運動相關腦磁波之腦部信號源動態造運動相關腦磁波之腦部信號源動態造
影影
Imaging Cortical Source Imaging Cortical Source
Dynamics of MovementDynamics of Movement--RelatedRelated
Neuromagnetic FieldNeuromagnetic Field
計畫編號:計畫編號:NSC 92NSC 92--22182218--EE--010010--005005
執行期限:執行期限:9292年年88月月11日至日至 9393年年77月月3131日日
主持人:陳麗芬主持人:陳麗芬 助理研究員助理研究員 國立陽明大學國立陽明大學 神經科學研究中心神經科學研究中心
特約助理研究員特約助理研究員 臺北榮民總醫院臺北榮民總醫院 教學研究部教學研究部
共同主持人:陳永昇共同主持人:陳永昇 助理教授助理教授 國立交通大學國立交通大學 資訊工程學系資訊工程學系
謝仁俊謝仁俊 主治醫師主治醫師 臺北榮民總醫院臺北榮民總醫院 教學研究部教學研究部
教授暨所長教授暨所長 國立陽明大學國立陽明大學 衛生資訊與決策研究所衛生資訊與決策研究所
計畫參與人員:劉宏毅、李宗展、黃勁霖、鄭志瑜計畫參與人員:劉宏毅、李宗展、黃勁霖、鄭志瑜
研究生研究生 國立交通大學國立交通大學 資訊工程學系資訊工程學系
MEG (MEG (MagnetoEncephaloGraphyMagnetoEncephaloGraphy))
VectorviewVectorview: 306 channels manufactured by : 306 channels manufactured by
ElektaElekta Inc. installed in Taipei Veterans Inc. installed in Taipei Veterans
General HospitalGeneral Hospital
MEG Source ModelingMEG Source Modeling
zz Forward problem:Forward problem:
zz Input: the positions, the amplitude and the Input: the positions, the amplitude and the
orientations of the source current dipolesorientations of the source current dipoles
zz Output: to estimate the measured data Output: to estimate the measured data
from MEG sensorsfrom MEG sensors
zz Inverse problem:Inverse problem:
zz Input: a set of measured dataInput: a set of measured data
zz Output: to estimate the parameters Output: to estimate the parameters
representing the source current dipoles, representing the source current dipoles,
including positions, the amplitude and the including positions, the amplitude and the
orientations orientations
Source model
Equivalent current dipole (ECD)
Parameters: location, amplitude,
and orientation
Fix, rotating, moving dipoles
Head model
related to the computation of volume
current
Spherical model
Boundary element model (BEM)
Finite element model (FEM)
Forward ModelForward Model
Source ModelSource Model
MaxwellMaxwell‘‘s equationss equations::
E : electric fieldE : electric field
: dielectric constant: dielectric constant
B : magnetic fieldB : magnetic field
: permeability: permeability
J : current densityJ : current density
: charge density: charge density
( )tE
t
B
∂∂+=×∇
=⋅∇
∂∂−=×∇
=⋅∇
εµ
ερ
00
0
0
??
?
?
?
ε 0
µ0
ρ
Source Model (cont.)Source Model (cont.)
zz Head is a conductor consisting ofHead is a conductor consisting of different different
tissuestissues..
zz Using GreenUsing Green’’s theory with boundary s theory with boundary
condition we can obtain the magnetic fieldcondition we can obtain the magnetic field
: outward directed unit normal vector: outward directed unit normal vector
+ : conductivity outside the surface+ : conductivity outside the surface
-- : conductivity inside the surface: conductivity inside the surface
d d : : rr -- rrqq
( ) ( ) ( ) ( ) ( ) r'r'nrrBrB d
S
ddV
i
i
m
i
ii ⎟⎠
⎞⎜⎝
⎛ ×−−= ∫∑
=
+−
∞
3
1
0
4
σσπ
µ
( ) ( )∫ ×=∞ G p ddd r'r'rB 304 Jπµ
( )rn i
Head ModelHead Model
zz BoundaryBoundary--elementelement--method model (BEM)method model (BEM)
zz Spherical head modelSpherical head model
rr: sensor position: sensor position
rrqq: position of dipole: position of dipole
qq: moment of dipole : moment of dipole
Head Model (cont.)Head Model (cont.)
zz Because the Because the radial magnetic field is zeroradial magnetic field is zero in in
the spherical head model, we can directly the spherical head model, we can directly
solve the MEG forward problem.solve the MEG forward problem.
