Comparison between realistic and spherical approaches in EEG forward modelling

Area of Science:

• Neuroscience
• Biophysics
• Medical Imaging

Background:

• Accurate head models are crucial for electroencephalography (EEG) to predict scalp voltages from intracranial current sources.
• The reliability of neural source reconstruction in EEG heavily depends on the accuracy of the forward model, regardless of the inverse method used.
• It remains unclear which brain regions are most affected by variations in head model geometry.

Purpose of the Study:

• To compare the accuracy of a realistic finite difference method-based head model against a standard four-layer spherical model for EEG source localization.
• To investigate the spatial variation of lead fields and identify brain regions most sensitive to head model geometry choices.
• To assess the impact of realistic versus spherical head models on the spatial discrimination of simulated neural sources.

Main Methods:

• Utilized simulated cortical sources within the MNI152 standard space.
• Employed a realistic head model based on the finite difference method and a four-layer spherical model.
• Analyzed lead field variations across the neocortex for a 62-channel EEG configuration using a point spread function.

Main Results:

• Realistic head geometry significantly improves the accuracy of neural source localization compared to spherical models.
• The benefits of realistic geometry are particularly pronounced for sources located in the temporal and occipital cortices.
• A mismatch in lead fields was observed between realistic and spherical head models, highlighting the importance of anatomical accuracy.

Conclusions:

• Realistic head models offer a substantial improvement in EEG source reconstruction, enhancing spatial discrimination capabilities.
• The choice of head model geometry critically impacts the reliability of EEG source localization, especially in specific cortical regions.
• Future EEG research should prioritize the use of anatomically realistic head models for more precise neural source imaging.