iconic grid – improving diagnosis of brain disorders

Download ICONIC Grid – Improving Diagnosis of Brain Disorders

If you can't read please download the document

Upload: ince

Post on 16-Mar-2016

49 views

Category:

Documents


1 download

DESCRIPTION

ICONIC Grid – Improving Diagnosis of Brain Disorders. Allen D. Malony University of Oregon. Professor Department of Computer and Information Science. Director NeuroInformatics Center Computational Science Institute. Outline. Brain, Biology, and Machine Initiative (BBMI) at UO - PowerPoint PPT Presentation

TRANSCRIPT

  • ICONIC Grid Improving Diagnosis of Brain DisordersAllen D. Malony

    University of OregonProfessorDepartment of Computer and Information ScienceDirectorNeuroInformatics CenterComputational Science Institute

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    OutlineBrain, Biology, and Machine Initiative (BBMI) at UONeuroinformatics researchDynamic brain analysis problemNeuroInformatics Center (NIC) at UONeuroinformatics technology and applicationsDense-array EEG and Electrical Geodesics, Inc. (EGI)Epilepsy and pre-surgical planning (Dr. Frishkoff)NIC research and developmentICONIC Grid HPC system at UOIBM HPC solutionsHPC/Grid computing for Oregons science industry

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Brain, Biology, and Machine InitiativeUniversity of Oregon interdisciplinary research in cognitive neuroscience, biology, computer scienceHuman neuroscience focusUnderstanding of cognition and behaviorRelation to anatomy and neural mechanismsLinking with molecular analysis and geneticsEnhancement and integration of neuroimaging facilitiesMagnetic Resonance Imaging (MRI) systemsDense-array EEG systemComputation clusters for high-end analysisEstablish and support UO institutional centers

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Brain Dynamics Analysis ProblemUnderstand functional activity of the human cortexDifferent cognitive research neuroscience contextsMultiple research, clinical, and medical domainsMultiple experimental paradigms and methodsInterpret with respect to physical and cognitive modelsRequirements: spatial (structure), temporal (activity)Imaging techniques for analyzing brain dynamicsBlood flow neuroimaging (PET, fMRI)good spatial resolution functional brain mappingtemporal limitations to tracking of dynamic activitiesElectromagnetic measures (EEG/ERP, MEG)msec temporal resolution to distinguish componentsspatial resolution sub-optimal (source localization)

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Integrated Dynamic Brain Analysis

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Experimental Methodology and Tool Integrationsource localization constrained to cortical surfaceprocessed EEGBrainVoyagerBESACT / MRIEEGsegmented tissues16x256 bits per millisec(30MB/m)mesh generationEMSEInterpolator 3DNetStation

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    NeuroInformatics Center (NIC) at UOApplication of computational science methods to human neuroscience problemsTools to help understand dynamic brain functionTools to help diagnosis brain-related disordersHPC simulation, large-scale data analysis, visualizationIntegration of neuroimaging methods and technologyNeed for coupled modeling (EEG/ERP, MR analysis)Apply advanced statistical analysis (PCA, ICA)Develop computational brain models (FDM, FEM)Build source localization models (dipole, linear inverse)Optimize temporal and spatial resolutionInternet-based capabilities for brain analysis services, data archiving, and data mining

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Funding SupportBBMI federal appropriationDoD Telemedicine Advanced Technology Research Center (TATRC)$40 million research attracted by BBMI$10 million gift from Robert and Beverly Lewis familyEstablished Lewis Center for Neuroimaging (LCNI)NSF Major Research InstrumentationAcquisition of the Oregon ICONIC Grid for Integrated COgnitive Neuroscience Informatics and ComputationNew proposalNIH Human Brain Project NeuroinformaticsGENI: Grid-Enabled Neuroimaging Integration

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Electrical Geodesics Inc. (EGI)EGI Geodesics Sensor NetDense-array sensor technology64/128/256 channels256-channel geodesics sensor netAgCl plastic electrodesCarbon fiber leadsNet StationAdvanced EEG/ERP data analysisStereotactic EEG sensor registrationResearch and medical servicesEpilepsy diagnosis, pre-surgical planning

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    EpilepsyEpilepsy affects ~5.3 million people in the U.S., Europe, & Japan

    EEG in epilepsy diagnosischildhood and juvenile absenceidiopathic (genetic)generalized or multifocal?

    EEG in presurgical planningfast, safe, inexpensive128/256 channels permit localization of seizure onset

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    EEG MethodologyElectroencephalogram (EEG)

    EEG time series analysisEvent-related potentials (ERP) Averaging to increase SNRLinking brain activity to sensorymotor, cognitive functions (e.g., visual processing, response programming)Signal cleaning (removal of noncephalic signal, noise)Signal decomposition (PCA, ICA, etc.)Neural Source localization

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    EEG Time Series - Progression of Absence SeizureFirst full spikewave

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Topographic Waveforms First Full Spike-Wave350ms interval

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Topographic Mapping of Spike-Wave ProgressionPalette scaled for wave-and-spike interval (~350ms)-130 uV (dark blue) 75 uV (dark red)1 millisecond temporal resolution is requiredSpatial density (256ch) to capture shifts in topography

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Animated Topography of SpikeWave DynamicsSpatial & Temporal Dynamics

    Linked NetworksFronto-thalamic circuit (executive control)Limbic circuit (episodic memory)

    Problem of SuperpositionHow many sources?Where are they located?

