Changes in cortical delta power during chronic invasive epilepsy monitoring

Abstract

Objective

Cortical delta band (1–4 Hz) activity is considered a biomarker for states of altered consciousness, with increased delta power observed during anesthesia, sleep, coma, and delirium. The current study sought to characterize delta power following electrode implantation with respect to patient demographics and clinical characteristics as well as type and duration of surgery.

Methods

Participants were 25 adult neurosurgical patients implanted with intracranial electrodes for clinical monitoring of their epilepsy. Resting state cortical activity was recorded at multiple occasions over the course of the monitoring period. The initial time point was defined as the first recording within 72 h following surgery. Analyses of cortical activity were conducted using a linear mixed effects modeling approach to account for within‐participant correlations and between‐participant heterogeneity.

Results

Throughout the monitoring period, delta power decreased in frontal, occipital, parietal, and temporal regions, indicating a global phenomenon. By contrast, beta (14–30 Hz) power remained stable. Delta power was higher following surgical cases that required craniotomy compared to stereoelectroencephalography cases. Surgery duration and anesthesia emergence duration were associated with higher delta power. Recordings from depth electrodes showed higher delta power compared to subdural electrodes. No significant effects of patients' age, sex, white blood cell count, antiseizure medication, and opioid medication dosage on postoperative delta power were found.

Significance

The results are consistent with a postoperative elevation in delta power that resolves over the course of the monitoring period and indicate an association between increased delta power and craniotomy surgery, as well as longer surgery and emergence durations. The current work provides a comprehensive analysis of surgical, clinical, and physiological factors, suggests risk factors, and lays fundamental groundwork for future studies.

Key points.

• Cortical delta, but not beta, activity decreases following intracranial electrode implantation surgery in epilepsy patients.

• No significant relationship was found between elevated delta power and sex, age, white blood cell count, or medication regimen.

• Craniotomy surgeries for electrode placement are associated with higher delta power than stereoelectroencephalography surgeries.

• Longer surgery and anesthesia emergence are associated with higher delta power.

1. INTRODUCTION

Elevated cortical delta (1–4 Hz; "slow‐wave") activity is a biomarker of reduced probability of consciousness in sleep, coma, vegetative states, and general anesthesia. ¹ , ² Elevated delta power, however, does not necessarily indicate loss of consciousness. ³ Increased delta power can be observed in waking consciousness, where it can indicate drowsiness and inattention. ⁴ , ⁵ , ⁶ More concerning is the association of elevated delta power with cognitive dysfunction, including dementia, sleep deprivation, and delirium. ³ , ⁷ , ⁸ , ⁹ Slow waves are postulated to disrupt information integration in the brain, leading to cognitive and perceptual impairments. ⁹ , ¹⁰ , ¹¹ Seizure activity, including interictal epileptiform discharges, is characterized by cortical slowing and increased delta power. ¹² , ¹³ Additionally, cortical injury (e.g., stroke, traumatic brain injury) and encephalopathy are also associated with elevated delta power, likely due to a neuroinflammatory response. ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ Elevated delta power predicts poor outcomes in hospitalized patients, including increased duration of stay and mortality rate. ¹⁸

Electroencephalographic (EEG) studies indicate increased delta power following major surgery. ¹⁹ , ²⁰ , ²¹ Surgery‐related factors may play a role in postoperative delta power elevation. Increased surgery duration is associated with greater rates of postoperative complications, including delirium, which is associated with global cortical slowing. ⁷ , ²² , ²³ Increased duration of anesthesia emergence is associated with a higher risk of postoperative complications. ²⁴ Antiseizure medications (ASMs) can slow cortical activity, ²⁵ as can some opioid analgesics. ²⁶

Previous research seeking to disentangle factors mediating elevated postoperative delta power and its recovery to baseline is limited, particularly following neurosurgery. ²⁷ The association of elevated delta power with unconsciousness, psychiatric disorders, cognitive and perceptual impairments, and cortical injury highlights a need to better understand this phenomenon. Our recent study reported that neurosurgical epilepsy patients who developed delirium following intracranial EEG (iEEG) electrode implantation had markedly higher cortical delta power postoperatively compared to those who did not. ⁷ It was also noted that patients generally had higher delta power early in the postoperative period.

