A Robust Automated Pipeline for Localizing SEEG Electrode Contacts

Abstract

Stereo-electroencephalography (SEEG) provides a powerful tool for preoperative evaluation of patients with drug-refractory epilepsy, and can supply more quantified and systematic information for epilepsy surgeons. Precise localization of the electrode contacts is fundamental for the clinical evaluation based on SEEG. The difficulties of segmentation of the electrode contacts include handling the metal artifacts in post-implantation CT images, separating electrodes close to each other, and identifying the contacts belong to the same electrode. Here, we developed a pipeline for automatic segmenting the SEEG electrode contacts from post-implantation CT volume data. The pipeline mainly includes morphological closing determination (MCD), threshold-reduction region growing (TRRG), electrode enter point repair (EEPR) based on electrode geometric information, such as directions, angles, and distances, interconnected electrodes determination and separation (IEDS), and craniocerebral interference removing (CCIR). The robustness and generality of our algorithm was validated on 12 subjects (135 electrodes, 1812 contacts). Compared to the manual segmentation (240 contacts), automatic localization was more precise and 9 times faster. Moreover, the sensitivity was as high as 99.55% ± 0.60%, while the positive predictive value reached 95.25% ± 1.38%. In addition, our algorithm successfully separated 14 sets of interconnected electrodes, verifying the stability of accurately segmenting close electrodes. To provide more useful information about electrode contacts, we fused post-implantation CT images with pre-implantation MRI T1 images, and provided the structural and functional brain region of each contact using AAL atlas and Brainnetome parcellation. We also developed a friendly MATLAB-based graphical user interface (GUI) in which the pipeline was implemented.