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Addressing Superposition: Brain Electrical FieldsBrain electrical fields are dipolarVolume conduction depth & location indeterminacy Highly resistive skull (CSF: skull est. from 1:40 to 1:80)Left-hemisphere scalp field may be generated by a right-hemisphere sourceMultiple sources superposition Radial source Tangential sourcesone and two sources varying depths

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Source LocalizationMapping of scalp potentials to cortical generatorsSignal decomposition (addressing superposition)Anatomical source modeling (localization)Source modellingAnatomical ConstraintsAccurate head model and physicsComputational head model formulationMathematical ConstraintsInverse solutions apply mathematical criteria such as smoothness (LORETA) to constrain the solution

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Dipole Sources in the CortexScalp EEG is generated in the cortexInterested in dipole location, orientation, and magnitudeCortical sheet gives possible dipole locationsOrientation is normal to cortical surfaceNeed to capture convoluted geometry in 3D meshFrom segmented MRI/CTLinear superposition

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Advanced Image SegmentationNative MR gives high gray-to-white matter contrastImage analysis techniquesEdge detection, edge merger, region growingLevel set methods and hybrid methodsKnowledge-basedAfter segmentation, color contrasts tissue typeRegistered segmented MRI

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Building Computational Brain ModelsMRI segmentation of brain tissuesConductivity modelMeasure head tissue conductivityElectrical impedance tomographysmall currents are injected between electrode pairresulting potential measured at remaining electrodesFinite element forward solutionSource inverse modelingExplicit and implicit methodsBayesian methodology

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Conductivity ModelingGoverning Equations ICS/BCSDiscretizationSystem of Algebraic EquationsEquation (Matrix) Solver Approximate SolutionContinuous SolutionsFinite-Difference Finite-Element Boundary-Element Finite-Volume SpectralDiscrete Nodal ValuesTridiagonal ADI SOR Gauss-Seidel Gaussian elimination (x,y,z,t) J (x,y,z,t) B (x,y,z,t)

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Alternating Direction Implicit (ADI) MethodFinite difference method C++ and OpenMP on IBM p655 running Linux305 seconds

    Chart2

    1111

    1.83150183151.86085343231.87773083452

    2.53164556962.60181582362.7032647123

    3.09597523223.20140440473.32169321184

    3.62318840583.82970599473.79220652945

    4.20168067234.37609075044.46454123986

    4.71698113214.82211538464.93499098427

    4.73933649295.17543859655.1073860668

    64x64x44

    128x128x88

    256x256x176

    ideal speedup

    number of processors

    speedup

    ADI speedup

    Sheet1

    MachineComp64x64x44128x128x88256x256x176

    timetime(s)speeduptimetime(s)speeduptimetime(s)speedup

    p655-1xlC1"0:10.00"101.00"1:40.3"100.31.00"26.00.43"15601.00

    p655-1xlC2"0:05.46"5.461.83"0:53.9"53.91.86"13:50.79"830.791.88

    p655-1xlC3"0:03.95"3.952.53"0:38.55"38.552.60"9:37.08"577.082.70

    p655-1xlC4"0:03.23"3.233.10"0:31.33"31.333.20"7:49.64469.643.32

    p655-1xlC5"0:02.76"2.763.62"0:26.19"26.193.83"6:51.37"411.373.79

    p655-1xlC6"0:02.38"2.384.20"0:22.92"22.924.38"5:49.42"349.424.46

    p655-1xlC7"0:02.12"2.124.72"0:20.8"20.84.82"5:16.11"316.114.93

    p655-1xlC8"0:02.11"2.114.74"0:19.38"19.385.18"5:05.44"305.445.11

    neuronicf77Fortran"0:57.24"57.24"12:40.55"760"2:08:44.92"7724.92

    neuronicg++Seq"0:13.83"13.83"2:36.44"156.44"41:05.81"2465.81

    p655-1xlCSeq"0:10.73"10.7301:48.8108.8"27:25.02"1645.02

    Sheet1

    64x64x44

    128x128x88

    256x256x176

    ideal speedup

    number of processors

    speedup

    ADI speedup

    Sheet2

    Running the ADI program for 200 iterations

    Sheet3

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Source Modeling with Standard Brain MRI ModelSource model for anterior negative slowwave (100-200 ms)Source model for first medial positive wave (216-234 ms)Source model for second medial positivewave (256-308 ms)