The present study examined factors contributing to variability of delta power initially and over the course of recovery following iEEG electrode implantation in neurosurgical epilepsy patients. The factors investigated include cortical location, age, sex, medication regimens, systemic inflammatory response as indexed by white blood cell (WBC) count, type and duration of surgery, duration of anesthesia emergence, and type of iEEG electrodes. Additionally, power in other frequency bands was examined to determine whether the decline in power over time was specific to the delta band.

2. MATERIALS AND METHODS

2.1. Participants

Participants were 25 adult neurosurgical patients (13 female, aged 19–56 years, median = 36 years). The patients were implanted with iEEG electrodes to localize seizure foci for treatment of drug‐resistant epilepsy. Data were collected at the University of Iowa Hospitals and Clinics between December 2013 and April 2024. The duration of clinical monitoring ranged from 8 to 22 days. iEEG recordings for the present study were conducted between days 1 and 19. The research protocol was approved by the University of Iowa Institutional Review Board. All participants provided written informed consent and could rescind it at any time without interrupting their clinical care. Participation in this study did not interfere with clinical data acquisition.

A summary of participant demographic and clinical information is provided in Table S1. ASM and pain medication regimens are summarized in Figure S1. ASM dosage was normalized to participants' home (presurgical) medication regimen (hospital dose/home dose) and averaged across all medications. In one participant, a new ASM was prescribed during monitoring and was excluded from the normalization score. Oxycodone and hydromorphone doses were converted into morphine equivalent using an online opioid conversion calculator. ²⁸ Medication dosages were changed over the course of the monitoring period as determined by the clinical care team.

WBC count, an approximation of systemic inflammation in the body, can be obtained from routine blood draws taken for clinical monitoring of complete blood counts. ²⁹ WBC counts were acquired from the first blood sample taken within 72 h following surgery, as noted in participants' clinical records. Two participants did not have WBC count measurements during this time.

Surgery duration was defined as the time between “anesthesia start” and “anesthesia stop” logged in the clinical records. Anesthesia type was defined by the agent used at the end of each surgical case (propofol or volatile anesthetics). “Emergence start” was defined as the time at which propofol infusion was halted or when concentration of volatile anesthetics began to decline rapidly from a steady state. “Emergence stop” was defined by the removal of the breathing tube, indicating the patient was able to breathe on their own.

2.2. Recording

Data were collected in a dedicated electrically shielded suite. Recordings were obtained using depth electrodes (stereotactic EEG [sEEG]) or a combination of subdural and depth electrodes (electrocorticography [ECoG]) in 11 and 14 participants, respectively. The sEEG approach involved implantation of depth electrodes via burr holes in the skull. ³⁰ ECoG involved placement of subdural electrode arrays over the pial surface of the cortex, exposed by a craniotomy and opening of the dura mater, ³⁰ , ³¹ with additional depth electrodes implanted to target structures such as amygdala and hippocampus.

Electrodes were manufactured by Ad‐Tech Medical. Depth arrays included 4–14 cylindrical contacts along the electrode shaft, with 2.2–10‐mm interelectrode distance. Subdural arrays consisted of platinum/iridium alloy discs (2.3‐mm diameter, 5–10‐mm interelectrode distance) embedded in a silicon membrane. A subgaleal electrode over the cranial vertex was used as a reference in all participants.

Electrode implantation approach and coverage were determined solely by clinical needs. Electrode contacts were excluded from analyses if they were identified as seizure foci, were localized outside cortical gray matter, or had excessive noise per visual inspection of the iEEG data. A total of 3199 recording sites from 25 participants were included in the present study. Coverage included frontal (n = 972), occipital (n = 119), parietal (n = 428), and temporal (n = 1675) lobe and putamen (n = 5; Figure S2).

No task (resting state) cortical activity was recorded at multiple points during the monitoring period (Figure S3). The recordings were opportunistic and varied in duration between 2.0 and 94.2 min (median = 10.2 min). The time after surgery in which the first recording was made varied due to clinical factors (i.e., transfer time from the intensive care unit to the epilepsy monitoring unit). All data were collected during daytime wakefulness, at least 4 h following the participants' most recent seizure. Data were acquired with a TDT RZ2 real‐time processor (Tucker‐Davis Technologies) or Neuralynx Atlas system (Neuralynx). Recorded data were amplified, filtered (.7–800‐Hz bandpass, 5‐dB/octave roll‐off for TDT; .1–500‐Hz bandpass, 12‐dB/octave roll‐off for Neuralynx), digitized at a sampling rate of 2034.5 (TDT) or 2000 Hz (Neuralynx), and downsampled to 1000 Hz.

