Safety and Utility of Hybrid Depth Electrodes for Seizure Localization and Single-Unit Neuronal Recording

Invasive monitoring remains a cornerstone in the diagnostic workup of patients with medically refractory epilepsy. Invasive monitoring using depth electrodes offers an opportunity to localize seizure onset zones in deeper structures such as the hippocampus, the amygdala or the insula, when they cannot be localized accurately with scalp electroencephalography (EEG) or other noninvasive methods. This technique, also referred to as stereo-EEG (SEEG), provides more anatomically precise ways to evaluate seizure network spread patterns [1]. However, the multiple brain penetrations associated with depth electrodes pose certain risks, such as intracerebral bleeding, infection, and cortical damage. Several reports document the benefits and risks of depth electrodes relative to subdural grid and strip monitoring. These reports demonstrate that, relative to grids and strips, depth electrodes are associated with lower complication rates, and better outcomes have been reported 1 year after resective surgery [2-7]. Thus, there is considerable justification to utilize depth electrodes instead of grids when clinically warranted.

A secondary benefit of depth electrode monitoring is that it permits the study of single-neuron activity during both normal behavior and ictal and interictal events [8]. For these applications, “hybrid” depth electrodes are used that contain additional microwire bundles that exit from the tip of the clinical macro-contact electrode (Fig. 1a). This modification provides a tremendous opportunity to study the human nervous system at a single-cell resolution in awake, behaving humans [8]. This setup has led to key new insights into the mechanisms of cognition, including long-term and working memory [9, 10], high-level visual representations [11-13], face processing and its impairment in autism [14, 15], processing of emotions [16, 17], and decision making [18]. To date, only one study has documented that the addition of microwires did not increase the risks of implantation [19]. However, when compared to standard depth electrodes, there is still a paucity of information on the safety and risks associated with hybrid depth electrodes. To address this concern, we performed a single-center retrospective analysis of our most recent 10-year case series with hybrid depth electrodes, seeking to compare the safety profile and ability of hybrid electrodes to localize seizures with well-reported results for non-hybrid electrodes.

The personal surgical database of a single surgeon (A.N.M.) was reviewed to identify all patients with medically refractory epilepsy who underwent stereotactic depth electrode insertion procedures between 2006 and 2017. The medical chart for each patient was reviewed to extract the following information: patient demographics (age, gender, race), duration of epilepsy until time of implant, indication for depth electrode monitoring, number and sites of electrodes implanted, types of electrode implanted, and final localization of seizures (if localized). Records were also reviewed to determine the length of implantation and any complications related to the implant procedure, recording period, or explant and postoperative course. The study was approved by the Institutional Review Boards of all participating hospitals.

The surgical methods employed remained essentially unchanged over the entire study period and are described in detail elsewhere [20]. Patients underwent a preoperative 1.5- or 3-T MRI of the entire brain with and without contrast, including SPGR (or equivalent) sequences in the axial and coronal planes. A CT angiogram with the stereotactic frame in place was acquired on the morning of surgery, and co-registered to the MRI during electrode targeting planning, which was performed on the same day as surgery. Both the CTA and MRI with contrast were used to avoid traversing sulci, and to identify surface arteries and veins, deeper vessels, and anatomic targets. Electrode targeting was planned with the Framelink® (Medtronic, Minneapolis, MN, USA) stereotactic software suite. Implantations were performed using a CRW (Integra, Plainsboro, NJ, USA) stereotactic frame with a custom-made Cosgrove Depth Electrode Insertion Kit (AdTech Medical, Racine, WI, USA). This resulted in orthogonal electrode placements (Fig. 1b), which were used for the great majority of electrode insertions. For nonorthogonal placements, the standard CRW arc was used.

Patients were implanted either under deep sedation with propofol or under general anesthesia (laryngeal mask). There were no anesthetic complications. For orthogonal targets, a standard targeting method was utilized as described in detail elsewhere [20]. The three types of electrodes used in this series were standard 1.1-mm-diameter non-hybrid depth electrodes (Spencer® SD08R-SP05X-000; Ad-Tech Medical, Racine, WI, USA), 1.3-mm-diameter hybrid depth electrodes [21] with associated internal microwires (Behnke-Fried BF08R-SP05X-000 and WB09R-SP00X-0B6; Ad-Tech Medical) [20] (Fig. 1a), and 0.8-mm-diameter non-hybrid Spencer® SEEG electrodes (RD06R-SP05X-000; Ad-Tech Medical). When hybrid Behnke-Fried electrodes were used, the macroelectrode was first inserted to target, and the microbundle was then passed through the hollow core of the macrobundle so that it protruded 4–6 mm from the end of the macroelectrode, after which the anchor bolt cap was tightened to hold the electrode assembly in place. Typically, we placed electrodes in bilateral symmetric mesial temporal and mesial frontal targets, with supplementary and nonorthogonal electrodes when clinically indicated. We obtained formal anteroposterior, lateral, and submental vertex plain skull films as well as postoperative MRI without contrast on all patients to determine final electrode placements (Fig. 1b).

