Abstract
Stereoelectroencephalography (SEEG) methodology, originally developed by Talairach and Bancaud, is progressively gaining popularity for the presurgical invasive evaluation of drug-resistant epilepsies.
To describe recent SEEG methodological implementations carried out in our center, to evaluate safety, and to analyze in vivo application accuracy in a consecutive series of 500 procedures with a total of 6496 implanted electrodes.
Four hundred nineteen procedures were performed with the traditional 2-step surgical workflow, which was modified for the subsequent 81 procedures. The new workflow entailed acquisition of brain 3-dimensional angiography and magnetic resonance imaging in frameless and markerless conditions, advanced multimodal planning, and robot-assisted implantation. Quantitative analysis for in vivo entry point and target point localization error was performed on a sub-data set of 118 procedures (1567 electrodes).
The methodology allowed successful implantation in all cases. Major complication rate was 12 of 500 (2.4%), including 1 death for indirect morbidity. Median entry point localization error was 1.43 mm (interquartile range, 0.91-2.21 mm) with the traditional workflow and 0.78 mm (interquartile range, 0.49-1.08 mm) with the new one (P < 2.2 × 10−16). Median target point localization errors were 2.69 mm (interquartile range, 1.89-3.67 mm) and 1.77 mm (interquartile range, 1.25-2.51 mm; P < 2.2 × 10−16), respectively.
SEEG is a safe and accurate procedure for the invasive assessment of the epileptogenic zone. Traditional Talairach methodology, implemented by multimodal planning and robot-assisted surgery, allows direct electrical recording from superficial and deep-seated brain structures, providing essential information in the most complex cases of drug-resistant epilepsy.
Surgery is an effective therapeutic option for the treatment of refractory epilepsy1 based on the concept of epileptogenic zone (EZ), which can be defined as “the area of cortex that is necessary and sufficient for initiating seizures and whose removal (or disconnection) is necessary for complete abolition of seizures.”2 In most cases, noninvasive presurgical investigations are sufficient for the definition of the EZ.3-5 As previously reported, in 25% to 50% of subjects, identification of the EZ entails the use of intracranial electroencephalography recording.6-10 Although electrodes arranged in subdural grids and strips offer the possibility to record only from brain surfaces, intracerebral electrodes enable direct recording from every cerebral structure, including depth of sulci and white matter, with a lower rate of complications.8,11-34 Intracerebral electrodes may be used with the main goal of lateralizing seizures (depth electrodes) or with the more ambitious aim of defining the EZ by stereoelectroencephalography (SEEG),11,12,35 a methodology developed by Talairach and Bancaud at Hô;pital Sainte Anne, Paris, France.36-39 The scientific interest in SEEG has grown throughout the years.40 The introduction of high-resolution imaging41,42 and robotic stereotactic systems8,19,42-44 greatly contributed to the technical evolution of this methodology.
Several phantom studies on stereotactic devices and systems are available,45-55 but many factors affect application accuracy in real surgery.56-58 There are a number of articles on the measurement of in vivo accuracy in deep brain stimulation procedures,57,59-67 biopsies,56,68 implantation of catheters,69 navigated craniotomies,70-73 and bitemporal recordings with depth electrodes,74-76 but to the best of our knowledge, no peer-reviewed studies of SEEG accuracy have been published.
The 3 goals of the present study are to describe recent SEEG methodological implementations carried out in our center, to evaluate safety, and to analyze in vivo application accuracy. The pertinent literature on the topic has been reviewed.
Patients and Methods
Five hundred consecutive SEEG implantations were performed between 1996 and 2011 for the presurgical evaluation of 482 patients with drug-resistant epilepsy (age, 2-56 years). Informed consent was obtained from all patients or their guardians. The total number of implanted electrodes was 6496. Analysis of the complication rate was performed on all 500 cases. For application accuracy analysis, complete data were available for the last 118 procedures (performed in 115 patients, with 1567 electrodes implanted). Thirty-seven of 118 procedures were performed according to the traditional workflow; the remaining 81 were performed according the new workflow.
