Limits...
Realistic and spherical head modeling for EEG forward problem solution: a comparative cortex-based analysis.

Vatta F, Meneghini F, Esposito F, Mininel S, Di Salle F - Comput Intell Neurosci (2010)

Bottom Line: In this paper we compare different spherical and realistic head modeling techniques in estimating EEG forward solutions from current dipole sources distributed on a standard cortical space reconstructed from Montreal Neurological Institute (MNI) MRI data.Computer simulations are presented for three different four-shell head models, two with realistic geometry, either surface-based (BEM) or volume-based (FDM), and the corresponding sensor-fitted spherical-shaped model.Realistic geometry turns out to be a relevant factor of improvement, particularly important when considering sources placed in the temporal or in the occipital cortex.

View Article: PubMed Central - PubMed

Affiliation: DEEI, University of Trieste, Via A. Valerio 10, 34127 Trieste, Italy. federica.vatta@deei.units.it

ABSTRACT
The accuracy of forward models for electroencephalography (EEG) partly depends on head tissues geometry and strongly affects the reliability of the source reconstruction process, but it is not yet clear which brain regions are more sensitive to the choice of different model geometry. In this paper we compare different spherical and realistic head modeling techniques in estimating EEG forward solutions from current dipole sources distributed on a standard cortical space reconstructed from Montreal Neurological Institute (MNI) MRI data. Computer simulations are presented for three different four-shell head models, two with realistic geometry, either surface-based (BEM) or volume-based (FDM), and the corresponding sensor-fitted spherical-shaped model. Point Spread Function (PSF) and Lead Field (LF) cross-correlation analyses were performed for 26 symmetric dipole sources to quantitatively assess models' accuracy in EEG source reconstruction. Realistic geometry turns out to be a relevant factor of improvement, particularly important when considering sources placed in the temporal or in the occipital cortex.

Show MeSH
Example of PSF values distribution versus distance from source, and fitting with biexponential Gaussian curve.
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC2821797&req=5

fig6: Example of PSF values distribution versus distance from source, and fitting with biexponential Gaussian curve.

Mentions: Following spatial smoothing, the quantitative evaluation of the PSF maps has been conducted by plotting, for each of the 78 analyzed sources, the obtained PSF values as function of the distance from the source position and fitting the map values with the corresponding best-fitting Gaussian-like function (biexponential Gaussian), as shown in Figure 6. The spatial extent of the PSF function, measured in mm, has been quantified by means of its full width at half maximum (FWHM) measure. The obtained PSF FWHMs have been reported and compared for all the 26 analyzed dipole sources for each source orientation, for the realistic BEM and FDM and the sensor-fitted spherical models. Table 3 summarizes the quantitative results of the performed analysis on the PSF maps. Basing upon a closer inspection of the PSF FWHM results presented in Table 3, it can be observed that the realistic FDM model presents an improvement over BEM in 68% of the total tested conditions (53 cases over 78), and specifically in 54% of the x-oriented sources (14 cases over 26), in 81% and 69% for the y- and z-oriented sources, respectively (21 and 18 cases over 26, resp.), and in 38% of the RMS (10 over 26). The realistic BEM presents an improvement over the spherical model in 62% of the total tested conditions (48 cases over 78), in 77%, 73%, 35%, and 77% of the situations for the x-, y-, and z-oriented sources and RMS, respectively. The improvement of FDM over the spherical model shows up in 88% of the analyzed situations for all the three source orientations, and in the 66% for the RMS. These trends are also confirmed by the mean ± SD values of the reported PSF FWHM results for the three models, shown in Table 3. Nine two-tailed paired t-tests have been performed to investigate differences between the spherical and the realistic models (pairs FDM versus BEM, FDM versus SPH, and BEM versus SPH) for the three source orientations. Statistically significant differences have been found in 7 out of the total 9 cases analyzed: for all source orientations in FDM versus SPH (x : p = 2.03 × 10−6; y : p = 1.98 × 10−4; z : p = 1.93 × 10−3), for the y- and z-oriented sources in FDM versus BEM (y : p = 1.69 × 10−3; z : p = 3.66 × 10−4), and for the x- and y-oriented sources in BEM versus SPH (x : p = 1.67 × 10−2; y : p = 2.83 × 10−2). The two-tailed paired t-tests performed on the RMS results showed significant differences in the FDM versus SPH pair (p = 1.36 × 10−2) and nonsignificant differences in the FDM versus BEM (p = 0.91) and in the BEM versus SPH (p = 5.63 × 10−2). In order to gather a broader evaluation of the PSF behavior on the overall brain cortex, we extended the evaluation of the FWHM PSF to all the 5000 points of the cortex surface. Figure 7 shows the differences between the FWHM RMS PSF maps between couples of different head models, to investigate the principal benefits or pitfalls given by the adoption of the different head models.


