The optimal montage to mark interictal epileptiform discharges and high-frequency oscillations in intraoperative electrocorticography

Abstract

Objective: In intraoperative electrocorticography (ioECoG), interictal epileptiform discharges (IEDs) and high-frequency oscillations (HFOs; ripples 80-250 Hz, fast ripples (FRs) 250-500 Hz) can be identified in average or bipolar montage. We studied how montage choice affects event identification.

Methods: Two reviewers independently marked IEDs and HFOs across three montages (average, horizontal- and vertical-bipolar) from 13 patients who were seizure-free after ioECoG-guided surgery. We analyzed the number of channels-with-events, total events count, events morphology (maximum-amplitude, duration, frequency), number of instances with overlapping events across multiple channels (event_instance), concordance of event_instances over montages, and percentage of channels-with-events in the resected-area.

Results: Bipolar montages yielded more channels-with-events, higher counts, and greater maximum-amplitude of IEDs and ripples compared to average montages. Average and horizontal-bipolar montages yielded more IED_instances than vertical-bipolar montages. Average montages detected the highest percentage of event_instances occurring only in this montage. Event duration, frequency, and percentage of channels-with-events in the resected-area did not differ across montages.

Conclusions: All three ioECoG montages are clinically useful to find epileptic events. The bipolar montage detects more events with greater amplitude, while the average montage uncovers a wider variety of unique events. Combining montages provides complementary information.

Significance: This study quantitatively revealed how different montages capture epileptiform events.

Keywords: Electrode grid; Epilepsy surgery; High-frequency oscillations; Intraoperative electrocorticography; Montage.