Three hundred fifty-four subjects underwent SEEG-guided tailored brain resections (344 of whom had a minimum follow-up of 12 months), 15 underwent SEEG-guided radiofrequency thermocoagulation, 20 are waiting for surgery, and 89 did not proceed to further treatment for epileptological/functional reasons or because of patient refusal. Three patients died while waiting for surgery, and 1 patient died 2 days after the implantation (see below).
Surgical Methodology
Traditional Talairach methodology, as described in detail previously,8,9 includes 2 surgical steps, stereotactic angiography and electrode implantation. This workflow was in use in our center until Fall 2009. Since then, using the updated workflow, we moved toward a 1-step surgical technique. Main planning and implantation steps are described below.
Imaging Acquisition
All patients are imaged by brain 3-dimensional (3-D) magnetic resonance imaging (MRI) and angiography in frameless and markerless conditions outside the operating room. The essential 3-D T1-weighted MRI scan is acquired on the sagittal plane and reformatted into axial slices with 560 × 560 matrix, 0.46 × 0.46 × 0.9-mm voxel, and no interslice gap. Other structural MRI, functional MRI, diffusion tensor imaging, computed tomography (CT) or CT-positron emission tomography scans are obtained when needed. We use the O-arm 1000 System (Medtronic; Minneapolis, Minnesota), a mobile cone-beam CT scanner that ensures 0.4 × 0.4 × 0.8-mm reconstructed voxels, to obtain 3-D digital subtraction angiography (DSA). For this purpose, we acquire a baseline data set and additional data sets during the selective injection of iodinate contrast medium into the arteries of interest. These 3-D data sets are obtained by selecting the preset high-definition protocol. The acquisition lasts for 24 seconds; the CT dose index is 23.56 mGy; and the dose-length product is 376.82 mGy·cm. The same protocol is subsequently repeated during the selective intra-arterial injection of iopamidole (300 mg/mL). The infusion rate is 2 mL/s. The infusion is manually started at the same time as the x-ray acquisition and lasts for 15 seconds, with a total volume of 30 mL (1.5 mL/s for 15 seconds in children). Both arteries and veins are enhanced in the same volume. In the case of anterior unilateral SEEG investigation, 1 vascular data set with selective injection of the corresponding internal carotid artery is enough. For investigation of the posterior regions, 1 more acquisition with injection in 1 vertebral artery is needed. In the case of bilateral anterior investigation, both internal carotid arteries are injected, and the maximum number of vascular acquisitions is 3 in the case of posterior bilateral electrodes implantation. Thus, the total number of 3-D acquisitions ranges from 2 to 4. Sedation is needed only for pediatric subjects.
Postprocessing
Imaging data sets are processed with Osirix,77,78 MRIcron and dcm2nii,79 FSL,80-82 Freesurfer,83,84 NIfTI2DICOM,85 and 3D Slicer.86-89 dcm2nii, mri_convert (Freesurfer) and NIfTI2DICOM are the utilities used to convert file formats. FLIRT (FSL) is used to register all data sets to 3-D DSA, which constitutes the reference space. Gray-scale thresholding vascular segmentation is obtained by registration and algebraic subtraction between baseline and iodine-enhanced data sets (Figure 1). Freesurfer is used to segment brain tissue and to separate the hemispheres. Because of the number of analysis steps, the first author developed a bash script to automate postprocessing workflow. At the end of the pipeline, the processed data are loaded into the stereotactic planning software.
O-arm 3-dimensional (3-D) rotational angiography (cross-eyed viewing). In A and B, 2 axial slices are extracted from 2 coregistered datasets without and with contrast medium injection into the right internal carotid artery. In C, the result of the algebraic subtraction (B − A) of the 2 data sets is depicted. D, a stereoscopic pair of images of 3-D volume rendering of the subtracted data set. The stereoscopic effect can be obtained looking at it with the cross-eyed viewing technique.
