Address correspondence to Ryder P. Gwinn, M.D., Swedish Neuroscience Institute, 550 17th St., Suite 500, Seattle, WA 98122, U.S.A. E-mail: email@example.com
Purpose: Cortical stimulation to abort seizures is under human investigation. Ideal electrode placement and stimulating parameters are unknown with poor understanding of tissue volume affected by stimulation or duration and nature of its effect on cortical activity. To help characterize this effect, we analyzed electrocorticography (ECoG) recorded adjacent to stimulated contacts during and after bipolar stimulation in patients undergoing functional cortical mapping with subdural electrodes.
Methods: We analyzed four functional mapping procedures in three patients. One row of contacts was chosen for bipolar stimulation at sequential distances. Stimulation parameters were those used for functional mapping. Pooled Teager energy (TE) and band power were calculated for: (1) baseline, (2) 5 s during stimulation, and (3) 5–15 s after the stimulus.
Results: Average TE increased during stimulation, falling with distance from the stimulus. Average poststimulus TE increased (284–905%) compared to baseline. Increased TE was observed: (1) up to 10 s after stimulation, (2) stimulation amplitudes of 4 mA or greater, and (3) up to 2 cm from the stimulus. There was no difference in poststimulus TE between the stimulated pair of contacts and outside the pair. Greatest increase in poststimulus signal power occurred in beta and gamma bands.
Conclusions: Human cortical stimulation of 50 Hz resulted in elevated ECoG energy measurements up to 10 s poststimulation. Contacts >2 cm from stimulated electrodes did not show significant response to stimulation. Separating contacts >2 cm on the cortical surface may not result in efficacious treatment of seizure activity using common stimulation amplitudes (2–10 mA).
Cortical stimulation has been performed for many years as both a diagnostic and therapeutic tool. Fritz and Hitzig first used cortical stimulation to map motor function in the dog in 1870. The first report of a physician stimulating the cortex of a human patient was probably by Robert Bartholow in 1874. A number of other surgeons began to use Faradic stimulation for functional localization over the following half century including Victor Horsley in 1887, William W. Keen in 1888, and Harvey Cushing in 1909 (Horsley, 1887; Keen, 1888; Cushing, 1909; Boling et al., 2002). Stimulation of the human brain was not routinely used in Neurosurgery, however, until Penfield and Jasper detailed the use of stimulation for both functional localization and the termination of seizures in 1954 (Penfield & Jasper, 1954). Chronic stimulation of the central nervous system (CNS) through implanted electrodes began in 1924 (Pachon & Delmas-Marsalet, 1924), and Irving Cooper implanted chronic radiofrequency stimulators in humans in 1972 to treat patients with epilepsy and movement disorders (Cooper, 1973). Today chronic stimulation of the cortical surface and deep brain structures is performed to treat movement disorders, epilepsy, stroke, psychiatric disorders, and pain.
Nathan et al. (1993) estimated the population of neurons directly stimulated with cortical surface electrodes by creating a finite element model of a hemisphere. Current density distribution resulting from stimulation was evaluated with this model. These evaluations estimated the volume conduction of electrical charge through the brain, accounting for changes in resistance between elements such as grey matter, white matter and cerebrospinal fluid (CSF), skull, and scalp. They concluded that high current densities were achieved between bipolar stimulating electrodes at separation distances of less than 1 cm, but poor current densities were seen with separations greater than 1 cm. This would imply that placing bipolar stimulating electrodes at separations of greater than 1 cm would be ineffectual in creating a uniform field between the stimulation electrodes.
Direct measurement of current in the human brain after stimulation is both difficult and limited as it does not provide information about the temporal or functional response of neurons. To begin to define the spatial-temporal profile of electrical stimulation on the human brain, we attempted to quantify the electrical changes recorded by electrode contacts near stimulated contacts in epilepsy patients undergoing functional mapping with intracranial grid and strip electrodes. These measured electrical changes reflect both the effects of stimulation due to possible volume conduction of the stimulus through CSF and tissue, as well as the response of neuronal populations beneath the stimulation and recording electrodes and connected populations. Neuronal responses to stimulation were measured using electrocorticography (ECoG) via subdural electrodes, and changes in signal energy were used as surrogate markers for changes in neuronal activity during and after stimulation.
We would ideally like to know exactly how stimulation alters neuronal behavior on a cellular and network level, both during and after stimulation. This is difficult, however, because the stimulus itself will be recorded in the electocorticogram, clouding our ability to accurately detect changes in neuronal behavior. For this reason, though we analyzed recordings taken during stimulation, the primary focus of this study was the poststimulation period which was subjected to detailed analysis.
