How to Recognize and Manage Missing Key Electrodes in Defining the Epileptogenic Zone

Abstract

Identifying and localizing the epileptogenic zone (EZ) is the cornerstone of successful epilepsy surgery, as it directly determines surgical efficacy and patient outcomes. Stereoelectroencephalography (SEEG) has emerged as a transformative tool for pinpointing the EZ, offering precise 3-dimensional recordings of seizure activity in individual patients. SEEG does not begin with electrode implantation, but with the formulation of a precise hypothesis regarding the epileptogenic networks involved. This hypothesis is based on data gathered during Phase I investigations, particularly the electroclinical correlations of seizure obtained from video-EEG monitoring. SEEG can be appropriately performed only when a prior estimation of EZ and propagation networks is available. The guiding principle of SEEG is therefore not exhaustive anatomical coverage, but rather the validation of a predefined hypothesis. Consequently, certain brain regions are deliberately not explored, there is always missing electrodes. However, in some cases, we encounter true “missing” electrodes, unanticipated gaps in electrode coverage, that can profoundly impact the diagnostic process. These missing electrodes may create critical gaps in the spatial sampling of epileptic networks, leading to incomplete localization of the EZ. This limitation can hinder the identification of seizure onset zone, propagation pathways, and network interactions, ultimately restricting surgical options and compromising patient outcomes. Current clinical practice highlights a significant knowledge gap in how to anticipate such issues during SEEG planning and how to optimally interpret data when electrode coverage is suboptimal. Key questions remain: How can electrode placement errors be identified early? And how can emerging technologies help compensate for these gaps? In this article, we aim to illustrate and discuss these critical issues.

Introduction

Since the stereoelectroencephalography (SEEG) beginning 65 years ago, the definition of the EZ is based on the recording of patients’ seizures in the 3-dimensional brain space. It remains a complex investigation and a hard-to-standardize procedure, since it must always be tailored to the patient's individual anatomo-electro-clinical correlations in the patient's seizures.^1,2 SEEG doesn’t start with electrode implantation but with the establishment of a precise hypothesis of epileptogenic networks involved based on the data collected in Phase I. In the SEEG method, signs are interpreted through a correlative process: the goal of video-EEG recording of seizures is to establish electrical–clinical correlations; imaging semiology describes the radiological signs of a putative epileptogenic “lesion” and as such it must be anatomically compared to the results of electrical–clinical correlations. Phase I is definitely the most important step in the SEEG method. ³ If you do not get from video-EEG clearly explicit and falsifiable hypotheses in terms of epileptogenic networks, an implantation plan cannot be designed. SEEG can be properly performed only when prior estimate of seizure onset and propagation networks is available. ³

Adjusting electrode trajectory is the essential task to be performed before SEEG. A correct estimation of the epileptogenic zone (EZ) localization can only be achieved by comparison of seizure onset with its early and late propagation: this is the process that allows to define the primary organization of the ictal discharge. If too many electrode contacts are recording white matter or are not optimally placed in propagation areas of the network, SEEG interpretation becomes arbitrary. Electrode implantation plan must take account of 3 factors: anatomical sampling, volume sampling and eventual surgical landmarks. The purpose of SEEG implantation is not exhaustive coverage, but rather the confirmation or refutation of specific hypotheses. Consequently, sampling bias is inherent to the methodology, and one or more electrodes are inevitably missing.

We also have to take into account that the main determinants of the seizure onset pattern are the underlying etiology and the spatial organization of the EZ, which is inherently influenced by the strategy and placement of electrodes. With regard to postsurgical seizure outcome, the spatial organization of the EZ appears to be the primary determining factor; however, the SOP also plays a contributory role, with the presence of low-voltage fast activity (FA) being associated with a more favorable prognosis. ⁴

Although formal confirmatory multicenter data are not yet available, most clinicians experienced in SEEG quietly agree on a shared reality: in roughly 5% of implantations, and likely more when we are fully honest, at least one electrode ends up missing, malpositioned, prematurely terminated, or simply not where we thought we placed it. This is not a theoretical issue but a lived one. Even a single absent or suboptimally positioned electrode can distort the reconstruction of seizure onset and early propagation, blur the true anatomical boundaries of the EZ, and influence critical decisions regarding resection, radiofrequency thermocoagulation, laser ablation, or ultimately the recommendation to avoid surgery altogether.