O-arm 3-dimensional (3-D) rotational angiography (cross-eyed viewing). In A and B, 2 axial slices are extracted from 2 coregistered datasets without and with contrast medium injection into the right internal carotid artery. In C, the result of the algebraic subtraction (B − A) of the 2 data sets is depicted. D, a stereoscopic pair of images of 3-D volume rendering of the subtracted data set. The stereoscopic effect can be obtained looking at it with the cross-eyed viewing technique.
Trajectory Planning
Stereotactic trajectories are planned in Voxim (IVS, Chemnitz, Germany), the software package supplied with the robotic system. Entry points (EPs) and target points (TPs) are defined for every trajectory by investigating multiplanar reconstructions, brain and vascular surface rendering, and images reformatted according to the planned vector (ie, parallel or orthogonal; Figure 2).
Stereotactic planning. Some screen grabs from Voxim, the stereotactic planning software supplied with the robotic system. A, in this screen capture, the projection of the trajectory (dotted yellow line) is visible on multiplanar reconstructions. The visualization fading in the sagittal view is 50% for angiography and 50% for magnetic resonance. The vector of the trajectory also is visible on the 3-dimensional view. This workspace allows planning trajectories, taking care of fundamental information. B, in this workspace, the surgeon checks for vessel conflicts on dedicated views in which slices are reformatted all along the trajectory or orthogonal to it. The visualization fading is set to 100% for vascular images.
Stereotactic planning. Some screen grabs from Voxim, the stereotactic planning software supplied with the robotic system. A, in this screen capture, the projection of the trajectory (dotted yellow line) is visible on multiplanar reconstructions. The visualization fading in the sagittal view is 50% for angiography and 50% for magnetic resonance. The vector of the trajectory also is visible on the 3-dimensional view. This workspace allows planning trajectories, taking care of fundamental information. B, in this workspace, the surgeon checks for vessel conflicts on dedicated views in which slices are reformatted all along the trajectory or orthogonal to it. The visualization fading is set to 100% for vascular images.
Electrode Implantation
Electrodes (Microdeep Intracerebral Electrodes-D08; Dixi Medical, Besançon, France; or Depth Electrodes Range 2069; Alcis, Besançon, France) are implanted under general anesthesia. Once registration between imaging and the Talairach frame space is performed (Figure 3), the robot (Neuromate, Renishaw-mayfield SA, Nyon, Switzerland) aligns the tool holder along the vectors of each planned trajectory to help fix the guiding screws to the skull. Subsequently, the frame is removed and the electrodes are placed under radioscopic control. Finally, a postimplantation 3-D CT is obtained with the O-arm for checking the correct positioning of the electrodes.
Registration step. A, the Neuromate frame adaptor was redesigned in 2009 to interface with the O-arm so that registration of the patient's head is possible. Special localizers for projective x-rays (laterolateral and anteroposterior) are mounted onto the Talairach frame, enabling the definition of the spatial position of x-ray source in the planning software. B, subsequently, a 2-dimensional–3-dimensional registration is possible, with a point-to-point manual rigid registration based on bone fiducial selection, so that all planned trajectories can be transformed into the frame space. The 3-dimensional data set with the skull markers is acquired by selecting the standard definition protocol (computed tomography dose index, 19.78 mGy; dose-length product, 316.43 mGy·cm).
Registration step. A, the Neuromate frame adaptor was redesigned in 2009 to interface with the O-arm so that registration of the patient's head is possible. Special localizers for projective x-rays (laterolateral and anteroposterior) are mounted onto the Talairach frame, enabling the definition of the spatial position of x-ray source in the planning software. B, subsequently, a 2-dimensional–3-dimensional registration is possible, with a point-to-point manual rigid registration based on bone fiducial selection, so that all planned trajectories can be transformed into the frame space. The 3-dimensional data set with the skull markers is acquired by selecting the standard definition protocol (computed tomography dose index, 19.78 mGy; dose-length product, 316.43 mGy·cm).
