Placement of Stereotactic Electroencephalography Depth Electrodes Using the Stealth Autoguide Robotic System: Technical Methods and Initial Results

Abstract

BACKGROUND:

Robotic systems are gaining acceptance as a preferred tool for the placement of electrodes for stereotactic electroencephalography (SEEG) studies.

OBJECTIVE:

To describe the technical methods for insertion of SEEG using the Medtronic Stealth Autoguide robotic system and detailed outcomes in the initial 9 patients implanted.

METHODS:

Nine patients underwent placement of electrodes for SEEG studies with the use of the Autoguide system. Patients had at least 10 electrodes placed. Targets were planned on a Stealth S8 planning station, and electrodes were placed under general anesthesia. A technique for placement is described in detail. Patient outcomes and accuracy of electrode placement were evaluated. Methods to improve accuracy were investigated. Comparison of postoperative MRIs with preoperative planning MRIs was performed to determine the accuracy of electrode placement.

RESULTS:

One hundred two electrodes were placed in 9 patients. Methods for placement and technical nuances are detailed. The distance from the planned target to the actual position of the electrode tip was measured in 8 of the 9 patients. The mean Euclidean distance was 4.67 ± 0.27 mm. There was 1 placement-related hemorrhage deficit in the first patient, and no deaths or infections. Adequate positioning of electrodes for seizure monitoring was obtained in all patients.

CONCLUSION:

Autoguide can be used for placement of electrodes for SEEG studies with acceptable degrees of patient safety, accuracy, and efficiency. Considering the cost of Autoguide compared with other robotic devices, it may be attractive option.

ABBREVIATIONS:

ACC

anterior cingulate cortex

CRW

Cosman-Roberts-Wells

DRF

digital reference frame

OF

orbitofrontal cortex

ROSA

robotic stereotactic assistant

SEEG

stereotactic electroencephalography

SMA

supplementary motor area.

Stereotactically implanted penetrating intracerebral electrodes are often used to localize the epileptogenic zone in patients with drug resistant epilepsy (DRE).¹ Stereotactic electroencephalography (SEEG) is a method for epileptogenic zone definition by means of intracerebral electrodes.^2,3 This technique was first developed in the 1960s^2,4,5 and since then has been shown to be safe^6,7 and effective for identifying seizure foci.

Traditionally, placement of SEEG electrodes required the use of stereotactic frames.⁸ Although accurate, this method is laborious and can limit trajectories. Modifications include the use of mounted sliding guide tubes for orthogonal trajectories.⁹ Robotic systems that integrate a preplanned stereotactic trajectory with a robotic manipulator arm to align the drill site and target site have been developed. The robotic stereotactic assistant (ROSA®) system (MedTech) is the most widely used robotic system for SEEG, with several reports in the literature.^10–12

In 2020, the Stealth Autoguide (Medtronic) was released. This system contains a compact robotic manipulator arm that integrates directly into the Medtronic StealthStation frameless navigation system. As one of the early adapters of this system, we have used it for the placement of multiple bilateral SEEG electrodes over the past 8 months. We now describe the technical methods for SEEG electrode insertion and results in this initial cohort.

METHODS

Patient Selection and Consent

According to the Cedars-Sinai Medical Center Institutional Review Board (IRB) guidelines, individual patient consent was not required for this study. IRB approval (IRB no. 49208) for a retrospective chart review was obtained.

Patients with DRE were referred for phase II monitoring with SEEG. Criteria for deciding on SEEG electrodes were based on multidisciplinary evaluation of all presurgical data, with targets defined based on a hypothesis of seizure onset zone.

Preoperative Assessment and Planning

For presurgical planning of electrode trajectories, we acquired multiple sequences of MRI of the brain (see Supplementary Material, http://links.lww.com/ONS/A90). In addition, computed tomography angiography (CTA) was obtained. These images sets are loaded into an S8 StealthStation (Medtronic) planning software suite and coregistered. A plan is created for each electrode target and trajectory (Figure 1). At our center, we routinely target bilateral orbitofrontal cortex, anterior cingulate cortex, supplementary motor area, amygdala, and hippocampus using Ad-Tech (Oak Creek) 1.2-mm diameter Behnke-Fried hybrid electrodes. Whenever possible, we aim for orthogonal trajectories. For other targets, such as insula, visual cortices, or focal dysplasia, nonorthogonal trajectories are commonly used, with Ad-Tech 0.8-mm diameter SEEG electrodes. After selecting the target and entry points, the MRI and CTA sequences are followed on the trajectory views and probe's eye views from the surface to the target, with plans adjusted to avoid vessels and to minimize traversing of sulci. The thickness of the skull at the trajectory entry site is also measured and noted.