If we are candid, the phrase “we really should have placed an electrode there” is not a rare confession murmured in SEEG planning meetings, it is practically a rite of passage in the field. It captures, with equal parts humor and discomfort, a problem that is real, recurrent, and far more consequential than any of us would prefer to admit.

The impact of missing electrodes is not uniform; rather, it varies across distinct clinical situations. In some cases, the absence of a single trajectory leaves a structurally abnormal region, such as an insular lesion, unsampled, creating uncertainty about its role in seizure generation. In others, an electrode that terminates in white matter may produce deceptively clean recordings that mislead the interpretation of EZ. Insufficient posterior sampling may obscure the true extent of a peri-rolandic or supplementary motor area (SMA) network, thereby altering decisions regarding resection versus neuromodulation. Advanced analytic techniques such as time–frequency analysis can be pivotal in accurately identifying the EZ. By providing a clear electrophysiological fingerprint, these methods help delineate the EZ with greater precision, reducing uncertainty and potentially eliminating the need for additional electrode implantation or repeat surgeries. ⁵

This article presents several representative examples of these scenarios, illustrating the diverse challenges posed by missing or suboptimal electrodes and the profound impact they can have on surgical strategy and patient outcomes. The potential contribution of emerging technologies, such as intracerebral forward and inverse modeling, will also be discussed.

How to Recognize the “Key Missing” Electrode

Some missing electrodes are immediately evident during the first postoperative evaluations, an electrode visibly short of its intended sulcus, one that veers into white matter, or a trajectory that does not precisely follow the preoperative plan. In these situations, the issue is recognized early, often within hours of implantation, and the team can begin weighing whether the missing information is clinically tolerable or whether additional sampling will be required to complete the diagnostic picture.

Other missing electrodes, however, reveal themselves only later, sometimes days into monitoring, when careful interpretation of ictal and interictal SEEG patterns exposes inconsistencies between the electrophysiological data and the original anatomo-electro-clinical hypothesis.

This late recognition often arrives quietly, not in the operating room but in multidisciplinary conferences, where the recordings stubbornly refuse to align with the preimplantation model. In these moments, one realizes that the implantation was technically correct but conceptually incomplete: a trajectory was omitted not by surgical error, but by a hypothesis that underestimated the true spatial organization of the epileptogenic network. Paradoxically, such incomplete but revealing explorations can be remarkably valuable. They force investigators to refine or even abandon earlier assumptions and can clarify why a prior hypothesis failed to account for the complexity of the patient's seizures.

Across true surgical omissions, subtle deviations from trajectory, and hypothesis-related sampling gaps, the timing of recognition plays an important role in determining next steps. Some missing electrodes can be compensated by anatomical extrapolation when the region lies within noneloquent cortex and the surrounding network is well understood. In eloquent regions, however, extrapolation is seldom safe, and the absence of a key trajectory may require reimplantation or a shift toward neuromodulation or minimally invasive approaches. Returning a patient to the operating room is never taken lightly. Reimplantation introduces new discomfort, prolongs monitoring, and carries cumulative risks of hemorrhage, infection, and anesthetic complications. These decisions require calm deliberation and collective judgment from epileptologists, neurosurgeons, neurophysiologists, imaging experts, and, increasingly, computational modelers.

Illustrative Scenarios of “Missing Electrodes”

Case 1

Stereotactic EEG implantation was associated with significant technical and surgical difficulties. As a result, the suspected lesion was not adequately sampled, leaving the precise relationship between the lesion and the patient's epileptogenic network uncertain.

Proposed Solution

To optimize diagnostic accuracy and clarify the lesion's involvement, the prompt addition of supplementary electrodes is strongly recommended. Early electrode placement is critical to avoid delays in understanding the epileptogenic network, minimize prolonged hospitalization, and reduce the risk of complications associated with extended monitoring.