Video 1 (Supplemental Digital Content 1, http://links.lww.com/NEU/A513) shows the main steps of the workflow.
Statistical Analysis
Proportions of complications with the 2 workflows were compared with a 2-way contingency table and 2-tailed Fisher exact test.
For surgical accuracy analysis, we measured the EP localization error (EPLE) and the TP localization error (TPLE). The former is defined as the unsigned euclidean distance (Figure 4) between planned EPs and the major axis of the implanted electrode (Figure 5). The latter is defined as the unsigned euclidean distance between planned TPs and the tip of the implanted electrode. In the sub-data set of 1050 electrodes implanted with the new workflow, several explanatory variables (listed in the Results section) were considered to investigate potential sources of errors and their possible association with EPLE. For bivariate analysis, the Pearson correlation test was used to analyze numerical variables, and the Kruskal-Wallis rank sum test was performed to analyze categorical (binomial or multinomial) variables. All bivariate tests should be understood as constituting exploratory data analysis; no adjustments for multiple testing were made. Multivariate analysis was performed by fitting a mixed-effects linear regression model with EPLE considered an outcome variable. Selection of the fixed-effect variables and the model simplification were obtained by considering the results of “best subsets” technique90 and manually adding them, one by one, in explorative temporary models. A random “surgical procedure identification” effect was included in the model to account for the number of electrodes implanted per patient, leading to correlated observations.
Euclidean distance. The euclidean distance between 2 points is defined as the square root of the sum of the squares of the differences between the corresponding coordinates of the points. In 3-dimensional euclidean geometry, the euclidean distance between 2 points a (ax, ay, az) and b (bx, by, bz) can be calculated with this formula. The schematic drawing clearly illustrates that the euclidean vector is by definition larger than any of the linear distances.
Euclidean distance. The euclidean distance between 2 points is defined as the square root of the sum of the squares of the differences between the corresponding coordinates of the points. In 3-dimensional euclidean geometry, the euclidean distance between 2 points a (ax, ay, az) and b (bx, by, bz) can be calculated with this formula. The schematic drawing clearly illustrates that the euclidean vector is by definition larger than any of the linear distances.
Entry point localization error measurement. The postimplantation intraoperative O-arm data set, after the FSL coregistration to the vascular reference space, is imported into Voxim to compare the planned and the real positions of implanted electrodes. Euclidean localization error is measured with an ad hoc trajectory created between the planned point (green arrow and dotted line) and the major axis of the implanted electrode (red arrow and continuous line). Only the projection on the coronal plane of the euclidean vector is depicted.
Entry point localization error measurement. The postimplantation intraoperative O-arm data set, after the FSL coregistration to the vascular reference space, is imported into Voxim to compare the planned and the real positions of implanted electrodes. Euclidean localization error is measured with an ad hoc trajectory created between the planned point (green arrow and dotted line) and the major axis of the implanted electrode (red arrow and continuous line). Only the projection on the coronal plane of the euclidean vector is depicted.
Values of P < .05 were considered as evidence of findings not attributable to chance. Statistical analysis was performed with R 2.14.91
The present paper was edited according to the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines.92,93
Measures
For patients implanted according to the traditional workflow, the intracranial position of implanted electrodes was assessed with postsurgical high-resolution, 3-D, T1-weighted MRI (26 procedures) or with postsurgical 3-D CT with 1-mm-thick slices (11 procedures). In all new-workflow procedures, the implantation was intraoperatively verified with cone-beam CT. For all 118 procedures, postimplantation images were automatically registered (FLIRT tool, mutual information, 6 df) to the reference data set. As a result of the absence of skull markers in preoperative images, the registration accuracy was verified only by careful visual inspection.94 After the registration, the postoperative actual position of the electrodes was compared with the preoperatively planned vectors in the imaging reference space. The values of EPLE, TPLE, and explanatory variables were obtained by 9402 manual measurements in the imaging or surgical space. In particular, localization errors were measured in the planning software as the euclidean distance between the planned EP and the major axis of the electrode (EPLE) or between the planned TP and the tip of the electrode (TPLE).
