HAL Id: hal-01095449
https://hal.inria.fr/hal-01095449
Submitted on 15 Dec 2014
HAL is a multi-disciplinary open access
archive for the deposit and dissemination of sci-entific research documents, whether they are pub-lished or not. The documents may come from teaching and research institutions in France or abroad, or from public or private research centers.
L’archive ouverte pluridisciplinaire HAL, est destinée au dépôt et à la diffusion de documents scientifiques de niveau recherche, publiés ou non, émanant des établissements d’enseignement et de recherche français ou étrangers, des laboratoires publics ou privés.
Distributed under a Creative Commons Attribution| 4.0 International License
Diffusion Magnetic Resonance information as a
regularization term for MEG/EEG inverse problem
Brahim Belaoucha, Anne-Charlotte Philippe, Maureen Clerc, Théodore
Papadopoulo
To cite this version:
Brahim Belaoucha, Anne-Charlotte Philippe, Maureen Clerc, Théodore Papadopoulo. Diffusion Mag-netic Resonance information as a regularization term for MEG/EEG inverse problem. The 19th International Conference on Biomagnetism, Aug 2014, Halifax, Canada. �hal-01095449�
Diffusion Magnetic Resonance information as a regularization term for
MEG/EEG inverse problem
B. Belaoucha, A. Philippe, M. Clerc, T. Papadopoulo
Project Team Athena, INRIA, Sophia Antipolis - Méditerranée, France
Contact: brahim.belaoucha@inria.fr, url: http://www-sop.inria.fr/athena
Several regularization terms are used to constrain the Magnetoencephalography (MEG) and the Electroencephalography (EEG) inverse problem. It has been shown that
the brain can be divided into several regions[1] with functional homogeneity inside each one of them. To locate these regions, we use the structural information coming
from the diffusion Magnetic Resonance (dMRI) and more specifically, the anatomical connectivity of the distributed sources computed from dMRI. To invistigate the
importance of the dMRI in the source reconstruction, we compare the solution based on dMRI-based parcellation to random parcellation.
1
Introduction
2
Methods:
(a) Destrieux(Dx) (b) Desikan-Killiany(DK) (c) Mindboggle(ML) (d) Random(R)
Fig.1: The different pre-clustering approaches used to cluster the cortex
3
Experiments & Results
Table.1: The final number of cortex regions vs value and pre-parcellation.
100 200 300 400 500 600 700 800 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Number of random clusters
Similarity measure SM(R,DX) SM(R,DK) SM(R,ML) SM(R,R) 100 200 300 400 500 600 700 800 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Number of random clusters
Similarity measure SM(R,DX) SM(R,DK) SM(R,ML) SM(R,R) 100 200 300 400 500 600 700 800 0.5 0.55 0.6 0.65 0.7 0.75 0.8 0.85 0.9 0.95 1
Number of random clusters
Similarity measure
SM(R,DX) SM(R,DK) SM(R,ML) SM(R,R)
Fig.2: SM values between the different parcellations of different subjects.
Subject 1 Subject 2 Subject 3 SM
3.2
Real data
3.1
Synthetic data
4
Conclusion
References
19th International Conference on Biomagnetism August 24-28,2014 Halifax, Canada
PSS-variance PSS-mean
using Random parcellation
PC-variance PC-mean using Random parcellation
PSS using dMRI parcellation
(Dx)
PC using dMRI parcellation (DX)
MNE PSS PC
PC dMRI
PSS dMRI
1 active dMRI region
1 random active region intersects
with 3 dMRI regions
10 15 30 Without noise 0 5 10 15 20 25 30 35 SNR Error MNE PC PSS
The reconstructed source error using MNE, PC, and PSS for different SNR, for constant active patch.
Because ML and DK
give bigger regions
than the DX, we
decided to use the
later
for
source
reconstruction.
0 1 2 3 4 5 6 −1 0 1 2Size of parcellation regions based on Destrieux (%),
0 1 2 3 4 5 6 −1
0 1 2
Size of parcellation regions based on Desikan−Killiany (%)
Number of regions (log
10 %) 0 1 2 3 4 5 6 −1 0 1 2
Size of parcellation regions based on Mindboggle (%) 0 1 2 3 4 5 6
−1 0 1 2
Size of parcellation regions based on Destrieux (%),
0 1 2 3 4 5 6 −1
0 1 2
Size of parcellation regions based on Desikan−Killiany (%)
Number of regions (log
10 %) 0 1 2 3 4 5 6 −1 0 1 2
Size of parcellation regions based on Mindboggle (%) 0 1 2 3 4 5 6
−1 0 1 2
Size of parcellation regions based on Destrieux (%),
0 1 2 3 4 5 6 −1
0 1 2
Size of parcellation regions based on Desikan−Killiany (%)
Number of regions (log
10 %) 0 1 2 3 4 5 6 −1 0 1 2
Size of parcellation regions based on Mindboggle (%)