2.3. Analysis

Montreal Neurological Institute (MNI) coordinates were obtained for each recording site using linear coregistration to the MNI152 T1 average brain, as implemented in the FMRIB Software Library (v5.0; FMRIB Analysis Group). Reconstruction of anatomical locations of the recording sites and their mapping onto MNI coordinates was performed using FreeSurfer (v5.3; Martinos Center for Biomedical Imaging) and in‐house software. ³² Analyses were carried out using MATLAB custom scripts (vR2023b, MathWorks). Cortical activity was characterized by spectral power (μV²), calculated using the MATLAB pspectrum function. For each recording, power was averaged in canonical frequency bands (delta: 1–4 Hz, theta: 4–8 Hz, alpha: 8–14 Hz, beta: 14–30 Hz, gamma: 30–70 Hz, high gamma: 70–150 Hz). The focus on delta power was motivated by its established clinical relevance, with beta band serving as a primary comparison.

A linear mixed effects (LME) modeling approach was used for analyses of iEEG power, with the random effect of participant modeled as a random intercept or random slope and intercept. In models that predicted changes in iEEG power as a function of time after surgery, random effects included slope and intercept of recording site nested within participant to account for the multiple recording sites and time points in each participant. A summary of all LME models used is provided in Table S2. To determine whether addition of fixed effects led to an improvement of the time course model fit, theoretical likelihood ratio tests were used (MATLAB lmecompare function). LME analysis results were reported as estimates of slopes (m), 95% confidence intervals (CIs), and associated p‐values. Comprehensive summaries of LME model results, including coefficient estimates, standard errors, t‐scores, p‐values, and CIs, are provided in Tables S3–S15.

3. RESULTS

3.1. Postoperative changes in iEEG power

Data from an exemplar participant, L423, shows a typical postoperative trajectory of delta power (Figure 1). Electrode coverage in this participant included frontal, occipital, parietal, and temporal cortex (Figure 1A). Power spectra from four recording sites demonstrate power elevation in delta and, to a lesser extent, theta band, but not in beta band in the early monitoring period (Figure 1B). Delta power, averaged across recording sites in each of the four cerebral lobes, decreased over the course of the monitoring period (Figure 1C). This pattern was seen in most participants (Figure S4).

FIGURE 1.

Exemplar data from participant L423. (A) Intracranial electroencephalographic electrode coverage is shown in the frontal, occipital, parietal, and temporal lobes (teal, yellow, gray, and red circles, respectively). Larger circles represent subdural array contacts; smaller circles connected with lines represent depth electrode contacts. White circles (“N/A”) denote sites that were excluded from analysis due to proximity to seizure foci or excessive noise or were identified as faulty contacts. Depth electrode insertion sites are represented by small black symbols. (B) Power spectra are shown from four recording sites (A–D) during each recording throughout the monitoring period. Recordings are identified by time since electrode implantation in days (d) and hours (h). Thin vertical lines denote canonical frequency bands (delta through high gamma). (C) Delta power averaged across recording sites in each of the four cerebral lobes, measured over the course of the monitoring period. g., gyrus.

Delta and beta power measured at each recording site in each participant is plotted as functions of time after surgery in Figure 2A. Delta power underwent a significant decrease over time (slope estimate m = −.0601, p < .0001) in contrast to beta power (m = −.00430, p = .519; Table S3). A secondary analysis revealed that power in theta band also underwent a significant decrease over time (slope estimate m = −.0418, p < .0001), although to a lesser extent compared to delta (slope difference estimate = .0143, p < .0001; Figure S5). Power in other frequency bands did not show comparable decreases over time.

FIGURE 2.

Delta and beta power throughout the monitoring period. (A) Summary of delta and beta power measurements (purple and blue symbols, respectively) over time in all recording sites and all 25 participants. Linear mixed effects (LME) model predictions of the time course of delta and beta power are plotted as solid lines; shading represents 95% confidence intervals. See Table S3 for LME model details. (B) LME model prediction of delta power in the four cerebral lobes over time (teal: frontal, yellow: occipital; gray: parietal, red: temporal). See Table S4 for LME model details.