Patients received intravenous cefazolin for 48 h after surgery. Antiepileptic drugs were tapered as indicated, and dressing changes were also performed typically once per week. When monitoring was complete the electrodes were removed in the operating room under sedation. Patients typically left the hospital the next morning.

Complete data were available for a total of 52 patients, 42 from Hospital 1 and 10 from Hospital 2, who underwent a total of 53 stereotactic placement of depth electrodes surgeries. Patient demographics are listed in Table 1. Briefly, there were 26 males and 26 females, with an average age at implant of 37 ± 14.1 years (range 16–70 years). The mean length of time with medically refractory epilepsy prior to depth electrode monitoring was 18 ± 15.2 years (range 1–53 years). Patient ethnicity could only be retrieved for Hospital 1 patients due to limited access to full demographic data at Hospital 2. Of the 42 Hospital 1 patients, 38.1% were Caucasian, 11.9% African American, 9.5% Asian, 33.3% Hispanic, and 7.2% other. As both hospitals were in similar catchment areas, demographics were likely very similar between the two centers.

The average time spent for continuous video EEG epilepsy monitoring was 15 ± 7.0 days, with an average hospital stay of 16 ± 6.8 days. The number of electrodes in each site over the entire patient population is listed in Table 2. The majority of patients underwent bilateral implant with electrodes targeting the mesial temporal areas (amygdala, hippocampus) and frontal (orbitofrontal gyrus, pre-supplementary motor area, anterior cingulate cortex) lobes. These areas were determined by clinical criteria alone. No electrodes were placed solely for research purposes. In some cases, electrodes were inserted in nonstandard target areas; typical indications for these electrode placements were suspected structural lesions (e.g., focal cortical dysplasia or tumor), or noninvasive monitoring suggesting potential involvement of that region. A total of 555 electrodes were implanted over the 11-year period reviewed (Table 2). On average, each patient had 10 electrodes. Of the 555 electrodes, 244 (44%) standard depth electrodes (Spencer®), 288 (52%) hybrid (Behnke-Fried) electrodes with associated internal microwires (Fig. 1a), and 23 (4%) smaller-diameter non-hybrid SEEG electrodes were implanted (Table 2). Of 53 cases, 43 (81.1%) included hybrid depth electrodes, with an average of 5 ± 3 hybrid depth electrodes used per case. As the hybrid electrodes were primarily used for research purposes, they were only implanted in the areas relevant to our research objectives. However, over time and as our studies expanded, hybrid electrodes replaced non-hybrid ones in many areas. For example, in cases early in our series hybrid electrodes were implanted solely in the amygdala and hippocampus, while for later cases (after 2009) hybrid electrodes were also implanted in the frontal lobe targets. SEEG electrodes were used in later (after 2014) cases, targeting the insula and parieto-occipital regions when indicated by noninvasive data.

The reason for depth electrode implantation was localization of epileptic foci. We therefore first categorized localization efficacy into three categories: not localized, partially localized, and localized. Poor localization occurred if the characteristics of the seizures captured suggested no clear epileptic focal point, there was a lack of seizure activity during monitoring, onset appeared not to arise from any of the contacts, or monitoring was ended prior to clear localization due to patient safety concerns. Partially localized seizures include those that were identified as possibly neocortical, but subdural grids were subsequently needed to further localize. Alternatively, areas of involvement were identified, but no clear localization of onset could be determined, thus further localization procedures were needed. Localized patients demonstrated stereotypical and reproducible seizure onset patterns, or multifocal independent onsets that no longer made the patient a resective surgery candidate. Of the 52 patients (53 cases), 32 (61.5%) had their epilepsy localized, 10 (19.2%) had their epilepsy partially localized, and 10 (19.2%) did not have localization.