FIGURE 1.

Intraoperative planning on the Medtronic StealthStation system with all 10 standard electrode plans pictured and 4 additional electrode plans based off of preoperative magnetoencephalography. ACC, anterior cingulate cortex; Amy, amygdala; Hipp, hippocampus; OF, orbitofrontal cortex; SMA, supplementary motor area.

Positioning

The patient is positioned supine on an operating room table and intubated. The table is rotated 180° away from the anesthesia machine (Figure 2). The entire head is shaved, and a Cosman-Roberts-Wells (CRW) stereotactic base frame is placed. The stereotactic frame provides maximal freedom of movement for the robotic arm compared with a 3-point head clamp, especially for bilateral electrode insertions. The patient is positioned in a “lounge chair” position with the back elevated approximately 30°. The CRW frame is secured to the Mayfield Ultra 360 adapter attached to the bed (Figure 3A). This specific Mayfield adapter is integral in the positioning of the patient so that an O-arm can be used without undraping the patient. The articulating arm of the Mayfield base is rotated so that the main axis of the arm is lateral, facilitating later use of the O-arm for radiographic confirmation of electrode placement (see Shahlaie et al¹³ for further details). The 3-way accessory attachment for the Autoguide is secured to the starburst base of the Mayfield attachment with the digital reference frame (DRF) of the Stealth system attached to one of the starburst side arms and positioned lateral and anterior to the patient's forehead on the left (Figure 3B). The presurgical images are then registered to the patient head using standard “tracer” algorithms. The registration accuracy of the Stealth navigation system should ideally be <2 mm. The Autoguide articulating arm is attached to the center starburst. The robotic manipulator base is then attached to the arm and connected to the control unit. The surgeon moves the Autoguide robotic base manually to confirm that the positioning will allow for targeting of all planned trajectories without repositioning of the arm.

FIGURE 2.

Diagram of the operating room setup with the Autoguide. The patient is rotated 180° from the anesthesiologist. The surgeon and assistant are at the patient's head with the scrub tech to the surgeon's right. By the patient's feet is the navigation system and tech, in easy view of the surgeon.

FIGURE 3.

A, Articulating arm of the Mayfield base is rotated so that the main axis of the arm is lateral, facilitating later use of the O-arm for radiographic confirmation of the electrode. B, Autoguide arm placement behind the patient for easy draping. The patient consented to publication of his/her image.

Draping, Autoguide Assembly, and Registration

The entire head is prepped with betadine solution. A U-drape is placed around the base of the CRW frame with the distal end extended over the patient's body. The Autoguide and articulating arm are then draped with a special clear bag drape. The remainder of the system is assembled per standard instructions. The Autoguide navigation tracker is registered on the StealthStation DRF by inserting its tip into one of the divots on the DRF similar to a handheld probe in a standard navigation system. Once registered, it is attached to the Autoguide base. The robot is now fully activated and connected to the navigation platform.

Electrode Placement

Electrodes were placed in 7 main steps:

1. Selecting the appropriate electrode plan on the navigation system.

2. Manually positioning the Autoguide robot to the proximity of the planned entry point and locking it in place (Supplementary Material Figure SDC1A and SDC1B, http://links.lww.com/ONS/A90).

3. Robotic fine-tuned movement of the Autoguide to the exact entry point (Supplementary Material Figure SDC1C, http://links.lww.com/ONS/A90).

4. Securing the drill guide.

5. Drilling the hole for the anchor bolt.

6. Inserting the anchor bolt.

7. Measuring electrode length and passing electrode to the target (Supplementary Material Figure SDC2, http://links.lww.com/ONS/A90).