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    UO ICONIC GridNSF Major Research Instrumentation (MRI) proposalAcquisition of the Oregon ICONIC Grid for Integrated COgnitive Neuroscience Informatics and ComputationPIsComputer Science: A. Malony, J. ConeryPsychology: D. Tucker, M. Posner, R. NunnallySenior personnelComputer Science: S. Douglas, J. CunyPsychology: H. Neville, E. Awh, P. WhiteComputational, storage, and visualization infrastructure

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    ICONIC GridSMPServerIBM p655GraphicsSMPSGI PrismSAN Storage System IBM SAN FSGbit Campus BackboneNICCISCISInternet 2SharedMemoryIBM p690DistributedMemoryIBM JS20CNIDistributedMemoryDell Pentium XeonNIC4x816162x82x16graphics workstationsinteractive, immersive vizother campus clusters5 TerabytesTape Backup

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    ICONIC Grid Hardwarep690 16 processorsp655 4 nodes 8 processors per nodeDell cluster 16 nodes 2 processors per nodeJS20 Blade 16 nodes 2 processors per nodeFAStT storage 5 TBSAN FSFibre ChannelFibre Channel

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Computational Integrated Neuroimaging System

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Leveraging Internet, HPC, and Grid ComputingTelemedicine imaging and neurologyDistributed EEG and MRI measurement and analysisNeurological medical servicesShared brain data repositoriesRemote and rural imaging capabilitiesNeet to enhance HPC and grid infrastructure in OregonBuild on emerging web services and grid technologyEstablish HPC resources with high-bandwidth networksCreate institutional and industry partnershipsCerebral Data Systems (UO partnership with EGI)Continue strong relationship with IBM and Life Sciences

    IBM TheatreSC 2004ICONIC Grid Improving Diagnosis of Brain Disorders

    Oregon E-Science Grid Region 3

    In cognitive neuroscience studies, it is has been shown that EEG and fMRI studies provide complimentary information that informs researchers about brain functions. We believe that the EEG can be used to compliment other brain imaging methods.Constraining the source solutions to the cortical surface is a major advantage for analyzing EEG and ERP effects. A working assumption is that sources are likely to be oriented normal to the cortical surface.

    Unifying Features of Idiopathic Generalized EpilepsiesNormal neurologically before onset of seizuresEEG changes appear generally well organized with otherwise normal backgroundLong term outlook for cognitive function is good

    Many of the epilepsies previously classified as generalized are not truly generalized. Idiopathic Generalized Seizures occur in neurologically normal childrenMost seizures or epilepsy syndromes previously classified as generalized or undetermined should be thought of as Epileptic EncephalopathiesEpileptic Encephalopathies may have an identifiable mechanism or respond to specific therapies

    Figure 1. Topographic waveform plot for the first full spike-wave in one patients seizure. Following the clinical EEG convention, negative is up. The 256 channels are arrayed in two dimensions as they would be seen looking down on the top of the head, with the nose at the top of the page, and with the lower channels unwrapped to the sides of the page. This 350 ms epoch captures the first wave and spike of this seizure, plus the initial onset of the second wave. Note that the prominent spike-wave pattern over medial, superior frontal sites inverts over lateral, inferior frontal sites, indicating neural sources in medial frontal cortex.Figure 1. Topographic waveform plot for the first full spike-wave in one patients seizure. Following the clinical EEG convention, negative is up. The 256 channels are arrayed in two dimensions as they would be seen looking down on the top of the head, with the nose at the top of the page, and with the lower channels unwrapped to the sides of the page. This 350 ms epoch captures the first wave and spike of this seizure, plus the initial onset of the second wave. Note that the prominent spike-wave pattern over medial, superior frontal sites inverts over lateral, inferior frontal sites, indicating neural sources in medial frontal cortex.Figure 2. Selected topographic maps for the spike-wave pattern shown in Figure 3. Palette is scaled for this wave and spike interval, from 130 uV (dark blue) to 75 uV (dark red). The interval characterized by each map is 1 ms, with the selections made to illustrate the major topographic transitions of this subjects spike-wave pattern.

    because the inverse problem

    anatomical constraints (head model)mathematical constraints (mathematical constraints such as smoothness)Figure 3. Distributed electrical source models (Low Resolution Electromagnetic Topography) for intervals of the spike-wave cycle shown in Figure 3 and 4. The source distributions are visualized in relation to a standard brain MRI model. The top row shows the source model that characterized the anterior negative slow wave (100 to 200 ms in Figure 4), which for this person engaged broad regions of the dorsolateral frontal lobes, particularly on the right. The middle row shows the first medial positive wave (216 to 234 ms), which was modeled by this method to include the medial as well as lateral walls of the frontal lobe. The bottom row shows the second medial positive wave (256 to 308 ms), which progressed anteriorly along the medial frontal lobe and engaged frontopolar cortex, again particularly on the right side.