This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. Primary support for the subproject and the subproject's principal investigator may have been provided by other sources, including other NIH sources. The Total Cost listed for the subproject likely represents the estimated amount of Center infrastructure utilized by the subproject, not direct funding provided by the NCRR grant to the subproject or subproject staff. The overall goal of this project is to develop and evaluate a new Bayesian reconstruction method for low-resolution MRI modalities that reduce artifacts and effectively increase resolution relative to standard Discrete Fourier Transform approaches. The new reconstruction method fully utilizes k-space data to reduce artifacts and the increase in resolution is achieved by incorporating high-resolution information from segmented structural MRI scans acquired in the same scanning session. The focus of the work within the Resource Research application will be to directly apply, extend and validate the Bayesian reconstruction methodology. Perfusion-Weighted MR Imaging (PWI) is the particular modality chosen for application.
The Specific Aims of the project are to 1) To apply the Bayesian low-resolution reconstruction algorithm to PWI;2) To validate the Bayesian algorithm relative to standard DFT reconstruction;3) To study the robustness of the Bayesian reconstruction method to miss-registration and segmentation error;and 4) To apply the Bayesian reconstruction methodology to a full clinical study. The Administrative Supplement project aims to AS1) extend methods to 3D GRASE and AS2) evaluate and validate K-Bayes for 3D GRASE.
Aim 1 : To apply the Bayesian low-resolution reconstruction algorithm to PWI. The low-resolution Bayesian reconstruction algorithm has been developed as a general procedure for reconstructing low-resolution MRI modalities. The algorithm will be adapted and applied to PWI datasets for which k-space data has been saved along with corresponding structural MRIs. The Bayesian model will be applied to data describing the change between tagged and untagged perfusion scans, i.e. the complex difference of the tagged and untagged conditions Aim 2: To validate the Bayesian algorithm relative to standard DFT reconstruction. Validation will be performed on both simulated and real data. The simulated data will mimic PWI as accurately as possible based on physical and biological knowledge. The real data will be acquired at higher-resolution than standard acquisition. Currently the volumetric PWI sequence implemented in the laboratory takes 34s to acquire. With improved acquisition sequences to be developed in the acquisition core, along with developments for parallel imaging described elsewhere in the reconstruction core, the inherent PWI resolution will be increased by a factor of two, i.e. to approximately 2x2x2 mm. This relatively high-resolution data can then be utilized as a gold standard that will be down-sampled to give low-resolution data by cutting out the central region of k-space.
Aim 3 : To study the robustness of the Bayesian reconstruction method to miss-registration and segmentation error. Co-registration error between the structural and perfusion MRIs, and segmentation error, are both potential confounding factors for the Bayesian reconstruction algorithm. Since the nature of the propagation of these errors is largely unknown, the Bayesian reconstructions will be examined in the presence of these errors and will be assessed based on metrics of overall error such as root mean square error. Both simulated and real data will be tested.
Aim 4 : To apply the Bayesian reconstruction methodology to a full clinical study. The Bayesian reconstruction algorithm as well as DFT will be applied to a small set of subjects with the objective of comparing perfusion in pathological conditions, e.g. Post traumatic Stress Disorder (PTSD) patients, who may present deactivation of certain brain regions relative to healthy controls. Statistical analysis to determine group differences will be performed based on the data from each reconstruction method. This will allow a quantitative assessment as to whether data reconstructed by the Bayesian algorithm method provides greater sensitivity and specificity in characterizing perfusion changes than does data reconstructed via DFT. Administrative Supplement Aims:
AS Aim 1 : To extend K-Bayes PMRI methodology to 3D k-space reconstruction of 3D GRASE. We will adapt K-Bayes to reconstruct 3D k-space 3D GRASE data. This requires extending the signal model of the parent proposal to 3D k-space acquisitions and then relating it to high-resolution anatomical information. Although K-Bayes modeling for 3D k-space is a relatively straightforward mathematical extension of the multi-slice model, careful algorithmic development, optimization, and programming is required to minimize increases in computational burden due to the additional 3rd k-space dimension. We propose to develop specialized (DFT-style) butterfly steps within the Expectation-Maximization (EM) algorithm to enable -order computations in contrast to the -order required for the standard algorithm. These butterfly steps need to be specially developed because the Fourier relationship is between non-equivalent dimension sizes: i.e., that of low k-space and high image space resolution.
AS Aim 2 : To validate K-Bayes methodology for 3D GRASE with test datasets. We will (a) determine the level of improvement of 3D GRASE K-Bayes over DFT and other reconstruction methods, and (b) investigate the propagation of error sources to quantifying the robustness of 3D GRASE K-Bayes. We propose to use a control-system-forward mechanism for validation using controlled simulations and optimized in vivo gold standards. We anticipate that for typical error levels, 3D GRASE K-Bayes will produce significantly more accurate and precise reconstructions than DFT-based methods.

Agency
National Institute of Health (NIH)
Institute
National Center for Research Resources (NCRR)
Type
Biotechnology Resource Grants (P41)
Project #
5P41RR023953-04
Application #
8362777
Study Section
Special Emphasis Panel (ZRG1-SBIB-J (40))
Project Start
2011-07-01
Project End
2012-06-30
Budget Start
2011-07-01
Budget End
2012-06-30
Support Year
4
Fiscal Year
2011
Total Cost
$35,549
Indirect Cost
Name
Northern California Institute Research & Education
Department
Type
DUNS #
613338789
City
San Francisco
State
CA
Country
United States
Zip Code
94121
Kuceyeski, A; Shah, S; Dyke, J P et al. (2016) The application of a mathematical model linking structural and functional connectomes in severe brain injury. Neuroimage Clin 11:635-647
Lam, Fan; Liu, Ding; Song, Zhuang et al. (2016) A fast algorithm for denoising magnitude diffusion-weighted images with rank and edge constraints. Magn Reson Med 75:433-40
Pannetier, Nicolas A; Stavrinos, Theano; Ng, Peter et al. (2016) Quantitative framework for prospective motion correction evaluation. Magn Reson Med 75:810-6
Kuceyeski, Amy; Navi, Babak B; Kamel, Hooman et al. (2016) Structural connectome disruption at baseline predicts 6-months post-stroke outcome. Hum Brain Mapp 37:2587-601
Friedman, Eric J; Young, Karl; Tremper, Graham et al. (2015) Directed network motifs in Alzheimer's disease and mild cognitive impairment. PLoS One 10:e0124453
Kuceyeski, Amy; Navi, Babak B; Kamel, Hooman et al. (2015) Exploring the brain's structural connectome: A quantitative stroke lesion-dysfunction mapping study. Hum Brain Mapp 36:2147-60
Ma, Chao; Liang, Zhi-Pei (2015) Design of multidimensional Shinnar-Le Roux radiofrequency pulses. Magn Reson Med 73:633-45
Zhao, Bo; Lu, Wenmiao; Hitchens, T Kevin et al. (2015) Accelerated MR parameter mapping with low-rank and sparsity constraints. Magn Reson Med 74:489-98
Lu, Zhao-Hua; Zhu, Hongtu; Knickmeyer, Rebecca C et al. (2015) Multiple SNP Set Analysis for Genome-Wide Association Studies Through Bayesian Latent Variable Selection. Genet Epidemiol 39:664-77
Raj, Ashish; LoCastro, Eve; Kuceyeski, Amy et al. (2015) Network Diffusion Model of Progression Predicts Longitudinal Patterns of Atrophy and Metabolism in Alzheimer's Disease. Cell Rep :

Showing the most recent 10 out of 120 publications