The location of recording sites (frontal, occipital, parietal, temporal) was examined as a fixed effect to identify potential regional differences in delta power changes over time (Figure 2B, Table S4). Addition of cerebral lobe as a fixed effect and including its interaction with time significantly improved the time course model fit (p < .0001). There was a significant decrease in delta power over time in frontal (m = −.0750, p < .0001), occipital (m = −.0523, p = .000166), parietal (m = −.0744, p < .0001), and temporal (m = −.0525, p < .0001) cortices. Pairwise comparisons revealed significantly steeper delta power slopes in frontal and parietal lobes compared to occipital and temporal lobes (p < .0001 for frontal vs. occipital, frontal vs. temporal, occipital vs. parietal, and parietal vs. temporal pairwise comparisons).

3.2. Participant demographics, clinical background, and medication

Previous work has shown an association of cortical delta power with age and sex. ³³ , ³⁴ Neither age nor sex had a significant effect on initial elevation in delta power in our cohort (p = .919 and p = .234, respectively; Table S5), and their inclusion did not improve the time course model fit (p = .411 and p = .406, respectively). Thus, there was not enough evidence to establish an association between age or sex and postoperative delta power elevation.

ASMs and opioid analgesics can increase cortical delta power. Participants' ASM dose typically started near their full home dose. Over the course of monitoring, patients were tapered off ASMs and opioids (see Figure S1). Neither drug type had a significant effect on the initial delta power elevation (m = −.0544, p = .855; m = .00113, p = .640, respectively; Table S6). Furthermore, neither ASMs nor pain medication were associated with a significant main effect on delta power in the time course model (ASM: m = .119, p = .279; pain medication: m = .00152, p = .391; Table S7).

3.3. Postoperative inflammation, recording electrode, surgery type, surgery, and emergence duration

Traumatic brain injury is associated both with inflammation and elevated delta power, ¹⁴ , ¹⁵ , ¹⁶ and high WBC count reflects increased systemic inflammation in the body. There was no significant effect of WBC count in the initial postoperative window on delta power (m = −.00139, p = .952; Table S8). Adding WBC count as a fixed effect did not improve the time course model fit (p = .409). This indicates that inflammation, measured by WBC count, was not associated with elevated delta power in our cohort.

The effect of electrode type (subdural vs. depth) was examined for potential differences in measured delta power between the two. This analysis was done on the data obtained in ECoG cases to avoid the confounding factor of surgery type (see below). Depth electrodes yielded significantly higher initial delta power compared to subdural electrodes (mean difference = .0952, 95% CI = .0285 to .162, p = .00516; Figure 3A, Table S9). To examine whether this estimate was sensitive to differences in electrode coverage, a secondary analysis was performed adjusting for cortical location (region of interest [ROI]) when estimating the difference between electrode types. Adjusting for ROI based on a parcellation previously used by Nourski et al., ³⁵ the estimated difference in initial delta power between depth and subdural electrodes was .0683 (95% CI = −.00929 to .146), that is, approximately 30% less than without this adjustment.

FIGURE 3.

Analysis of effects of recording electrode type and surgery type on postoperative delta power. (A) Comparison of initial delta power in electrocorticography (ECoG) cases with electrode type (subdural or depth). In each violin plot, symbols represent individual recording sites, white circles denote medians, horizontal lines denote means, bars denote first and third quartiles, and whiskers show the range of lower and higher adjacent values (i.e., values within 1.5 interquartile ranges below first or above third quartile, respectively). See Table S9 for linear mixed effects (LME) model details. (B) Comparison of initial delta power recorded with depth electrodes following stereotactic electroencephalography (sEEG) and ECoG surgery (blue and red symbols, respectively). See Table S10 for LME model details. (C) LME model prediction of delta power recorded from depth electrodes over time for sEEG and ECoG cases (blue and red plots, respectively). Shading represents 95% confidence intervals. See Table S11 for LME model details.

ECoG surgery requires a craniotomy for implantation of subdural arrays, whereas in sEEG cases depth electrodes are implanted via burr holes. ECoG cases (median duration = 515 min, range = 450–647 min) took longer than sEEG cases (median duration = 345 min, range = 263–451 min; p < .0001). To examine the effect of surgery duration on postoperative delta power, we first investigated the effect of surgery type. As an effect of surgery may be attributed to the type of electrode used, we examined the effect of surgery type on delta power measured using depth electrodes only. Comparison of initial delta power between ECoG and sEEG cases revealed a significant effect of surgery type, wherein ECoG cases were associated with higher initial delta power (mean difference = .384, 95% CI = .132 to .636, p = .00283; Figure 3B, Table S10). With adjustment for ROI, the estimated difference for ECoG versus sEEG cases was .309 (95% CI = .0475 to .571), approximately 20% less than without adjustment.

