Source Analysis and Multimodal Neuroimaging · Source Analysis and Multimodal Neuroimaging ... • Brain activity is distributed in 3- ... fMRI has high spatial resolution.

Source Analysis and

Multimodal Neuroimaging

Bin He Department of Biomedical Engineering Institute for Engineering in Medicine

University of Minnesota

Brain Activity is Spatio-temporal Process

•  Brain activity is distributed in 3-dimensional space and evolves in time.

•  High spatial resolution imaging modalities needed.

•  High temporal resolution imaging modalities needed.

•  Noninvasive imaging modalities needed for human brain imaging. 100 ms

Spatial & Temporal Resolution of Neuroimaging Modalities

He & Liu, IEEE Rev BME, 2008

Strategies

 Increase temporal resolution of fMRI

 Enhance spatial resolution of EEG/MEG  Multimodal imaging combining fMRI

and electromagnetic imaging

Problem Formulation

Brain Electric Source

Volume Conductor E/MEG

Equivalent Brain Source

Volume Conductor

Model E/MEG

Inverse Problem/ Source Analysis

Forward Problem

Why Source Analysis?

  The head volume conductor smears the EEG/MEG distribution over the scalp, thus it is difficult to relate scalp EEG/MEG to intra-cranial sources by only inspecting scalp EEG/MEG.

Electrophysiological Neuroimaging •  Neuroimaging aimed at improving substantially

the spatial resolution of conventional EEG/MEG through deconvolution, and use of anatomic information on the brain and head as obtained from magnetic resonance imaging.

•  We shall abbreviate it as EEGI below.

From Sensor to Source Space

Michel & He, Niedermeyer's Electroencephalography, 2011

Issues in EEGI

•  Scalp Mapping •  Equivalent Source Modeling •  Volume Conductor Modeling •  Inverse Source Solutions •  Validation and Biomedical Applications

Electrophysiological Mapping of Binocular Rivalry

Zhang, Jamison, Engel, He, He, Neuron, 2012

Low density / high density array

32 electrodes 64 electrodes 128/256 electrodes

Effect of electrode numbers - Patient data

Mean and standard deviation of localization errors in clinical data analysis

Lu et al., Clin. Neurophysiol., 2012

Issues in EEGI

•  Scalp Mapping •  Equivalent Source Modeling •  Volume Conductor Modeling •  Inverse Source Solutions •  Validation and Biomedical Applications

Electric Source Modeling

He et al, IEEE Trans BME, 2011

Equivalent Source Modeling

•  Point Source Models - Moving Dipole - Fixed Dipole - Multipole •  Distributed Source Models

- 2D Cortical Potential - 2D Cortical Dipole Layer - 3D Dipole Distribution - 3D Current Distribution

3D Source Models

Dipole Source Model

Distributed Source Model

•  Amplitude, location, orientation

•  Nonlinear problem

•  Modeling focal and compact source

•  Amplitude

•  Linear problem

•  Modeling distributed source network

Issues in EEGI

•  Scalp Mapping •  Equivalent Source Modeling •  Volume Conductor Modeling •  Inverse Source Solutions •  Validation and Biomedical Applications

Head Conductor Modeling

•  Homogeneous Sphere Model •  Inhomogeneous 3-concentric-spheres Model •  Inhomogeneous 4-concentric-spheres Model •  Homogeneous Realistic Geometry (RG) Model •  Piece-wise Homogeneous RG Model (BEM) •  Inhomogeneous RG Model (FEM)

RG-Head based EEG Inverse Solution

He et al., IEEE Trans on BME, 1987

Brain-Head Modeling

He et al., NeuroImage, 2002

Issues in EEGI

•  Scalp Mapping •  Equivalent Source Modeling •  Volume Conductor Modeling •  Inverse Source Solutions •  Validation and Biomedical Applications

Forward and Inverse Problems

Forward Problem

Inverse Problem

Source Model

-  current dipoles

-  discrete and distributed source models

Volume Conductor Model

-  sphere head model

-  boundary element model

-  finite element model Source Imaging

-  discrete source imaging

-  distributed source imaging

Solving inverse problem

ill-posedness

regularization

Moving Dipole Localization

Dipole Source Model

Comparator

Volume Model

Parameter Modifier

Initial Parameters

Inverse Solution

Subject Mapping

System

EEG/MEG

Linear Inverse Problem

s

x

x - Scalp EEG/MEG

s - Unknown Source Vector

A - Transfer Matrix

b - Noise

•  Tikhonov regularization:

k : truncation parameter

s(λ) = (AtA+λR)−1Atx

where

Regularization Approach

•  Truncated singular value decomposition:

s(k) = A#x =VΣk−1Utx

L-Curve Approach

sFA φ−

F large k

small λ

small k large λ

Hansen 1992

Issues in EEGI

•  Scalp Mapping •  Equivalent Source Modeling •  Volume Conductor Modeling •  Inverse Source Solutions •  Validation and Biomedical Applications

Electrophysiological

Source Imaging

MRI

Source Imaging

EEG

Cortical Imaging from EEG

Bai, He, et al., Brain Topography, 2011

Cortical Imaging of Epileptic Activity

Lai, He, et al., NeuroImage, 2011

Sparse Source Imaging from EEG

Ding & He, Human Brain Mapping, 2008

Spatio-temporal Seizure Imaging

Yang, He, et al., NeuroImage, 2011

EEG Imaging of Seizure Sources

Yang, He, et al., NeuroImage, 2011

Beyond EEGI

 Increase temporal resolution of fMRI

 Enhance spatial resolution of EEG/MEG  Multimodal imaging combining fMRI

and electromagnetic imaging

Multimodal Neuroimaging

  Electrophysiological and hemodynamic measurements represent complementaty responses of brain activation.

  fMRI has high spatial resolution.   EEG has high temporal resolution.   Integration of fMRI with EEG represents a natural

approach to further improving spatio-temporal resolution, but a new challenge due to the very different time scales of these measurements.

EEG/fMRI & Brain Activity EEG

BOLD-fMRI activations Scalp potential maps

Neural activation

Neurovascular coupling

fMRI and EEG

He et al., IEEE TBME, 2011

Origin of BOLD

Logothetis et al., Nature, 2001

Simultaneous EEG-fMRI Recordings

Im, He et al., J Neurosci. Meth., 2006

Spatiotemporal Neuroimaging by fMRI-EEG Integration

EEG fMRI

fMRI-constrained cortical source imaging

Statistical Parametric Mapping

Cross-modal Relationship

Liu and He, NeuroImage, 2008

fMRI/EEG Multimodal Imaging

(Grill-Spector and Malach, Annu. Rev Neurosci, 2004)

Liu & He, NeuroImage, 2008

BOLD vs VEP Power

Liu, He, et al, NeuroImage, 2010

Functional Connectivity Imaging in Source Space

Babiloni, He, et al, NeuroImage, 2005

eConnectome

  MATLAB-based, GUI   Temporal Analysis   Scalp Topographical Analysis   Spectral Analysis   Cortical Current Density Imaging   Functional Connectivity Mapping   EEG, V1.0, 2010   ECoG, V1.0, 2010   MEG V1.0, 2011   Open-source, GPL

Electrophysiological Connectome

http://econnectome.umn.edu

He et al., J Neurosci. Methods, 2011; Dai et al., Brain Topography, 2012

Thank you