The entire process from start to finish can typically be performed in 15 to 20 minutes per electrode site (see Supplementary Material, http://links.lww.com/ONS/A90 for details on operative time). When docking the guide tube, predicted error of the trajectory is measured on the StealthStation. We require an error less than 1 mm and will readjust or repeat the docking process if error exceeds this. A detailed description of the electrode placement process, including technical nuances noted over the course of our experience, is provided in the Supplementary Material, http://links.lww.com/ONS/A90.

Once completed, an O-arm spin (Figure 4) is performed to confirm placements. A full head wrap is applied. We obtained skull x-rays and postoperative MRI scan to confirm electrode placements and assess for surgical complications.

FIGURE 4.

Intraoperative O-arm images merged to preoperative MRI with plans overlaid. A, Coronal view with bilateral supplementary motor area plans, B, coronal view with bilateral anterior cingulate cortex plans, C, coronal view with bilateral amygdala plans, D, axial view with bilateral supplementary motor area plans, E, axial view with bilateral anterior cingulate cortex plans, and F, axial view with bilateral amygdala plans. Note that O-arm overlays are shown here for illustrative purposes, but actual errors between planned and final trajectories were measured by comparing plans from preoperative MRI with final electrode placements visualized on postoperative MRIs.

Measurement of Lead Tip Location Relative to Planned Target Location

To determine the accuracy of the Autoguide-based placement, we measured the difference between the planned position of the electrode tip and the actual final position as seen on postoperative imaging. For each electrode, the plan coordinates were considered to be 0, 0, 0 and the electrode tip distance from target was recorded in the lateral (Δx), anterior-posterior (Δy), and vertical (Δz) directions. In the x-axis, medial was positive and lateral was negative. In the y-axis, anterior was positive and posterior was negative. In the z-axis, superior was positive and inferior was negative. The postoperative MRIs were merged with the scans used for planning trajectories. The 3-dimensional distance between the tip location and the preoperative plan, the Euclidian distance = $\sqrt{\left[{\left(\Delta \text{x}\right)}^2+{\left(\Delta \text{y}\right)}^2+{\left(\Delta \text{z}\right)}^2\right]}$, was calculated. Mean and standard errors were calculated.

RESULTS

Patient Population

Nine patients (6 women and 3 men; mean age was 36 ± 10.8 years) underwent bilateral surgery according to the following breakdown: 4 patients with 10 electrodes, 2 patients with 14 electrodes, 2 patients with 11 electrodes, and 1 patient with 12 electrodes. Ninety electrodes were implanted in the bilateral orbitofrontal, anterior cingulate, and supplementary motor area, amygdala, and hippocampus (1 electrode per side per target per patient); 4 electrodes were implanted in the bilateral insular cortices of 1 patient; and 8 electrodes were placed based on magnetoencephalography dipole clusters. Three electrodes for 1 patient were subsequently removed on the discovery of a postoperative intraparenchymal hemorrhage, and 1 patient's preoperative plans were not saved and therefore not available for analysis, resulting in a total of 77 electrodes included in analysis.

Accuracy of Electrode Placement

The average differences in the x, y, and z coordinates from preoperative planning as measured on postoperative MRI are given in Table 1. The average differences in the x, y, and z directions for all 77 electrodes were −0.2 ± 0.27, −0.4 ± 0.37, and 1.4 ± 0.36 (mean ± standard error) mm, respectively. There was no preferential direction in which electrodes were systematically displaced. The mean Euclidean difference was 4.67 ± 0.27 mm. All electrodes were deemed accurately placed enough to allow for proper sampling of the brain region targeted and did not need to be repositioned.

TABLE 1.

Differences Between Intraoperative Planning and Postoperative MRI Measurements of DBS Electrode Tip Locations

