Presurgical evaluation pathway — оцінка придатності до хірургічного лікування епілепсії. From pharmacoresistance to resection, laser ablation, or neurostimulation.
The 2010 ILAE consensus defines drug-resistant epilepsy as failure of adequate trials of two tolerated, appropriately chosen and used antiepileptic drug (AED) schedules — whether as monotherapies or in combination — to achieve sustained seizure freedom. This single definition is the gatekeeper for the entire surgical evaluation pathway: no candidacy discussion begins until pharmacoresistance is confirmed.
An adequate AED trial requires three conditions simultaneously: the drug must be appropriately chosen for the patient's seizure and epilepsy type (a narrow-spectrum sodium-channel blocker is the wrong choice for generalized myoclonic epilepsy, for instance), it must be titrated to a maximally tolerated or guideline-recommended dose, and it must be taken consistently for a sufficient duration to judge efficacy — typically several half-lives plus enough time to capture the patient's typical inter-seizure interval.
Crucially, "failure" must reflect the drug not working, not the patient not taking it, or not tolerating it, or being on the wrong drug for their syndrome. This distinction separates true pharmacoresistance from pseudo-resistance — a category that includes misdiagnosis (psychogenic non-epileptic seizures mimicking epilepsy), poor adherence, drug-drug interactions lowering effective levels, and syndrome-inappropriate drug selection. Every candidacy workup begins by re-litigating the medication history to rule out pseudo-resistance before labeling a patient drug-resistant.
The Kwan & Brodie cohort (2000) established the now-canonical statistic: after failure of two appropriately chosen AEDs, the probability that a third or fourth drug will achieve sustained seizure freedom drops to roughly 1–4%. Continuing to cycle through medications beyond this point yields diminishing returns while the patient continues to accrue seizure-related harm — accidents, cognitive decline, psychosocial disability, and a meaningfully elevated risk of Sudden Unexpected Death in Epilepsy (SUDEP), which rises with seizure frequency and nocturnal generalized tonic-clonic seizures.
Despite this evidence, the "epilepsy treatment gap" remains large: population studies consistently show that patients who meet drug-resistance criteria wait a median of 10–20 years before being referred to a comprehensive epilepsy center, if they are referred at all. Multiple randomized and cohort studies demonstrate that earlier surgical intervention correlates with better seizure and quality-of-life outcomes, in part because prolonged uncontrolled epilepsy is associated with progressive structural and functional network changes that may themselves worsen surgical prognosis.
Once two appropriately chosen, adequately dosed AEDs have failed, the expected marginal benefit of a third trial is small — this is precisely the threshold at which international guidelines recommend referral to a comprehensive epilepsy center for presurgical evaluation, not another medication switch.
Once pharmacoresistance is confirmed, the comprehensive presurgical workup assembles independent lines of evidence to localize a single, resectable epileptogenic zone. No single test is sufficient on its own — the strength of the case comes from multiple modalities converging on the same anatomical target.
The cornerstone of the workup is prolonged inpatient video-EEG monitoring, typically lasting 5–10 days. AEDs are often reduced or withdrawn under close observation to provoke the patient's habitual seizures. Simultaneous video and scalp EEG recording allows two parallel analyses: the ictal semiology (the observable behavioral sequence — aura, automatisms, posturing, evolution) and the ictal EEG onset pattern (which scalp electrodes show the earliest rhythmic change at seizure onset).
Semiology alone carries substantial localizing value refined over decades of epilepsy surgery literature: unilateral dystonic posturing lateralizes contralaterally, early oroalimentary automatisms suggest mesial temporal onset, and versive head turning just before secondary generalization often lateralizes contralaterally to the seizure focus. The EEG onset zone, meanwhile, provides an electrophysiological correlate that must agree with — or at minimum not contradict — the semiological localization.
A dedicated epilepsy-protocol MRI (typically 3-Tesla, with thin-slice T1 and FLAIR sequences plus formal hippocampal volumetry) is fundamentally different from a routine clinical MRI. It is read specifically to detect subtle findings that general radiology reads often miss: mesial temporal sclerosis (hippocampal atrophy and T2/FLAIR hyperintensity), focal cortical dysplasia (blurring of the gray-white junction, cortical thickening, the "transmantle sign"), cavernomas, and low-grade tumors such as DNET or ganglioglioma.
A clearly identified, surgically accessible lesion that matches the electro-clinical localization is the single strongest positive predictor of good surgical outcome. Conversely, an MRI-negative workup does not exclude surgery but substantially raises the bar for the remaining modalities to converge, and often mandates invasive intracranial EEG before a resection can be safely offered.
