Limits...
Source activity correlation effects on LCMV beamformers in a realistic measurement environment.

Belardinelli P, Ortiz E, Braun C - Comput Math Methods Med (2012)

Bottom Line: Results show that high correlation affects mostly dipoles placed at small distances (1, 5 centimeters).In this case the sources merge.If the dipoles lie 3 centimeters apart, the beamformer localization detects attenuated power amplitudes and blurred sources as the correlation level raises.

View Article: PubMed Central - PubMed

Affiliation: MEG Center, University of Tübingen, Otfried Mueller Street 47, Tübingen, Germany.

ABSTRACT
In EEG and MEG studies on brain functional connectivity and source interactions can be performed at sensor or source level. Beamformers are well-established source-localization tools for MEG/EEG signals, being employed in source connectivity studies both in time and frequency domain. However, it has been demonstrated that beamformers suffer from a localization bias due to correlation between source time courses. This phenomenon has been ascertained by means of theoretical proofs and simulations. Nonetheless, the impact of correlated sources on localization outputs with real data has been disputed for a long time. In this paper, by means of a phantom, we address the correlation issue in a realistic MEG environment. Localization performances in the presence of simultaneously active sources are studied as a function of correlation degree and distance between sources. A linear constrained minimum variance (LCMV) beamformer is applied to the oscillating signals generated by the current dipoles within the phantom. Results show that high correlation affects mostly dipoles placed at small distances (1, 5 centimeters). In this case the sources merge. If the dipoles lie 3 centimeters apart, the beamformer localization detects attenuated power amplitudes and blurred sources as the correlation level raises.

Show MeSH
Localization results of the two different dipole couples: couple #1 and #2 (left and right, 3 cm distance, in the higher panel) and couple #1 and #3 (both left, 1,5 cm distance, in the lower panel).
© Copyright Policy - open-access
Related In: Results  -  Collection


getmorefigures.php?uid=PMC3351244&req=5

fig4: Localization results of the two different dipole couples: couple #1 and #2 (left and right, 3 cm distance, in the higher panel) and couple #1 and #3 (both left, 1,5 cm distance, in the lower panel).

Mentions: Since the localization results of LCMV and DICS appear extremely similar (as one should expect), we will focus on DICS results. The power mappings of dipoles oscillating with low, medium, and high correlation are shown in Figure 4. In presence of a low correlation level, we obtain a good localization result for both couples of dipoles (absolute maxima on dipole sites, Figure 4(a) (couple #1 + #2) and Figure 4(d) (couple #1 + #3)). Performances decrease only slightly for a correlation level of 0.55 (Figures 4(b) and 4(e)). The localizations are marginally more blurred than in the previous case, and the relative power is faintly reduced. In the case of the close dipoles #1 and #3 one could get the deceptive idea that a 55% source correlation level provides for better results than 15% (Figures 4(d) and 4(e), lower panel, XY plane). This is only because source #3 is detected as more blurred, and its presence is perceivable in the lower XY plane where the absolute maximum of source #1 is found. In the case of high correlation, the two sources are still recognizable for the couple #1 + #2 (Figure 4(c)) whereas couple #1 + #3 is detected as a single source (Figure 4(f)).


Source activity correlation effects on LCMV beamformers in a realistic measurement environment.

Belardinelli P, Ortiz E, Braun C - Comput Math Methods Med (2012)

Localization results of the two different dipole couples: couple #1 and #2 (left and right, 3 cm distance, in the higher panel) and couple #1 and #3 (both left, 1,5 cm distance, in the lower panel).
© Copyright Policy - open-access
Related In: Results  -  Collection

Show All Figures
getmorefigures.php?uid=PMC3351244&req=5

fig4: Localization results of the two different dipole couples: couple #1 and #2 (left and right, 3 cm distance, in the higher panel) and couple #1 and #3 (both left, 1,5 cm distance, in the lower panel).
Mentions: Since the localization results of LCMV and DICS appear extremely similar (as one should expect), we will focus on DICS results. The power mappings of dipoles oscillating with low, medium, and high correlation are shown in Figure 4. In presence of a low correlation level, we obtain a good localization result for both couples of dipoles (absolute maxima on dipole sites, Figure 4(a) (couple #1 + #2) and Figure 4(d) (couple #1 + #3)). Performances decrease only slightly for a correlation level of 0.55 (Figures 4(b) and 4(e)). The localizations are marginally more blurred than in the previous case, and the relative power is faintly reduced. In the case of the close dipoles #1 and #3 one could get the deceptive idea that a 55% source correlation level provides for better results than 15% (Figures 4(d) and 4(e), lower panel, XY plane). This is only because source #3 is detected as more blurred, and its presence is perceivable in the lower XY plane where the absolute maximum of source #1 is found. In the case of high correlation, the two sources are still recognizable for the couple #1 + #2 (Figure 4(c)) whereas couple #1 + #3 is detected as a single source (Figure 4(f)).

Bottom Line: Results show that high correlation affects mostly dipoles placed at small distances (1, 5 centimeters).In this case the sources merge.If the dipoles lie 3 centimeters apart, the beamformer localization detects attenuated power amplitudes and blurred sources as the correlation level raises.

View Article: PubMed Central - PubMed

Affiliation: MEG Center, University of Tübingen, Otfried Mueller Street 47, Tübingen, Germany.

ABSTRACT
In EEG and MEG studies on brain functional connectivity and source interactions can be performed at sensor or source level. Beamformers are well-established source-localization tools for MEG/EEG signals, being employed in source connectivity studies both in time and frequency domain. However, it has been demonstrated that beamformers suffer from a localization bias due to correlation between source time courses. This phenomenon has been ascertained by means of theoretical proofs and simulations. Nonetheless, the impact of correlated sources on localization outputs with real data has been disputed for a long time. In this paper, by means of a phantom, we address the correlation issue in a realistic MEG environment. Localization performances in the presence of simultaneously active sources are studied as a function of correlation degree and distance between sources. A linear constrained minimum variance (LCMV) beamformer is applied to the oscillating signals generated by the current dipoles within the phantom. Results show that high correlation affects mostly dipoles placed at small distances (1, 5 centimeters). In this case the sources merge. If the dipoles lie 3 centimeters apart, the beamformer localization detects attenuated power amplitudes and blurred sources as the correlation level raises.

Show MeSH