Multi-Equivalent Dipoles Fitting for ... · Multi-Equivalent Dipoles Fitting for...
1
Multi-Equivalent Dipoles Fitting for Magnetoencephalography Research Ma L 1 , Wang Y 2 , Simon JZ 1,3,4,5 1 Bioengineering Program, 2 Cognitive Neuroscience of Language Laboratory, 3 Department of Electrical & Computer Engineering, 4 Neuroscience and Cognitive Science Program, 5 Department of Biology University of Maryland at College Park References 1. Hamalainen, M., R. Hari, R. J. Ilmoniemi, J. Knuutila, and O. V. Lounasmaa. Magnetoencephalography – theory, instrumentation, and applications to noninvasive studies of the working human brain. Reviews of Modern Physics. P. 413-497. Vol. 65, No. 2, April 1993. 2. Baillet, S., J. C. Mosher, and R. M. Leahy. Electromagnetic brain mapping. IEEE Signal Processing Magazine. P. 14-30, Nov. 2001. 3. Mosher, J. C., R. M. Leahy, and P. S. Lewis. EEG and MEG: Forward solutions for inverse methods. IEEE Transactions on Biomedical Engineering. P. 245-259. Vol. 46, No. 3, March 1999. Motivation System Diagram Conclusions _ Using the proposed method the two simultaneously fit dipoles represent well the current sources in the brain. _ Speed up by factor of 8 already in progress. Computational Sensorimotor Systems Lab ♣ Why fit two dipoles? MEG responses to auditory stimuli typically give independent spatial field distributions on each hemisphere of the brain, generated by two independent neural sources. ♣ Why simultaneously fit two dipoles? Previously, two dipoles were fit separately, so different sequences of dipole fitting gave inconsistent results. τ The whole-head magnetic field is calculated using the spherical head model. r: sensor locations; r q : dipole locations; q: dipole moment; A: lead field matrix; S: dipole magnitudes. τ Least-Squares (LS) estimation is used to achieve a minimum of the cost function. M: m agnetic signals from MEG. τ Pyramid search: coarse search in the whole head, then fine search in region of interest (ROI). τ By searching finite pair sources in the right and left hemispheres simultaneously, the pair of source locations minimizing the value of the cost function give the corresponding two equivalent dipoles locations. Approach ω The simultaneous fitting of two source dipoles significantly improved the accuracy compared to the separate fitting and does not have the sequence problem. ω Comparing the measured magnetic fields from the MEG signals with the calculated magnetic fields from the two equivalent dipoles, the small residue indicate the two simultaneously fit dipoles represent well the current sources in the brain. ω The coarse and fine searches require 16 hours and 2 hours each. Head matrix Sensor matrix Calculate the whole-head lead field using the spherical head model Lead field filter Define the measure of fit using Least-Squares (LS) estimation Find source locations minimizing the cost function Region of interest (ROI) Dipole magnitudes Dipole locations # of iteration ≤ M # of iteration > M )M) AA (I M) A(A M AS M J T + + − = − = − = 2 2 LS The neurons of the brain communicate using electrical signals. These signals give rise to minute magnetic fields in the brain. Magnetoencephalography (MEG) uses ultra-sensitive detectors to measure these magnetic fields from outside the head. Among the many emerging technologies for measuring brains in action, MEG is perhaps the most exciting, because it pinpoints brain activity with great precision in both space (1cm) and time (1ms). Introduction MEG signals are recorded in a magnetically shielded room with a 160 channel whole-scalp Neuromag-160 neuromagnetometer, which is sensitive to small magnetic field as 10fT. Simultaneous fitting Separate fitting (R then L) Separate fitting (L then R) Magnetic Fields Results Dipole Fit Residue amplified by 2 Measurement )) ( ( ) ( ( ) ( 4 2 0 q q q q q r r, r r q r q r r, r r, b(r) F F F ∇ × − × = ⋅ π μ )) ( ( ) ( 2 r r r r, q q ⋅ − + = r rd d F q q r r d r r d r r, ) ) ( 2 ( ) 2 2 ) ( ( ) ( 2 d r d r d d r d F ⋅ + + − + + ⋅ + = ∇ T AS b(r) = Sensors & Equivalent Dipoles
Transcript of Multi-Equivalent Dipoles Fitting for ... · Multi-Equivalent Dipoles Fitting for...