To compare the seizure-localizing capabilities of non-hybrid versus hybrid electrodes, we analyzed the localization outcome in patients who were implanted with at least 4 hybrid electrodes (as this was the minimum number of hybrid electrodes used in all cases involving hybrid electrodes) and compared it to patients who only had non-hybrid electrodes implanted. We found that the probability of epilepsy localization was not different between the two groups (p = 1, Fisher exact test 2 × 2 contingency table). When restricting the localization criterion even further to localized, partially localized, or not localized, we found that 72.5% of patients who had hybrid electrodes implanted had their seizures localized, 15.0% had them partially localized, and 12.5% had their seizures not localized (Fig. 2). In patients with only non-hybrid electrodes implanted, 44.4% of patients had their seizures localized, 44.4% had them partially localized, and 11.1% had their seizures not localized (Fig. 2). Like our previous comparison, we found that there was no statistical difference in the probability of having seizures localized, partially localized, or not localized between patients who had at least 4 hybrid electrodes implanted and patients with only non-hybrid electrodes used for monitoring (p = 0.15, Fisher exact test 2 × 3 contingency table). Patients who were categorized as not localized due to lack of seizure activity during monitoring, or monitoring was ended prior to clear localization due to patient safety concerns, were not included in these comparisons.

After depth electrode monitoring and any additional diagnostic tests were completed, common procedures that followed were resective surgery or MRI-guided laser ablation (Visualase®; Medtronic) if localization was determined, subdural grid placement for further localization, vagal nerve stimulator (Cyberonics Inc., Houston, TX, USA) implantation for patients with nonlocalized epilepsy, responsive neural stimulation (NeuroPace®, Mountain View, CA, USA) implant, or in some cases, no further surgical treatment (Table 3).

Of 53 procedures and 555 depth electrode insertions, operative or postoperative complications were noted in 11 (20.8%) procedures. This translated to an overall per-electrode complication rate of 2.3%. We quantified each type of complication observed on a per-electrode basis (Table 4). For 7 electrodes (1.3%), complications arose from difficulties involving bolt or electrode insertion or removal procedures. For example, during the insertion procedure, part of an anchor bolt fractured and was left in place in the skull or later removed during electrode removal. In one instance, a hybrid electrode sheared off during removal and was later removed during the scheduled resection surgery. It is important to note that hardware breakages, which did not result in brain damage or were otherwise clinically insignificant, made up most of our complication rate and when removed our complication rates were 11.3% per case and 1.1% per electrode. There were no instances of electrode-related infection, either during implantation and continuous EEG monitoring, or after removal.

We documented 6 occurrences of postoperative bleeding from either an electrode insertion or removal (1.1%) in 5 patients. Of these 6 occurrences, 4 involved small or asymptomatic bleeds that did not have a neurological consequence and no surgical intervention was needed. These included 1 patient noted to have a small subarachnoid bleed and a small subdural hematoma (Fig. 3a) after electrode removal, 1 patient with a small left intraventricular bleed (Fig. 3b), and 1 patient with diffuse subarachnoid bleed after implantation (Fig. 3c), without clinical sequalae or underlying vascular lesion identified by MR angiogram. All 3 patients were asymptomatic from these episodes, with blood detected on routine postsurgical imaging. Postoperative bleeding that required surgical intervention occurred in 2 cases (0.4%). One patient (Fig. 3d), was noted to have left-sided hemiparesis and a dilated right pupil in the recovery room 30 min after removal of depth electrodes. A CT revealed an acute right subdural hematoma, which was urgently evacuated via craniotomy with full recovery. At surgery to remove the hematoma no clear source of bleeding was identified. The patient made a rapid recovery and subsequently underwent temporal lobectomy 6 weeks later. A second patient (Fig. 3e, f) developed a left anterior temporal intraparenchymal hematoma that was identified shortly after implantation, while still in the recovery room. The patient became acutely aphasic and somnolent and underwent emergency removal of the electrodes and evacuation of the hematoma. At surgery active bleeding from a small artery in the anterior temporal lobe was identified as the suspected source of bleeding, although it was not clear if this was caused by a hybrid electrode or an SEEG electrode, both placed within 1 cm of each other. The patient’s aphasia progressively improved almost back to normal by 4 months after surgery, and she has remained seizure free. One patient developed a small lacunar left thalamic infarction with residual sensory deficit after deplantation. As none of the electrodes were near the site of infarct, this was suspected to be related to the patient’s history of thrombophilia rather than electrode insertion or removal and thus did not contribute to our overall per-electrode complication rate.