Electrode number	Electrode location	Patient	ΔX	ΔY	ΔZ	Mean |Δ|	Euclidean distance
1	L OF	A	−3.0	−2.1	−2.6	2.57	4.49
2	L ACC	A	−2.1	−3.9	−3.8	3.27	5.84
3	L SMA	A	−2.2	−3.2	−2.5	2.63	4.62
4	R Hipp	A	0.0	0.0	−1.7	0.57	1.70
5	L Hipp	A	−0.6	−5.8	−0.8	2.40	5.89
6	R Amy	A	−4.3	−1.2	0.0	1.83	4.46
7	L Amy	A	1.4	−2.2	−1.6	1.73	3.06
8	R OF	B	0.9	−1.2	1.5	1.20	2.12
9	L OF	B	0.0	1.9	0.7	0.87	2.02
10	R ACC	B	−3.2	−2.0	1.9	2.37	4.22
11	L ACC	B	−1.1	−4.5	0.0	1.87	4.63
12	R SMA	B	−0.3	−1.6	0.0	0.63	1.63
13	L SMA	B	1.3	−1.6	2.0	1.63	2.87
14	R Hipp	B	0.6	−1.7	−2.2	1.50	2.84
15	L Hipp	B	0.4	0.9	−0.4	0.57	1.06
16	R Amy	B	2.1	−1.3	1.1	1.50	2.70
17	L Amy	B	−1.1	−3.4	−1.0	1.83	3.71
18	R OF	C	0.3	4.3	6.7	3.77	7.97
19	L OF	C	−1.0	2.3	0.7	1.33	2.60
20	R ACC	C	1.0	1.6	0.0	0.87	1.89
21	L ACC	C	0.6	0.8	−0.6	0.67	1.17
22	R SMA	C	1.1	−1.7	0.0	0.93	2.02
23	L SMA	C	−0.9	0.0	0.7	0.53	1.14
24	R Hipp	C	−0.5	3.9	0.0	1.47	3.93
25	L Hipp	C	0.7	1.3	−0.4	0.80	1.53
26	R Amy	C	0.6	1.4	0.0	0.67	1.52
27	L Amy	C	3.6	0.0	0.0	1.20	3.60
28	R OF	D	−8.6	−2.2	3.8	4.87	9.66
29	L OF	D	2.0	−0.5	1.8	1.43	2.74
30	R ACC	D	−0.3	−5.2	3.8	3.10	6.45
31	L ACC	D	1.3	3.0	−0.9	1.73	3.39
32	R SMA	D	−7.3	−3.7	6.8	5.93	10.64
33	L SMA	D	−2.9	−0.2	2.9	2.00	4.11
34	R Hipp	D	2.8	−6.1	4.3	4.40	7.97
35	L Hipp	D	5.7	−0.7	3.0	3.13	6.48
36	R Amy	D	−6.5	−6.7	7.3	6.83	11.85
37	L Amy	D	4.6	−1.3	1.7	2.53	5.07
38	R OF	E	−1.2	−1.0	5.4	2.53	5.62
39	L OF	E	0.0	−1.7	4.7	2.13	5.00
40	R ACC	E	0.0	−2.5	5.8	2.77	6.32
41	L ACC	E	−1.1	−2.3	6.9	3.43	7.36
42	R SMA	E	1.5	3.7	−5.6	3.60	6.88
43	L SMA	E	0.3	7.4	−4.6	4.10	8.72
44	R Hipp	E	0.0	−1.5	3.4	1.63	3.72
45	L Hipp	E	0.2	3.1	0.0	1.10	3.11
46	R Amy	E	−1.2	−1.4	3.7	2.10	4.13
47	L Amy	E	−0.7	1.8	−0.3	0.93	1.95
48	R OF	F	−1.2	−1.7	2.6	1.83	3.33
49	L OF	F	−1.7	−0.2	2.1	1.34	2.71
50	R ACC	F	0.0	−2.7	3.7	2.13	4.58
51	L ACC	F	−0.8	−1.5	3.5	1.93	3.89
52	R SMA	F	−0.8	0.5	7.2	2.83	7.26
53	L SMA	F	−2.9	1.4	5.0	3.10	5.95
54	R Hipp	F	−1.0	0.0	4.4	1.80	4.51
55	L Hipp	F	−0.7	−1.4	2.1	1.40	2.62
56	R Amy	F	−0.7	−0.3	3.3	1.43	3.39
57	L Amy	F	0.3	−2.9	3.4	2.20	4.48
58	R OF	G	1.7	2.9	−2.4	2.33	4.13
59	L OF	G	2.3	3.1	−3.8	3.07	5.42
60	R ACC	G	0.0	5.0	−3.8	2.93	6.28
61	L ACC	G	0.0	7.0	−2.7	3.23	7.50
62	R SMA	G	2.2	5.6	−2.0	3.27	6.34
63	L SMA	G	0.8	7.9	0.5	3.07	7.96
64	R Hipp	G	0	1.8	−0.4	0.73	1.84
65	L Hipp	G	1.2	1.5	1.2	1.30	2.26
66	R Amy	G	−2.8	3	−2.7	2.83	4.91
67	L Amy	G	0	2.4	−3	1.80	3.84
68	R OF	H	1.9	−6.1	7.8	5.27	10.08
69	L OF	H	0	3.8	3.7	2.50	5.30
70	R ACC	H	−4.6	−4.7	0.3	3.20	6.58
71	L ACC	H	0.6	−1	4.4	2.00	4.55
72	R SMA	H	0	−5.9	4.9	3.60	7.67
73	L SMA	H	3.3	−1.2	2.5	2.33	4.31
74	R Hipp	H	−2.6	−3.8	6.5	4.30	7.97
75	L Hipp	H	2.7	0	3.6	2.10	4.50
76	R Amy	H	0	−5	4.7	3.23	6.86
77	L Amy	H	1.1	0	2.2	1.10	2.46
Mean			−0.2	−0.4	1.4	2.28	4.67
Standard error			0.27	0.37	0.36	0.14	0.27