RESULTS
The New Workflow
The procedure was successfully completed in 100% of cases. Moreover, several qualitative results can be highlighted. Tridimensional anatomic reconstructions allowed fast trajectory planning, with clearer topographic information. Freesurfer computed excellent cortical surface reconstructions in a vast majority of cases except for those with gross structural abnormalities such as previous large resections. Visual inspection allowed verification of the high degree of accuracy of all automatic registrations performed by FLIRT (FSL). Our homemade method for postprocessing angiographic 3-D images led to satisfactory multiplanar reconstructions (useful mainly in deep-seated regions such as the insular vascular lamina) and to surface renderings of the vascular tree (useful mainly to look for avascular EPs). In a few cases affected by movement artifacts, the isosurface 3-D rendering was noisy. Nevertheless, trajectories could be successfully planned with only conventional multiplanar reconstructions. This new angiographic technique was progressively introduced, and in a number of procedures, traditional 2-dimensional DSA was also obtained. In these cases, we could qualitatively compare 3-D DSA with traditional stereoscopic angiography without evidence of a clinically relevant lack of information. Despite the number of postprocessing steps, command-line interface of the used packages allowed the development of a bash script for automatizing the workflow, with a dramatic reduction in procedural error risks. Freesurfer takes about 14 to 16 hours on Apple Mac Pro (2 × 2.26-GHz Quad-Core Intel Xeon). The rest of the processing, run in parallel with Freesurfer, lasts about 1 hour.
Systematic use of the robot allowed targeting of every intracranial structure, with unlimited options and unrestricted obliquities. The mobile CT scanner allowed intraoperative imaging and immediate check of the positioning of the electrodes soon after implantation (high-definition protocol).
The general anesthesia duration for the surgical implantation was not significantly different. Its median value was 315 minutes with the traditional workflow and 330 minutes with the new one (P = .78).
Accurate preimplantation and postimplantation 3-D imaging, integrated in multimodal interactive scenes created with 3D Slicer, made it easier to understand the electrode positions. This tool led to a computer-aided representation of the brain anatomy and individualized electrode placement for each patient. This patient-specific 3-D representation facilitated the definition of the EZ, the multidisciplinary discussions, and the planning of resective surgery.
Complications
Complications are detailed in Table 1, and the relative literature review is summarized in Table 2. Of 419 traditional procedures, 5 intracranial bleedings requiring surgical treatment (2 with permanent hemiplegia) occurred. In all 4 cases of subdural or intraparenchymal hemorrhage, the bleeding source was spatially correlated with the cortical EP of an electrode. Moreover, a 3-year-old child who developed postimplantation severe hyponatremia and massive cerebral edema died 2 days after implantation.
Global complication rates with the traditional and new workflows were 19 of 419 (4.5%) and 5 of 81 (6.2%; P = 1), respectively. Major complication rates were 11 of 419 (2.6%) and 1 of 81 (1.2%) with the traditional and the new workflows, respectively (P = .7).
In Vivo Application Accuracy
EPLE and TPLE values are reported in Table 3. EPLE was > 2 mm in 39 of 1050 electrodes (3.7%) implanted with the new workflow and in 153 of 517 electrodes (29.5%) implanted with the traditional workflow (P < 2.2 × 10−16). It was > 3 mm in 6 of 1050 electrodes (0.5%) implanted with the new workflow and in 59 of 517 electrodes (11.4%) implanted with the traditional workflow (P < 2.2 × 10−16).
Along the intracranial track, 120 of 1567 electrodes (7.7%) bent, with only 61 (3.9%) with a TPLE > 3 mm (Figure 6).
