Source Analysis Head Models

Head Models

BESA Research provides head models that are based either on multi-shell spherical head models, standardized realistic head models using finite elements (FEM), or individual realistic FEM models that were created in BESA MRI. Standard FEM models are currently only available for EEG.

It is highly recommended that you carefully read the specifications of the different head models below to understand what assumptions you are making in your data analysis when selecting a particular head model.

The Head Model chapter is subdivided into four sections:* Individual FEM model: Realistic head model for EEG and MEG

• Standardized FEM model: Realistic approximation for EEG
• Age-appropriate template models: Realistic head models from infancy to adulthood
• Spherical and Ellipsoidal Head Models for EEG
• Spherical Head Model for MEG

To select a different head model, use the Head Model Selection list in the parameter box. If you right click onto the Head Model dropdown list in the parameter box or on the EEG/MEG button in the channel box, a special Head Model Selection popup menu will appear.

Individual FEM Model

The head model is the link between the dipole sources and the EEG / MEG signals. It describes the electrical conductivity distribution inside of the head. By segmenting the different tissues having different conductivities based on the subject's MRI an individual head model can be constructed which takes into account anatomical differences between subjects. Using the finite element method (FEM) it is then possible to accurately simulate the EEG / MEG signal generated by a given source distribution.

By using an individual FEM model the inverse solution can become more precise (Vanrumste et al. 2002, Roth et al. 1993, Cuffin 1996, Haueisen 1997). This is especially true (a) for sources in brain regions that are not described well by a spherical head model (e.g. basal temporal lobe); (b) when individual heads show deviations from the norm (e.g. lesions). Thus, if the research target is to achieve maximal localization precision, an individual realistic head model is strongly recommended.

BESA MRI creates four-layer FEM models which differentiate between the tissues scalp, skull, cerebrospinal fluid (CSF), and brain, and which allow arbitrarily complex geometries. Default conductivity values of 0.33 S/m for scalp and brain tissue, 0.0042 S/m for skull tissue, and 1.79 S/m for CSF tissue are used. Different conductivity values can be chosen by the user when creating the models in BESA MRI.

In contrast to three-layer realistic head models, four-layer FEM models gain additional precision as the CSF layer is associated with yet another conductivity value than the other head tissues (Baumann et al. 1997). Neglecting CSF for EEG forward modeling can lead to larger localization errors (Ramon et al. 2003, Wendel et al. 2008, Lanfer et al. 2012).

The FEM modeling in BESA MRI was developed in cooperation with the research group around Dr. Carsten Wolters (Münster).

Using Individual FEM Models in BESA Research

Individual FEM models can be used in BESA Research after coregistration has been performed in BESA MRI. Detailed instructions on how to perform the coregistration can be found in the BESA help or in the quick guide which is available on the BESA homepage (www.besa.de/tutorials/quickguides).

In the coregistration process BESA MRI will generate a leadfield table, and a description of the source space. The leadfield table contains the simulated EEG potentials or MEG signals for sources in x-, y-, and z-direction distributed on a regular grid covering the entire source space. The source space specifies the locations where BESA Research is allowed to place dipole sources.

After successfully running the coregistration the coregistration dialog in BESA Research indicates that an individual FEM model is defined. The model can now be used in the Source Analysis module. To do so select the entry Individual FEM from the head model selection dropdown list. BESA Research will now load the leadfield table and the source space definitions. During dipole fitting BESA Research computes the EEG potentials or MEG signals for any given dipole source employing cubic Bezier-spline interpolation. This way, the individual FEM model can be used for the source analysis like any of the other head models.

References:

Baumann, S.B., D.R. Wozny, S.K. Kelly, and F.M. Meno. “The Electrical Conductivity of Human Cerebrospinal Fluid at Body Temperature.” IEEE Transactions on Biomedical Engineering 44, no. 3 (March 1997): 220–223. doi:10.1109/10.554770.

Cuffin, B. N. “EEG Localization Accuracy Improvements Using Realistically Shaped Head Models.” IEEE Transactions on Biomedical Engineering 43, no. 3 (March 1996): 299–303. doi:10.1109/10.486287.

Haueisen, J., C. Ramon, M. Eiselt, H. Brauer, and H. Nowak. “Influence of Tissue Resistivities on Neuromagnetic Fields and Electric Potentials Studied with a Finite Element Model of the Head.” Biomedical Engineering, IEEE Transactions on 44, no. 8 (1997): 727–35. doi:10.1109/10.605429.

