Systematic comparisons of interpolation techniques in topographic brain mapping

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Topographic mapping of brain potential data often relies on interpolation techniques.
• Nearest neighbor (local) and spline (global) methods are commonly employed for this purpose.
• Quantitative EEG (qEEG) analysis requires accurate spatial representation of brain activity.

Purpose of the Study:

• To quantitatively evaluate the performance of local (nearest neighbors) and global (planar and spherical splines) interpolation techniques for topographic mapping of electroencephalogram (EEG) data.
• To assess the suitability of these methods for increasing spatial resolution and in advanced qEEG analyses.

Main Methods:

• Cross-validation was used, predicting potentials at each electrode site using data from other sites in a 31-electrode montage.
• Errors were quantified using measures of inaccuracy, precision, bias, and tolerance.
• The evaluation included background EEGs from normal volunteers and patients with neurological conditions.

Main Results:

• None of the evaluated interpolation techniques performed adequately.
• Errors increased significantly for localized EEG components, potentially without limit.
• Global spline techniques outperformed the local nearest-neighbor method.
• Optimal orders were identified as 2 for nearest-neighbor and 3 for splines.

Conclusions:

• Interpolation techniques are not recommended for use with standard electrode densities (e.g., International 10-20 system) in EEG topographic mapping.
• These methods should not be used to enhance spatial resolution or in sophisticated qEEG analyses, such as radial-current density estimation.
• The findings highlight limitations of current interpolation methods for accurate brain potential mapping.