Center for Scientific Computation in Imaging


CSCI Research

Active Grants

NIH R01AG054049: New Quantitative Neuroimaging Metrics of Structural and Functional Connectivity of the Locus Coeruleus as a Novel Biomarker of Alzheimer's Disease Pathogenesis and Progression

The original theory offered by Braak & Braak (1991)—that neurofibrillary tangle pathology proceeds along well-defined predilection sites beginning in the medial temporal cortex—has been modified by the same author to suggest that the pathologic process instead commences in the lower brainstem (Braak et al 2011). The first visible pathologic changes are now thought to occur in the locus coeruleus (LC) and then spread via its axonal projections to transentorhinal/entorhinal cortex (TEC). We propose to study LC change using a novel computational morphology method, combined with novel methods of measuring white matter microstructural tractography and functional connectivity to TEC. These new methods are designed to overcome major shortcomings in current neuro-­MRI analysis methods that limit the ability to detect subtle structural and functional changes associated with early AD. Such alterations across the aging-­MCI-­AD continuum, as well as in those cognitively normal individuals with risk factors for AD (e.g., CSF AD biomarkers; apolipoprotein E ee4 carriers), would provide significant advances in our understanding of the pathogenesis of AD across clinical transition points and perhaps during this ‘silent’ period (i.e., prior to the occurrence of traditional AD biomarker positivities). Using our newly developed diagnostic and MRI metrics, we propose to quantify variations in LC morphology and its projections to TEC (which we term the LC-­TEC system).

We recently received supplemental funding under this award to augment the structural morphology estimates with our Joint Estimation Diffusion Imaging (JEDI) method to provide improved sensitivity to the assessment of gray matter (GM) characteristics, and sub-voxel diffusion anisotropy, facilitating assessment of GM tissue status, and improving anisotropy estimates in white matter (WM). This supplement also augments the functional modes and connectivity estimates with functional tractography enhanced by averaged structural GM/WM tissue constraints derived from our JEDI method to provide improved sensitivity to assessment of functional modes and connectivity.

Principal Investigators:  Dr. Lawrence Frank; Dr. Mark Bondi (Psychiatry, UCSD)

NSF Award 2114860: Collaborative Research: Detection and Estimation of MultiScale Complex Spatiotemporal Processes in Tornadic Supercells from High Resolution Simulations and Multiparameter Radar

Supercell thunderstorms are prolific producers of tornadoes, and are entirely responsible for the strongest, most long-lived tornadoes that are responsible for the bulk of storm-related death and destruction. Theoretical, observational, and numerical research on supercells has improved our understanding of these powerful storms, including the environments favorable for supercells of varying strengths, as well as the internal processes associated with tornado formation. Despite these advances, tornadoes produced from supercells elude accurate prediction and continue to cause fatalities annually within the United States. Before significant improvements in forecasting tornadoes can be made, the factors that determine whether a supercell will produce a long-lived violent tornado, versus no / weak tornado, must be elucidated. In the proposed work, ensembles of tornado-resolving supercell simulations in varied environments will be conducted using the Bryan Cloud Model (CM1) and analyzed in novel ways on the NSF-sponsored Frontera supercomputer. Entropy Field Decomposition (EFD), a Bayesian analysis technique that excels in the automated detection of statistically significant patterns in large volumes of spatiotemporal data, will serve as a unifying framework for all data sources used in this proposal. These data sources include raw model data, radar-simulated model data, and rapid-scan radar observational data of tornadic and nontornadic supercells. The utility of EFD to detect features/processes that elude traditional meteorological analysis techniques will be a centerpiece of this proposal.

To our knowledge, this is the first large set (~100) of tornado-resolving ensemble simulations and detailed intercomparisons between mobile radar data and tornado-resolving simulations. EFD is a new technique to the field of meteorology that can greatly aid in the automated analysis and comparison of large amounts of spatiotemporal data originating from both observations and simulations without the need for large amounts of training data. Project findings can potentially improve understanding and prediction of supercell thunderstorms and benefit society by mitigating loss of life.

Principal Investigators: Dr. Lawrence Frank; Dr. Vitaly Galinsky; Dr. David Bodine (University of Oklahoma Norman); Leigh Orf (University of Wisconsin-Madison)

NIH R01AR070830: Non-invasive measurements of muscle microstructure assessed by diffusion tensor imaging

Skeletal muscle is important for function, metabolism, and cardiovascular health, and therefore, injury and disease adversely affect quality of life and healthcare costs. Currently, muscle biopsy is the gold standard for diagnosis and monitoring muscle disease and recovery, but is invasive, costly, and prone to sampling errors, and not conducive to serial monitoring of muscle health. It is also semi- quantitative, and often difficult to extrapolate to the entire muscle, limiting its scientific value, thus a new approach is required. MRI has been used to noninvasively quantify changes in volume, fat distribution, and water content in muscle. Diffusion tensor imaging (DTI) is a version of MRI that measures anisotropic diffusion of water, which is related to tissue microstructure, but tends to yield non-specific changes regardless of the injury or disease state. The key reason for this lack of specificity is that the explicit relationships between microstructure and diffusion have not been rigorously studied, nor carefully calibrated. To address this gap in knowledge, the purpose of this proposal is to compare muscle microstructure and MRI diffusion properties of muscle in novel and tightly controlled computer simulations, precision engineered phantoms, and animal models of muscle injury and disease. Our central hypothesis is that DT-MRI can be directly related to muscle microstructural changes, when appropriate pulse sequences are used to uncouple complex pathology. Aim #1 will use computer-based simulations of muscle structure and biochemistry to carefully understand how diffusion is related to multiple muscle microstructural changes. Aim #2 will utilize 3D precision-engineered models to relate diffusion to muscle structure in real-world DT-MRI experiments. These experiments will be integrated into a final in vivo set of experiments (Aim #3), which are designed to test the accuracy of DT-MRI to uncouple complex microstructural changes in the presence of muscle atrophy, inflammation, and degeneration. These experiments will elucidate the understudied relationships between microstructure and diffusion in muscle. The long-term goal is to serially quantify muscle microstructure non-invasively. This approach is innovative in that it combines state-of-the art imaging, simulation, nanofabrication, and morphology methods to generate a clinically meaningful measurement tool.