The effects of electrode and surgery type (see Figure 3A,B) warranted focus on depth electrodes only and the inclusion of surgery type as a fixed effect in subsequent analyses. Adding surgery type as a fixed effect to the time course model led to a significant fit improvement (p = .0124). ECoG cases were associated with higher delta power early in the monitoring period (Figure 3C, Table S11), consistent with the initial delta power difference noted above. Although delta power in ECoG cases was higher initially, it decreased at a faster rate compared to sEEG cases, and predictions for delta power in sEEG and ECoG cases were equal at 13.9 days after surgery (see Figure 3C).

Longer surgery duration can be associated with greater elevation of delta power. ²² Surgery duration was found to be associated with a higher initial delta power for sEEG cases (m = .00347, p = .00157) but not for ECoG (m = .00126, p = .303), the latter possibly reflecting a ceiling effect (Figure 4A, Table S12).

FIGURE 4.

Analysis of effects of surgery duration and emergence duration on initial postoperative delta power. (A) Linear mixed effects (LME) model prediction of initial delta power recorded from depth electrodes for stereotactic electroencephalography (sEEG) and electrocorticography (ECoG) cases (blue and red plots, respectively) with surgery duration and type as fixed effects. Shading represents 95% confidence intervals. See Table S12 for LME model details. (B) LME model prediction of initial delta power recorded from depth electrodes for sEEG and ECoG cases (blue and red plots, respectively) with anesthesia emergence duration and type as fixed effects. See Table S15 for LME model details.

There was a difference in the timing of the first recordings for ECoG and sEEG cases, with the latter typically being made earlier in the postoperative period (see Figure S3). To address the possibility that this difference affected the results, ECoG/sEEG comparisons presented in Figures 3B and 4A were replicated following exclusion of data recorded within the first 36 h of electrode implantation surgery (i.e., the first block in participants L585 through R764) and participants for whom the first block was recorded after 60 h (R376, L403, L409, and L416). This was intended to reduce the variability in timing of initial delta power measurements across participants and mitigate the potential association of timing with surgery type. This secondary analysis yielded the same findings both for comparison of initial delta power between the two surgery types and the effect of surgery duration. Specifically, ECoG cases were associated with higher initial delta power (m = .336, p = .0238; Figure S6a, Table S13). Surgery duration was associated with higher initial delta power for sEEG (m = .00341, p = .00197) but not ECoG cases (m = .000936, p = .462; Figure S6b, Table S14).

Duration of anesthesia and use of inhalational anesthetics can predict inadequate awakening, which includes delayed emergence (sedation or unresponsiveness 30–60 min following termination of the anesthetic) and delirium, a condition characterized by cortical slowing. ²³ , ²⁴ , ³⁶ One participant met criteria for delayed emergence (duration = 49 min; see Table S1). There was no significant correlation between surgery duration and emergence duration in this participant cohort (p = .333, R ² = .0783). Emergence duration was found to be significantly associated with a higher initial delta power for ECoG (m = .0516, p < .0001) and sEEG cases (m = .0108, p = .00125), with a significant interaction between surgery type and emergence duration (p = .000972; Figure 4B, Table S15).

4. DISCUSSION

Intracranially recorded cortical activity was examined in the initial postoperative period and over the course of monitoring. The study primarily focused on delta power due to its established clinical relevance in epilepsy and disorders of impaired cognition. The comparison with beta band was motivated by the utility of delta–beta ratio as an index of arousal in scalp EEG studies. ³⁷ The current study reveals that delta, but not beta, power decreased over time following surgery throughout the brain, presumably reflecting postoperative recovery from an initial elevation. ²¹ This change in delta power was parallelled to a lesser extent by the decline in theta power. Power in other frequency bands did not show comparable decreases over time, ruling out a broadband power decrease due to changes in electrode impedance.