A 20-year-old left-handed female with drug-resistant focal epilepsy presented after 2 episodes of status epilepticus. Autoimmune workup was negative. Her seizure semiology included an initial sensation of heat accompanied by nausea and vomiting, followed by impaired awareness. Brain MRI revealed a T2-hyperintense lesion of uncertain significance in the left anterior insula. The working hypothesis localized the EZ to the left mesial temporal–perisylvian region.

SEEG implantation targeted the temporal–perisylvian area; however, no electrode was placed within the insular lesion (Figure 1). SEEG findings indicated seizure onset in the left amygdala, but the relationship between the amygdala and the insular lesion remained unclear. Consequently, an additional electrode was implanted in the anterior insula to sample the lesion (Figure 2). Fortunately, this electrode showed no ictal involvement. This critical information guided the decision to avoid insular intervention. The patient underwent a targeted resection of the left amygdala and hippocampus and has remained seizure-free for 2 years.

Figure 1.

SEEG preimplantation using Talairach stereotaxic method. Letters represent electrodes targeting various regions based on the hypothesis.

SEEG, stereoelectroencephalography.

Figure 2.

Sagittal T1-weighted brain MRI showing post-implantation electrode placement. The arrow indicates the additional electrode (W') placed within the insular lesion

Case 2

This case underscores the critical importance of accurate electrode implantation and precise targeting of deep brain regions such as the orbitofrontal and mesial frontal areas. Inaccurate placement can lead to incomplete or misleading data, resulting in erroneous conclusions about the epileptogenic network and, ultimately, inappropriate surgical decisions.

Proposed Solutions

Comprehensive SEEG expertise and strong collaboration between epileptologists and neurosurgeons are essential. Furthermore, distinguishing the true EZ from mere propagation patterns is crucial.

A 58-year-old male with refractory focal epilepsy originating from the right hemisphere was referred from an outside hospital after recurrent seizures despite prior intracranial monitoring and subsequent responsive neurostimulation (RNS) targeting the right entorhinal and hippocampus. The initial evaluation failed to accurately localize the EZ due to improper electrode implantation. Following RNS placement, his seizures worsened.

After comprehensive reevaluation at our center, his RNS was removed and a second SEEG implantation was performed (Figure 3). Interestingly, SEEG revealed abundant low-amplitude FA in the right orbitofrontal region throughout the recording (Figure 4). Two habitual seizures were triggered by stimulating electrode O (orbitofrontal). Retrospective review of the prior implantation showed that the orbitofrontal electrode had terminated in white matter, leading to an erroneous conclusion regarding the EZ.

Figure 3.

T1-weighted brain MRI showing the post-implantation map. The arrows point to electrodes O and J in the deep regions of the right orbitofrontal cortex (electrode O is located very anteriorly).

Figure 4.

20-s typical SEEG seizure on bipolar montage. Ictal high-frequency activity was seen in medial leads of electrodes O and J (orbitofrontal).

SEEG, stereoelectroencephalography.

Case 3

The EZ was identified; however, its posterior boundary could not be delineated, leaving the full extent of the zone undetermined.

Proposed Solutions

The prompt addition of supplementary electrodes is recommended to accurately define the posterior border of the EZ and guide the surgical resection plan.

A 28-year-old right-handed male with a history of childhood right peri-rolandic resection following intracranial monitoring at an outside hospital presented for reevaluation. The working hypothesis localized the EZ to the right SMA and pre-SMA (Figure 5). SEEG delineated the EZ within the right SMA and mid-cingulate region; however, the posterior border of the potential resection remained unclear. To address this, 3 additional electrodes were implanted in the primary motor cortex and dorsal postcentral gyrus (Figure 6). These electrodes demonstrated significant involvement at seizure onset, revealing that the EZ was more extensive than initially suspected. Based on these findings, the patient underwent RNS implantation targeting the right SMA rather than resection.

