Advanced Electroclinical Evaluation in Drug-Resistant Epilepsy: Integration of Subcortical Source Modeling and 3D Electrode Mapping into Electrical Source Imaging

Drug-resistant focal epilepsy affects approximately 30% of people with epilepsy, and accurate identification of the epileptogenic zone (EZ) remains a critical determinant of surgical eligibility. Electrical Source Imaging (ESI), which reconstructs the cerebral sources of electrical activity from scalp EEG signals, has been increasingly proposed as a complementary tool to the multimodal neuroimaging techniques currently employed in presurgical evaluation. The aim of this work was to extend a previously developed ESI pipeline through the introduction of two methodological advancements: the implementation of a three-dimensional, patient-specific localization of EEG electrodes, and the integration of subcortical structures into the source model. The pipeline was developed in MATLAB using the FieldTrip toolbox and applied to six patients enrolled in the epilepsy surgery programme of the Level III Regional Epilepsy Centre, Modena Academic Hospital, who underwent multichannel EEG and T1-weighted MRI acquisition. Patient-specific head model was built using the T1-weighted MRI, while source space was defined using three different source models: a purely cortical model, and two extended cortico-subcortical models including the hippocampus, amygdala and thalamus, represented respectively as surfaces or as a surfaces and volumes structures, alternatively. EEG electrode positions were refined through three-dimensional patient-specific head surface scanning and compared with the standard Micromed configuration. Source localization was performed using the eLORETA algorithm, with subsequent parcellation according to the Desikan-Killiany atlas. Analytical feasibility was assessed by evaluating the overlap between regions showing the highest mean squared current density, the clinically presumed EZ and, where available, the surgical resection area. The agreement between the identified regions and the clinically presumed EZ was satisfactory in approximately half of the cohort, whereas the remaining patients showed partial overlap with the expected lobar localization. Among operated patients, the overlap between reconstructed sources and the surgical cavity appeared to correlate with postoperative outcome, reaching the highest degree in the patient achieving complete seizure freedom. Subcortical structures included in the extended source models were identified among the regions with the highest mean squared current density in almost all patients; in one case, the detected subcortical structure was located within the resected area, while in another it represented the region with the highest mean squared current density, exceeding all cortical regions. The comparison between 3D-scan patient-specific and standard electrode configuration showed comparable results, with only minor and-non-systematic differences. Despite the preliminary nature of these findings and the limited single-centre cohort, the results suggested that the integration of patient-specific electrode localization and cortico-subcortical source models into an automated ESI pipeline is technically feasible and may represent a reliable and valuable tool for more accurate forward problem formulation. However, ESI should not be considered an independent criterion for surgical decision-making, particularly with low-density electrode configurations, but rather as a complementary method to support a more accurate localization of the EZ.