Verification of implanted trajectories. A, coronal and axial reconstruction showing optimal matching between the planned trajectory and the implanted electrode. B, a case of minor bending of the electrode along its intracerebral track, with a target point localization error < 2 mm. C, a case of major deformation, with a target point localization error > 3 mm.
Verification of implanted trajectories. A, coronal and axial reconstruction showing optimal matching between the planned trajectory and the implanted electrode. B, a case of minor bending of the electrode along its intracerebral track, with a target point localization error < 2 mm. C, a case of major deformation, with a target point localization error > 3 mm.
Bivariate analysis between explanatory variables and localization errors is reported in Tables 4 and 5. The output of the mixed-effects linear models fitted on the sub--data set of 1050 electrodes implanted with the new workflow is reported in Table 6. Finally, Table 7 summarizes the results of the present study and reviews the literature for articles reporting in vivo application euclidean error for stereotactically implanted devices (electrodes, catheters, biopsy needles).
DISCUSSION
The 3 goals of the present study are to describe recent SEEG methodological implementations carried out in our center, to evaluate safety, and to analyze in vivo application accuracy.
The sample size of the examined series, the largest among similar studies published, is the most obvious strength of this study. The retrospective nature of the study is the most important limitation. It must also be stressed that the literature review was narrative and not systematic.
The New Workflow
Compared with the previously reported traditional SEEG workflow,8,9 the new workflow enables performance of the procedure with only 1 surgical step. With the traditional workflow, the stereotactic brain angiography was acquired in the operating room, under general anesthesia, after the fixation of the Talairach frame. On the contrary, 3-D DSA is now obtained outside the operating room, without any frames or markers, with the patient awake (except for pediatric subjects).
High-quality open-source, freely available software packages made the postprocessing step highly robust. We were able to develop our own script for automatizing all operations, reducing both human time and the risk of errors. The automatic pipeline is time consuming, but it can be unlinked from surgery by running it a few days before. Although quantitative data were not available for this retrospective analysis, we can state that stereotactic planning for SEEG is now less time consuming. Tridimensional brain and vascular surface reconstructions accelerated patient-specific anatomic study, especially in cases with abnormal gyral pattern. Moreover, 3-D reconstructions can be freely rotated, thus facilitating the elaboration of avascular trajectories that address the epileptological strategy, without the constraints of the Talairach traditional surgical workflow.
To the best of our knowledge, the use of the O-arm system for 3-D DSA has not been reported in peer-reviewed articles. This technique allowed displaying cortical vascular anatomy in multiplanar reconstructions and 3-D rendering for optimal computer-aided planning. However, we had to integrate the features of this equipment by using additional software packages. Moreover, this mobile cone-beam CT scanner was used intraoperatively for patient registration and for postimplantation evaluation, as previously described for deep brain stimulation procedures,95,96 thus replacing time-consuming, MR-based postoperative electrode imaging.
Three-dimensional imaging and the Neuromate, a robotic stereotactic device developed in Grenoble with the essential contributions of Professor A.L. Benabid,43 advantageously complemented each other. Throughout the years, we moved completely to the use of this system, abandoning the constraining Talairach double grid.
Radiation exposure is a sensitive topic. We acquire 6 CT data sets in the most complicated cases, 4 data sets for 3-D DSA (baseline, left and right internal carotid artery, and vertebral artery injection), and 2 data sets for intraoperative use (1 at the beginning of the procedure for registration purposes, the last 1 for verifying the implantation). Radiation exposure of each run amounts to about one-third of a standard brain CT (3.5-mm-thick slices). Thus, overall exposure corresponds to about 2 standard CT scans, which is quite reasonable.
Software tools such as FSL, Freesurfer, and 3D Slicer were useful not only for improving multimodal planning but also, and even more so, for refining the interpretation of video-SEEG monitoring data and outlining surgical resections.