Multi-Equivalent Dipoles Fitting for Magnetoencephalography ResearchMa L1, Wang Y2, Simon JZ1,3,4,5
1Bioengineering Program, 2Cognitive Neuroscience of Language Laboratory, 3Department of Electrical & Computer Engineering,4Neuroscience and Cognitive Science Program, 5Department of Biology
University of Maryland at College Park
References1. Hamalainen, M., R. Hari, R. J. Ilmoniemi, J. Knuutila, and O. V. Lounasmaa.
Magnetoencephalography – theory, instrumentation, and applications to noninvasive studies ofthe working human brain. Reviews of Modern Physics. P. 413-497. Vol. 65, No. 2, April 1993.
2. Baillet, S., J. C. Mosher, and R. M. Leahy. Electromagnetic brain mapping. IEEE SignalProcessing Magazine. P. 14-30, Nov. 2001.
3. Mosher, J. C., R. M. Leahy, and P. S. Lewis. EEG and MEG: Forward solutions for inversemethods. IEEE Transactions on Biomedical Engineering. P. 245-259. Vol. 46, No. 3, March 1999.
Motivation
System Diagram
Conclusions_ Using the proposed method the two simultaneously fit dipoles
represent well the current sources in the brain.
_ Speed up by factor of 8 already in progress.
Computational
Sensorimotor Systems
Lab
♣ Why fit two dipoles?MEG responses to auditory stimuli typically give independent spatialfield distributions on each hemisphere of the brain, generated by twoindependent neural sources.
♣ Why simultaneously fit two dipoles?Previously, two dipoles were fit separately, so different sequences ofdipole fitting gave inconsistent results.
τ The whole-head magnetic field is calculated using the sphericalhead model.
r: sensor locations; rq: dipole locations; q: dipole moment; A: lead field matrix; S: dipolemagnitudes.
τ Least-Squares (LS) estimation is used to achieve a minimum of thecost function.
M: m agnetic signals from MEG.
τ Pyramid search: coarse search in the whole head, then fine searchin region of interest (ROI).
τ By searching finite pair sources in the right and left hemispheressimultaneously, the pair of source locations minimizing the value ofthe cost function give the corresponding two equivalent dipoleslocations.
Approach
ω The simultaneous fitting of two source dipoles significantlyimproved the accuracy compared to the separate fitting and doesnot have the sequence problem.
ω Comparing the measured magnetic fields from the MEG signalswith the calculated magnetic fields from the two equivalent dipoles,the small residue indicate the two simultaneously fit dipolesrepresent well the current sources in the brain.
ω The coarse and fine searches require 16 hours and 2 hours each.
Head matrix Sensor matrix
Calculate the whole-head lead field using thespherical head model
Lead field filter
Define the measure of fit using Least-Squares(LS) estimation
Find source locations minimizingthe cost function
Region of interest(ROI)
Dipole magnitudes
Dipole locations
# of iteration ≤ M# of iteration > M
)M)AA(IM)A(AMASMJ T ++ −=−=−=22
LS
The neurons of the brain communicate using electrical signals. Thesesignals give rise to minute magnetic fields in the brain.Magnetoencephalography (MEG) uses ultra-sensitive detectors tomeasure these magnetic fields from outside the head. Among themany emerging technologies for measuring brains in action, MEG isperhaps the most exciting, because it pinpoints brain activity with greatprecision in both space (1cm) and time (1ms).
Introduction
MEG signals are recorded in a magnetically shielded room with a 160channel whole-scalp Neuromag-160 neuromagnetometer, which issensitive to small magnetic field as 10fT.
Simultaneousfitting
Separate fitting(R then L)
Separate fitting(L then R)
Magnetic Fields
Results
Dipole FitResidue
amplified by 2
Measurement
))(()(()(4 2
0qqqq
qrr,rrqrqrr,
rr,b(r) FF
F∇×−×= ⋅
πµ
))(()( 2 rrrr, qq ⋅−+= rrddF qq rrd
rrd
rr, ))(
2()22)(
()(2
drdrd
dr
dF
⋅++−++⋅+=∇
TASb(r) =Sensors &
Equivalent Dipoles
