pSpikes as a Noninvasive Biomarker for Epileptogenic Zone Localization

Epilepsy is defined clinically by recurrent seizures that arise without provocation, yet behind that definition lies enormous diversity in how the condition manifests and how it is treated. For some patients, medication is a reliable lifeline, suppressing seizures for years at a time. However, for nearly one third remain resistant to every drug combination available. For these individuals, the option of surgery often comes to the table—not as a casual suggestion, but as the only path that still offers a chance of long-term seizure freedom. Surgery requires a clear map of the epileptogenic zone (EZ), the smallest volume of cortex that, if removed or disconnected, can abolish seizures and a mistake in judgment can be costly and leaving part of the EZ behind cause the seizures to persist; on the other hand, removing too much can cause vital cognitive or motor functions to be irreversibly impaired.

At present, mapping the EZ is done with a mixture of tools but none of which is ideal. Intracranial EEG is still considered the most definitive method. Depth electrodes or cortical grids can capture activity directly from regions suspected of generating seizures, and when they succeed, the information is invaluable. Yet the procedure is invasive, risky, and inherently limited by the fact that only selected regions can be sampled. Safer approaches such as high-density scalp EEG and magnetoencephalography provide whole-brain coverage with millisecond precision, but spatial resolution suffers. Signals recorded at the scalp are smeared and blended by the electrical properties of skull and scalp tissue, meaning that what is observed on the surface may not faithfully represent the true cortical generators. One pragmatic solution has been to analyze interictal spikes—the sharp waveforms that pepper the EEG between seizures. They are relatively easy to detect, abundant in most patients, and provide at least a starting point for localization. The problem is that spikes are not specific. Many reflect irritative tissue rather than the true seizure onset zone, and occasionally they appear in brain regions far removed from the actual focus. Clinicians have long known this, which is why they interpret spikes with caution, aware that they can mislead as often as they guide.

A more recent candidate biomarker has been high-frequency oscillations, or HFOs. Work with intracranial recordings suggests that ripples and fast ripples are closely tied to epileptogenic tissue. The promise is enticing, but the translation to scalp EEG has been limited. These oscillations are faint, easily drowned out by noise, and difficult to separate from normal rhythms or technical artifacts. As a result, the field has faced with the gap: noninvasive methods that are rich in data but poor in specificity, and invasive approaches that are highly specific but risky and limited. To this account, a new research paper published in Epilepsia and conducted by Colton Gonsisko, Zhengxiang Cai, Xiyuan Jiang, Andrea Duque Lopez, Gregory Worrell, and Bin He from the Department of Biomedical Engineering at Carnegie Mellon University and Mayo Clinic. The researchers developed an advanced noninvasive method to localize epileptogenic brain regions by combining a novel biomarker—spikes coinciding with high-frequency oscillations (pSpikes)—with a powerful source imaging algorithm known as FAST-IRES. This approach allows individual scalp-recorded spikes to be traced back to their cortical generators with high accuracy, without the need for averaging large number of spikes. In doing so, they provided a clinically practical tool that outperforms conventional spike analysis and holds direct promise for guiding epilepsy surgery.

The research team drew upon recordings from twenty-four patients with drug-resistant focal epilepsy who underwent high-density scalp EEG as part of presurgical evaluation. Each patient generated hundreds of interictal spikes, which were carefully classified into four groups. The rarest, pSpikes, were those that coincided with high-frequency oscillations; others included nSpikes with irregular high-frequency fluctuations, rSpikes without oscillatory activity, and the pooled category aSpikes, representing all detected discharges. The researchers then applied a sophisticated source imaging method, the FAST-IRES algorithm, which had been designed to extract not only the location but also the spatial extent of the cortical generators from single events. By doing so, they could test in a direct and quantitative way whether pSpikes offered an advantage in pinpointing the epileptogenic zone. The authors found that when imaging pSpikes in patients who later became seizure-free after surgery, the estimated sources aligned closely with the resected tissue. Localization errors averaged less than seven millimeters, markedly smaller than the errors seen with other spike classes, which clustered around fifteen millimeters. This difference may appear modest in absolute terms, but in surgical planning such precision can mean sparing or removing critical brain regions. Moreover, the authors evaluated sensitivity and precision and found pSpikes captured more of the resected zone while avoiding spurious spread to non-epileptogenic tissue which suggests that these events were far more specific markers of the true pathological substrate.

Additionally, the team examined patients who have multiple spike morphologies and in these cases, conventional spike imaging often produced confusing or even contradictory maps, sometimes pointing toward contralateral hemispheres. However, pSpikes consistently traced back to the surgical target, offering clarity where other signals misled. The value of this distinction was highlighted by contrasting seizure-free and non–seizure-free groups. In patients who failed surgery, pSpike imaging pointed to regions outside the resection which implied that residual epileptogenic cortex had been left behind. Thus, the new method can both predict success and also expose the reasons for surgical failure.

In conclusion, Professor Bin He and colleagues successfully provided a clinically practical tool that outperforms conventional spike analysis and holds direct promise for guiding epilepsy surgery. They identified those spikes that carry high-frequency oscillations and drew out a signal that is more tightly linked to the epileptogenic zone than the mixed population of spikes that clinicians usually examine. Another aspect that stands out is the ability to work with single events rather than massive averages. Conventional wisdom has held that spikes recorded on the scalp are too noisy to trust unless multiple spikes are averaged to get rid of noise. Yet the group’s use of the FAST-IRES algorithm challenges that assumption. They were able to take an individual pSpike, even one buried in a background of scalp noise, and still recover a source estimate that matched clinical ground truth. This kind of result opens a path toward shorter hospital stays and less burdensome monitoring, and it may make presurgical testing accessible for patients who would otherwise never qualify for invasive evaluation. It is a reminder that engineering advances, when carefully tuned to clinical problems, can change what is considered possible in practice.

The implications reach further than surgical planning alone. The fact that pSpikes map more reliably to epileptogenic tissue points to something deeper about how seizures begin. The temporal coincidence of a spike and an oscillation may reflect a pathological state that is closer to seizure onset than either feature alone. If that is correct, then pSpikes may one day play a role not only in mapping but also in prediction, in guiding neuromodulation therapies, or even in shaping how we think about the mechanisms of ictogenesis. Particularly telling was the observation that in patients who did not become seizure-free, pSpike imaging pointed to regions outside the resection. That kind of feedback has clear potential to guide second interventions or refine decision-making after surgery. What comes next is validation. Larger, prospective trials will be needed to know whether this new strategy can be implemented into routine workflows. But if it holds up, the approach could redefine noninvasive care in epilepsy.