Data was collected from four functional mapping sessions in three epilepsy patients with intracranial electrodes placed for seizure localization. All three patients were females with ages 18–29 years. Two patients had prior surgery for a right frontal cortical dysplasia and a left temporoparietal ganglioglioma. The third patient had nonlesional epilepsy and underwent first a subdural strip survey study, and then a more focused study using subdural grids. Final pathology revealed only gliosis.
Cortical strips and/or grids contained recessed platinum disk contacts with 1 cm center-to-center separation (Ad-Tech, Racine, WI, U.S.A.). These platinum electrodes are recessed within their silastic housing to ensure that there is no edge effect or signal derived from non-neuronal structures. Each contact has a total diameter of 4 mm, with a 2.3 mM exposed surface. Referential and ground electrodes are placed intracranially with contacts faced up against the dura to be electrically neutral as possible. All intracranial EEG (icEEG) recordings were examined prior to the start of the experiment to be sure that the subdural strip was performing well. During the experiment the icEEG from all contacts of the strip were continuously monitored. Following data collection, the icEEG from all electrode contacts was carefully evaluated to identify and remove artifacts. One strip or one row of a grid was chosen for stimulation and analysis after all clinical information had been collected during intracranial monitoring.
In the first case, habitual antiepileptic medications (AEDs) had not been reinstituted, in the second case one of two AEDs had been restarted. In the third and fourth cases, all AEDs had been reinstituted prior to functional mapping. No benzodiazepines were used during the course of the experiments.
The time to last spontaneous seizure was greater than 6 h in all cases. In one patient, two seizures were evoked during the course of function mapping. In this case, the reported study was conducted 1 h after the second seizure was evoked. The patient and the patients EEG were examined after the conclusion of the mapping session and a decision was made to proceed with this experiment as no effect of the seizure could be discerned.
The longest strip or grid row was chosen which was farthest from a seizure focus or cortex likely to produce seizures or clinical symptoms when stimulated (Fig. 1). The recordings from the electrode strip used for the experiment were examined before, during, and after the experiment to ensure that good icEEG recordings were obtained. The information for this study was acquired over a less than 2 h time frame using reference and ground electrodes that were subject to the same overall conditions as the test electrodes to minimize the chance that drift would be a factor influencing the results of the study. Stimulation parameters were identical to those used for the functional mapping: 50 Hz, 200 usec pulse width, 5-s train duration (Grass S-12 Stimulator, Grass Technologies, West Warwick, RI, U.S.A.). At each bipolar pair stimulation intensity began at 2 mA and increased to 10 mA in 2 mA increments, unless there were symptoms elicited by the stimulation in which case no further stimulation was performed at the bipolar pair. Central contacts in the strip/row were stimulated first. The bipolar intercontact distance was increased by 1 cm until the outermost contacts were stimulated. For example, the two contacts highlighted in Fig. 1 would be stimulated first. Following this, one of the previous pairs was kept while the next further contact out was chosen to complete the new bipolar pair. This procedure was repeated until the outer most contacts were reached.
A period of 20–45 s was observed between one stimulation and the other. Preliminary evidence indicated that ECoG changes may persist 10–15 s after stimulation, but patients had already undergone up to 2 h of functional mapping, so efforts were made not to unduly burden patients with long delays between stimuli. We therefore chose a target of 30 s between stimuli with a range between 20 and 45 s observed. A long-term potentiating (or depressing) effect of stimulation cannot be ruled out, though one has not been demonstrated during clinical mapping procedures. ECoG was recorded continuously from contacts in the strip not currently stimulated (Bio-logic Systems Corp., Mundelein, IL, U.S.A.). ECoG data was then transferred to a workstation for analysis with custom written programs in MATLAB (The MathWorks, Natick, MA, U.S.A.). Intracranial EEG was sampled at 256 Hz.
Prestimulus (0–5 s prior to stimulus onset), stimulus, and poststimulus time periods (0–5, 5–10, 10–15 s) were determined through visual inspection of the ECoG record (Fig. 2). The ECoG data were segmented into 1-s segments. Artifacts and after-discharges were delineated in the data records and 1-s segments with artifacts and after-discharges were removed from the analysis. A total of 633 stimulations were included in the study. Activity was measured from all contacts on the strip or row electrode other than the bipolar pair being stimulated.