Figure 5.

Right side: preimplantation map targeting right SMA/pre SMA areas. Left side: postimplantation MRI brain T1 with electrodes placed in the targeted regions.

SMA, supplementary motor area.

Figure 6.

20-s typical SEEG seizure on bipolar montage. Preictal spiking followed by high-frequency activity was seen in mesial contacts of electrodes H, M, A, and B. The arrows point to the ictal pattern in the added electrodes (A and B). Right lower corner: T1 MRI brain with implanted electrodes. Two added electrodes A and B were highlighted.

SEEG, stereoelectroencephalography.

Case 4

Accurate characterization of seizure semiology and rigorous anatomo-electro-clinical correlation are essential during the presurgical evaluation phase to formulate a sound hypothesis regarding the EZ. In this case, although a clear EZ pattern was identified, its posterior extent remained uncertain, posing a challenge for defining surgical boundaries.

Proposed Solutions

Time–frequency analysis proved invaluable by revealing a robust EZ fingerprint, which allowed us to confidently use the less prominent electrode as the posterior limit for resection. This approach minimized the need for additional electrode implantation.

A 23-year-old left-handed male presented with a long-standing history of poorly characterized episodes described as a “dreamy state” and dizziness since childhood. He had a left temporal lobe cavernoma and previously underwent subdural grid placement followed by cavernoma resection. Seizures recurred shortly after surgery. Based on detailed presurgical evaluation, including semiology, the working hypothesis localized the EZ to the left superior temporal gyrus. The patient subsequently underwent SEEG implantation targeting the left temporal–perisylvian region. Interictal and ictal activity was confined to 2 electrodes (U’ and T’) in the left posterior superior temporal gyrus (Figure 7). Time-frequency analysis (Figure 8) demonstrated a characteristic EZ fingerprint as described by Grinenko et al, ¹ with preictal spiking followed by double narrow-band FA and suppression of low-frequency components, most prominently in electrode U’ and less so in T’. This critical information provided confidence to proceed with resection of the region between U’ anteriorly and T’ posteriorly. The patient has remained seizure-free for 4 years postoperatively.

Figure 7.

30-second typical SEEG seizure recorded on a bipolar montage. Pre-ictal spiking followed by low-voltage fast activity was observed in electrodes U′ and T′, with no involvement of the mesial temporal regions. The lower right corner shows a post-implantation T1-weighted brain MRI with implanted electrodes. The arrows indicate the transition from the pre-ictal to the ictal rhythm in electrodes U′ and T′.

Figure 8.

Nihon Koden wide band EEG analysis shows frequent preictal high amplitude broad-band spiking in electrode U’, and then prolonged narrow-band high- frequency activity and suppression of low frequencies. Abscissa is time in seconds (120 s); ordinate is frequency in Hertz.

How Can Emerging Technologies Help Compensate for These Gaps?

In some cases, nonoptimal implantation may occur. As signal processing becomes increasingly integrated into routine clinical practice, it is being used to support clinicians in refining the definition of the EZ, particularly through the analysis of ictal and interictal biomarkers.^5–10 These approaches are now emerging as potential solutions to address the undersampling problem inherent to SEEG.

It can help estimating the quality of the implantation. In their article, Jaber et al ¹¹ proposed a framework to evaluate the goodness of the implantation. Working on the reconstruction of sources during an auditory task, Lopez-Madrona et al ¹² also showed that it was possible to estimate the quality of the implantation, looking at the confidence interval of the dipole reconstruction (Figure 9).

Figure 9.

Techniques to evaluate the quality of the implantation. (A) Permutation strength proposed to evaluate the quality of the implantation. (B) Have a look at the spatial map of the source localization.

