Additional file 1 of A network-based EEG source imaging framework for noninvasive localization of epileptogenic zones in MRI-negative focal drug-resistant epilepsy

Supplementary Material 1: Supplementary Table 1. Desikan-Killiany Atlas. Supplementary Table 2. Baseline Characteristics of Patients. Supplementary Table 3. Diagnostic Outcome Indicators. Supplementary Fig. 1. Preoperative assessment flowchart. Note: Stage 1a: All patients initially undergo standard non-invasive evaluations. If findings are conclusive, surgical intervention may proceed directly. Stage 1b: For inconclusive or discordant results, additional advanced non-invasive investigations may be performed to refine EZ localization or assess surgical risk. The selection of modalities depends on clinical context and institutional resources. Stage 2: If stage 1b results remain non-localizing or suggest that the EZ is adjacent to eloquent cortex, SEEG combined with cortical stimulation is employed to delineate both the EZ and functional areas. This study involved a retrospective collection of patients’ scalp EEG data and the localization of EZ using ESI, which were subsequently evaluated for concordance with postoperative resection sites and clinical prognosis. Supplementary Fig. 2. Stages of seizure. Note: The horizontal axis represents time, and the vertical axis denotes EEG channel labels. The timeline includes the interictal, preictal, ictal, and postictal phases. In this study, a 10-min segment of EEG data preceding seizure onset—highlighted by red brackets and indicated with a red star—was selected for further analysis. Supplementary Fig. 3. EEG preprocessing. Note: (a) A 10-min EEG segment preceding seizure onset was selected; (b) electrode channel positions were accurately aligned according to the 10–20 system; (c) a band-pass filter (0.5–80 Hz) and a notch filter (48–52 Hz) were applied to remove baseline drift and power line interference, respectively; (d) independent component analysis (ICA) was conducted to isolate noise components; (e) artifacts and identified interference signals were removed; and (f) the cleaned EEG data were segmented into 2-s epochs, ensuring minimal residual noise and optimal signal quality for subsequent analysis. Supplementary Fig. 4. Head models. Note: (a) Acquisition of T1-weighted magnetic resonance imaging data; (b) Image format conversion; (c) Brain surface reconstruction and segmentation quality check; (d) Import of CAT12 segmentation results into Brainstorm and generation of three orthogonal views to assess data quality; (e) Generation of a Boundary Element Method (BEM) head model; (f) Construction of a three-layer head surface comprising scalp, skull, and brain tissue; (g) EEG electrode localization; and (h) Integration of electrode position data for final construction of an individualized realistic head model. Supplementary Fig. 5. Flowchart for ESI and brain network analysis. Note: (a) EEG segments were selected and preprocessed; (b) preprocessed data were projected into source space using sLORETA to obtain the spatiotemporal distribution of cortical current density; (c) representative source signals were extracted from different brain regions; (d) a dynamic directed functional network was constructed using DTF; (e) graph theoretical analysis was applied to identify and localize potential EZs; and (f) identified EZs were compared with postoperative resection areas for validation. Supplementary Fig. 6. Schematic diagram of degree centrality metrics. Note: Each circle represents a distinct brain region node within the epilepsy network. Blue arrows denote unidirectional causal connection weights between nodes. Taking Node 1 as an example: DCin refers to the number of incoming connections it receives (2), DCout indicates the number of outgoing connections it sends (1), and the total DC (degree centrality) is the sum of both, totaling 3. Supplementary Fig. 7. Flowchart of SEEG implantation and surgical treatment. Note: (a) SEEG implantation planning was conducted based on noninvasive preoperative evaluation results, followed by electrode implantation surgery; (b) continuous SEEG monitoring was performed to capture both interictal and ictal EEG data; (c) an individualized surgical resection plan was developed according to the SEEG findings; (d) postoperative histopathological diagnosis was obtained; and (e) postoperative cranial CT scans were conducted at regular intervals to assess resection extent and detect potential complications. Supplementary Fig. 8. Multiband PSD analysis of scalp EEG source signals in 15 epilepsy patients. Note: The figure illustrates the power distribution across multiple frequency bands (δ, θ, α, β, low-gamma, and high-gamma) in the EEG signals, highlighting the spectral characteristics of brain activity. The X-axis represents frequency (Hz), and the Y-axis denotes power spectral density (dB/Hz).