The primary measure employed for this study was Teager energy (TE). TE is a weighted measure of signal energy (E ∝ w2A2, where A is signal amplitude and w is frequency) such that high-frequency signals contribute more than low frequency signals (Kaiser, 1990). TE was selected because it was considered, through earlier work, to be a better measure of cortical excitability than the conventional measure of signal energy (E ∝ A2). Essentially, an attempt was made to detect seizures with the use of signal energy. It was argued that a measure of the energy of icEEGs should attribute ictal segments with greater energy than interictal segments. Eleven datasets, each dataset containing a seizure and interictal data, were collected from five patients. The conventional measure of signal energy was compared to TE. It was found that TE was better than the conventional measure of energy in delineating interictal and ictal data, and thus for detecting seizures. As TE performed better than the conventional measure of signal energy for detecting seizures and distinguishing seizure EEG from background EEG, it was considered to be a better measure of excitability of intracranial EEGs (Zaveri, 1993; Zaveri et al., 1993). Furthermore, this weighted energy measurement is also in keeping with a growing understanding that physiologically important neuronal activity occurs at higher frequencies (Bragin et al., 1999, 2002, 2004; Worrell et al., 2004; Bragin et al., 2005; Jirsch et al., 2006; Le Van Quyen et al., 2006; Rampp & Stefan, 2006). TE was calculated for all artifact and after-discharge free 1-s segments of ECoG. The energy estimates were obtained after filtering the ECoG between 0 and 50 Hz. This step was performed to limit the possibility of introducing of high-frequency noise into energy estimates. In addition to TE, the power for the delta, theta, alpha, beta, and gamma frequency bands, a frequency band centered around the stimulation frequency (45–55 Hz), and a high frequency band (65–128 Hz) was obtained for each artifact free 1-s segment of ECoG by use of the fast Fourier transform. The spectral measures were obtained to better understand which frequencies contribute to observed changes in TE. The stimulus and poststimulus energy and power measures, for each electrode contact, were divided by the prestimulus measures to obtain percent changes from baseline. Average and standard deviations from pooled samples were determined for the stimulus and three indicated poststimulus time periods from the four intracranial EEG studies. Analysis was performed on normalized responses to stimulation. The normalization was performed with respect to a baseline recording prior to simulation for each electrode contact. Results are relative changes with respect to the prestimulation baseline. In this way, any change in electrode characteristics during the recordings would not influence results unless the changes to electrode characteristics are dynamic and occur over the same time period as the analysis (over seconds).
The data from each electrode contact was tabulated in the following manner. The contact was first categorized as to whether it lay between the two stimulated contacts or if it lay outside the stimulated contacts to determine whether there was a stronger response between stimulated contacts than outside of stimulated pairs. The first set of contacts was called “inner contacts” and the second set was called “outer contacts.” Data from the inner and outer electrode contacts were also pooled to form a third combined data category. In addition to classifying contacts as to whether they were inner or outer contacts, the distance between each contact and the closest stimulated contact was determined.
Three sets of results were generated. The first is a comparison of data from the prestimulus baseline to the intrastimulus data. This analysis was performed on pooled data (both inner and outer contacts) at distances ranging from 1 to 5 cm, and at stimulation intensities from 2 to 10 mA. The second set of results compared poststimulus data to prestimulation baseline for each of the three poststimulus time periods and at all intensities. The third set of results report a comparison of TE between the “inner contacts” and data for “outer contacts.” This comparison was performed for the 1-cm distance for all stimulation intensities, and for the first poststimulus time period.
Intrastimulus Teager energy
Changes in TE values compared to baseline were computed during the application of a 50 Hz stimulus. These changes are shown in Fig. 3. The maximum increases in energy during stimulation occurred at contacts closest to the stimulation contacts. Significant increases in energy were seen at all stimulation intensities (2–10 mA) for contacts from 1 to 4 cm from a stimulated contact. There were insufficient samples at 5-cm distance to make statistical comparisons. Individual patients often did not uniformly show proportionately equivalent drops in energy with each more distant contact from a stimulation site. The lack of a graded decay with distance may reflect inhomogeneity in charge density and volume conduction due to changes in resistivity between gyri and sulci or the thickness of the CSF layer below an electrode in the individual instance. The group profile, which is the average of stimulations delivered at multiple stimulation sites for all patients, demonstrates a graded decrease with distance which is maintained at different stimulation intensities.
