Accuracy of intracranial electrode placement for stereoelectroencephalography: A systematic review and meta-analysis

Summary

Objective

Stereoelectroencephalography (SEEG) is a procedure in which electrodes are inserted into the brain to help define the epileptogenic zone. This is performed prior to definitive epilepsy surgery in patients with drug-resistant focal epilepsy when noninvasive data are inconclusive. The main risk of the procedure is hemorrhage, which occurs in 1–2% of patients. This may result from inaccurate electrode placement or a planned electrode damaging a blood vessel that was not detected on the preoperative vascular imaging. Proposed techniques include the use of a stereotactic frame, frameless image guidance systems, robotic guidance systems, and customized patient-specific fixtures.

Methods

Using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines, a structured search of the PubMed, Embase, and Cochrane databases identified studies that involve the following: (1) SEEG placement as part of the presurgical workup in patients with (2) drug-resistant focal epilepsy for which (3) accuracy data have been provided.

Results

Three hundred twenty-six publications were retrieved, of which 293 were screened following removal of duplicate and non–English-language studies. Following application of the inclusion and exclusion criteria, 15 studies were included in the qualitative and quantitative synthesis of the meta-analysis. Accuracies for SEEG electrode implantations have been combined using a random-effects analysis and stratified by technique.

Significance

The published literature regarding accuracy of SEEG implantation techniques is limited. There are no prospective controlled clinical trials comparing different SEEG implantation techniques. Significant systematic heterogeneity exists between the identified studies, preventing any meaningful comparison between techniques. The recent introduction of robotic trajectory guidance systems has been suggested to provide a more accurate method of implantation, but supporting evidence is limited to class 3 only. It is important that new techniques are compared to the previous “gold-standard” through well-designed and methodologically sound studies before they are introduced into widespread clinical practice.

Stereoelectroencephalography (SEEG) is a procedure that was developed by Talairach et al.¹ and is undertaken as part of the presurgical evaluation of patients in whom noninvasive investigations cannot accurately define the epileptogenic zone (EZ). The EZ can be defined as the “minimal area of the cortex that must be resected to produce seizure freedom.”² As part of the investigations prior to epilepsy surgery, patients undergo detailed noninvasive clinical, neurophysiologic, neuropsychological, neuropsychiatric, and multimodal imaging investigations.³ If these noninvasive investigations are concordant and the EZ can be accurately determined, such as in most cases of hippocampal sclerosis, then the patient can safely undergo surgery with good clinical outcomes.⁴ In cases where noninvasive investigations are nonconcordant, invasive intracranial recordings are required, which may take the form of subdural grid, SEEG electrode insertion, or both.⁵ A recent meta-analysis has highlighted that the main complications associated with SEEG include intracranial hemorrhage, infection, implant malfunction, and malposition.⁶ Before SEEG electrode insertion, trajectories are carefully planned with prior knowledge of the critical neurovascular structures.^7,8 Computer-aided planning has been employed in this regard to determine the safest trajectories that maximize gray matter sampling while ensuring a safe distance from vasculature.^9,10 Understanding the accuracy of the implantation method is necessary to incorporate a safe threshold away from blood vessels during trajectory planning. Cardinale et al., following a prospective analysis of 500 patients in which 6,496 electrodes were implanted, calculated a safe distance of 2.88 mm based on the mean entry point error (0.86 mm) with the addition of three standard deviations (SD; 3 × 0.54 mm) and the probe radius (0.4 mm).¹¹ This therefore provides a 99% estimate of confidence that a safe trajectory can be implanted should any vessels be greater than this distance away. Accuracy of SEEG implantations is therefore paramount for electrode implantation, as the corridors for implantation between cerebral vasculature are narrow, especially when multiple electrodes are implanted. Another potential consequence of inaccurate electrode placement is the inability to achieve electrophysiologic recordings from the intended anatomic brain region. Target points for SEEG electrodes are chosen based on the hypothesis generated from the summation of information provided by the noninvasive investigations. The SEEG recordings help to define the EZ and hence, the region for resection that will result in seizure freedom. Electrode malposition therefore exposes patients to the risks of SEEG unnecessarily, and of failure to achieve identification of the EZ. The published literature describes a number of different techniques including the use of a stereotactic frame, frameless image guidance, robotic trajectory guidance, and custom patient specific fixture systems. A review of the history of SEEG techniques and those used in high-volumes centers has been published recently.¹² We aimed to undertake a meta-analysis of all the published literature in patients with refractory focal epilepsy who have undergone SEEG implantation to determine which technique is most accurate when compared to the preoperative planned trajectories. We hope this will guide surgeons in determining a safety threshold when planning SEEG trajectories.

