Temporal stability of intracranial electroencephalographic abnormality maps for localizing epileptogenic tissue

Abstract

Objective: Identifying abnormalities on interictal intracranial electroencephalogram (iEEG), by comparing patient data to a normative map, has shown promise for the localization of epileptogenic tissue and prediction of outcome. The approach typically uses short interictal segments of approximately 1 min. However, the temporal stability of findings has not been established.

Methods: Here, we generated a normative map of iEEG in nonpathological brain tissue from 249 patients. We computed regional band power abnormalities in a separate cohort of 39 patients for the duration of their monitoring period (.92-8.62 days of iEEG data, mean = 4.58 days per patient, >4800 hours recording). To assess the localizing value of band power abnormality, we computed $D_{\mathrm{RS}}$ -a measure of how different the surgically resected and spared tissue was in terms of band power abnormalities-over time.

Results: In each patient, the $D_{\mathrm{RS}}$ value was relatively consistent over time. The median $D_{\mathrm{RS}}$ of the entire recording period separated seizure-free (International League Against Epilepsy [ILAE] = 1) and not-seizure-free (ILAE $>$ 1) patients well (area under the curve [AUC] = .69). This effect was similar interictally (AUC = .69) and peri-ictally (AUC = .71).

Significance: Our results suggest that band power abnormality D_RS, as a predictor of outcomes from epilepsy surgery, is a relatively robust metric over time. These findings add further support for abnormality mapping of neurophysiology data during presurgical evaluation.

FIGURE 1

Computing band power abnormality in a sample time window and region in Patient 1. (A) Sample 30‐s time window of intracranial electroencephalographic (iEEG) data in a subset of Patient 1's contacts. All contacts within a sample region, l.superiortemporal₂, are highlighted in blue. (B) From the 30 s of iEEG data, the relative log band power of each of the four contacts in the sample region was computed. (C) Averaging relative log band power across all of the region's contacts produces the region's relative band power. (D) Relative log band power was also computed in a separate cohort of 249 subjects, yielding a normative map of this measure. (E) Patient 1's regional relative band power was then z‐scored relative to the normative map. The region's abnormality was defined as the maximum absolute z‐score (here, 2.07) across the five frequency bands. (F) The process is repeated for all regions. (G) The abnormality values are normalized so their sum equals 1 and plotted for resected and spared regions. This process was repeated for all time windows in each patient.

FIGURE 2

Time‐varying abnormalities and $D_{\mathrm{RS}}$ in sample Patients 1 (A–D) and 2 (E–H). (A, E) Heatmap of regional maximum absolute band power abnormalities, with each column corresponding to a 30‐s time window in the patient's intracranial electroencephalographic recording. Abnormalities in each time window are normalized to sum to 1, thus showing each region's contribution to the total abnormality in that time window. Resected regions are outlined with a black box. (B, F) Time‐varying $D_{\mathrm{RS}}$ computed from band power abnormalities. Histogram to the right of each plot shows the distribution of $D_{\mathrm{RS}}$ in each recording, with the median $D_{\mathrm{RS}}$ marked with a bold horizontal line. The circle and dashed vertical line mark a sample time window that had a $D_{\mathrm{RS}}$ equal to the patient's median $D_{\mathrm{RS}}$. $D_{\mathrm{RS}}=.5$ is also shown with a dashed black line for reference. (C, G) Normalized abnormalities of spared and resected regions in the sample time window with $D_{\mathrm{RS}}$ equal to the patient's median $D_{\mathrm{RS}}$. Quartiles of the abnormality distributions are marked with dashed lines. (D, H) The same abnormalities on a brain surface from top and side views. ILAE, International League Against Epilepsy.

FIGURE 3

Variability in $D_{\mathrm{RS}}$ across patients. (A) Distribution of $D_{\mathrm{RS}}$ in each patient's intracranial electroencephalographic recording, as shown in sample patients in Figure 2B,F. Bold vertical lines show the median of each distribution. Number of days of data used to compute each distribution is also provided. (B) Localizing percentage of time windows (i.e., percentage of time windows with $D_{\mathrm{RS}}$ $\le$ .5) in our cohort. (C) Comparison of median $D_{\mathrm{RS}}$ in patients who were seizure‐free (International League Against Epilepsy [ILAE] = 1) versus not seizure‐free (ILAE = 2–5) after surgery. Quartiles of the $D_{\mathrm{RS}}$ distributions are marked with dashed lines. (D) Receiver operator characteristic curve using median $D_{\mathrm{RS}}$ as a binary classifier of patient surgical outcome. AUC, area under the curve.

FIGURE 4

$D_{\mathrm{RS}}$ in interictal and peri‐ictal periods. (A, B) Interictal and peri‐ictal $D_{\mathrm{RS}}$ in sample Patient 3. Seizure times are marked with vertical dashed red lines. (A) Heatmap of time‐varying regional band power abnormalities (normalized to sum to 1). Resected regions are outlined with a black box. (B) Time‐varying $D_{\mathrm{RS}}$, colored by whether the time window was interictal (gray) or peri‐ictal (orange). Histogram to the right shows the distribution of $D_{\mathrm{RS}}$ and median value (bold horizontal lines) in each time period. (C, D) Interictal and peri‐ictal $D_{\mathrm{RS}}$ across patients. (C) Median peri‐ictal $D_{\mathrm{RS}}$ versus median interictal $D_{\mathrm{RS}}$ of each patient, colored by patient surgical outcome. (D) Comparison of median interictal (left) and peri‐ictal (right) $D_{\mathrm{RS}}$ in patients who were seizure‐free versus not seizure‐free after surgery. Quartiles of the $D_{\mathrm{RS}}$ distributions are marked with dashed lines. AUC, area under the curve; ILAE, International League Against Epilepsy.