Sensitivity of beamformer source analysis to deficiencies in forward modeling

Area of Science:

• Neuroscience
• Biophysics
• Medical Imaging

Background:

• Beamforming techniques are increasingly used for electroencephalography (EEG) and magnetoencephalography (MEG) source analysis.
• The accuracy of these methods heavily relies on the precision of the leadfield, which is influenced by volume conductor modeling.

Purpose of the Study:

• To systematically investigate the impact of inaccurate volume conductor models on EEG-based synthetic aperture magnetometry (SAM) beamformer performance.
• To analyze reconstruction errors in peak location, amplitude, and width due to model inaccuracies.

Main Methods:

• Developed a realistic human head finite element model from multimodal MRI data.
• Conducted theoretical analysis and computer simulations to assess beamformer performance.
• Varied parameters including geometry, anisotropy, sensor noise, and sensor coverage.

Main Results:

• Localization errors up to several centimeters can occur depending on source position, sensor coverage, and model accuracy.
• Beamformer peak amplitude and width are significantly affected by the interplay between noise and volume conductor model accuracy.
• Beamformers attempt to find the best leadfield fit within their scanning space, a principle applicable to other localization methods.

Conclusions:

• Realistic volume conductor models are essential for reliable EEG/MEG source reconstruction using beamforming.
• High signal-to-noise ratio benefits beamformer performance but necessitates accurate modeling.
• The study highlights the critical need for precise anatomical and biophysical head models in neuroimaging analysis.