42
Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information Science Don Tucker Electrical Geodesics, Inc

Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Embed Size (px)

Citation preview

Page 1: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Neuroinformatics for Telemedicine and Medical Services

Neuroinformatics CenterUniversity of Oregon

Allen D. MalonyDepartment of Computerand Information Science

Don TuckerElectrical Geodesics, Inc.

Page 2: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

NeuroInformatics Center (NIC) at UO

Computational science applied to human neuroscience Tools to help understand dynamic brain function Tools to help diagnosis brain-related disorders HPC simulation, complex data analysis, medical services

Integration of neuroimaging methods and technology Coupled measures and evaluation (EEG/ERP, MR) Advanced statistical signal analysis (PCA, ICA) Advanced image analysis (segmentation, anatomy) Computational head modeling (electromagnetics, FDM) Source localization modeling (dipole, linear inverse)

Internet-based capabilities for brain analysis services, data archiving, and data mining

Page 3: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

NIC Organization

Allen D. Malony, Director Don M. Tucker, Associate Director Sergei Turovets, Computational Physicist Bob Frank, Senior Data Analyst Kai Li, Computer Scientist Chris Hoge, Computational Software Engineer Matt Sottile, Computer Scientist, CIS Department Dejing Dou, Computer Scientist, CIS Department Gwen Frishkoff, Neuro Scientist, Wisconsin Medical C. Brad Davidson, Systems administrator Adnan Salman, Ph.D. student, Computer Science Jason Sydes, Ph.D. student, Computer Science

Page 4: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

NIC Scientists/Students and Projects Sergei Turovets, Ph.D., Physics

Physics-based computational models of human head tissues for developing electrical and optical probes of brain activity

Kai Li, Ph.D., Computer ScienceTissue segmentation of neurological images, cortical surface extraction of the individual brain, and brain visualization

Bob Frank, M.S., MathematicsStatistical analysis of neurophysiological recordings to separate brain from non-brain signals in scalp and source space

Chris Hoge, M.S., Computer ScienceComputational software development, high-performance signal analysis tools, application server environments, EEG and MRI databases, neuroinformatics workflow

Matt Sottile, Ph.D., Computer ScienceEEG data modeling methods for detecting spike and seizure

Dejing Dou, Ph.D., Computer ScienceData mining and ontologies

Gwen Frishkoff, Ph.D., Psychology, Wisconsin Medical CollegeAutomated component separation, ERP pattern classification, and neuro electromagnetic ontologies

Adnan Salman, Ph.D. student, Computer ScienceOptimization solutions for finding human head tissues conductivity

Jason Sydes, M.S. student, Computer ScienceEEG and MRI database development

Page 5: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Dense-array EEG signal analysis and decomposition Artifact cleaning and component analysis

Automatic brain image segmentation Brain tissue identification Cortex extraction

Computational head modeling Tissue conductivity estimation Source localization

Statistical analysis to detect brain states Discriminant analysis Pattern recognition

Electromagnetic databases and ontologies HPC, data management, workflow, services delivery

Neuroinformatic Challenges

Page 6: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Observing Dynamic Brain Function

Brain activity occurs in cortex Observing brain activity requires

high temporal and spatial resolution Cortex activity generates scalp EEG EEG data (dense-array, 256 channels)

High temporal (1msec) / poor spatial resolution (2D) MR imaging (fMRI, PET)

Good spatial (3D) / poor temporal resolution (~1.0 sec) Want both high temporal and spatial resolution Need to solve source localization problem!!!

Find cortical sources for measured EEG signals

Page 7: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Electromagnetics Modeling / Source Localization

Page 8: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Computational Head Models Source localization requires modeling

Full physics modeling of human head electromagnetics

Step 1: Head tissue segmentation Obtain accurate tissue geometries

Step 2: Numerical forward solution 3D numerical head model Map current sources to scalp potential

Step 3: Conductivity modeling Inject currents and measure response Find accurate tissue conductivities

Step 4: Source optimization Create cortex dipoles / lead field matrix

Page 9: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Brain Tissue Segmentation (K. Li)

Exploit various prior knowledge Structural, geometrical, morphological, radiological

Segmentation workflow Classifying voxel types of entire image then extract brain

Two core segmentation techniques Relative thresholding for voxel classification Morphological image analysis for brain extraction