Lanfer, B., I. Paul-Jordanov, M. Scherg, and C. H. Wolters. “Influence of Interior Cerebrospinal Fluid Compartments on EEG Source Analysis.” In Proceedings BMT 2012. Vol. 57. Jena: De Gruyter, 2012. doi:10.1515/bmt-2012-4020.

Ramon, Ceon, P. Schimpf, J. Haueisen, M. Holmes, and A. Ishimaru. “Role of Soft Bone, CSF and Gray Matter in EEG Simulations.” Brain Topography 16 (2003): 245–248. doi:10.1023/B:BRAT.0000032859.68959.76.

Roth, Bradley J., Marshall Balish, Alexander Gorbach, and Susumu Sato. “How Well Does a Three-sphere Model Predict Positions of Dipoles in a Realistically Shaped

Head?” Electroencephalography and Clinical Neurophysiology 87, no. 4 (October 1993): 175–184. doi:10.1016/0013-4694(93)90017-P.

Vanrumste, Bart, Gert Van Hoey, Rik Van de Walle, Michel R. P. D’Havé, Ignace A. Lemahieu, and Paul A. J. M. Boon. “Comparison of Performance of Spherical and Realistic Head Models in Dipole Localization from Noisy EEG.” Medical Engineering & Physics 24, no. 6 (July 2002): 403–418. doi:16/S1350-4533(02)00036-X.

Wendel, K, N G Narra, M Hannula, P Kauppinen, and J Malmivuo. “The Influence of CSF on EEG Sensitivity Distributions of Multilayered Head Models.” IEEE Transactions on Bio-Medical Engineering 55, no. 4 (April 2008): 1454–1456. doi:10.1109/TBME.2007.912427.

Standardized FEM Model: Realistic Approximation for EEG

The standardized FEM model has been created from an averaged head using 50 individual MRIs in Talairach space. The averaged head is used for the standard MRI displays and shows a smoothed brain. Major sulci can be identified. In the 3D mapping window, the averaged skin surface is used for the voltage and CSD maps and the smoothed averaged cortical surface is used for the surface image maps.

The standardized FEM model provides a realistic approximation to the averaged head in Talairach space and uses three compartments: brain/CSF, skull, scalp. The interfaces between the compartments brain, skull and scalp were segmented from the average MRI employing threshold segmentation, manual segmentation and smoothing steps. The skull thickness was modeled to be approximately half of the distance between the brain and the scalp surface, at maximum 6 mm.

It is a common feature to all realistic head models that little is known about the different brain tissues, their conductivities and anisotropies. Especially at the lower parts of the brain, it is extremely difficult to segment bone, CSF and cavities from structural MRI scans. Therefore, all so-called realistic head models

make simplifying assumptions on these tissues and round off and smooth their geometries. For numerical reasons, the BESA Research standardized FEM model does not explicitly model the CSF layer. However, the smoothing effect of the CSF layer is partly compensated for by assuming an anisotropic skull conductivity: The tangential conductivity within the skull is modeled to be 3 times larger than the radial conductivity across the skull.

Bone conductivities are age dependent, with children having higher bone conductivities as compared to adults. For that reason, BESA Research provides several standardized head models with different conductivity ratios (cr) of radial skull conductivity and brain conductivity. The adult FEM model with a cr of 80 produces comparable locations in depth to the 4-shell ellipsoidal model with default parameters. Precise conductivity values are generally not known. Rough estimates have been obtained by comparing the depth of sources in subjects of different age groups. For that purpose, in combined EEG/MEG recordings of SEP/SEFs and epileptic spikes, the dipole fit results to the EEG and the MEG data were compared and matched. This resulted in the approximate age recommendations given in the Head Model Selection list for the different FEM models:

Age	Recommended cr
3-4	10
4-6	15
6-8	20
8-10	30
10-12	40
12-14	50
14-16	60
16-20	70
adult	80 (default)

Practical implementation of the FEM model

When selecting a realistic approximation model from the Head Model Selection list or from the Head Model Selection popup menu, a precalculated leadfield table is loaded. Each table contains the forward solution for a standard set of electrode locations and orthogonal dipoles at fixed grid points in the brain. Different tables have been precalculated for the various conductivity ratios using the same standard MRI average of 50 subjects.  Individual realistic models will simply need separate tables and MRI volume and surface files.

