Probabilistic forward model for electroencephalography source analysis.

Area of Science:

• Neuroscience
• Biophysics
• Medical Imaging

Background:

• Accurate source localization using electroencephalography (EEG) and magnetoencephalography (MEG) relies on precise head models, including skull conductivity.
• Magnetic Resonance Imaging (MRI) excels at soft tissue anatomy but lacks skull resolution, while Computed Tomography (CT) provides this but is not routine.
• Current noninvasive methods for mapping skull conductivity in 3D are insufficient.

Purpose of the Study:

• To develop a probabilistic forward modeling framework to quantify uncertainties in EEG/MEG source localization.
• To investigate the feasibility of estimating skull conductivity directly from EEG data.
• To assess the impact of skull conductivity uncertainties on source localization accuracy.

Main Methods:

• Introduced a probabilistic forward modeling approach to propagate parameter uncertainties into source localization errors.
• Explored simultaneous optimization of dipole parameters and skull conductivity values using EEG data.
• Utilized Cramer-Rao bounds to evaluate the reliability of conductivity estimation and its effect on localization.

Main Results:

• The proposed probabilistic framework allows for the propagation of uncertainties in head model parameters, such as skull conductivity, into source localization estimates.
• Simultaneously optimizing for dipole parameters and skull conductivity from EEG data, as previously suggested, did not improve source localization accuracy.
• This joint optimization approach failed to yield reliable estimates of skull conductivity.

Conclusions:

• Skull conductivity must be accurately measured independently or uncertainties must be explicitly incorporated into source location estimates.
• Probabilistic forward modeling is crucial for managing uncertainties in EEG/MEG source localization when precise conductivity values are unavailable.
• The study highlights limitations in current noninvasive methods for determining skull conductivity and its impact on neuroimaging.