Ordinary least squares dipole localization is influenced by the reference

Area of Science:

• Neuroscience
• Biophysics
• Signal Processing

Background:

• Accurate equivalent source estimation is vital in fields like EEG and MEG.
• The choice of reference electrode significantly impacts signal quality and analysis outcomes.
• Understanding reference effects on noise correlations is key to improving source localization accuracy.

Purpose of the Study:

• To investigate how common and average referencing techniques differentially affect equivalent source estimation accuracy.
• To elucidate the role of noise correlations in mediating the impact of referencing methods.
• To compare the performance of ordinary least squares (OLS) and generalized least squares (GLS) estimation with different referencing strategies.

Main Methods:

• Empirical data analysis.
• Analytical derivations.
• Computational simulations.
• Comparison of Ordinary Least Squares (OLS) and Generalized Least Squares (GLS) estimation techniques.

Main Results:

• Common and average referencing methods yield different accuracies for equivalent source estimates.
• The influence of the reference on noise correlations is the primary mechanism behind this differential effect.
• OLS estimation with average referencing is preferable to common referencing but remains sub-optimal.
• GLS estimation, which accounts for correlated noise, provides optimal accuracy with both average and common referencing, leading to comparable results.

Conclusions:

• Reference selection critically influences source estimation, mediated by noise correlations.
• For OLS, average referencing is the recommended approach, though not ideal.
• GLS offers a superior solution for accurate source estimation, regardless of the referencing method used, by properly modeling noise correlations.