorithms is the use of a unified systems approach to exploit information redundancy available in MR imaging. There are ..."
Abstract

Cited by 1 (1 self)
 Add to MetaCart
This thesis describes a number of algorithms related to the acquisition, reconstruction and postprocessing of Magnetic Resonance data. The basic theme underlying each of these algorithms is the use of a unified systems approach to exploit information redundancy available in MR imaging. There are three basic contributions. The first concerns the development of a new motion correction algorithm for TimeResolved MR Angiography. Motion artifacts in angiography data are very difficult to remove without affecting vascular evolution. Our algorithm uses successive POCS iterations to remove unwanted artifacts without degrading quality. Doubleblind testing has indicated significant improvement over angiograms created manually by experienced radiologist. In summary, our method seeks to exploit temporal redundancy to remove motion artifacts. The second contribution is our recent work on Parallel MR imaging in presence of sensitivity errors using a Maximum Likelihood technique. It can be shown that standard phased array reconstruction using popular parallel imaging methods is inappropriate in presence of errors in measuring sensitivity maps of coils. Since
An Analysis Approach for HighField fMRI Data from Awake NonHuman Primates
"... fMRI experiments with awake nonhuman primates (NHP) have seen a surge of applications in recent years. However, the standard fMRI analysis tools designed for human experiments are not optimal for analysis of NHP fMRI data collected at high fields. There are several reasons for this, including the t ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
(Show Context)
fMRI experiments with awake nonhuman primates (NHP) have seen a surge of applications in recent years. However, the standard fMRI analysis tools designed for human experiments are not optimal for analysis of NHP fMRI data collected at high fields. There are several reasons for this, including the trialbased nature of NHP experiments, with intertrial periods being of no interest, and segmentation artefacts and distortions that may result from field changes due to movement. We demonstrate an approach that allows us to address some of these issues consisting of the following steps: 1) Trialbased experimental design. 2) Careful control of subject movement. 3) Computerassisted selection of trials devoid of artefacts and animal motion. 4) Nonrigid betweentrial and rigid withintrial realignment of concatenated data from temporally separated trials and sessions. 5) Linear interpolation of intertrial intervals and highpass filtering of temporally continuous data 6) Removal of interpolated data and reconcatenation of datasets before statistical analysis with SPM. We have implemented a software toolbox, fMRI Sandbox
Multishot versus SingleShot Pulse Sequences in Very High Field fMRI: A Comparison Using Retinotopic Mapping
"... Highresolution functional MRI is a leading application for very high field (7 Tesla) human MR imaging. Though higher field strengths promise improvements in signaltonoise ratios (SNR) and BOLD contrast relative to fMRI at 3 Tesla, these benefits may be partially offset by accompanying increases i ..."
Abstract

Cited by 1 (0 self)
 Add to MetaCart
(Show Context)
Highresolution functional MRI is a leading application for very high field (7 Tesla) human MR imaging. Though higher field strengths promise improvements in signaltonoise ratios (SNR) and BOLD contrast relative to fMRI at 3 Tesla, these benefits may be partially offset by accompanying increases in geometric distortion and other offresonance effects. Such effects may be especially pronounced with the singleshot EPI pulse sequences typically used for fMRI at standard field strengths. As an alternative, one might consider multishot pulse sequences, which may lead to somewhat lower temporal SNR than standard EPI, but which are also often substantially less susceptible to offresonance effects. Here we consider retinotopic mapping of human visual cortex as a practical test case by which to compare examples of these sequence types for highresolution fMRI at 7 Tesla. We performed polar angle retinotopic mapping at each of 3 isotropic resolutions (2.0, 1.7, and 1.1 mm) using both accelerated singleshot 2D EPI and accelerated multishot 3D gradientecho pulse sequences. We found that singleshot EPI indeed led to greater temporal SNR and contrasttonoise ratios (CNR) than the multishot sequences. However, additional distortion correction in postprocessing was required in order to fully realize these advantages, particularly at higher resolutions. The retinotopic maps produced by both sequence types were qualitatively comparable, and showed equivalent test/retest reliability. Thus, when surfacebased analyses are planned, or in other circumstances where geometric
A consistent and stable approach to generalized sampling
 Journal of Fourier Analysis and Applications
, 2014
"... Abstract We consider the problem of generalized sampling, in which one seeks to obtain reconstructions in arbitrary finite dimensional spaces from finitely samples taken with respect to an arbitrary orthonormal basis. Typical approaches to this problem consider solutions obtained via the consistent ..."
Abstract