We compared cases in which 4 or more hybrid electrodes were used (“hybrid cases”) to cases where no hybrid electrodes were used (“standard cases”). When comparing the per-electrode complication rates between hybrid and standard electrode cases, we found that 2 (0.7%) complications occurred with standard electrode cases and 5 (1.7%) complications occurred with hybrid cases This difference was not statistically significant. This excludes any complications specifically involving bolt insertion and removal as these do not depend on electrode type, and the same bolt was used for both electrode types (except for SEEG electrodes for which 1 bolt-related complication was reported). Additionally, when looking at the specific characteristics of each complication such as location of bleed and what vascular structures were identified as the source of bleeding, we determined that no complications could be directly or indirectly attributed to the use of hybrid electrodes. More specifically, the 2 clinically significant bleeds did not occur near the site of the microwires, and the 4 instances of nonclinically significant bleeds did not occur proximal enough to the microwire locations to be directly attributed to the hybrid electrodes. Since the microwire bundle is inserted only after the main macroelectrode is inserted, only bleeds that occurred at the most medial aspect of the electrode would be directly related to the microwires, and no bleed of that sort was identified.

The yield of extracellular single-neuron action potentials recorded from the hybrid microwires was variable and improved over time as techniques and equipment improved. Early in our experience single-unit data could be collected from an average of 3–4 (37%) microwire bundles, while later single units were typically achieved in 70–80% of microwires. Units could reliably be recorded for over 2 weeks. An average of 3–7 isolated single units could be obtained from each microwire, with extreme variability ranging from 0–16 units per bundle. The amygdala tended to yield the most reliable recordings, followed by the anterior cingulate cortex, pre-supplementary motor area, and hippocampus. High-quality local field potentials could be recorded from microwires in essentially every bundle unless the entire bundle had high impedances, which occurred in < 5% of all electrodes implanted.

To effectively manage medically refractory epilepsy, precise localization of the epileptogenic zone is crucial. Invasive electrode monitoring provides an opportunity to more accurately localize the zone of seizure onset, especially in cases of nonlesional epilepsy. Depth electrodes offer distinct advantages in these settings, allowing for monitoring of medial structures, white matter, and lateral cortical structure simultaneously. Several series have documented that depth electrode monitoring is both safe and effective, with overall per-case complication rates ranging from 1 to 26% [2-6, 19, 22, 23]. Our series reports similar results, with an overall per-case complication rate of 20.8% and a per-electrode rate of 2.3%. The majority of our reported complications were minor and related to hardware breakage. This may suggest an error in surgical technique and indeed we are now very careful when tightening in the anchor bolts to avoid shearing loads. No instances of anchor bolt breakage have been noted in the past 18 months.

Overall hemorrhagic complications were extremely low, at 0.4% per electrode. Only 2 out of 52 patients required surgical intervention. None of the complications that occurred could be directly attributed to the use of hybrid electrodes containing microwires. Further, it seems improbable that the 0.2-mm difference in diameter between the hybrid (1.3 mm) and standard (1.1 mm) electrodes would result in a meaningful change in hemorrhage rate, a finding supported by our data.

Our reported hemorrhage rate with hybrid electrodes was slightly higher (1.7%) than that with standard electrodes (0.7%). This difference was not statistically significant. As we implanted more hybrid electrodes (52%) than standard (44%) or SEEG electrodes (4%), we might expect to see a slightly higher bleeding rate with hybrids. If we assume the risk for bleeding is identical for all electrode types, we would theoretically have expected to see a 1.0% bleed rate with standard electrodes, a 1.2% rate with hybrids, and a 0.1% rate for SEEG electrodes. These results closely approximate our findings and suggest that there does not appear to be any difference in complication rate between standard and hybrid electrodes.

Of note, since we used identical anchor bolts for our standard and hybrid electrodes, and hardware breakage represented our most common complication, it is possible that there would be a reduction in complications if only SEEG anchor bolts, which are smaller, were used. However, as only 4% of our electrodes were SEEG we have insufficient data to support his view. A larger, multicenter database incorporating significantly large numbers of patients with both SEEG and hybrid electrodes is needed to more definitively answer this question.