Formal neuropsychological testing assesses memory (verbal vs. visuospatial, often differentially lateralizing), language, executive function, processing speed, and general intellectual ability. A pattern of material-specific memory impairment (e.g., verbal memory deficits with a left temporal focus) provides an independent, non-invasive lateralizing signal that should agree with imaging and EEG.
Functional imaging adds a metabolic and perfusion dimension: interictal FDG-PET frequently shows focal hypometabolism at the epileptogenic zone, with reported sensitivity of 70–90% in MRI-negative temporal lobe epilepsy — making it particularly valuable when structural imaging is unrevealing. Ictal SPECT, especially when subtracted from an interictal scan and co-registered to MRI (SISCOM), captures the transient hyperperfusion that accompanies seizure onset, offering a further independent localizing data stream.
Concordance is the organizing question of the entire multidisciplinary epilepsy surgery conference: does semiology, scalp EEG localization, the MRI lesion, and the neuropsychological deficit pattern all converge on the same anatomical region? The answer determines both the confidence of the surgical plan and the probability of postoperative seizure freedom.
At epilepsy surgery conference, each case is reviewed against four independent data streams: (1) ictal semiology from video-EEG, (2) scalp EEG ictal onset localization, (3) the structural MRI lesion, and (4) the neuropsychological lateralization pattern. When all four streams agree — for example, a patient with dystonic posturing lateralizing to one side, ictal EEG onset over the ipsilateral temporal region, a hippocampus that is atrophic and hyperintense on the same side, and verbal memory deficits consistent with that side being dominant — the case is described as highly concordant.
This is not a formality: concordance is the single best-validated predictor of surgical outcome in the epilepsy literature. Unilateral temporal lobe epilepsy with hippocampal sclerosis and full four-way concordance carries some of the highest reported seizure-freedom rates of any epilepsy surgery, commonly 70–80% Engel class I outcomes at one to two years.
Partial concordance (two or three of four modalities agreeing) still permits surgery in many cases, but with lower expected seizure-freedom rates and a materially higher likelihood that intracranial EEG (stereo-EEG or subdural grids) will be required to resolve the discrepancy before a resection can be safely planned. Poor concordance — modalities actively conflicting, such as a right-sided MRI lesion with left-lateralizing semiology — is a major red flag: it may indicate a multifocal epileptogenic network, an inaccurate lesion (incidental finding unrelated to the seizures), or a seizure onset zone that scalp EEG simply cannot resolve.
In discordant cases, the presurgical team faces a fork: invest in invasive intracranial monitoring to directly sample the suspected region and its neighbors, or conclude that the case is not currently amenable to focal resection and redirect the patient toward neuromodulatory therapy instead.
Multiple outcome series have shown a roughly graded, monotonic relationship between the number of concordant modalities and postoperative seizure freedom: each additional agreeing data stream meaningfully raises the probability of an Engel class I outcome, while each conflicting stream lowers it. This graded relationship is why presurgical conferences do not treat candidacy as a binary yes/no decision but as a probability estimate that is continuously updated as each test result arrives — video-EEG, MRI, PET, neuropsych, and, when needed, intracranial EEG all feed the same running estimate of expected surgical benefit.
A single "positive" test is never sufficient. It is the pattern of agreement — or disagreement — across semiology, EEG, imaging, and neuropsychology that the surgical conference ultimately votes on, because concordance is the strongest available proxy for a genuinely focal, resectable epileptogenic zone.
Even a well-localized, highly concordant epileptogenic zone is not automatically resectable — the final gate is risk. Before committing to surgery, the team must know exactly which hemisphere and which cortical regions are responsible for language and memory, and how close the proposed resection margin sits to that eloquent tissue.
The Wada test (intracarotid sodium amobarbital procedure) remains the historical gold standard for lateralizing language and memory function. A short-acting barbiturate is injected selectively into one internal carotid artery, transiently anesthetizing that hemisphere while the patient is tested for language production and memory encoding. If language function fails during injection of one hemisphere but is preserved during injection of the other, language is lateralized to the failing side. The procedure is repeated on the contralateral side to assess memory reserve in the hemisphere that would remain after resection — a critical question for dominant temporal lobe surgery, where the contralateral (non-resected) hippocampus must be able to support memory on its own.