The decrease in delta power across the cortex indicates a global phenomenon; faster decline in frontal and parietal cortex suggests that delta power may have been most elevated above baseline immediately after surgery in these two regions. The global workspace theory and the integrated information theory are two prominent theories of consciousness that emphasize the contributions of frontal and posterior parietal cortex, respectively, in the conscious experience. ³⁸ , ³⁹ With that in mind, significantly steeper delta power decrease in frontal and parietal lobes compared to occipital and temporal lobes may have implications for disordered consciousness following surgery. These results also confirm previous findings that delta power can be substantially elevated in patients who are conscious and responsive. ³ , ⁷ Thus, delta power should be used with caution when evaluating the level of consciousness in unresponsive patients, such as those with disorders of consciousness. These results also motivate follow‐up studies in which patients with elevated delta power are evaluated carefully for any accompanying mild cognitive impairment.

Elevated delta power, as measured with magnetoencephalography and scalp EEG, can be associated with seizure foci and epileptiform activity in patients with epilepsy. ¹² , ¹³ In the present study, all but one participant (L405) had seizure foci located in the temporal lobe. Thus, it is possible that the highest elevation in delta power seen in temporal cortex could in part reflect the presence of the seizure focus within the same cerebral lobe. However, we observed that delta power was elevated globally in the brain following surgery.

Age and sex were not associated with initial postoperative elevation of delta power in our cohort. Resting state brain activity is known to undergo changes across development, including a generalized cortical slowing with age. ³³ Our oldest participant was 56 years old, and it is possible that the limited range precluded identification of age effects.

ASMs and opioid analgesics did not have a significant effect on initial delta power or its change over time. These results are consistent with literature on a number of commonly prescribed ASMs, including levetiracetam, brivaracetam, perampanel, and lamotrigine which have not been found to increase delta power. ⁴⁰ , ⁴¹ , ⁴² By contrast, carbamazepine, oxcarbazepine, phenytoin, and topiramate can increase delta power. ²⁵ , ⁴³ , ⁴⁴ The effects of valproate are inconclusive, with an overall reduction in delta power reported in some epilepsy patients and an increase in others. ⁴⁵ , ⁴⁶ Given this variability, a relationship between ASM dosage and delta power may have been obscured in the current study by examining the percentage of all participants' home ASMs as a single fixed effect. In contrast to our current findings, previous studies reviewed in Malver et al. ²⁶ showed that opioid analgesics administered postoperatively were associated with increased delta power. However, Lötsch et al. ⁴⁷ demonstrated that morphine administration in men led to a decrease in delta power followed by an increase of delta over time. Fluctuating opioid effects may have limited our ability to identify a relationship between opioid dosage and delta power.

Inflammation can be associated with postoperative elevation of delta power in adults. ²¹ , ²⁷ Elevated postoperative delta activity was common in pediatric patients with lower levels of inflammatory markers (IL‐6 and IL‐8 proinflammatory cytokines) and oxidative stress, indexed by plasma ascorbate, following cardiopulmonary bypass surgery. ²⁰ Inflammation occurs around iEEG electrode insertion sites; inflammation due to traumatic brain injury is accompanied by an increase in delta power. ¹⁴ , ¹⁵ WBC count is a nonspecific marker of systemic inflammation that can be measured from routine clinical whole blood draws. ⁴⁸ High initial WBC count was hypothesized to be associated with elevated delta power; however, no significant effect of WBC count was found. ASMs can cause reduced WBC counts, particularly when the medication is effective in controlling seizures for a given patient or multiple ASMs are used in conjunction. ⁴⁹ , ⁵⁰ Thus, it is possible that participants' ASMs affected WBC count, obscuring its association with delta power. Other possible confounds include the continuous administration of antibiotic and anti‐inflammatory medications to participants, age, ethnicity, obesity, and timing of the whole blood draw with respect to the time of the initial recording. ⁵¹

We observed higher initial delta power with a steeper decline over time in recordings made with depth electrodes compared to those with subdural electrodes. Change in delta power measured by depth and subdural electrodes may be related to electrode location. Additionally, depth electrodes are implanted directly into the brain parenchyma and thus may be associated with more brain tissue damage. Depth electrodes reach mesial structures, whereas subdural electrodes remain on the brain surface. In slow‐wave sleep, delta waves originate in deeper cortical structures before traveling to the hemispheric convexity. ⁵²