Moreover, in case of nonoptimal implantation, some techniques were demonstrated powerful to overcome this problem. Gavaret et al ¹³ used simultaneous recordings to reconstruct sources on MEG-negative recordings based on SEEG interictal spikes. Source localization starts also to be used directly on SEEG recordings and showed a potential interest.^14–18 ⁠⁠Recent techniques combining advanced signal analysis such as Independent Component Analysis and source localization seems promising to disentangle mixed activities at sensors level and allow relevant source reconstruction.^12,19 Finally, the venue of large-scale brain modeling could also become of importance in the near future. Jirsa et al proposed a virtual brain based on a neural mass model mimicking the functioning of the seizing brains.^20–23 It's now under study to confirm its added value to clinical practice^24,25 (Figure 10).

Figure 10.

Examples of signal analysis techniques used to help overcome the undersampling limitation of SEEG (the different approaches are illustrated in different patients). (A) Use of simultaneous recordings. A simultaneous MEG/SEEG recording is shown. SEEG was used to trigger spike detection in MEG recordings, enabling dipole localization using MEG. Source localization targeted an anatomical region that was not sampled by SEEG. (B) Source localization supported by independent component analysis (ICA) can be used to extract spike or discharges and localize them within the brain, thereby increasing the effective field of view of SEEG. Orange and yellow correspond to the generator find with the ICA. (C) The use of computational models, such as the virtual brain modeling, which integrates structural and functional data with mathematical models, allows estimation of epileptogenicity in any brain region. The yellow volume corresponds to the concordant clinical epileptogenic zone (EZ) defined by SEEG recordings and by the model, while the red volume represents the additional regions identified by the model. The patient was seizure-free after resection of both regions.

SEEG, stereoelectroencephalography.

Discussion

The phenomenon of “missing electrodes” in SEEG encompasses a wide spectrum of technical, anatomical, and interpretative challenges.

Semiology serves as a critical source of behavioral data, offering valuable clues to the EZ and forming an essential component of the SEEG approach. Furthermore, anatomo-electroclinical correlations are fundamental to presurgical epilepsy evaluation and constitute the core methodology underlying SEEG interpretation. ²⁶

Moreover, time-frequency analysis has become a cornerstone in identifying the EZ and its spatial extent. Current evidence highlights 4 key components of the EZ fingerprint: preictal spiking, narrow-band high-frequency activity, suppression of low-frequency rhythms, and rebound of broad-band slower activity. Incorporating the significance of FA, regardless of its precise etiology or generator, particularly in motor epilepsy ²⁷ —adds a critical dimension to EZ characterization as we demonstrated in Case 3.

We have illustrated that SEEG must always be grounded in a strong, well-constructed hypothesis. Explorations performed with weak or vague preimplantation models, essentially “fishing expeditions,” risk missing critical regions, misinterpreting patterns, and subjecting patients to unnecessary invasive procedures without the promise of meaningful diagnostic yield. The lesson is not that incomplete hypotheses are acceptable, but that even the best-informed hypotheses occasionally evolve under the weight of new evidence. SEEG should remain hypothesis-driven, deliberate, and strategically anchored in the patient's imaging, semiology, and electrophysiology.

The cases presented in this article exemplify the diversity of these scenarios. A structurally conspicuous insular lesion left unsampled required an additional electrode to rule out a false therapeutic target; a presumed frontal onset proved illusory when a misplaced electrode created deceptively clean recordings; and insufficient posterior sampling in a peri-rolandic case transformed a seemingly focal onset into a broader network more suitable for neuromodulation than resection. These are familiar problems in SEEG practice, not rare curiosities, and they underscore the need for both meticulous planning and flexible interpretative strategies.

Emerging multimodal methods may help prevent both true and perceived electrode omissions. ¹¹ In the future, integration with computational models of seizure propagation may allow presurgical simulation of likely network pathways, guiding more precise targeting. ²³

Ultimately, the challenge of missing electrodes, whether detected early, recognized late, or revealed through careful interpretation, embodies the dynamic nature of SEEG-guided epilepsy surgery. It demands both discipline and adaptability: discipline in constructing robust hypotheses and executing precise implantations, and adaptability in reinterpreting data as new patterns emerge. When these principles are balanced, SEEG becomes not merely a diagnostic test, but a powerful, evolving scaffold upon which safe and effective surgical strategies can be built.