The standardized FEM model currently uses 81 standard electrodes of the 10-10 electrode system. The locations of these electrodes have been obtained by projecting the default spherical coordinates of these electrodes (cf. Default.ecd) onto the averaged skin surface. These locations are very similar to the averaged locations obtained using the individual landmarks to construct the 10-10 electrode system on each individual MRI surface according to the distance rules. The center of the equivalent sphere is at Talairach location (0, -17.6, -1.0). It is located approximately 5 mm anterior to the posterior commissure.

The leadfield of a given dipole inside the head is computed by first interpolating the pre-calculated leadfields of the neighboring grid points using a second order Bezier interpolation. This leadfield, obtained for the 81 standard electrodes, is then interpolated to the actual electrode positions of the data set under analysis.

Age-appropriate template models: Realistic head models from infancy to adulthood

It is often not possible to record MRI images from all subjects, especially so if the subjects are children or infants. Therefore, BESA Research provides the possibility to use age-appropriate realistic templates for source analysis. The templates were derived by non-linear averaging across real brains in according age-ranges from infancy to adulthood.

Template brains used for age-appropriate models were kindly provided by John E. Richards, University of South Carolina, USA.

If you would like to publish results obtained with the use of the age-specific models, please reference the publications below:

• Richards, J.E., & Xie, W. (2015). Brains for all the ages: Structural neurodevelopment in infants and children from a life-span perspective. In J. Bensen (Ed.), Advances in Child Development and Behavior (Vol 48, Chapter 1, pps 1-52)
• Richards, J.E., Sanchez, C., Phillips-Meek, M., & Xie, W. (2015). A database of age-appropriate average MRI templates. Neuroimage,doi:10.1016/j.neuroimage.2015.04.055. 

In detail, the template brains used are described in the following publications:

• 0-4 years:

• NIH Pediatric MRI Database (NIHPD; Almli, C. R., Rivkin, M. J., & McKinstry, R. C. (2007). The NIH MRI study of normal brain development (objective-2): Newborns, infants, toddlers, and preschoolers. Neuroimage, 35(1), 308-325)
• Sanchez, C.E., Richards, J.E., & Almli, C.R. (2011). Neurodevelopmental MRI brain templates for children from 2 weeks to 4 years of age, Developmental Psychobiology
• Richards, J.E. (2009). Attention in the brain and early infancy. In S.P. Johnson (Ed.), Neoconstructism: The new science of cognitive development
• Richards, J.E. (2010). What's inside a baby's head? Structural and functional brain development in infants. International Conference on Infant Studies, Baltimore, MD, March, 2010.
• Richards, J.E., & Xie, W. (2015). Brains for all the ages: Structural neurodevelopment in infants and children from a life-span perspective. In J. Bensen (Ed.), Advances in Child Development and Behavior (Vol 48, Chapter 1, pps 1-52).;
• Richards, J.E., Sanchez, C., Phillips-Meek, M., & Xie, W. (2015). A database of age-appropriate average MRI templates, Neuroimage, doi:10.1016/j.neuroimage.2015.04.055
• Fillmore, P.T., Richards, J.E., Phillips-Meek, M.C., Cryer, A., & Stevens, M. (2015). Stereotaxic MRI brain atlases for infants from 3 to 12 months of age. Developmental Neuroscience, doi:10.1156/000438749

• 6-18 years:

• NIHPD (Evans, A. C. (2006). The NIH MRI study of normal brain development. Neuroimage, 30(1), 184-202.)
• Sanchez, C.E., Richards, J.E., & Almli, C.R. (2010). Age-specific MRI brain templates for healthy brain development from 4 to 24 years, Unpublished ms.
• Sanchez, C.E., Richards, J.E., & Almli, C.R. (2012). Age-specific MRI templates for pediatric neuroimaging. Developmental Neuropsychology, 37, 379-399.
• Richards, J.E., & Xie, W. (2015). Brains for all the ages: Structural neurodevelopment in infants and children from a life-span perspective. In J. Bensen (Ed.), Advances in Child Development and Behavior (Vol 48, Chapter 1, pps 1-52)
• Richards, J.E., Sanchez, C., Phillips-Meek, M., & Xie, W. (2015). A database of age-appropriate average MRI templates. Neuroimage, doi:10.1016/j.neuroimage.2015.04.055.