Principal Investigators: Dr. Samuel Ward (Orthopaedics/Radiology, UCSD); Dr. Lawrence Frank

NVIDIA Academic Hardware Grant

The NVIDIA Academic Hardware Grant Program endeavors to advance education and research by enabling groundbreaking, innovative, and unique academic research projects with world-class computing resources.

This project is using novel space-time analysis methods to study the generation and maintenance of violent, long-track tornadoes. By comparing space-time storm modes estimated from very high spatial and temporal resolution numerical simulations generated on a supercomputer and their simulated radar signatures with actual mobile Doppler radar collected during severe weather field projects, we will gain a better understanding of the physical processes that lead to tornado genesis and maintenance and enhance the ability of forecasters to detect their signatures in radar data earlier, leading to more timely and accurate public forecasting of these dangerous and costly severe weather events. Fast, efficient visualization of very large multivariate time dependent volumetric data is critical to every stage of this program, and requires the most advanced visualization software and hardware.

Principal Investigators: Dr. Lawrence Frank

Recently Completed Grants

NSF ACI-1560405: INSPIRE: Quantitative Estimation of Space-Time processes in volumetric data (QUEST)

This project involves the development of advanced computational methods for accurate quantitative characterization of parametric features embedded within temporally varying high-resolution 3D voxel- based digital imaging modalities. We will develop a numerical implementation of a general analysis of spatio-temporal data based upon our recently developed entropy field decomposition (EFD) theory, a probabilistic method efficiently incorporating prior information from multiple data modalities, including numerical simulations, to automatically discern subtle space-time patterns in data from complex physical processes. The project is motivated by our discovery that two important but seemingly disparate scientific problems: characterizing structure-function relationships in the human brain from magnetic resonance imaging (neuro-MRI) data and tornadogenesis and maintenance in severe thunderstorms as observed by mobile Doppler radar (MDR) systems and simulated with numerical models may be commonly approached using our methods. Neuro-MRI data from the Human Connectome Project, combined with numerical simulations of neuro-MRI signals using our MR diffusion simulation platform, DiffSim, and MDR data from the Doppler- On-Wheels system, combined with tornado simulations using the CM1 model, is the focus of the general cross-disciplinary methodology we propose.

Principal Investigators: Dr. Lawrence Frank; Dr. Joshua Wurman (CSWR, Boulder, CO); Dr. Leigh Orf (UW-Madison, WI)

NSF ACI-1440412: SI2-SSE: Wavelet Enabled Progressive Data Access and Storage Protocol (WASP)

The major goal of this project was to develop a common software framework for supporting a multi-scale progressive data refinement method, based upon the representation of data as a wavelet expansion, that enables interactive exploration of very large data sets for the bio- and geo-sciences communities. This allows the multi-scale analysis, storage, and visualization of 'big data' collected in a wide range of disciplines and on a multitude of platforms, from high end computing systems, to personal laptop computers used by students and researchers out in the field.

Principal Investigators: Dr. Lawrence Frank; John Clyne (NCAR, Boulder, CO)

NIH R01MH096100: Diffusion Imaging in Gray Matter

The major goals of this project were to extend idealized theoretic models of brain white matter (WM) and grey matter (GM) to more realistic physiological models through numerical simulations; 2) Perform high field experiments on well-characterized WM and GM phantoms to validate both our extended theoretical and simulation models, and on excised normal rat brains to assess variations from idealized models; 3) Develop a clinical double pulsed field gradient (dPFG) pulse sequence strategy to be tested on normal humans. Our central hypothesis is that the dPFG method is sensitive to sub-voxel tissue structure manifest in three measurable forms of diffusion anisotropy: microscopic anisotropy (uA), compartment shape anisotropy (CSA), and ensemble anisotropy (EA), which can be used in-vivo for the quantitative assessment of GM architecture in humans.

Principal Investigator: Dr. Lawrence Frank

NSF DBI-1147260: Collaborative Research: Shape Analysis for Phenomics with 3D Imaging Data (SAPID)

The major goal of this project was to develop the Shape Analysis for Phenomics with 3D Imaging Data (SAPID) Toolkit, or STK, for quantitative morphological analysis of 3D volumetric imaging data, primarily focusing on MRI and CT data acquired from fishes and other zoological specimens. Our specific aims include; 1) Development of semiautomated geometric morphometry systems for noisy 3D MRI data based on diffeomorphic spatial normalization and geometric metamorphosis methods, and production of software useable by the biological community; 2) Development of a robust and efficient automated system for signature-based shape analysis for 3D noisy MRI image data; and, 3) Application of these new shape description methods to two classic and outstanding problems in evolutionary and comparative morphology.

Principal Investigator: Dr. Lawrence Frank

Research Partners

National Science Foundation
National Institutes of Health
Department of Veteran's Affairs
National Center for Atmospheric Research
San Diego State University
NVIDIA Corporation
Institute for Engineering in Medicine, UC San Diego
Scripps Institution of Oceanography, UC San Diego
Birch Aquarium, UC San Diego
University of California, San Diego
Center for Functional MRI, UC San Diego
Muscle Physiology Laboratory, UC San Diego