Compared to ECoG, sEEG is considered less extensive and better tolerated by patients, as reflected by shorter recovery and lower postoperative complication rates. ²² , ³¹ Delta power was higher following ECoG than sEEG surgery, as measured in depth electrodes. Delta power also underwent a steeper decline over time in ECoG patients than in sEEG patients. It is possible that these findings could also reflect differences in recovery following ECoG surgery associated with infection, inflammation, and the extent of the surgery. ²² , ³¹

Longer surgery duration is associated with greater risk of postoperative complications, including infection, venous thromboembolism, bleeding, and necrosis. ²²  In this study, longer sEEG cases were associated with higher initial delta power, an effect that reached a plateau for the ECoG cases. Delayed emergence from general anesthesia is a concern for delirium, a disorder associated with elevations in cortical slow‐wave activity. ²⁴ , ⁵³ Emergence duration following ECoG cases with gas anesthesia was associated with higher initial delta power than sEEG cases with propofol anesthesia in the current study. Abnormalities in perioperative medication metabolization following cessation of general anesthesia are thought to contribute to delayed emergence, including medications reported to increase delta activity such as benzodiazepines and some opioids. ²⁴ , ²⁶ Multiple risk factors can contribute to delayed emergence, including comorbidities, genetic variation, volatile gas anesthesia, drug interactions, sleep disorders, and metabolic causes. ⁵³ This highlights the difficulty in distinguishing between these factors and a need to better understand how these factors interact with emergence duration to play a role in delta power elevation.

A caveat of all iEEG studies is generalization of conclusions to a more general patient population. Efforts were made to minimize potential impacts of the participants' epilepsy on research findings, such as recording at least 4 h after the most recent seizure. A caveat of this study is the assumption that delta power decreases over the course of the monitoring period toward a preoperative level. It is important to note that we did not have preoperative delta power measurements and could not continue to monitor it following electrode explantation. Therefore, we did not make assumptions as to whether delta power reached preoperative levels by the end of the monitoring period. Another caveat of this study is the possibility of confounding fixed effects; ECoG cases were longer on average than sEEG cases and included subdural electrode placement, whereas sEEG cases only included depth electrodes. For this reason, we only analyzed delta power recorded by depth electrodes to examine the effect of surgery. Similarly, to examine the effect of electrode type, we only analyzed delta power recorded in ECoG cases. Future studies may seek to disentangle the possible confounding factors contributing to elevated delta power.

The current study lays a foundation for future work to investigate additional contributions to postoperative elevation of delta power. Investigation of the patients' personal history including substance use, history of anesthesia, and conditions that could impact metabolism of anesthetic agents can help better understand causes of elevated delta power. ²⁴ Perioperative factors including the specific combination of medications in the operating room and their interactions can contribute to high delta power. Further research in this area may inform outcomes associated with elevated delta power, including postoperative and subsyndromal delirium. ⁵⁴

5. CONCLUSIONS

Cortical delta, but not beta power, undergoes a decrease over time following iEEG electrode implantation in neurosurgical epilepsy patients. Longer surgery duration and surgery requiring a craniotomy are associated with higher initial delta power postoperatively. The increase in delta power resolves over the course of the monitoring period, and its time course depends on surgery type. Depth electrodes yield higher delta power measures than subdural electrodes. These findings are relevant both for interpretation of intracranially recorded electrophysiological data as part of clinical monitoring and for basic research.

AUTHOR CONTRIBUTIONS

Emily R. Dappen: Data curation; formal analysis; investigation; project administration; visualization; writing—original draft preparation; writing—review & editing. Bryan M. Krause: Formal analysis; methodology; software; writing—review & editing. Rashmi N. Mueller: Investigation; writing—review & editing. Matthew I. Banks: Conceptualization; funding acquisition; supervision; writing—review & editing. Kirill V. Nourski: Conceptualization; data curation; formal analysis; funding acquisition; investigation; methodology; project administration; resources; software; supervision; visualization; writing—original draft preparation; writing—review & editing.

CONFLICT OF INTEREST STATEMENT

None of the authors has any conflict of interest to disclose. We confirm that we have read the Journal's position on issues involved in ethical publication and affirm that this report is consistent with those guidelines.

Supporting information

Data S1.

Dappen ER, Krause BM, Mueller RN, Banks MI, Nourski KV. Changes in cortical delta power during chronic invasive epilepsy monitoring. Epilepsia. 2025;66:2995–3005. 10.1111/epi.18419

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the corresponding author upon reasonable request.