Robotic Surgery in Pediatric Epilepsy : Current Evidence, Clinical Impact, and Future Directions

Robotic systems

The most commonly used robotic platforms in epilepsy surgery include ROSA^® (Zimmer Biomet) [32], Neuromate^® (Renishaw) [12], and Stealth Autoguide^TM (Medtronic) [22]. All systems provide frameless trajectory guidance using preoperative imaging datasets combined with rigid robotic arms capable of submillimetric positional accuracy. Although hardware architectures differ, the fundamental workflow remains similar : preoperative trajectory planning, patient-to-image registration, robotic alignment, percutaneous drilling and anchor bolt placement, electrode or catheter insertion, and postoperative imaging verification. ROSA^® is currently the most widely reported platform in neurosurgery literature. It integrates optical tracking with automated trajectory alignment and supports rapid multi-trajectory execution, making it well suited for high-density SEEG implantation. Neuromate^® has a longer historical footprint in functional neurosurgery and remains used in selected centers. Stealth Autoguide^TM represents a compact robotic guidance system integrated into navigation systems and has been increasingly evaluated in pediatric cohorts.

Preoperative planning

Trajectory planning is typically performed using fused MRI and vascular imaging (computed tomography angiography, magnetic resonance angiography, or digital subtraction angiography). Advanced planning software enables automated vessel segmentation and collision detection, reducing the risk of vascular injury. In pediatric patients, vessel caliber may be small and cortical veins relatively superficial, emphasizing the importance of meticulous trajectory optimization [12,84]. Electrode trajectories are designed to maximize gray matter sampling while minimizing cortical transgression length and vascular proximity. For deep targets such as insula, cingulate cortex, mesial temporal structures, and hypothalamic lesions, robotic alignment allows steep oblique trajectories that are difficult to achieve reproducibly using manual frameless techniques [1,12,27,39,40].

Registration methods

Registration accuracy directly impacts stereotactic precision. Common methods include bone fiducials, stereotactic frame fiducials, surface laser scanning, and skin fiducials registration. Pediatric patients present unique challenges due to smaller cranial surface area, thinner bone, and limited tolerance for rigid fixation. Surface registration may be less reliable in very young children, whereas the placement of bone fiducials requires skin and skull puncture with associated anesthesia time. Although no study has established a definitive lower age limit or minimum skull thickness threshold for SEEG, including its registration process, a minimal acceptable skull thickness of 2 mm has been pragmatically used, generally corresponding to a minimum age of 18 months [83]. In general, registration accuracy improves from adhesive skin fiducials, to laser facial surface matching, and to implanted skull fiducials or the use of stereotactic frame fiducials [1,40] (Fig. 1).

Fig. 1.

Four registration methods are available for robot-assisted stereoelectroencephalography. Skin fiducials (A), surface matching (B), bone fiducials (C), and stereotactic frame registration (D) are illustrated. Registration accuracy increases from (A) to (D), with the latter two methods providing the higher accuracy. This figure was generated with assistance from ChatGPT.

Intraoperative workflow

Once registration is completed, the robot automatically aligns the guidance sleeve to each planned trajectory. The surgeon performs drilling, anchor bolt placement, and electrode insertion manually through the robotic guide. Depending on the specific surgical procedure, larger drill diameters may be required, and anchor bolt placement is not necessary. Robotic repositioning between trajectories typically requires less than 30 seconds, significantly reducing cumulative setup time compared with frame-based systems. Usually frame-based techniques require manual adjustment of five coordinates—x, y, z, ring angle, arc angle for each trajectory—which is more time-consuming and introduces additional opportunities for submillimetric error. In pediatric cases with large numbers of electrodes, this efficiency advantage becomes clinically meaningful by reducing anesthesia duration and surgeon fatigue. Post-implantation verification is commonly performed using intraoperative computed tomography (CT) or postoperative CT/MRI fusion to confirm electrode position and rule out hemorrhage.

Indications

Robot-assisted SEEG is particularly advantageous in pediatric patients with complex epileptogenic networks where extensive three-dimensional sampling is required. Typical indications include MRI-negative epilepsy, suspected deep or insular onset, multilobar or bilateral involvement, prior failed resection, and discordant noninvasive data. SEEG allows hypothesis-driven sampling of distributed networks while preserving cortical integrity. Compared with subdural grids, SEEG provides superior access to deep structures and bilateral hemispheric coverage without the morbidity associated with craniotomy. Robotic assistance further enhances the feasibility of high-density implantation by reducing procedural burden and maintaining consistent accuracy across numerous trajectories [1,12,18,33,40,60] (Fig. 2).