• 20-24 years:

• Sanchez, C.E., Richards, J.E., & Almli, C.R. (2012). Age-specific MRI templates for pediatric neuroimaging. Developmental Neuropsychology, 37, 379-399. Fillmore, P.T., Phillips-Meek, M.C., and Richards, J.E. (2013), Age-specific MRI brain and head templates for healthy adults from 20 through 89 years of age. Frontiers in Aging Neuroscience.6, doi: 10.3389/fnagi.2015.00044
• Richards, J.E., & Xie, W. (2015). Brains for all the ages: Structural neurodevelopment in infants and children from a life-span perspective. In J. Bensen (Ed.), Advances in Child Development and Behavior (Vol 48, Chapter 1, pps 1-52),
• Richards, J.E., Sanchez, C., Phillips-Meek, M., & Xie, W. (2015). A database of age-appropriate average MRI templates. Neuroimage, doi:10.1016/j.neuroimage.2015.04.055.
• Work from the IXF and OASIS MRI projects

Spherical and Ellipsoidal Head Models for EEG

4 shell ellipsoidal

The basis of this model is the multi-shell spherical head model with the fast algorithm by P. Berg and M. Scherg (A fast method for forward computation of multiple-shell spherical head models. Electroenceph. clin. Neurophysiol., 1994: 90, 58-64). The 4 homogeneous shells are the brain, the CSF, the bone and the skin. You can modify the thicknesses of the shells and their conductivities in the parameter box. The center of the sphere is determined from the 3D electrode locations. If you use standard electrodes without 3D coordinates (see chapter [Special-Topics.chm::/Electrode_conventions.htm Electrodes].), the multi-shell model is spherical. If 3D electrode coordinates have been defined, the spherical shells are distorted into an ellipsoid that best fits the 3D electrode cloud. The transformation from spherical to ellipsoidal is achieved via a 3*3 distortion-rotation matrix calculated by a least squares fit of the true 3D electrodes with their projection onto the best fitting sphere. Bone thickness and conductivity is age dependent, with children having higher bone conductivities as compared to adults. Precise conductivity values are generally not known. Rough estimates have been obtained by comparing the depth of sources in subjects of different age groups. For that purpose, in combined EEG/MEG recordings of SEP/SEFs and epileptic spikes, the dipole fit results to the EEG and the MEG data were compared and matched. This resulted in the following recommendations for bone thickness and conductivity depending on the subject's age:

 

Age	Bone thickness	Bone Conductivity
3	4.7 mm	0.040
4	4.8 mm	0.030
5	4.9 mm	0.022
6	5.0 mm	0.018
7	5.2 mm	0.016
8	5.4 mm	0.014
9-10	5.7 mm	0.012
11-12	6.0 mm	0.010
13-14	6.3 mm	0.008
15-16	6.7 mm	0.006
adult	7.0 mm (default)	0.0042 (default)

3 shell Ary approximation

The forward solution is calculated using a homogeneous head model after transforming the eccentricity of the dipole according to a non-linear equation (Scherg 1990; Ary 1981). Essentially, this model uses the similarity of the voltage topography of a dipole within a homogeneous sphere at a deeper location (e.g.

60% eccentricity) with the topography of a dipole in a 3-shell spherical model at a shallower location (e.g. 84% eccentricity).

 

Homogenous sphere

This model uses a dipole in a sphere with homogeneous conductivity. The homogeneous model is most suitable for the calculation of epicortical potentials.

 

Polynomial 4 shells

This model employs the widely-used polynomial expansion to calculate an iterative forward model using multiple spherical shells. This model is much slower than the fast 4-shell ellipsoidal head model, but has been provided for the purpose of comparison.

Spherical Head Model for MEG

BESA Research currently uses the spherical head model by Sarvas (1987). If no coregistration with the individual MRI has been performed in BESA Research (i.e. no individual MRI has been specified, see chapter Integration with MRI and fMRI), you can define the center for the sphere using the individual MRI by seeding a dipole at the assumed center using BrainVoyager. Then perform the Write Sphere Center File (*.cot) entry of the Head Box popup menu.  

For additional information please see the BrainVoyager help system at www.brainvoyager.com.