Some studies investigating complication rates of invasive electrodes compare depth electrodes to subdural grids by reviewing cases in which both types of electrodes were implanted concurrently [2-6]. Placantonakis et al. [4] reported 26 patients where subdural strips were used, and depth electrodes placed in mesial lobes were added in 50% of those patients. They noted no complications involving the depth electrodes. Similarly, Hedegärd et al. [2] reported minor complications in 13 of 271 (4.8%) cases of invasive monitoring. However, this complication rate includes both grid and depth electrodes. Of their 14 depth electrode cases, 1 had a complication, resulting in a per-case complication rate of 7.1%. In the present series we review a larger pool of procedures utilizing only depth electrodes. Thus, we present a more realistic evaluation of depth electrode risks and complications.

Schmidt et al. [5] retrospectively reviewed 317 electrode implantation procedures with a total of 768 depth electrodes and showed a case-based complication rate of 25.5%. The majority of patients underwent 3 or 4 electrode implants per patient, which makes it difficult to compare to our series (average 10 electrodes per patient) or other SEEG studies with a higher number of implants. However, it is interesting to note that our case-base complication rate is similar. This observation suggests that safety is not significantly compromised by adding more electrodes, although the limits of how many electrodes would significantly increase risk remains to be determined.

Our study reports complications not noted by others. For example, we include technical complications such as bolt or electrode insertion and/or removal difficulties, which were not considered in the previous studies. When removing this category from our analysis, we found a 11.3% per-case and 1.1% per-electrode complication rate, which compares favorably to the existing studies that exclude this criterion. Furthermore, a recent meta-analysis of depth electrode monitoring cases [3] found that in 2,624 patients, the most common complication was hemorrhage followed by infection. However, we found our most common complication to be hardware malfunction or breakage involving an anchor bolt or an electrode during insertion or removal procedures, which pose minimal threat to patients and are easily resolvable. We report no cases of infection.

An important goal of this study was to demonstrate the safety of hybrid depth electrodes with associated microwires during monitoring, which has not been previously documented for a subject pool of this size. Hefft et al. [19] reviewed 25 cases using both hybrid and standard depth electrodes and reported no clinically significant complications associated with hybrid depth electrodes. Likewise, when examining our cases, we were unable to conclude that hybrid depth electrodes (as compared to standard depth electrodes) had a direct causal association with the complications.

Using hybrid depth electrodes allows investigators to simultaneously study single-neuron recordings and human behavior, providing novel information about the neurophysiological underpinnings of cognition [8, 19, 24, 25]. In addition, some studies indicate that single-unit recordings may have clinical utility [26-30]. Such examples include using single-neuron activity to predict seizure onset zones [26] and investigate the mechanisms underlying regional seizure spread [27], as well as characterizing the functional connectivity in epileptic networks [28] and the neural oscillatory signatures of the epileptogenic brain [29]. Steinmetz et al. [30] used single-neuron activity to measure the degree of functional connection between neurons within epileptic hypothalamic hamartomas and identified different electrophysiological phenotypes. A complete data set of single-unit activity and behavioral data is available for download [31]. These studies only begin to reveal the significant clinical utility that hybrid depth electrodes might offer.

Our results confirm the safety and utility of depth electrodes for intracranial epilepsy monitoring. They also support the view that hybrid and non-hybrid depth electrodes appear to be equally safe. Patients undergoing depth electrode monitoring have an increased likelihood of progressing to resective surgery and better outcome at 1 year when compared to other intracranial monitoring methods [7]. In line with this report, we found that 61.5% of our patients were able to have their seizures fully localized, with no difference in localization accuracy between hybrid and non-hybrid depth electrodes. These finding are likely reflective of the fact that patients with mesial temporal epilepsy are most likely to undergo depth electrode monitoring and are also most likely to have successful localization and surgical outcomes. They do not suggest that depth electrodes are superior at localizing seizures than subdural grids, which are more commonly used for neocortical epilepsy. Further experience with high-density SEEG will be needed to clarify that issue.

The localization accuracy and safety of hybrid electrodes appear to be equal to those of non-hybrid electrodes. Further, they provide a unique opportunity to investigate single-neuron activity during intracranial monitoring.

This work was supported by NIMH/NINDS grants U01 NS103792 and R01 MH110831-01 (U.R.), as well as a grant from the McKnight Neuroscience Foundation (A.N.M. and U.R.).

The study was approved by the Institutional Review Boards of all participating hospitals.

None of the authors have any disclosures or conflicts of interest to report.