Reconstruction of electroencephalographic data using radial basis functions

Area of Science:

• Biomedical Engineering
• Neuroscience
• Signal Processing

Background:

• Scalp potential interpolation is crucial for accurate electroencephalography (EEG) data analysis.
• Existing methods like nearest-neighbor averaging and spherical splines have limitations in predictive accuracy.
• The application of multiquadric methods, common in sphere-like surface reconstruction, to EEG has not been extensively investigated.

Purpose of the Study:

• To introduce and evaluate a novel multiquadric interpolation method for scalp potential interpolation in EEG.
• To compare the predictive performance of the multiquadric method against established techniques.
• To assess the computational efficiency of the multiquadric method relative to spherical splines.

Main Methods:

• A cross-validation approach was employed to assess interpolation accuracy.
• Two error measures were calculated: maximal error and mean square error.
• The methods were tested using 30-channel EEG data from 10 healthy volunteers.

Main Results:

• The multiquadric interpolation method demonstrated superior performance for both maximal error and mean square error.
• The multiquadric method was found to be computationally easier to implement than spherical splines.
• The findings indicate significant advantages of multiquadrics in EEG reconstruction.

Conclusions:

• Multiquadric interpolation presents a robust and efficient alternative for EEG scalp potential reconstruction.
• This method enhances the accuracy of EEG data analysis compared to conventional techniques.
• Further investigation into multiquadric applications in neuroscience is warranted.