Page 10: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Cortical Surface Reconstruction

Performed after brain tissue segmentation Use the marching cube isosurface algorithm Guarantee topology correctness Application to surface tessellation and dipole creation

Page 11: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Skull Warping

Skull atlas and T1 MRI (inputs) Performed after brain extraction Transformation sequences

Scaling the brain volume Rigid transformation

(translation, rotation) Affine transformation

(translation, rotation,scaling, and shearing)

Deformation field based transformation

For head model construction when individual skull unavailable

Skull warping

Page 12: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

BrainK Segmentation and Cortex Extraction

GUI-based application or command-line programs Fully automatic processing in both modes

Segmentation tuning with global parameters Allows segmentation editing

segmentation result

corticalsurface

visualizationworkflow

parameters

Page 13: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Modeling Head Electromagnetics (S. Turovets) Head volume conduction (isotropic Poisson equation)

Need to model tissue and skull anisotropy With existing principal axes, the tensor is symmetrical with

6 independent terms: ij = ji

Numerical implementation so far dealt with the orthotropic case: ii are different, all other components of ij = 0, ij.

Complete anisotropic forward solver with arbitrary ij implemented

()=S in , - scalar function of (x,y.z)

-() n = 0 on

(ij)=S in , ij - tensor function of (x,y.z)

-ij() n = 0 on

Page 14: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Conductivity Optimization (A. Salman)

Correct source inverse solutions depend on accurate estimates of head tissue conductivities Scalp, skull, brain, CSF, …

Design as a conductivity search problem Estimate conductivity values Computer forward solution and compare to measured Iterate until error threshold is obtained (global minimum)

Use electrical impedance tomography methods Multiple current injection pairs

(source, sink) Parallelized conductivity search

currentsource

currentsink

measurementelectrodes

Page 15: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Conductivity Scanning and Registration (EGI)

EGI geodesic sensor net integration Scanning current injection hardware EEG and bounded EIT data acquisition

EGI photogrammetry system Machine vision for sensor position registration EGI tool for validation, correction, and visualization

Page 16: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Directed Components Analysis (R. Frank)

EEG artifacts lead to errors in EEG analysis EEG artifact identification and removal

Apply signal analysis algorithms (PCA, ICA) Directed components analysis (DCA)

DCA dynamically models the artifactual and non-artifactual (cortical) activity in the recorded EEG

Artifactual and cortical models facilitate extraction of artifactual activity while preserving cortical activity Target application to eye blink removal

Computationally efficient implementation permits real-time artifact extraction

Page 17: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

DCA Extraction of Eye Blink Intensity

Inner product of DCA spatial filter and EEG scalp topography at each time sample extracts temporal stream of eye blink intensity

EEG Data Window

EEG1 …EEG2 EEG3 EEGn

Spatial Filter

K*EB Blink Intensity

……

Topographies

(time slices)

Channels

(time series)

or

… …

Page 18: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

DCA Extraction of Eye Blink Artifacts

Extraction is applied to overlapping EEG data windows permitting updating of artifactual and cortical models

Blink - Free EEG

“Blinky” EEG= - B

T *Blink

Intensity

Page 19: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Neural ElectroMagnetic Ontologies (NEMO)

How can EEG data be compared across laboratories? Need a system for representation, storage, mining, and

dissemination of electromagnetic information Need standardization of methods for measure generation

and classification of information Identification and labeling of components Patterns of interest

NEMO will address issue by providing Spatial and temporal ontology database Use for data representation, mining, and meta-analysis Components in average EEG and MEG (ERPs)

D. Dou, G. Frishkoff, R. Frank

Page 20: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

NEMO: Neuro ElectroMagnetic Ontologies

Page 21: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

ERP Pattern Analysis and Classification

Extract ERP patterns onto PCA / ICA components

P100

N100

fP2

P1r /N3

P1r /MFN

P300

100ms

170ms

200ms

280ms

400ms

600ms

scalp data

Page 22: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Statistical Metric Generation

Supports Knowledge Discovery through Data (KDD) component for NEMO EEG ontology development

Quantifies attributes of PCA / ICA components Spatial, temporal and functional attributes

Metrics may combine with expert-defined rules to automatically match components to ERP patterns

expertrules

Measures Truth table

Page 23: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Computational Integrated Neuroimaging System