Poststimulation Teager energy
Average TE was calculated over a 5 s period following stimulation and compared to a baseline period 5 s immediately prior to stimulation. These comparisons were made in contacts adjacent to stimulated contacts. The first comparisons were made without distinguishing adjacent contacts within a bipolar stimulated pair from those outside a bipolar pair. Observations from the same distance to the nearest stimulated contact were pooled. A subsequent comparison was made examining differences in response to stimulation whether the contact was between a stimulated pair of electrodes or outside a bipolar pair.
Results from the first time period following stimulation, for the pooled contacts, are shown in Fig. 4. The profile of the change in TE values in the first poststimulation period, as a function of distance from the stimulation site and stimulation strength, is similar to the profile obtained during the stimulation period (Fig. 3). During the first 5 s period following stimulation, the TE was 905% of baseline with a 10 mA stimulus at a distance of 1 cm from a stimulated contact. This is a considerably smaller increase than that seen during stimulation, but reflects changes in neuronal activity uncontaminated by the stimulus. The average ratio of poststimulation energy to stimulation energy, for the stimulations which resulted in significant changes was 0.016. Increasing current produced a clear dose response effect at this distance. The effect of stimulation on poststimulus TE fell with decreasing stimulus amplitude, distance from a stimulated contact and as a function of time. Stimulation with as little as 4 mA showed significant increases in TE, but only at 1 cm from the stimulated contact, and only in the first 5 s following stimulation. At a distance of 2 cm from a stimulated contact, there were significant increases in energy at 8 and 10 mA current.
The second poststimulation time epoch (5–10 s) showed significantly increased TE with 8 and 10 mA, but only at 1 cm from the closest stimulated pair (Fig. 5). More than 2 cm from or more than 10 s after stimulation, no significant changes in TE were seen in the pooled group.
Inside versus outside contacts
After determination that the significant changes in poststimulus energy take place only within 2 cm of a stimulus, a comparison was made to determine if the magnitude of response was greater for those contacts between a bipolar pair of stimulating electrodes compared to those outside. Pooled observations of average TE change for contacts 1 and 2 cm from inside a bipolar pair were not significantly higher than those 1 and 2 cm outside, indicating there was no additive effect of “surrounding” tissue at recording contacts within a pair of bipolar stimulating electrode contacts. Results of the comparisons for a distance of 1 cm from the stimulation site are shown in Table 1 (the text refers to 1–2 cm, and the Table to 1 cm).
Table 1. Change in average Teager energy, relative to baseline, in the 5 s period following stimulation for contacts inside the stimulated pair (inner contacts) and outside the pair (outer contacts)
Contacts were within 1 cm of a stimulated contact for this comparison. There was no significant difference in the Teager energy measured within a stimulated bipolar pair compared to Teager energy measured outside the stimulated bipolar pair. Results are an average from all patients.
Poststimulation spectral analysis
TE measurements reflect the energy of all component frequencies of an ECoG sample, but do not delineate the relative contribution of power at different frequencies. To help define these relative contributions, average power measurements were made for delta, theta, alpha, beta, gamma, stimulation frequency (45–55 Hz), and high-frequency (65–128 Hz) bands for parameters where the greatest TE changes were noted. This was the first 5 s following stimulation at a distance of 1 cm from a stimulus. Ratios of these estimates compared to baseline estimates are shown in Fig. 6. Increases were seen in all frequency bands at the highest levels of stimulation (8 or 10 mA); however, the higher frequency bands, particularly beta, were disproportionately increased compared to the lower frequencies, particularly at the lower stimulus amplitudes. Interestingly, the largest response poststimulation was not in the gamma, stimulation, or high frequency bands, but rather the beta band.
A relationship between stimulation strength and distance to stimulation was also observed in the beta and gamma frequency bands and not observed in the lower frequency bands. Figures depicting these measurements are included as supplemental information to this manuscript (see Supplementary Figs. S1–S7). Taken together these measures suggest that the TE measure corresponds in large part to the changes in beta and gamma frequencies, and to a much lower extent the changes in the lower frequencies (delta, theta, and alpha). The TE was calculated for ECoG bandlimited to 0.1–50Hz. The contribution of the stimulation and high-frequency band, hence, were not included in the TE measurements as the filtering performed on the ECoG precluded their contribution to TE. The signal power measurements suggest, however, that the primary changes to the poststimulation signal have been included within the TE measure. The changes in the higher frequencies were smaller, and not including these signals in the TE computation may not have resulted in a considerable loss of information.
