Predicting monopolar local field potential power from bipolar recordings in deep brain stimulation

Area of Science:

• Neuroscience
• Biomedical Engineering

Background:

• Deep brain stimulation (DBS) is a key therapy for neurological disorders like Parkinson's disease (PD).
• Modern DBS devices record local field potentials (LFPs) using bipolar configurations to reduce noise and artifacts.
• Bipolar recordings can attenuate important physiological signals, limiting precise LFP analysis.

Purpose of the Study:

• To develop a model for estimating monopolar LFP power from bipolar recordings.
• To provide a method for more spatially precise LFP signal analysis in DBS.

Main Methods:

• A retrospective study analyzed 64 PD patients undergoing STN or GPi DBS implantation.
• Local field potentials (LFPs) were recorded, and power spectral density (PSD) was calculated for monopolar and bipolar signals.
• A linear regression model was developed and validated to predict monopolar power from bipolar recordings.

Main Results:

• The developed model accurately estimated monopolar LFP power from bipolar recordings, with high R-squared values (e.g., 0.9015 for C0).
• The model demonstrated strong generalizability across different DBS targets (STN, GPi) and validation sets.
• Root Mean Square Errors (RMSE) were low, indicating reliable predictions.

Conclusions:

• Monopolar LFP power can be accurately estimated from bipolar recordings using a linear regression model.
• This hardware-agnostic approach enhances spatial signal disambiguation for DBS.
• The findings support improved DBS programming and adaptive stimulation strategies.