Methods

The meta-analysis was registered with the international prospective register of systematic reviews (PROSPERO) database and was assigned the registration number CRD42016047839 through which the review protocol can be reviewed.

Using the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines,¹³ a structured search of the PubMed, Embase, and Cochrane databases was undertaken. The last date of the search was September 16, 2016. Eligibility for inclusion in the metaanalysis include peer-reviewed publications in which fulll-ength English-language manuscripts were available through electronic indexing comprising:

1. Preclinical or clinical studies of patients with refractory focal epilepsy

2. Undergoing SEEG implantation as part of presurgical evaluation

3. The technique for insertion has been described

4. Postimplantation imaging has been performed (computed tomography [CT] or magnetic resonance imaging [MRI])

5. The method for measurement of deviation from the planned trajectory has been described

6. The accuracy of the implantation has been measured from the postoperative imaging

Two independent researchers applied the search criteria using the search terms:

1. ((drug resist*) OR refractory) AND epilepsy

2. (((stereoelectroencephalography) OR stereo EEG) OR SEEG) OR depth electrode)

In total 328 studies were identified. Following removal of duplicate and non–English-language, studies 296 manuscript titles and abstracts were screened. After applying the eligibility criteria, 35 articles that were analyzed. A comparison of the articles for inclusion between the two independent researchers was undertaken and revealed high concordance between the identified studies. Any discrepancy was resolved through mutual review and involvement of the senior author. The remaining 17 studies were included in the qualitative and 15 in the quantitative synthesis (see Fig. 1).

Figure 1.

Summary of search strategy. Adapted from Moher et al.²⁹ Epilepsia © ILAE

Data extraction was performed using a table with a predefined set of criteria. The risk of bias and methodological quality of the included studies was calculated using the methodological index for nonrandomized studies (MINORS) in which rating scores out of 16 and 24 for noncomparative and comparative studies, respectively, are generated.¹⁴ Low scores suggest methodologically flawed studies. There was good internal consistency between the ratings from the two independent assessors as defined by a Cronbach’s alpha of 0.86. Mean accuracy of implantation results for entry point or target point error were combined using an inverse variance method and stratified by technique. Studies were weighted from random-effects analysis. Statistical analysis was performed using SPSS 24 and Stata (Version 14).

Results

Study quality

From the 17 studies included in the qualitative synthesis, one study was preclinical, one study contained a combination of preclinical and clinical results, and the remaining studies were all clinical. In the majority of studies (11/17) no comparison between different techniques of implantation was undertaken. From the remaining six studies, five compared outcome results to retrospective data sets (historical cohorts) and the single preclinical study compared two robotic trajectory guidance systems prospectively. One of the studies by Gonzalez-Martinez et al.¹⁵ used previously published data as a historical comparison for a prospective study and therefore appears twice (once for the stand-alone results and again for the comparison). Two studies were removed from the quantitative analysis because the method used to assess accuracy was deemed sufficiently different to prevent any meaningful results comparison (see Table 1).

Table 1. Summary of data synthesis.