( ) ( ) ( ) ( )( )( )qqqqq FFF rr,rrqrqrr,rr,rB ∇⋅×−×= 2 04π
µ
( ) ( )( )
( ) ( ) qq
qq
d
ddd
d
ddF
ddF
rrr-rrr
r
rr,
rrrrrr,
⎟⎠
⎞⎜⎝
⎛ ⋅++⎟⎟⎠
⎞
⎜⎜⎝
⎛ ++⋅+=∇
⋅−+=
222
2
2
Source LocalizationSource Localization
zz BeamformerBeamformer
–– A kind of spatial filterA kind of spatial filter
–– Widely applied in radar Widely applied in radar
and sonarand sonar
–– Regarded as a Regarded as a
weighted virtual sensorweighted virtual sensor
zz LeadfieldLeadfield
–– A kind of simple spatial A kind of simple spatial
filterfilter
Linearly Constrained Minimum Linearly Constrained Minimum
Variance BeamformerVariance Beamformer
zz To provide a solution to inverse problemTo provide a solution to inverse problem
zz To pass brain activity with unit gain at a To pass brain activity with unit gain at a
specified location while attenuating specified location while attenuating
activity originating at other locationsactivity originating at other locations
subject tosubject to [van [van VeenVeen, 97], 97]
W: filter, W: filter, qqoo: source, y: filter output, H: forward solutiony: filter output, H: forward solution
trtr CCyy: variance of y: variance of y
y
qo
C
w
trmin 1=oo qq hw
LCMV Beamformer (cont.)LCMV Beamformer (cont.)
zz Solving by Lagrange multipliers yieldsSolving by Lagrange multipliers yields
x: measurementx: measurement
zz Refinement Refinement -- tradeoff between spatial specificity tradeoff between spatial specificity
and noise sensitivity and noise sensitivity [Robinson, 98][Robinson, 98]
μμ: Backus: Backus--Gilbert regularization parameter (Gilbert regularization parameter (--1<1<μμ<<∞∞))
ΣΣ: variance matrix of sensor noise: variance matrix of sensor noise
oo
o
o
qx
T
q
x
T
q
q
hCh
Ch
w 1
1
−
−
=
[ ]
[ ] oo
o
o
qx
T
q
x
T
q
q
hCh
Ch
w 1
1
−
−
∑
∑
+
+= µ
µ
Source Strength ReconstructionSource Strength Reconstruction
zz Estimating the expectation value of the Estimating the expectation value of the
source powersource power
zz NormalizationNormalization
Q: noise covariance matrixQ: noise covariance matrix
[ ]{ }11^ trVar −−= ooo qxTqq hCh
[ ]{ }
[ ]{ }11
11^
tr
tr
Var −−
−−
=
oo
oo
oN
q
T
q
qx
T
q
q
hQh
hCh
CorticalCortical--based Neuromagnetic based Neuromagnetic
Functional ImagingFunctional Imaging
zz Networks of cortical neural cell assemblies are Networks of cortical neural cell assemblies are
the main generators of MEG/EEG signals. the main generators of MEG/EEG signals.
((BailletBaillet et al. 2001)et al. 2001)
Our ApproachOur Approach
zz AnatomyAnatomy--constrainedconstrained
–– cortical surfacecortical surface
–– normal vectornormal vector
zz Reduce time complexity Reduce time complexity
from O(nfrom O(n44) to O(n) to O(n22))
MRI (Magnetic Resonance Imaging )MRI (Magnetic Resonance Imaging )
Installed in Taipei Veterans General HospitalInstalled in Taipei Veterans General Hospital
Cortical Surface ReconstructionCortical Surface Reconstruction
zz ToolsTools::
1.FreeSurfer (MIT & 1.FreeSurfer (MIT & HavardHavard University)University)
2.Surefit (University of Washington)2.Surefit (University of Washington)
The MRI volume. The skull-stripped
volume.
The left
hemisphere’s
orig surface.
RenderingRendering
VTK (Visualization Toolkit)VTK (Visualization Toolkit)
zz A free and open source system for 3D computerA free and open source system for 3D computer
graphics. graphics.
zz Higher level of abstraction than rendering Higher level of abstraction than rendering
libraries like OpenGL.libraries like OpenGL.
zz A wide variety of visualization algorithms A wide variety of visualization algorithms
including scalar, vector, texture etc.including scalar, vector, texture etc.