ACC, anterior cingulate cortex; Amy, amygdala; DBS, deep brain stimulation; Hipp, hippocampus; OF, orbitofrontal cortex; SMA, supplementary motor area; Hipp, hippocampus.

Delta value (Δ) of x, y, and z. Mean |Δ| calculated as mean of absolute value of delta x, absolute value of delta y, and absolute value of delta z for each electrode.

Complications

There was 1 postoperative hemorrhage in the first patient implanted. The patient emerged from anesthesia with left-sided hemiplegia and was taken back for hematoma evacuation. Three of his 10 electrodes were removed, but he was able to complete video electroencephalography monitoring. At 6 months after surgery, he had regained full movement of his left leg and partial movement of his arm. No other complications were noted.

DISCUSSION

Accuracy of Autoguide in SEEG Placement

The mean Euclidean distance for the 77 electrodes measured was 4.67 ± 0.27 mm. The mean error in each axis was less than 1.4 mm. This degree of accuracy is approximately 2.5 to 3 mm greater than what has been reported with other robotic systems,^12,14 but nonetheless still reasonable for SEEG electrode targeting. Error can be due to a combination of factors including (1) inherent inferior accuracy of frameless stereotaxy compared with frame-based methods, (2) adjustment in trajectory introduced by manual tapping of the Autoguide drill guide into the bone, (3) the flexibility of the SEEG electrodes, and (4) human error when measuring the depth of the electrode or tightening the anchor bolt into place. In our view, the frameless registration method and the docking of the Autoguide drill guide into the bone are likely important sources of error. With tapping of the drill guide into the bone, we noted variation from the target up to 2.5 mm. Care is taken during this step to try to limit the error to less than 1 mm, but nonetheless, this is likely the primary source of error. This flexibility is largely absent in frame-based methods or more rigid robotic systems, such as ROSA® or Neuromate (Renishaw), where bone anchoring is not required, and registration can be performed more similar to frame-based systems. It is possible that the requirement of the tapping tube with Autoguide is partially responsible for lesser accuracy when compared with the other robotic systems. Furthermore, the flexibility of SEEG electrodes as compared with a biopsy needle or deep brain stimulation electrode likely contributes to some variation, and we have observed this finding with other systems used over the years. However, trajectories were never so inaccurate as to prevent useful interpretation of acquired electroencephalography data or require repositioning of electrodes. Overall, our results indicate that the Autoguide system is useful and reliable for placement of SEEG electrodes using both orthogonal and nonorthogonal trajectories.