Publication	Minors	Study type	Frame/frameless	Technique	Number of patients	Total number of electrodes	Error measure/grading scale	EP error (mm) from plan	TP error (mm) from plan	Angle error (deg) from plan	Complications	Comments
Dorfer et al.23 (Level 3)	15	Preclinical phantom	Frameless	iSYS1 vs. Vertek arm (Medtronic) Prospective	1 phantom	5 iSYS1 vs. 5 Vertek arm	Lateral deviation	iSYS1 = 0.6 ± 0.4 Vertek arm = 1.4 ± 0.5 (mean ± SD)	iSYS1 = 0.8 ± 0.7 mm Vertek arm = 1.4 ± 0.7 mm (mean ± SD)	Not specified	N/A	Reduction in time for alignment to trajectory with iSYS1
		Clinical	Frameless	iSYS1 Prospective	16	93	Lateral deviation	iSYS1 = 1.54 ± 0.8 iSYS1 + K-wire = 1.18 ± 05 (mean ± SD)	iSYS1 = 1.82 ± 1.1 iSYS1 + K-wire = 1.66 ± 1.12 (mean ± SD)	Not specified	0	Further improvement in iSYS1 accuracy following modification of technique with K-wire (p = 0.021)
				Vertek arm (Medtronic) Retrospective	Not specified	Not specified	Lateral deviation	Vertek arm = 3.5 ± 1.5 (mean ± SD)	Vertek arm = 3.0 ± 1.9 (mean ± SD)	Not specified	Not specified	Comparison to retrospective Vertek arm cohort using skin fiducials where iSYS1 insertions used bone fiducials Reduction in EP (60%) and TP (40%) error with iSYS1 compared to Vertek probe technique in historical controls
Roessler et al. (2016)26 (Level 3)	10	Clinical	Frameless combined with intraoperative MRI	Brainlab navigation of reduction tube and immobilization with two fixation arms (Lyla retractors) Prospective	6	58	Euclidean distance	1.4 ± 1.2 (mean ± SD)	3.2 ± 2.2 (mean ± SD)	Not specified	Nil	Mean time for electrode implantation 12 min (50% quicker than frame-based SEEG) Intraoperative MRI registration accuracy 1–2 mm Frameless system 28% cheaper compared to frame-based system Relatively short electrode lengths (mean = 37.3 mm)
Narvaez-Martinez et al.17 (Level 4)	9	Clinical	Frameless combined with O-arm	Vertek arm (Medtronic) Prospective	10	69	Not specified	Not specified	1.39 (mean)	Not specified	Nil	Mean time for electrode implantation 34.7 min
González-Martínez et al.22 (Level 3)	12	Clinical	Frameless (Leksell frame used as fixation device)	ROSA Robotic assistant device Prospective	100	1,245	Euclidean distance	1.2 (interquartile range 0.78–1.83)	1.7 (interquartile range 1.20–2.30)	Not specified	Total complication rate 4% per patient (3% asymptomatic and 1% symptomatic hemorrhage) Risk of major hemorrhagic complication 0.08% per electrode	Mean time for electrode implantation 30 min using the ROSA Registration accuracy <0.75 mm EP and TP target errors measured in 500 consecutive electrode insertions using the ROSA
	9		Frame-based	Leksell frame Retrospective	100	1,310	Euclidean distance	1.1	Not specified	Not specified	Total complication rate 3% per patient (2% asymptomatic and 1% transient symptomatic hemorrhage) Risk of hemorrhagic complication 0.2% per electrode	ROSA implantations were a means of 222 min shorter than frame-based implantations (p < 0.001) No significant difference between mean entry point error and complication rate between ROSA and historical frame-based control
Verburg et al.25 (Level 4)	8	Clinical	Frameless	VarioGuide (BrainLab) Retrospective	7	89	Euclidean distance	Not specified	3.5 Median (95% CI 2.9–3.9)	Not specified	One patient required decompressive hemicraniectomy following symptomatic hemorrhage (14%) Risk of hemorrhagic complication 1.1% per electrode	Laser surface registration accuracy not specified Poor fixation of single electrode resulted in 13.7 mm maximum error of electrode
Hou et al.21 (Level 3)	16	Clinical	Frameless	Navigus tool Prospective	36	173	Lateral deviation	Not specified	2.03 ± 0.98 (mean ± SD)	Not specified	Nil	Mean time for electrode implantation of 19.4 min in frameless compared to 34.5 min in frame-based group (p < 0.05) Surface tracing registration accuracy not specified Electrode implantation accuracy worse in the temporal compared to frontal lobe (p < 0.05)
			Frame-based	Leksell frame Retrospective	28	62	Lateral deviation	Not specified	1.79 ± 0.82 (mean ± SD)	Not specified	Hematoma rate 3% in frame-based group (all asymptomatic)	Mean time for electrode implantation of compared to 34.5 min in frame-based group (p < 0.05) Surface tracing registration accuracy not specified Electrode implantation accuracy worse in the temporal compared to frontal lobe (p < 0.05)
Meng et al.28 (Level N/A)	18	Preclinical	Frameless	Robot arm with Polaris Prospective	1 phantom	12	Euclidean distance	Not specified	1.16 ± 0.38 (mean ± SD)	Not specified	N/A	Scalp fiducial marker based registration accuracy not specified
			Frameless	Robot arm with Optotrak Prospective	1 phantom	12	Euclidean distance	Not specified	0.62 ± 0.23 (mean ± SD)	Not specified	N/A	Accuracy errors are for trajectory alignment not following electrode or bolt placement. Drilling and placement errors are not measured Trajectory location time < 6 min for 12 electrodes Robotic device not clinically available
