Electroencephalography for Radiation-Free Orientation Estimation of Directional Deep Brain Stimulation Leads

Abstract

Background: Programming segmented deep brain stimulation (DBS) leads requires knowledge of their rotational orientation, but current postoperative workflows rely mainly on radiographic imaging. Objective: To develop and evaluate a radiation-free electroencephalography (EEG) framework for directional lead orientation estimation, in a realistic phantom and postoperative human data. Methods: A realistic head phantom containing a directional DBS lead was mounted on a rotating tray and stepped through known orientations in 5{degrees} increments to evaluate EEG-based orientation estimation. Stimulation artefact topographies were compared with forward-model predictions using Pearson correlation, and a tri-montage approach combined estimates from all three same-ring electrode pairs to improve accuracy. Robustness was tested by adding brain noise and reducing the electrode count. The method was then applied to a patient with bilateral directional DBS leads implanted in the globus pallidus. Robustness in the human data was assessed by reducing the sampling rate and the number of EEG electrodes. Results: In the phantom, the best simplified forward model was a tangential BEM dipole positioned midway between the active contacts. The tri-montage reduced the maximal mean orientation error to 0.59{degrees} (95% CI: 4.4{degrees}). A conservative tray-position analysis yielded a pooled mean error of 2.05{degrees} (95% CI: 3.20{degrees}). In the human dataset, EEG-derived orientations differed from imaging-based references by about 8{degrees} for the right lead and 15{degrees} for the left lead, and remained stable after downsampling and montage reduction. Conclusions: Phantom validation established the key modelling assumptions, and postoperative human recordings showed that EEG can recover clinically plausible directional lead orientations under realistic conditions.