MaterialsMaterials
zz MEG recordingsMEG recordings
–– WholeWhole--head 306 channels MEG system (head 306 channels MEG system (VectorviewVectorview, , NeuromagNeuromag, ,
Finland)Finland)
–– SelfSelf--paced right index finger lifting movement around every 8 paced right index finger lifting movement around every 8
secondsseconds
–– 250Hz sampling rate250Hz sampling rate
–– ~100 epochs (each from ~100 epochs (each from --4 sec to 4 sec)4 sec to 4 sec)
zz MEG preprocessing MEG preprocessing
−− SSP (Signal Space Projection)SSP (Signal Space Projection)
−− EOG (ElectroEOG (Electro--OculoGramOculoGram) rejection ) rejection
−− Baseline correctionBaseline correction
−− Synchronized averagingSynchronized averaging
−− BandpassBandpass filteringfiltering
zz MRI scanningMRI scanning
–– Siemens MRISiemens MRI
–– Field of view: 230 x 230 x 192 (mm)Field of view: 230 x 230 x 192 (mm)
–– Image volume size: 256 x 256 x 128Image volume size: 256 x 256 x 128
Signal
Preprocessing
Cortical surface
Reconstruction
Source
Localization
Workflow
Visualization
Neuromagnetic Imaging of Brain Activity
•(a) Estimated power-SPM of brain activity
(d) Somatosensory representation
• (b) Location of peak pSPM in MRI
(c) Reconstructed source waveform
¾ 右圖解:(a)為左腦皮質使用光束構成法
進行活動源估算的活動顯著性(ρ)區域分
佈;(b)、(c)分別為(a)圖中具最顯
著性皮質位置(圖中白點)之相對磁振造
影影像及重建神經活動序列信號s,該位置
與文獻上手指運動相關皮質區(d)相吻合
• Motor loop
• Reconstructed spatiotemporal dynamics of brain activity
Neuromagnetic Imaging of Brain Activity
¾ 右圖是從展示影片中 -280ms 至 640ms 擷取出的 24 張畫面,紅圈
標示的位置即是活化訊號最強烈區,與相關腦神經生理知識中有關
感覺-運動系統 (sensorimotor system)控制肢體運動時腦部活化
區域之動態機轉(mechanism)(上圖)是非常吻合的,由此可顯示本
軟體系統能正確可靠地分析腦部活動動態分佈。
ReferencesReferences
zz Mosher J. C., Leachy R. M., and Lewis P. S. Mosher J. C., Leachy R. M., and Lewis P. S. ““EEG and MEG: Forward Solutions for Inverse EEG and MEG: Forward Solutions for Inverse
Methods,Methods,”” IEEE Transaction on Biomedical Engineering,IEEE Transaction on Biomedical Engineering, vol.46, pp.245vol.46, pp.245--259, 1999. 259, 1999.
zz Robison S. E., and Robison S. E., and VrbaVrba J. J. ““Functional Functional NeuroimagingNeuroimaging by Synthetic Aperture by Synthetic Aperture MagnetometryMagnetometry,,”” CTF CTF
System Inc. Port System Inc. Port CoquitlamCoquitlam, Canada, 1998. , Canada, 1998.
zz SekiharaSekihara K. et al. K. et al. ““Reconstructing Reconstructing SpatioSpatio--Temporal Activities of Neural Sources Using an MEG Temporal Activities of Neural Sources Using an MEG
Vector Beamformer Technique,Vector Beamformer Technique,”” IEEE Transaction on Biomedical Engineering,IEEE Transaction on Biomedical Engineering, vol.48, no.7, vol.48, no.7,
pp.760pp.760--771, 2001.771, 2001.
zz Avilla S. et al. Avilla S. et al. ““The VTK UserThe VTK User’’s Guide,s Guide,”” KitwareKitware, Inc.,, Inc., 2003.2003.
zz Van Essen D. C. et al. Van Essen D. C. et al. ““An Integrated Software System for SurfaceAn Integrated Software System for Surface--based Analyses of Cerebral based Analyses of Cerebral
Cortex,Cortex,”” Journal of American Medical Informatics Association.Journal of American Medical Informatics Association. ((Special issue on the Human Brain Special issue on the Human Brain
ProjectProject)) vol.8, pp.443vol.8, pp.443--459, 2001459, 2001
zz Van Van VeenVeen B. D. et al. B. D. et al. ““Localization of Brain Electrical Activity via Linearly ConstrainLocalization of Brain Electrical Activity via Linearly Constrained Minimum ed Minimum
Variance Spatial Filtering,Variance Spatial Filtering,”” IEEE Transaction on Biomedical Engineering,IEEE Transaction on Biomedical Engineering, vol.44, no.9, pp.867vol.44, no.9, pp.867--880, 880,
1997.1997.
zz Schroeder W., Martin K, and Schroeder W., Martin K, and LorensenLorensen B.B. ““The Visualization Toolkit,The Visualization Toolkit,”” 3rd edition, 3rd edition, KitwareKitware, Inc.,, Inc.,
2002.2002.
zz Van Van VeenVeen B. D., and Buckley K. B. D., and Buckley K. ““Beamforming: a Versatile Approach to Spatial Filtering,Beamforming: a Versatile Approach to Spatial Filtering,”” IEEE IEEE
ASSP Magazine,ASSP Magazine, vol.5, pp.4vol.5, pp.4--24, 1988.24, 1988.
zz Gross J. et al.Gross J. et al. ““Dynamic Imaging of Coherent Sources: Studying Neural InteractionDynamic Imaging of Coherent Sources: Studying Neural Interactions in the Human s in the Human
Brain,Brain,”” Proceedings of the National Academy of Sciences,Proceedings of the National Academy of Sciences, vol.98, no.2, pp.694vol.98, no.2, pp.694--699, 2001.699, 2001.