Nowell et al.10 (Level 4)	12	Clinical	Frameless	Vertek arm (Medtronic) Prospective	22	187	Lateral deviation	Not specified	3.66 ± 2.21 (mean ± SD)	Not specified	Hematoma rate of 4.5% (asymptomatic)	Scalp fiducial marker based registration accuracy not specified Intraparenchymal electrode deviation reported Median time for implantation 137 min (range 80–167 min) Eight electrodes failed to reach their neurophysiological target
Balanics et al.16 (Level 4)	10	Clinical	Frameless	Custom patient specific frame Prospective	4	53	Lateral deviation for Entry point and Euclidean distance for target point	0.68 mean (interquartile range 0.30–0.98)	1.64 mean (interquartile range 0.84–2.50)	Not specified	Nil	Low number of patients Only orthogonal and up to 2 oblique trajectories possible No adjustable components Requires 2 stage procedure (1 week apart)
Cardinale et al.11 (Level 3)	16	Clinical	Frame-based	Talairach frame Retrospective	37	517	Euclidean distance	1.43 median (interquartile range 0.91–2.21) Error >2 mm in 29.5% >3 mm in 11.4%	2.69 median (interquartile range 1.89–3.67)	Not specified	Overall major complication rate 2.4% per patient and 0.03% per electrode. Four major hemorrhagic complications with frame-based system and none with frameless	Significant improvement in both EP and TP accuracy with NeuroMate implantations over the traditional Talairach frame (p < 2.2 × 1016) Likely that postoperative hemorrhage rate was underestimated as post-op MRI or fan-beam CT were not undertaken Higher risk of bleeding from EP error Angiographic data used for trajectory planning
			Frameless	Neuromate robotic stereotactic device (Renishaw) Prospective	81	1,050	Euclidean distance	0.78 median (interquartile range 0.49–1.08) EP error >2 mm in 3.7% >3 mm in 0.5%	1.77 median (interquartile range 1.25–2.51)	Not specified
Munyon et al.30 (Level 4)	8	Clinical	Frame-based	With subdural grid placement Retrospective	6	31	Euclidean distance	Not specified	1.0 ± 0.15 mm (mean ± SD)	Not specified	Not specified	Center of planned craniotomy equidistant from fixation posts No significant difference between accuracy of electrode placement when craniotomy for subdural grid was performed Intraoperative brain shift not accounted for
Ortler et al.20 (Level 3)	11	Clinical	Frameless	Vogele-Bale-Hohner (maxillary fixation) system Retrospective	3	6	Lateral deviation	2.17 ± 2.19 mm (mean ± SD)	2.43 ± 0.98 mm (mean ± SD)	Not specified	Small subcortical haemorrhage	Only intrahippocampal depth electrodes placed in the longitudinal axis of the hippocampus included
			Frame-based	Fischer – Leibinger Stereotactic frame Retrospective	6	11	Lateral deviation	1.37 ± 0.55 mm (mean ± SD)	1.80 ± 0.39 mm (mean ± SD)	Not specified	Nil	Entry point error values taken at 4 cm behind tip of electrode for uniformity not at skull surface Aiming device calculated accuracy accepted when alignment to plan <0.5 mm and <1 deg
Mascott31 (Level 4)	8	Clinical	Frameless	SureTrak (Medtronic) with variety of systems incl. Leyla retractor arm (Aesculap) and Vertek arm (Medtronic)	7	42	Euclidean distance	Not specified	3 ± 1.5 mm (mean ± SD) 2.4 ± 1 (mean ± SD) following depth correction	Not specified	Not specified	No rigid (bolt) fixation of electrode to skull resulting in depth inaccuracies Bone fiducials used to provide consistent registration accuracy
Mehta et al.27 (Level 4)	12	Clinical	Frameless	Guide Frame-DT (Medtronic)	20	41	Lateral deviation and Anatomical Target Localization grading system	Not specified	3.1 ± 0.5 mm (mean ± SD) from planned trajectory 0.4 ± 0.9 mm from edge of anatomical structure	Not specified	Nil	Guidance based on MRI alone (not MRI and CT fusion) Skin fiducials used Electrode placement through craniotomy at same time as grid placement (no rigid skull fixation) Occipitotemporal approach statistically more inaccurate (p < 0.05)
Van Roost et al.19 (Level 4)	9	Clinical	Frame-based	Fischer – Leibinger Stereotactic frame	164	212	Anatomical Target Localization	Not specified	Target accuracy: Hippocampal head 97% (63% direct and 34% marginal) Hippocampal body 96% (58% direct and 38% marginal) Amygdala 75% (2% direct and 73% marginal)	Electrode inclination to hippocampus in AP orientation 93% correct, 6% too steep and 1% too flat	Subcortical hemorrhage 2.12% Infection 2.12% Risk of permanent neurological deficit 0.7%	Only intrahippocampal depth electrodes placed in the longitudinal axis of the hippocampus included Anatomical target localization accuracy provided not deviation from plan
Davies et al.18 (Level 4)	7	Clinical	Frameless	Freehand neuronavigation guided following craniotomy for grid placement	12	15	Distance from hippocampus	Not specified	0.8 mm (range 0–5 mm)	Not specified	Nil	Only intrahippocampal depth electrodes placed orthogonally into the temporal lobe included Distance to edge of anatomical structure measured on postoperative MRI not compared to preoperative plan SEEG electrode placement following craniotomy for grid placement No correction for brainshift