The spatial dispersion of cortical stimulation and its subsequent effect on cortical activity has been largely unstudied despite the rapid development of devices that employ electrical impulses in an attempt to control the function of neuronal tissue. Presumably, introduction of charge to a population of neurons during stimulation produces a change in the behavior of those neurons, which we can then measure by changes in the resulting EEG activity. The model based evaluations of Nathan et al. predicted that the bipolar electrode stimulation results in poor current density distributions at clinically relevant strengths with separation distances greater than 1 cm (Nathan et al., 1993). Our results indicate that the current spread in patients undergoing cortical mapping can be detected at least 4 cm from a stimulated electrode, but measurements during stimulation are subject to a number of criticisms.
ECoG waveforms observed at nearby contacts during stimulation have typically been considered “stimulus artifact” and the deviation of activity at these contacts from baseline may reflect a number of electrical sources. These sources may include volume conduction of the electrical stimulus through cortical tissue or CSF, activity of underlying neurons during stimulation, or it may be the result of artifactual currents induced in the recording circuitry because of the stimulation. The first two sources reflect the local electrical environment that surrounds a recording electrode contact, including the neurons beneath that contact. Currents induced in recording circuitry, however, are true artifact and would lead to misleading interpretation of current flow across the surface of the brain. There is no easy way to accurately remove this potentially misleading influence; however, one would predict that such artifact would influence all adjacent electrode contacts, wires, and amplifiers regardless of the cortical location of the electrode contacts. We observed the amplitude of the measured response to be correlated to cortical distance. This is not suggestive of electronic cross-talk in the instrumentation system.
Energy measurements made after the stimulus has terminated should reflect only the activity of the underlying neuronal tissue and may be most useful in determining both the spatial and temporal dispersion of cortical stimulation. Our findings are the first to indicate that with parameters typically used in cortical mapping of patients with subdural electrodes, there is increased activity in neuronal populations at least 2 cm from stimulated electrodes that lasts as long as 10 s after stimulus termination. There was a relationship between the current applied and the magnitude of the energy increase after stimulation with greater currents resulting in increased energy. Furthermore, spectral analysis indicates that higher frequencies disproportionately contribute to the overall increase in energy, though the frequency of the stimulation (50 Hz) was higher than the band with the greatest increase in power (beta). The difference between the frequency of stimulation and the frequency of the greatest response suggests that there is a nonlinear transmission of power between stimulation and measurement sites. This spatial nonlinearity is not surprising given our understanding of the cortex.
These findings help substantiate that the cortical stimulation can change the electrical activity of underlying tissue with a predictable spatial and temporal profile. This may help to guide the placement of electrodes for therapeutic delivery of cortical stimulation, but it is important to consider the differences in stimulus parameters used in cortical mapping and those used for therapeutic intervention, particularly in epilepsy. It has been established that 50 Hz stimulation produces an excitatory response in biological systems and often results in self sustaining afterdischarges and even clinical seizures (Penfield & Jasper, 1954; Lesser et al., 1984, 1999; Blume et al., 2004). Stimulation at higher or even lower frequencies may result in very different spatiotemporal responses in neuronal tissue, and further investigations of a similar nature should be undertaken using parameters more typically used in therapeutic interventions. Furthermore, this study examined cortical response to stimulation only as far as 5 cm from a cortical stimulus. Propagation of a stimulus along network pathways may alter more distant neuronal populations and these remain to be studied.
In conclusion, this study shows that TE measurements of ECoG can be used to compare activity levels in human brain tissue before and after stimulation. In three patients with unique pathology and different electrode placements, 50 Hz stimulation caused a measurable increase in average ECoG energy during, and up to 10 s after stimulation. The increase in ECoG power after stimulation was more easily seen in frequencies >8 Hz. Energy fell as a function of time and distance from stimulation, and was no longer present more than 2 cm or 10 s after stimulation.
Contacts more than 2 cm from a bipolar stimulation electrode contact, inside or outside the bipolar pair, showed no significant poststimulation response and suggest that there is no value to separating bipolar stimulation contacts by more than this distance. As predicted by a model based evaluation, separating bipolar stimulating contacts >2 cm on the cortical surface may not allow effective current densities to be evenly distributed between the bipolar pair at commonly used amplitudes (2–10 mA). Future studies should examine local changes in ECoG activity with electrode sizes, configurations and stimulation frequencies currently used in therapeutic trials to confirm cortical response profiles with these parameters.
Work was partially funded by NIH grant R01 NS044102.
Conflict of interest: 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. None of the authors have any conflict of interest to report regarding this work.