Complications
Previous studies reported a low incidence of complications for SEEG procedures.11,12,19,32 Indeed, only 2 procedures were complicated by permanent neurological deficits in our series. Moreover, there was 1 death in a 3-year-old child. This event was likely due to severe hydroelectrolytic imbalance, probably not directly related to the surgical procedure itself. Neither emergency investigations nor the autoptic examination could ascertain the cause of this fatal occurrence.
For the purpose of this study, we classified a complication as major when causing a permanent deficit, when a surgical treatment was required (hematoma evacuation, anticipated removal of implanted devices, etc), or for potential severity of the adverse event (meningitis, encephalitis, osteomyelitis, etc). According to these criteria, we observed 12 major complications (2.4%, 0.03% of implanted electrodes) and 12 minor complications (2.4%). Most likely, the incidence of minor intracranial bleedings has been underestimated in this series because blood-sensitive postoperative imaging (MR or conventional fan-beam CT) was not obtained in all cases. In fact, in the 419 implantations performed with the traditional workflow, electrodes were imaged by MR in 407 cases and by CT in the remaining 12 (electrode metal artifacts could hide small intracranial bleedings). We performed intraoperative postimplantation imaging with the blood-insensitive O-arm scans in all 81 new-workflow surgeries. Therefore, minor bleedings were detected only by subsequent standard CT or MR studies when occasionally executed.
The area of higher risk of bleeding is the electrode cortical EP, as suggested by the 4 cases of major hemorrhage that occurred in the traditional-workflow period. No major bleeding was observed with the new workflow, probably owing to the lower EPLE.
Thanks to the results of this study, we can estimate with 99% confidence a safe entry region according to the following formula: mean EPLE + 3 SD + probe radius = radius of safe entry region (0.86 + 0.54 × 3 + 0.4 mm = 2.88 mm).
The global complication rate was lower compared with that of subdural grids and strips (Table 2). Fountas34 provides an exhaustive systematic literature review of the complications of subdural grids implantation. Our low rate of complications is probably due to both accurate planning and surgery and the valuable creation of Dixi and Alcis electrodes with a 0.8-mm diameter, smooth surface, and blunt tip.
In Vivo Application Accuracy
Unlike stereotactic phantom studies, we measured in vivo application accuracy, intended as “a measure of the accuracy of the devices when used in the real-word setting.”45 The number of similar studies is quite small, and we were able to find only 2 meeting posters that addressed SEEG accuracy.97,98 With the new workflow, we obtained a significant reduction of EPLEs and TPLEs. In Table 7, our values are compared with those of other reports of similar measurements (in vivo euclidean errors for implantation of stereotactic devices). Localization errors of the present series were the best despite the use of stylet-free semirigid electrodes, implanted without the aid of a guiding tube. Mean EPLE, a value not affected by any eventual intracranial deviation, is even better than that obtained with the same robotic system in phantom conditions.50 It must be highlighted that Li and coworkers50 scanned the phantom with 2-mm-slice-thickness CT; the lower resolution of their experimental conditions explains this result.
Explanatory variables significantly associated with EPLE at multivariate analysis were the reference data set, owing to the lower distortion99 and the smaller voxel size of O-arm data sets compared with MRI; skin-skull distance and incidence angle of the trajectory (the thicker the extracranial tissues and the wider the angle, the more difficult the control of the possible drill bending); and the temporal pole as entry zone because of the tridimensional curvature of the bony surface.
Explanatory variables significantly associated with TPLE at multivariate analysis were the EPLE (and, implicitly, all variables associated with it), the occurrence of intracranial deviation of the electrode, and the length of the intracranial path.
Generalizability
The external validity of safety and accuracy data is likely high, provided that SEEG is performed under similar conditions. The main elements that should guarantee the generalizability of the results are the multimodal and high-resolution imaging, including an informative angiographic data set; careful planning; the proper use of accurate and precise stereotactic devices; and the use of well-manufactured, small-diameter, blunt-tip, smooth-surface electrodes.