Calculated MINORS scores were a median 9 of 16 for noncomparative and 15.5 of 24 for the comparative studies, suggesting that studies had significant methodological flaws. Included studies provided level 3 evidence for individual case–control studies and level 4 evidence for case series. No randomized control trials in this area were identified. No studies included blinding or provided a prospective power calculation. Follow-up periods were deemed adequate for the purposes of accuracy determination as all measurements were derived from the post-operative imaging. From the comparative studies, control groups were rarely adequately balanced with regard to baseline characteristics.

Accuracy measurement

No consistent means of measuring accuracy within the published studies was identified. Error between the planned and implanted trajectories was measured using Euclidian distance in 8 of 17 studies and lateral deviation in 5 of 17. A single study¹⁶ combined both measures using lateral deviation for the entry point and Euclidian distance for the target point, and one study did not specify how the errors were measured.¹⁷

Accuracy data

From all the studies accuracy data has been provided for 13 different implantation systems (5 frameless, 3 frame-based, 3 robotic trajectory guidance, and one patient-specific custom-frame system; Fig. 2). Two studies were excluded from the quantitative analysis, as the method of accuracy was determined as distance from the edge of an anatomic structure opposed to distance from the planned trajectory.^18,19

Figure 2.

Forest plot showing (A) entry point (B) target point accuracy based on operative implantation technique. Mean (solid diamond) and 95% confidence interval (solid line) provided with percentage weighting based on inverse variance method. Group (subtotal) and overall mean with 95% confidence interval for mean (hollow diamond) provided with statistic (I²) and p-value for heterogeneity showing significant heterogeneity between robotic and frameless studies preventing meaningful comparison. Epilepsia © ILAE

The combined accuracy of the:

1. Frameless systems were entry point (EP) error mean 2.45 mm (0.39–4.51 95% confidence interval [CI]) and target point (TP) error mean 2.89 mm (2.34–3.44 95% CI).