… …

raw

storageresources

virtualservices

compute resources

Page 24: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

CDS Medical Services Software Layers (C. Hoge)

Page 25: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

CDS Work Flow System Architecture

Page 26: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Application Server User Package

Page 27: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Application Server Job Management Package

Page 28: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

MR/EEG Database Schema (J. Sydes)

Support experiment information hierarchy Uses XCEDE

Support for workflows MR image analysis Head model building NEMO

Supports workflow provenance

MySQL DB engine

Page 29: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Companion Slides

Page 30: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Skull and Brain Anisotropy Parameterization

Total number of unknowns for skull conductivity N = 1 + 2N

MRI DT brain map (Tuch et al, 2001)

r t

Linear relation betweenconductivity and diffusiontensor eigenvalues Parameterization:

= K (d - d0)

Page 31: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

bEIT Procedure

3D subject geometry CT-registered with MRI Individual subject CT is not required

Sensor positions are acquired by photogrammetry system and registered with head model

Finite difference method solution for particular set of tissue conductivities

Search for conductivity solutions

Page 32: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Tissue Conductivity Estimation (Results)Table 1. The Human Subject Experiments Results, Conductivity in S/m

Subj./site MRI/CT Age/Gend. Skull Scalp Brain11/26-90 Yes 40/M .020 .46 .2311/46-109 .01012/26-2 Yes 38/M .01112/15-72 .016 .52 .36

Page 33: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

ODESSI (A. Salman)

Open Domain-enabled Environment for Simulation-based Scientific Investigations

Page 34: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Domain Problems in Neuroscience for ODESSI Scientific investigations

Verification forward and inverse solvers (4-shell sphere) grid convergence, time step, special step

Validation phantoms, animal experiments, data from surgery

Uncertainly and sensitivity analysis sensor position, injected current, tissue conductivities, …

Parameter sweep Optimization, extract optimal parameters to fit the data Comparative analysis

Data management Geometry MRI/CT, current injection, results

Page 35: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Investigations in Conductivity Modeling Forward solver parameter

tuning Time-step: controls the

speed of reaching the steady state

Convergence tolerance: sets the level of convergence

Geometry resolution error assessment High resolution (1mm),

accurate but inefficient Low resolution (2mm),

efficient but less accurate

Page 36: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Investigations in Conductivity Modeling

Sensitivity Analysis How the uncertainty in each electrode’s potential can be

apportioned to uncertainties to inputs Potentials at the electrodes is:

Insensitive to variation in CSF tissues conductivity Highly sensitive to scalp and skull conductivities Sensitive to brain conductivity

Page 37: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Eye blink topography

EEG

Eye blink topography

Source model

Cortical topography

EEG

Cortical topography

Source model

Data Model

Matrix of eye blink topography

(KEB)

and cortical eigenvectors

(K1 .. KCT)

Spatial Filter Generation

Matrix pseudo inverse

Extract temporal evolution of eye blink intensity

Refine Extracted Eye Blink Intensity

Nullspace filtering Frequency filtering

Extract Eye Blinks

DCA Eye Blink Removal Sequence

Page 38: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

DCA Derived Spatial Filter DCA spatial filter extracts intensity of activity correlated to the

blink topography that is not described by the principal eigenvectors of the cortical EEG covariance matrix KEB: Blink topography (Artifactual Model)

KC1, KC2, …, KCT: Cortical eigenvectors (Cortical Model)

K*EB, K*C1, …, K*CT: Pseudo-inverse of blink and cortical eigenvectors (Spatial filter: K*EB)

Blink Topography + Cortical Eigenvectors

KEB …KC2KC1 KCT

Matrix Pseudo - Inverse

K*EB

K*C1

K*C2

K*CT

Page 39: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

NIC Computational Services Architecture

interfaceadaptors

Page 40: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Sub-schema: Experiment Hierarchy / Provenance

Experiment Hierarchy Provenance

Page 41: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

Sub-schema: Experiment Hierarchy (condensed)

Based upon XCEDE

Page 42: Neuroinformatics for Telemedicine and Medical Services Neuroinformatics Center University of Oregon Allen D. Malony Department of Computer and Information

XCEDE

XML-based Clinical and Experimental Data Exchange Developed by BIRN