Fig. 1

Schematic illustration of the ioECoG recording situations, the three montages with their corresponding channels, and three representative patient examples. A. Schematic representation of a standard electrode grid configured with AVG, HBIP, and VBIP montages. After opening the dura, the grid is placed in various positions per stage, and ioECoG is recorded in each position for 2–5 min. Based on the standard 4x5 grid configuration, the AVG consist of 20 channels, the HBIP of 16 channels, and the VBIP of 15 channels. An example of data from a single channel, showing events in the IED, ripple, and fast ripple (FR) band, is presented for each montage. AVG (in red): average reference signal is calculated by summing the voltages from all electrodes (excluding the mastoid reference, and bad channels) and dividing this by the total number of electrodes. The potential for each electrode is calculated by subtracting the average reference signal from the voltage recorded at that electrode. HBIP (in blue): for each neighboring pairs of adjacent electrodes that are transversely aligned − left to right in horizontal direction over the grid −, the bipolar potential is calculated as the voltage recorded at the left electrode subtracted from the voltage recorded at the right electrode (e.g., 01-02, 02-03, 03-04, 04-05). VBIP (in green): for each neighboring pair of adjacent electrodes that are longitudinally aligned − top to bottom in vertical direction over the grid −, the bi-polar potential is calculated as the voltage recorded at the lower electrode subtracted from the voltage recorded at the upper electrode (e.g., 01-06; 06-11; 11-16). B. Patient example (Pat.nr 13) showing IEDs in the frontal lobe for the three montages. Bi) intra-operative photographs (before-resection, pre-resection recording, and after-resection) are matched to determine channel positions and resected tissue (also for Ci) and Di)). Bii) patient characteristics (Pat.nr, histopathology, anatomical location, 1y seizure outcome) and schematic visualization of the pre-resection grid position including the resected tissue area to label channel as either resected or non-resected (also for C ii) and D ii)). Biii) in the pre-recording, IEDs were observed on AVG channel 09, 14, 15 and 20; on HBIP channel 08–09, 09–10, 13–14, 14–15, and 19–20; on VBIP channel 04–09, 09–14, 14–19, 10–15, 15–20. Instantaneous artifacts were observed on AVG channel 05, 12, 17; on HBIP channel 04–05, 11–12, 12–13, 16–17, 17–18; on VBIP channel 05–10, 07–12, 12–17. C. Patient example (Pat.nr 02) showing ripples in the parietal lobe for the three montages. Ciii) in the pre-recording, ripples were observed on AVG channel 03; on HBIP channel 02–03, 03–4; on VBIP channel 03–08. D. Patient example (Pat.nr 01) showing FRs in the temporal-parietal-occipital region for the three montages. Diii) in the pre-recording, FRs were observed on AVG channel 07; on HBIP channel 06–07, and 07–08; on VBIP channel 02–07, 07–12. Instantaneous high-frequency noise were observed on AVG channel 20; on HBIP channel 19–20; on VBIP channel 15–20. Abbreviations: AVG = average montage; HBIP = horizontally-oriented bipolar montage; VBIP = vertically-oriented bipolar montage; IED = interictal epileptiform discharge; FR = fast ripple; ioECoG = intraoperative electrocorticography; Pat.nr = patient number; FCD = focal cortical dysplasia; Engel 1A = seizure free ≥ 1 year after surgery. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Comparison of montage performance across AVG, HBIP, and VBIP montages. A. Number of channels-with-events. Ai) at recording- and Aii) patient- group level, BIP montages (both H&V) yielded significantly more channels with ripples than an AVG montage. (recording group-level: HBIP vs. AVG p < 0.0001, VBIP vs. AVG p < 0.0001; patient group-level: HBIP vs. AVG p = 0.003, VBIP vs. AVG p = 0.005). B. Total events count. Bi) at recording group-level, BIP montages (both H&V) yielded significantly more IEDs and ripples compared to an AVG montage (HBIP vs. AVG IEDs p = 0.0001, ripples p < 0.0001; VBIP vs. AVG, IEDs p = 0.008, ripples p < 0.0001); Bii) at patient group-level, BIP montages (both H&V) yielded more significantly ripples compared to an AVG montage (HBIP vs. AVG p = 0.002, VBIP vs. AVG p = 0.02); an HBIP montage yielded significantly more IEDs compared to an AVG montage (p = 0.02). C. Number of event_instances. At recording-group level, both AVG and HBIP montages yielded significantly more IED_instances compared to a VBIP montage (AVG vs. VBIP p = 0.0002; HBIP vs. VBIP p = 0.05). D. Concordance of event_instances over montages. Scaled Venn plots per event type displaying the number (% of total event_instances) of concordant event_instances at each inter- and divergence section across the three montages. An AVG montage yielded the highest percentage of unique event_instances: 17 % IEDs, 16 % ripples, and 30 % FRs. A VBIP montage yielded the lowest percentage of unique event_instances with 3 % IEDs, and 1 % ripples. E. Event morphology. Ei) at recording group-level, BIP montages (both H&V) yielded a significantly higher maximum-amplitude for IEDs and ripples than those in an AVG montage (HBIP vs. AVG IEDs p = 0.02, ripples p = 0.0001; VBIP vs. AVG IEDs p = 0.0003, ripples p = 0.003); A VBIP montage yielded a significantly higher maximum-amplitude for FRs in comparison to those in an AVG montage (p = 0.002). Eii) duration and Eiii) frequency analysis showed no significant differences for IEDs and HFOs among the three montages at recording group-level. F. Percentage of channels in pre- and intermediate-resection recordings with events located within the resected-area: There were no significant differences in IEDs and HFOs among the three montages. Statistical significances at the patient group-levels (Aii) and Bii)) were determined using the Friedman test, with the Nemenyi test applied as a post-hoc analysis. Statistical significances at the recording group-levels were determined by a GLMM (model 1 for panel E; model 2 for panel Ai), Bi), C, F) with Tukey adjustment. The inner horizontal line marks the median, and the inner box indicates the standard deviation (25th to 75th percentiles). Abbreviations: AVG = average montage; HBIP = horizontally-oriented bipolar montage; VBIP = vertically-oriented bipolar montage; IED = interictal epileptiform discharge; FR = fast ripple; GLMM = generalized linear mixed model; HFOs = high frequency oscillations; NS = not significant.