Realistic and spherical head modeling for EEG forward problem solution: a comparative cortex-based analysis.

Vatta F, Meneghini F, Esposito F, Mininel S, Di Salle F - Comput Intell Neurosci (2010)

Example of PSF values distribution versus distance from source, and fitting with biexponential Gaussian curve.
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC2821797&req=5

fig6: Example of PSF values distribution versus distance from source, and fitting with biexponential Gaussian curve.
Mentions: Following spatial smoothing, the quantitative evaluation of the PSF maps has been conducted by plotting, for each of the 78 analyzed sources, the obtained PSF values as function of the distance from the source position and fitting the map values with the corresponding best-fitting Gaussian-like function (biexponential Gaussian), as shown in Figure 6. The spatial extent of the PSF function, measured in mm, has been quantified by means of its full width at half maximum (FWHM) measure. The obtained PSF FWHMs have been reported and compared for all the 26 analyzed dipole sources for each source orientation, for the realistic BEM and FDM and the sensor-fitted spherical models. Table 3 summarizes the quantitative results of the performed analysis on the PSF maps. Basing upon a closer inspection of the PSF FWHM results presented in Table 3, it can be observed that the realistic FDM model presents an improvement over BEM in 68% of the total tested conditions (53 cases over 78), and specifically in 54% of the x-oriented sources (14 cases over 26), in 81% and 69% for the y- and z-oriented sources, respectively (21 and 18 cases over 26, resp.), and in 38% of the RMS (10 over 26). The realistic BEM presents an improvement over the spherical model in 62% of the total tested conditions (48 cases over 78), in 77%, 73%, 35%, and 77% of the situations for the x-, y-, and z-oriented sources and RMS, respectively. The improvement of FDM over the spherical model shows up in 88% of the analyzed situations for all the three source orientations, and in the 66% for the RMS. These trends are also confirmed by the mean ± SD values of the reported PSF FWHM results for the three models, shown in Table 3. Nine two-tailed paired t-tests have been performed to investigate differences between the spherical and the realistic models (pairs FDM versus BEM, FDM versus SPH, and BEM versus SPH) for the three source orientations. Statistically significant differences have been found in 7 out of the total 9 cases analyzed: for all source orientations in FDM versus SPH (x : p = 2.03 × 10−6; y : p = 1.98 × 10−4; z : p = 1.93 × 10−3), for the y- and z-oriented sources in FDM versus BEM (y : p = 1.69 × 10−3; z : p = 3.66 × 10−4), and for the x- and y-oriented sources in BEM versus SPH (x : p = 1.67 × 10−2; y : p = 2.83 × 10−2). The two-tailed paired t-tests performed on the RMS results showed significant differences in the FDM versus SPH pair (p = 1.36 × 10−2) and nonsignificant differences in the FDM versus BEM (p = 0.91) and in the BEM versus SPH (p = 5.63 × 10−2). In order to gather a broader evaluation of the PSF behavior on the overall brain cortex, we extended the evaluation of the FWHM PSF to all the 5000 points of the cortex surface. Figure 7 shows the differences between the FWHM RMS PSF maps between couples of different head models, to investigate the principal benefits or pitfalls given by the adoption of the different head models.

Bottom Line: In this paper we compare different spherical and realistic head modeling techniques in estimating EEG forward solutions from current dipole sources distributed on a standard cortical space reconstructed from Montreal Neurological Institute (MNI) MRI data.Computer simulations are presented for three different four-shell head models, two with realistic geometry, either surface-based (BEM) or volume-based (FDM), and the corresponding sensor-fitted spherical-shaped model.Realistic geometry turns out to be a relevant factor of improvement, particularly important when considering sources placed in the temporal or in the occipital cortex.

View Article: PubMed Central - PubMed

Affiliation: DEEI, University of Trieste, Via A. Valerio 10, 34127 Trieste, Italy. federica.vatta@deei.units.it

ABSTRACT
The accuracy of forward models for electroencephalography (EEG) partly depends on head tissues geometry and strongly affects the reliability of the source reconstruction process, but it is not yet clear which brain regions are more sensitive to the choice of different model geometry. In this paper we compare different spherical and realistic head modeling techniques in estimating EEG forward solutions from current dipole sources distributed on a standard cortical space reconstructed from Montreal Neurological Institute (MNI) MRI data. Computer simulations are presented for three different four-shell head models, two with realistic geometry, either surface-based (BEM) or volume-based (FDM), and the corresponding sensor-fitted spherical-shaped model. Point Spread Function (PSF) and Lead Field (LF) cross-correlation analyses were performed for 26 symmetric dipole sources to quantitatively assess models' accuracy in EEG source reconstruction. Realistic geometry turns out to be a relevant factor of improvement, particularly important when considering sources placed in the temporal or in the occipital cortex.

Show MeSH