Task-based functional MRI (typically verb generation, semantic decision, or word-generation paradigms) has increasingly supplanted the Wada test as a first-line, non-invasive method for language lateralization, showing concordance with Wada results in roughly 85–90% of cases in validation studies. Memory fMRI paradigms (scene encoding, word-pair encoding) are used similarly to probe hippocampal activation asymmetry. Where fMRI and Wada disagree, or where fMRI activation is ambiguous or bilateral, the invasive Wada test — or direct intracranial mapping — is still used to resolve the lateralization question before finalizing the surgical plan.
Once language and memory are lateralized, the risk calculus becomes spatial: how close does the proposed resection margin come to eloquent cortex — classically Broca's and Wernicke's areas, primary motor and sensory strips, and the dominant hippocampus? For lesions or epileptogenic zones abutting these regions, awake craniotomy with direct electrical cortical stimulation mapping is used intraoperatively: the patient performs language or motor tasks while the surgeon stimulates cortex adjacent to the planned resection, allowing margins to be tailored in real time to spare functional tissue. For dominant-hemisphere temporal lobe surgery specifically, verbal memory decline is a well-documented risk (on the order of 30–40% of patients show measurable verbal memory decline), which is why preoperative memory lateralization and counseling are integral to informed consent.
Surgical planning is never purely anatomical — it is anatomical risk weighed explicitly against expected seizure-freedom benefit. A resection that is oncologically or electrographically "complete" but sacrifices language or leaves a patient functionally amnestic is not considered a surgical success.
The presurgical evaluation converges on a single decision: is this patient a favorable candidate for a focal, curative-intent procedure, or should they be directed toward a neuromodulatory alternative? The landmark randomized evidence base — and three decades of outcome registries — now make this decision one of the most evidence-grounded in all of epilepsy care.
Patients with confirmed drug resistance and high concordance across semiology, EEG, MRI, and neuropsychology — with an acceptable risk profile relative to eloquent cortex — are the ideal candidates for definitive surgery. The randomized controlled trial by Wiebe et al. (NEJM, 2001) remains foundational: it demonstrated that anterior temporal lobectomy produced dramatically superior seizure freedom compared with continued medical therapy in patients with temporal lobe epilepsy, ending the debate about whether surgery should be offered earlier rather than as a last resort.
Laser interstitial thermal therapy (LITT) has emerged as a minimally invasive alternative to open resection for well-localized lesions such as mesial temporal sclerosis or small focal cortical dysplasias, delivering stereotactically guided thermal ablation through a small burr hole. Seizure-freedom rates are somewhat lower than open lobectomy but the reduced surgical morbidity, shorter recovery, and better-preserved neuropsychological outcomes make it an increasingly favored option for appropriately selected concordant cases.
When concordance is poor, the epileptogenic zone is multifocal or non-localizable, or the risk to eloquent cortex is unacceptable relative to expected benefit, resective surgery is not offered — but the patient is not without options. Neurostimulation devices are designed precisely for this population: Vagus Nerve Stimulation (VNS) delivers intermittent electrical stimulation to the left vagus nerve and is indicated broadly for drug-resistant epilepsy regardless of focality, with roughly 50–60% of patients achieving a ≥50% seizure reduction at one to two years (freedom is rare).
Responsive Neurostimulation (RNS) and Deep Brain Stimulation (DBS, targeting the anterior nucleus of the thalamus) are reserved for patients with one or two localizable but non-resectable foci; RNS in particular closes the loop by detecting abnormal electrographic activity and delivering targeted stimulation in real time, with responder rates that climb over years of therapy, commonly reaching around 70% at long-term follow-up even though outright seizure freedom remains uncommon with neurostimulation compared with resection.
In practice, the resection-versus-neurostimulation decision sits on a continuum defined jointly by drug-resistance confirmation, concordance strength, and eloquent-cortex risk — exactly the two-slider model this simulator represents. A patient with only one AED failure is not yet a surgical candidate at all; a patient with confirmed pharmacoresistance and high concordance is steered toward resection or ablation; a patient with confirmed pharmacoresistance but poor or partial concordance is steered toward neurostimulation while further localization (often intracranial EEG) is considered. This staged, evidence-weighted pathway is what allows modern epilepsy surgery to offer some of the highest success rates of any procedure in neurology while keeping poorly-localized or high-risk patients safely out of the operating room.
Across multiple randomized trials and long-term registries, appropriately selected surgical candidates achieve seizure freedom at rates unmatched by any medication regimen — yet only a small fraction of eligible drug-resistant patients are ever referred for evaluation, making the presurgical pathway itself one of the biggest untapped levers in epilepsy care.