2. Frame-based systems were EP error mean 1.43 mm (1.35–1.51 95% CI) and TP error mean 1.93 mm (1.05–2.81 95% CI).

3. Robotic trajectory guidance systems were EP error 1.17 mm (0.80–1.53 95% CI) and TP error 1.71 mm (1.66–1.75 95% CI).

Discussion

Accuracy measures

EP error is the difference in the actual from the planned position at which the electrode passes through the skull. This can be affected by misregistration of the neuronavigation system, inaccurate alignment, and deflection during drilling. TP error is the difference in the actual from the planned position of the electrode at the target site. TP accuracy is affected by the angle at which the electrode passes through the skull (even when the EP is accurate), deflection of the electrode at the dura or within the brain, rigidity of the electrode, and depth to which the introducer is inserted. The choice of insertion technique has a greater effect on the EP error but the stability of the system will also affect the angle of entry, which in turn has a direct impact on the TP accuracy. The EP and TP accuracies are based on the segmentation of the electrode positions on the postoperative CT scan and have been measured in a variety of ways, although Euclidean distance and lateral deviation were most commonly used. Comparison of accuracies between the two methods can lead to inaccuracy as the Euclidean distance takes into account depth inaccuracies, whereas lateral deviation does not. Given that Euclidean distance was used in 8 of 17 and lateral deviation in 5 of 17 studies, this introduces significant heterogeneity and prevents meaningful comparisons between studies using different accuracy measures. Given that none of the compared techniques for the implantation of SEEG electrodes directly affect depth error, as this is surgeon controlled, some authors advocate the use of lateral shift over Euclidean distance. We were unable to consider studies that used lateral deviation and Euclidean distance separately due to the small number in the literature, and have therefore opted to amalgamate them while recognizing the imprecision that this introduces. A uniform rating scale is required to facilitate accurate comparisons between different studies. There is a large variation in the number of patients and electrodes in the published studies ranging from six electrodes in three patients²⁰ to 1,050 electrodes in 81 patients.¹¹ To account for this, the studies in the meta-analysis were weighted using an inverse variance method. The overall incidence of hemorrhage from SEEG electrode implantation is estimated to be 0.18% per electrode.⁶ Given the relatively small numbers of studies and variable complication reporting in some studies, we are unable to correlate accuracy with hemorrhage rate.

Frame-based systems

Five studies provided accuracy data for the Leksell, Fischer-Leibinger, and Talairach frame-based systems. All studies were retrospective and data were provided as historical control groups for the comparison to frameless^20,21 and robotic trajectory guidance systems, ROSA^15,22 and Neuromate,¹¹ providing level 3 evidence. Hou et al.²¹ used a frameless system involving the Navigus tool in a prospective cohort of 36 patients in whom 173 electrodes were implanted compared to historical use of the Leksell frame in 28 patients for the insertion of 62 electrodes. Surface-tracing registration was used for the frameless system and did not reveal any significant difference in the overall electrode accuracy between the frameless and Leksell frame accuracies. The use of surface tracing is thought to be less accurate to bone fiducials and could have reduced the accuracy of the frameless implantation technique. There was a significant reduction in the time taken for electrode implantation from 34.5 to 19.4 min using the frameless system, compared to frame-based. This represents the only published study in which the baseline characteristics of the case and control groups have been matched. Ortler et al.²⁰ compared the Fischer-Leibinger frame in six patients with the frameless Vogele-Bale-Hohner maxillary fixation system in three patients for the purpose of bilateral longitudinal hippocampal electrode insertion. There was no difference in accuracy found between the two systems, with the Fischer-Leibinger and Vogele-Bale-Hohner systems providing EP errors of 2.17 ± 2.19 mm (mean ± SD) and 1.37 ± 0.55 mm (mean ± SD), respectively, and TP errors of 2.43 ± 0.98 mm (mean ± SD) and 1.80 ± 0.39 mm (mean ± SD), respectively. The overall number of patients in the study was very small, and there was a lack of a prospective power calculation. As such it is likely that the study was inadequately powered to detect a clinically significant difference.

Cardinale et al.¹¹ compared a historical cohort of 37 patients who had undergone 517 electrode insertions using the Talairach stereotactic frame, with 81 patients undergoing 1,050 electrodes using the NeuroMate robotic trajectory guidance system. There was a significant improvement in both the EP and TP accuracy with the NeuroMate robotic system over the historical cohort of patients implanted with the Talairach frame (p < 2.2 × 10¹⁶). EP error reduced from a median of 1.43 mm (interquartile range [IQR] 0.91–2.21) to 0.78 mm (IQR 0.49–1.08). In a similar study by Gonzalez-Martinez et al.,²² the implantation of 1,245 electrodes in 100 patients using the ROSA robotic trajectory guidance system was compared with a historical cohort of 100 patients implanted with 1,310 electrodes using the Leksell frame. EP error was not significantly different between the two methods. No TP error was provided for the Leksell frame historical cohort. Historical comparison data in this study was provided as a means of reference and not for formal statistical comparison. The calculated heterogeneity statistic for EP accuracy between frame-based systems was 0%. Excluding the small study by Ortler et al.,²⁰ the remaining studies had very tight CIs suggesting valid comparisons can be made between frame-based techniques.