Fig. 2.

This figure demonstrates extensive bilateral stereoelectroencephalography electrode implantation using both orthogonal and oblique trajectories for comprehensive three-dimensional sampling, including deep brain regions such as the mesial temporal structures and insula.

Accuracy and efficiency

Accuracy is commonly quantified using entry-point error (EPE), target-point error (TPE), or radial error (RE) measured on postoperative imaging. Systematic reviews and meta-analyses across adult and mixed populations consistently report robotic systems achieving mean EPE approximately 1.0–1.5 mm and TPE approximately 1.5–2.0 mm. These values are comparable or superior to frame-based stereotaxy and significantly better than freehand navigation techniques [12,31,32,66,85]. Pediatric-specific series mirror these findings. Reported mean RE typically range between 1.5 and 2.0 mm, with minimal outliers exceeding clinically meaningful thresholds. Importantly, accuracy appears preserved across age groups when appropriate registration methods are used, suggesting that pediatric anatomical constraints do not fundamentally compromise robotic performance [1,40]. Comparative pediatric studies between robotic platforms have demonstrated similar accuracy profiles, although workflow efficiency, setup complexity, and footprint may differ. Learning curve analyses indicate that operative time per electrode decreases substantially after approximately 20 cases, supporting scalability for high-volume programs [4,45,64,67,76].

One of the most tangible advantages of robotics lies in workflow efficiency. Multi-trajectory implantation using framebased systems requires repeated manual repositioning and verification, which can significantly prolong operative time. Robotic repositioning is automated and rapid, enabling consistent throughput even in cases requiring 12–20 electrodes [32,46,52]. Reported pediatric series demonstrate average implantation times ranging from 8–15 minutes per electrode after the initial learning phase, with total operative times often reduced by 30–40% compared with historical frame-based cohorts. Reduced anesthesia exposure is particularly relevant in younger children and those with comorbidities [40,56].

Clinical impact

The clinical utility of robotic-assisted SEEG extends beyond technical precision, serving as a critical diagnostic bridge to definitive surgical intervention. Recent pediatric cohorts demonstrate that SEEG achieves clinically decision-guiding seizure focus localization in approximately 80–90% of cases, facilitating tailored resections, ablations, or neuromodulations. Post-SEEG surgical outcomes show seizure freedom rates (Engel class I) ranging from approximately 50% to 75%, which are statistically comparable to historical subdural grid monitoring but achieved with significantly lower major morbidity (<2%) and enhanced three-dimensional sampling capabilities [27,32,40,43,55,69,70,81,82] (Table 1). Furthermore, integrated robotic workflows enable a seamless intra-procedural transition from diagnostic mapping to stereotactic interventions, such as MRgLITT or SEEG-guided RF-TC, within a unified planning and navigational environment [36,55].

Table 1.

Comparison of SDG and SEEG

Complications

Hemorrhage remains the most clinically significant complication of SEEG; however, recent pediatric robotic series report favorable safety profiles, with symptomatic hemorrhage rates occurring in only 0–1% of cases. While total radiographic hemorrhage rates may reach up to 3%, these events are predominantly small, asymptomatic, and typically identified only through routine postoperative imaging. Infection rates are similarly low, generally remaining below 1–2%, which underscores the minimally invasive nature of the robotic approach. Although hardware-related issues—such as electrode bending, bolt loosening, or inaccurate seating—can occur, particularly in pediatric patients with exceptionally thin skulls, these risks are effectively mitigated through meticulous bolt selection and precise drilling techniques. Importantly, current clinical evidence indicates no significant increase in complication rates when utilizing robotic guidance compared to traditional frame-based stereotactic methods [8,20,32,40,55].

Limitations of current evidence

Despite encouraging results, several limitations persist. Most pediatric robotic SEEG studies are retrospective, single-center experiences with heterogeneous reporting standards. Definitions of accuracy metrics, hemorrhage classification, and seizure outcome vary widely, limiting meta-analytic robustness. Direct comparisons between robotic and frame-based or other prior stereotactic approaches in pediatric populations remain scarce. Cost-effectiveness analyses remain sparse and institution-dependent [1,47,49,56,94].