Cited by 1 (1 self)
 Add to MetaCart
(Show Context)
Abstract We consider the problem of generalized sampling, in which one seeks to obtain reconstructions in arbitrary finite dimensional spaces from finitely samples taken with respect to an arbitrary orthonormal basis. Typical approaches to this problem consider solutions obtained via the consistent reconstruction technique of Eldar et al and also solutions of overcomplete linear systems. However, the consistent reconstruction technique is known to be nonconvergent and illposed in important cases, such as the recovery of wavelet coefficients from Fourier samples, and whilst the latter approach presents solutions which are convergent and numerically stable when the system is sufficiently overcomplete, the solution becomes inconsistent with the original measurements. In this paper, we consider generalized sampling via a nonlinear minimization problem and prove that the minimizers present solutions which are convergent, numerically stable and consistent with the original measurements. We also provide analysis in the case of reconstructing in compactly supported wavelets from Fourier samples. We show that for wavelets of sufficient smoothness, there is a linear relationship between the number of wavelet coefficients which can be accurately recovered and the number of Fourier samples.
Reconstruction and Intensity Inhomogeneity Correction in Phased Array MR Imaging
"... All intext references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately. ..."
Abstract
 Add to MetaCart
(Show Context)
All intext references underlined in blue are linked to publications on ResearchGate, letting you access and read them immediately.
Thesis and Dissertation Repository. Paper 1040. Fast Absolute Quantification of In Vivo Water and Fat Content with Magnetic Resonance Imaging
, 2012
"... Part of the Biological and Chemical Physics Commons, and the Engineering Physics Commons This Dissertation/Thesis is brought to you for free and open access by Scholarship@Western. It has been accepted for inclusion in Electronic Thesis ..."
Abstract
 Add to MetaCart
(Show Context)
Part of the Biological and Chemical Physics Commons, and the Engineering Physics Commons This Dissertation/Thesis is brought to you for free and open access by Scholarship@Western. It has been accepted for inclusion in Electronic Thesis
Expertise reduces neural cost but does not modulate repetition suppression Expertise Reduces Neural Cost
"... The extent to which repetition suppression is modulated by expertise is currently unknown. We used eventrelated fMRI to test whether architecture students would respond faster to buildings and would exhibit stronger repetition suppression in the fusiform gyrus (FG) and parahippocampa cortex (PHC) t ..."
Abstract
 Add to MetaCart
(Show Context)
The extent to which repetition suppression is modulated by expertise is currently unknown. We used eventrelated fMRI to test whether architecture students would respond faster to buildings and would exhibit stronger repetition suppression in the fusiform gyrus (FG) and parahippocampa cortex (PHC) than students from other disciplines. Behaviorally, we found shorter response latencies with target repetition in all subjects. Moreover, the repetition of targets and distracters was associated with decreased neural responses in the FG and PHC in all subjects. In control, but not in architecture students, reaction times during the first repetition of the target were correlated with activation in the cuneus, lingual gyrus, inferior parietal lobule, insula, and anterior cingulate cortex, indicating that the nonexperts had to recruit additional regions in order to perform the task. Our findings suggest that due to their expertise, architects were able to encode and detect building stimuli at a lower neural cost.
Fast Joint design of RF and Gradient waveforms for MRI parallel excitation
, 2012
"... could not have done the work in this thesis without a tremendous amount of help, support, and sacrifice from many people. First of all, I would like to thank my cochairs, Professor Douglas Noll and Professor Jeffrey Fessler for so much support both in research and life. I think the best teachers ar ..."
Abstract
 Add to MetaCart
(Show Context)
could not have done the work in this thesis without a tremendous amount of help, support, and sacrifice from many people. First of all, I would like to thank my cochairs, Professor Douglas Noll and Professor Jeffrey Fessler for so much support both in research and life. I think the best teachers are those who just not only tell the good lessons, but also show them by their own examples, and I am so lucky to have such special ones as my cochairs. I deeply appreciate that they granted me a chance to join their lab and waited for a long time with incredible patience until I could settle down to MRI research. I hope that someday I would be mature enough to pass along the great lessons I learned from them to others around me. I would also like to present profound gratitude to Professor Anna Gilbert for her great contribution in developing the core of my RF pulse design algorithms introduced in this thesis. Working with Anna was a very pleasant experience, and taught me a lot about how to enjoy research. Professor Thomas Chenevert and Professor Clayton Scott also deserve my gratitude for serving on my committee and returning very helpful suggestions for my thesis. I also deeply thank Korea Foundation of Advanced Studies