Frameless systems

The frameless systems included in the analysis include the Vertek arm (Medtronic),^17,23,24 Varioguide (Brain-Lab),^25,26 Navigus tool (Medtronic),²¹ and the Guide Frame-DT (Medtronic).²⁷ A single study compared the use of the iSYS1 robotic trajectory guidance system for the insertion of 93 electrodes in 16 patients with a historical cohort using the Vertek arm frameless technique.²³ The number of patients and baseline characteristics of the historical cohort was not specified. There was a 40% reduction in the EP error from 3.5 ± 1.5 mm (mean ± SD) with the Vertek arm to 1.54 ± 0.8 mm (mean ± SD) with the iSYS1 robotic trajectory guidance system. TP error was reduced by 20% from 1.82 ± 1.1 mm (mean ± SD) to 3.0 ± 1.9 mm (mean ± SD). Historical comparison data in this study were provided as a means of reference and not for formal statistical comparison. All other studies using frameless systems were case-series in which accuracy data were measured and therefore provides level 4 evidence. The calculated heterogeneity statistic for frameless techniques included in the meta-analysis was 98.9%, suggesting that significant heterogeneity exists between individual studies that prevents any meaningful comparisons between the different frameless techniques. Combined accuracy data are provided for different frameless techniques, but the significant heterogeneity between the studies prevents any meaningful conclusions from being drawn.

Robotic guidance systems

The robotic trajectory guidance systems include the ROSA,²² NeuroMate,¹¹ and iSYS1.²³

As stated previously, comparisons between the robotic trajectory guidance systems has been with retrospective frame-based and frameless systems. A single preclinical prospective comparison between a robotic arm using different guidance systems (Polaris and Optotrak) has been published.²⁸ Twelve electrodes were inserted into a single phantom using each technique. This device, however, is not clinically available and therefore there are no clinical publications of its use to date. There have been no prospective clinical comparisons of robotic trajectory guidance systems with other techniques or between robotic trajectory guidance systems. The calculated heterogeneity statistic for robotic techniques included in the meta-analysis was 99.4%, suggesting that significant heterogeneity exists between individual studies that again prevents any meaningful comparisons between the different robotic techniques. Combined accuracy data are provided for different robotic techniques, but the significant heterogeneity between the studies prevents any meaningful conclusions from being drawn.

Conclusion

The accuracy of SEEG electrode implantation using a variety of techniques has been published. Studies to date are mostly single-center case series providing level 4 evidence. Some studies have provided comparisons between different implantation techniques, but all clinical comparisons have been of retrospective cohorts (level 3), with variable study quality. Calculated heterogeneity statistics suggest that meaningful comparisons between studies can only occur between different frame-based techniques and not between frameless or robotic techniques. The lack of a uniform measure of accuracy likely contributes to this heterogeneity and reduces the validity of the pooled data such that no meaningful conclusions can be drawn. There is some limited evidence suggesting that robotic trajectory guidance systems may provide greater levels of accuracy compared to both frameless and frame-based systems, but the studies are of low quality and provide low levels of evidence. There is therefore a need for high quality prospective control trials between different SEEG implantation techniques to define which methods provide the highest levels of accuracy.

Key Points.

• Currently used surgical techniques for SEEG include frame-based, frameless, and robotic applications

• A PRISMA systematic review and meta-analysis of the literature revealed 15 studies eligible for quantitative analysis

• Studies supporting accuracy of implantation techniques are limited to class 3 evidence with significant heterogeneity preventing meaningful comparison

• There is a need for well-designed prospective control studies comparing different SEEG implantation techniques to guide future clinical practice

Biography

Vejay N. Vakharia is a senior clinical research associate in clinical and experimental epilepsy.