Rationale for stereotactic ablation in pediatric epilepsy

Stereotactic ablation techniques have emerged as minimally invasive alternatives to open resection for selected epileptogenic substrates. In pediatric patients, deep-seated lesions such as hypothalamic hamartomas, periventricular nodular heterotopia, mesial temporal sclerosis, and focal cortical dysplasia located near eloquent cortex pose significant surgical challenges. Open approaches in these regions can be associated with a higher risk of neurovascular injury, endocrinologic disturbance, and postoperative functional morbidity, including physical and cognitive sequelae. RF-TC and MRgLITT allow targeted destruction of epileptogenic tissue while minimizing collateral damage and hospitalization. Robotic guidance enhances the feasibility of these procedures by enabling reproducible trajectory alignment, especially for steep, oblique, or multi-segment trajectories that may be difficult to achieve reliably using conventional frameless navigation [14,16,21,53]. Furthermore, the advantages of robot-assisted ablation techniques have enabled the application of RF-TC and MRgLITT to procedures traditionally performed through open disconnection surgery, such as corpus callosotomy and functional hemispherotomy [11,13,15,57].

SEEG-guided RF-TC

SEEG-guided RF-TC allows focal coagulation through implanted depth electrodes and can be performed without additional surgical access (since RF-TC is not approved for use with all depth electrode platforms, dedicated RF probes may be required when compatible systems are unavailable). In pediatric epilepsy, RF-TC has been explored both as a palliative therapy and as a prognostic tool to predict surgical benefit. Robotic implantation improves electrode spatial accuracy and consistency, indirectly enhancing RF-TC targeting precision [17,34,75]. Recent pediatric series suggest that RF-TC may be a viable curative treatment option for patients with selected focal epilepsies, particularly in periventricular nodular heterotopia and focal cortical dysplasia type 2b [53,78]. However, in the overall focal epilepsy patient population, long-term seizure freedom after standalone SEEG-guided RF-TC remains uncommon (typically ~15–25% at ≥12 months in published series and pooled analyses), supporting its role primarily as an adjunct or bridge to definitive surgery rather than a uniformly curative modality [9,34,72] (Fig. 3).

Fig. 3.

A : This patient had a long-term epilepsy-associated tumor involving the left precuneus. Despite resection of the most active epileptogenic zone, including the tumor and surrounding cortex as defined by stereoelectroencephalography (SEEG), seizures persisted. B : The lingual gyrus where rare epileptiform discharges were identified during SEEG, was treated with radiofrequency thermocoagulation (red circle) using the previously planned SEEG trajectory, allowing precise targeting without additional trajectory planning.

Robot-assisted MRgLITT

MRgLITT combines stereotactic laser catheter placement with real-time MRI thermometry to achieve controlled thermal ablation. Robotic systems are increasingly used for laser probe insertion due to the need for precise trajectory alignment and minimal angular deviation over long intracranial distances. Pediatric experience with robot-assisted LITT has been most extensively reported in hypothalamic hamartoma (HH), a deep midline lesion classically associated with gelastic seizures and high surgical morbidity using open approaches. Recent pediatric series employing robot-assisted, real-time MRI-guided LITT have demonstrated favorable seizure outcomes, with Engel class I rates approaching 50–70% in carefully selected patients. These studies emphasize that robotic alignment provides sufficient accuracy even in very young patients with small cranial dimensions and narrow working corridors. Importantly, transient neurological deficits, including oculomotor palsy and endocrine disturbances, remain possible and appear correlated with lesion size and proximity to critical hypothalamic structures rather than robotic technique itself. Beyond HH, robotic LITT has been applied to mesial temporal epilepsy, focal cortical dysplasia, and periventricular lesions in pediatric cohorts, although reported numbers remain limited [3,10,41,50,93]. Advantages include reduced blood loss, shorter hospital stay, and faster recovery compared with open surgery. However, ablation volume constraints, heat sink effects near cerebrospinal fluid spaces, and difficulty achieving complete disconnection in large or irregular lesions remain important limitations [10].

Safety considerations

Thermal injury to adjacent eloquent structures remains the principal risk of stereotactic ablation. Robotic accuracy minimizes off-target catheter deviation but does not eliminate risks associated with thermal spread, vascular proximity, or tissue heterogeneity. MRI thermometry and conservative ablation thresholds are essential, particularly in pediatric patients with smaller brain volumes and developing neural circuits [10,79].