Safety of SEEG Electrode Placement While Using Autoguide

We noted 1 postoperative intraparenchymal hemorrhage of 102 electrode placements (0.98%). On further investigation, we noted that the drill was set to the depth of the bone measured by the CTA. This depth overestimated the bone thickness, and the drill likely injured the underlying dura and brain. As noted in Supplemental Material Table S1, http://links.lww.com/ONS/A90, the measurement of bone thickness on CTA was not reliable when compared with measurement on computed tomography (CT) or MRI. Because measuring bone thickness only on CT, we did not have any other hemorrhagic complications. The risk of significant intracranial bleeding from SEEG electrode placement is well-documented with Carlson et al¹⁵ noting a hemorrhagic complication rate of 0.4% per electrode in 555 electrodes. A large meta-analysis indicated a 1% risk of hemorrhage.¹⁶ Ollivier et al¹⁷ reported a 1.5% rate of symptomatic hemorrhage. Thus, based on our limited experience to date, there does not seem to be a difference in the hemorrhage rate with the Autoguide system.

Modifications Suggested Based on Initial Experience

Based on this initial experience with Autoguide, we have developed several modifications that increase the accuracy of the system, reduce risk of complications, and provide for more smooth placements. These modifications are described in the Supplementary Material, http://links.lww.com/ONS/A90 methods, and interested readers are referred there for more details.

Autoguide Compared With Other Robotic Systems

Other institutions have used the ROSA® robotic system for placement of SEEG electrodes and found that most electrodes were within the range of 1.89 to 3 mm on average from the target.^14,18,19 Similar results have been reported with other systems including other robotic systems and frame-based systems.^9,20 This suggests that these other robotic systems or frame-based methods may be more accurate; however, there are many confounding factors, including surgeon technique, types of electrodes used, and accuracy of measurement methods. In this series, we used the StealthStation, frameless navigation with intraoperative patient registration with either fiducials or facial tracing. Furthermore, the need to tap the guide sleeve into the skull also altered the trajectory slightly. We suspect that these methods may make Autoguide slightly more inaccurate compared with other robotic systems and frame-based stereotaxy. Methods to reduce these errors are described in the Supplementary Material, http://links.lww.com/ONS/A90, based on experience with these first 9 cases. However, in our experience, this degree of accuracy is acceptable for SEEG studies and biopsy. We have also used the device for placement of responsive neural stimulation electrodes (NeuroPace) with good results.

One comparison of frame-based vs frameless methods in SEEG electrode insertion found that when retrospectively looking at electrodes placed in the hippocampus and amygdala, 8 of the 45 electrodes placed with the frameless technique compared with 0 of the 36 electrodes placed with the frame-based technique missed their target and were not clinically useful.²¹ Although we found that all of our electrodes were clinically useful, it is important to note the difference in accuracy with frame-based vs frameless technology used with robotic placement.

There is a large cost differential between the ROSA® averaging more than $700 000 per unit as compared with the Autoguide, which costs between $125 000 and $200 000. Furthermore, the ROSA® system has add-on costs and an expensive service agreement for software updates. Thus, the highly attractive price point for the Autoguide may make it far easier to be adapted by hospitals.

CONCLUSION

In our experience, the use of the Autoguide provides a reliable and cost-efficient robotic tool for placement of SEEG electrodes. Increased experience with this device and reports from larger series will be important to determine the ultimate utility of the Autoguide system for SEEG electrode placement.

Funding

This study was funded by National Institutes of Health U01 NS117839 (to Dr Mamelak).

Disclosures

The authors have no personal, financial, or institutional interest in any of the drugs, materials, or devices described in this article.

Supplemental Digital Content

Supplementary Material. Detailed description of the technique used to implant intracerebral electrodes using the Autoguide system is provided in the supplemental digital content (SDC). Technical nuances of the system and insights gained from the initial cases are provided, with the aim of facilitating use for other new adapters. Furthermore, greater details about some of the methods for data collection and analysis used, as well as interpretation of the data are found in the SDC.

COMMENT

The authors should be commended for publishing a case series cataloging surgical data related to depth electrode placement using the Medtronic Autoguide robot. The detailed workflow and process improvement based on some initial difficulties may allow programs using similar resources to “make new mistakes.” The reported accuracies are more divergent than most comparable series using alternative robotic guidance platforms, but the authors are committed to iterative refinement of the technique to make the technology safer. As an early career neurosurgeon and developer of new programs myself, I applaud the honesty and detail in this report. I think the conclusions from retrospective “technical note” case series are more operational when published along with relevant standard work.

Jeffrey Steven Raskin

Indianapolis, Indiana, USA