Future Perspectives
The number of implanted electrodes with an EPLE > 3 mm is small. Nonetheless, although very low, the risk of this major electrode displacement at the EP represents a threat for the safety of the procedure. From our surgical experience, we believe that the most important variable determining large errors should be the bending of the drill. This error could probably be decreased with the development of a robot able to actively drill the skull under the control of an adequate feedback system.
On the basis of the analysis of our data, we are currently working on the development of an SEEG automatic planner.
Last, but not least, a sensitive topic for future discussion could be the clinical use of “only research software,” not approved for clinical/surgical use. Packages such as FSL, Freesurfer, and 3D Slicer fully demonstrated their value, even in terms of geometric accuracy for stereotactic planning. As a matter of fact, we cannot ignore that these software programs have been developed by the most acknowledged research groups from all over the world. It is obvious that the surgeon must be aware of the features and limits of the available tools (a statement that also applies to certified software).
CONCLUSION
SEEG is a safe and accurate methodology that is gaining popularity for invasive electroencephalography recordings aimed to identify the epileptogenic zone. The traditional Talairach methodology, recently updated by the use of the most advanced multimodal planning tools and robot-assisted surgery, allows one to directly record electric activity from every brain structure, providing valuable information in the most complex cases of refractory epilepsy.
Acknowledgments
We would like to thank Alberto Torresin, MSc, Paola Colombo, MSc, Marco Ciboldi, MSc, Gianni Origgi, MSc, Alessandro Peroni, Gianfranco De Gregori, MSc, and all their coworkers for their invaluable contribute to our daily work.
REFERENCES
COMMENTS
In this article, the authors described their large experience with the stereoelectroencephalography (SEEG) methodology, reporting accuracy results with a newly developed method of SEEG implantation and comparing it with the more traditional method of implantation (using the Talairach frame and telestereoangiography). In addition to the technical description, morbidity rates were compared between the 2 methods.
The SEEG methodology, originally developed in France by Talairach and Bancaud, is a precise, safe, and highly efficient method for the localization of the hypothetical epileptogenic zone in patients with difficult-to-localize seizures in which a deeply located focus is strongly suspected. Its application in epilepsy centers outside Europe, especially in the United States, has been recently rediscovered, maybe because of new, innovative techniques of implantation and novel imaging modalities, making the procedure less complex and more precise, as nicely described here. Nevertheless, specific indications for SEEG vs subdural grids/strips methodologies are undefined. Hence, the main future challenge will be to validate the SEEG method by long-term seizure outcome results and to compare those results with the those obtained from the subdural method in an attempt to better refine indication criteria for the different methods of invasive monitoring that are currently available. The authors should be congratulated for this important contribution.
Jorge Gonzalez-Martinez
Cleveland, Ohio
Pioneered in the late 1950s by Talairach and Bancaud, stereoelectroencephalography (SEEG) remains a crucial tool for exploration of some drug-resistant partial epilepsies, allowing an optimal delineation of the epileptogenic focus. The original methodology used a stereotactic ventriculography to draw the anterior commissure--posterior commissure line superimposed to a stereotactic angiography. The anatomical targets were reached orthogonally by use of the Talairach atlas. Nowadays, this methodological ground remains unchanged, but the indirect targeting provided by the use of the atlas has been replaced by a direct, magnetic resonance imaging--based targeting. However, the need for a perfect stereotactic registration remains, as well as that for a minimal risk of vascular injury. These goals may be reached, from now on, thanks to several modern technological tools such as robots, 3-dimensional imaging, flat-panel x-ray detectors, and appropriate software, with every system having its own advantages and drawbacks.
The methodology described in this study, coming from a very experienced group, while being safe and accurate, appears to be particularly innovative. Therefore, it represents a significant contribution to the literature related to the field of SEEG, which is undoubtedly destined to continue expanding worldwide.
Marc Guenot
Lyon, France
ABBREVIATIONS
- DSA
digital subtraction angiography
- EP
entry point
- EPLE
entry point localization error