Evolving role of neuromodulation

Neuromodulation therapies, including vagus nerve stimulation (VNS), DBS, and RNS are increasingly utilized in children with DRE who are not candidates for curative resection or ablation. While VNS is U. S. Food and Drug Administration-approved for pediatric use, DBS and RNS remain largely off-label in pediatric populations and are primarily supported by institutional series [44,59,80,87]. Robotic assistance offers theoretical advantages for neuromodulation implantation by improving lead placement accuracy, enabling complex multi-target trajectories, and potentially reducing operative variability. Studies of anterior thalamic nucleus DBS have demonstrated that closer proximity of the active electrode contacts to the intended anatomical target is significantly associated with improved seizure reduction and responder status [35,37]. Robotic stereotactic systems have been shown to enhance targeting precision and reproducibility compared with conventional techniques, thereby improving the reliability of electrode placement. Although robotic assistance primarily improves technical accuracy, these improvements may translate into optimized stimulation delivery and enhanced clinical outcomes by ensuring consistent engagement of therapeutic neural targets [30,54]. However, pediatric-specific evidence directly linking robotics to improved neuromodulation outcomes remains limited.

DBS

Systematic reviews and clinical series have demonstrated that DBS can achieve meaningful seizure reduction in carefully selected pediatric populations, although complete seizure freedom remains uncommon. Yan et al. [90] reported in a systematic review that more than half of pediatric patients experienced significant seizure reduction, with seizure freedom achieved in a minority of cases. Similarly, broader reviews of pediatric neurostimulation have emphasized that DBS offers a viable palliative treatment option in children with refractory epilepsy who lack conventional surgical options, with an acceptable safety profile and relatively low rates of permanent neurological complications [80].

Among available DBS targets, the centromedian nucleus (CM) and anterior nucleus (ANT) of the thalamus have been most extensively studied due to their central roles in thalamocortical network modulation. Recent pediatric and mixed-population cohorts have demonstrated that CM-DBS can significantly reduce seizure frequency and severity while improving quality of life measures, particularly in generalized or multifocal epilepsy syndromes. Meta-analytic and cohort studies have consistently shown that thalamic DBS is associated with sustained seizure reduction over long-term follow-up, although clinical outcomes vary depending on epilepsy subtype, target selection, and electrode positioning [2,58,68,91]. These findings support the role of DBS as a network-modulating therapy that may alter pathological synchronization within distributed epileptogenic circuits. Evidence from adult randomized trials, including the pivotal SANTE (Stimulation of the Anterior Nucleus of the Thalamus for Epilepsy) trial [29], has further established the efficacy and long-term safety of ANT-DBS, demonstrating progressive seizure reduction over extended follow-up periods [73,74]. Although pediatric use remains off-label, current evidence supports DBS as an important adjunctive neuromodulation strategy for pediatric DRE, particularly in patients who are not candidates for resective surgery, while highlighting the need for further prospective studies to optimize target selection, electrode placement accuracy, and patient selection criteria.

Robotic assistance may be particularly advantageous in DBS procedures targeting thalamic nuclei, such as ANT and CM, where precise localization of small, deep-seated targets and reproducible trajectory alignment across bilateral implantations are essential. While adult functional neurosurgery literature supports robotic accuracy in DBS implantation, pediatric extrapolation requires caution due to anatomical variability and growth considerations.

RNS

RNS is a closed-loop neuromodulation therapy that delivers stimulation contingent on detected epileptiform activity, while simultaneously providing chronic ambulatory intracranial electroencephalography that can inform longitudinal seizure-network management. In the United States (US), the RNS System is indicated for adults with focal (partial-onset) DRE, and pediatric use is therefore typically considered off-label and limited to highly selected cases at experienced centers [77]. Adult randomized and long-term prospective studies demonstrate durable seizure reduction over time with acceptable device-related risks, including infections and explantations as the most common serious adverse events, supporting its role as an established option for nonresectable focal epilepsy or patients with eloquent/complex networks [7,59,62]. In China, RNS has also entered early clinical evaluation through domestically developed systems, with first-in-human reports and early multicenter experience published [86,92].

Pediatric clinical reports (primarily retrospective series) suggest that RNS can be feasible in adolescents and selected younger children, often in scenarios such as MRI-negative focal epilepsy, multifocal/bilateral networks, or cases in which resection/ablation is limited by functional cortex. Across reported pediatric cohorts, outcomes are usually presented as median seizure-frequency reduction and responder rates over intermediate followup, and complication profiles resemble adult experience (notably infection/wound complications and hardware revisions), though sample sizes remain modest and heterogeneity is substantial [26,61]. Importantly, current pediatric data do not yet permit strong conclusions regarding comparative effectiveness versus other neuromodulation strategies, nor do they isolate the incremental value of robotics. Nevertheless, robotic stereotaxy is increasingly adopted for depth lead placement—particularly when targeting deep structures or bilateral networks.