Very high-frequency activity adds a sharper signal to intraoperative epilepsy mapping
Intraoperative electrocorticography has long relied on interictal spikes to guide resection in drug-resistant focal epilepsy, yet spikes alone can be imprecise. This multicenter preprint argues that very high-frequency activity—defined here as 500 to 2,000 Hz—may mark a more biologically meaningful subset of those discharges, especially when the activity travels across contacts within a tight temporal window. The central idea is straightforward: a spike accompanied by distributed, ultra-fast activity may be less like an isolated electrical event and more like a footprint of epileptogenic tissue actively propagating through a local network.
The study drew on a retrospective two-center cohort of patients who underwent electrocorticography followed by resection between January 2016 and May 2025. Across 56 patients, investigators identified 132,359 spikes, and 1,925 were vHFA-positive. Mean age was 21.2 years, and 26 participants were female. The sheer imbalance between total spikes and the small subset carrying very high-frequency activity is clinically important: the biomarker is selective, not ubiquitous, which is exactly what gives it potential value in surgical localization.
A single temporal detail stands out. The vHFA peak preceded the associated spike peak by 8.0 milliseconds, with a 95% confidence interval of 5.0 to 16.0. In epilepsy surgery, millisecond-scale ordering matters because it hints at physiology rather than artifact. Early high-frequency activity may reflect the initial breakdown of local inhibitory control or the first phase of synchronized neuronal firing before the slower spike becomes visible on standard electrocorticographic review.
Co-occurring spikes carried the strongest localization signal
The analysis separated spikes into two categories: co-occurring, when discharges appeared across contacts within 50 milliseconds, and isolated, when they did not. vHFA-positive spikes were much more likely to fall into the co-occurring category even after adjustment for anatomy and inter-electrode distance, with an odds ratio of 3.43 and a 95% confidence interval of 2.82 to 4.17. In practical terms, the presence of very high-frequency activity marked spikes that were spatially distributed across the cortex rather than confined to a single contact.
That pattern matters because propagation is often the problem surgeons are trying to solve. A spike that travels across nearby contacts may reveal the network edge of the epileptogenic zone more clearly than a solitary discharge, especially in complex focal epilepsy where the true source can be obscured by spread. The higher odds linked to co-occurring events suggest that vHFA is not just a companion signal; it may identify the spikes most likely to reflect epileptogenic tissue with surgical relevance.
Resection aligned more closely with vHFA-positive spikes than with spikes overall
Among the 35 seizure-free patients, contacts with higher rates of vHFA-positive co-occurring spikes were far more likely to be included in the resected region. The odds ratio was 11.92 per occurrence per minute, with a 95% confidence interval of 5.62 to 25.32. For comparison, the association for all spikes was much weaker, with an odds ratio of 1.09 per occurrence per minute and a 95% confidence interval of 1.06 to 1.11.
The contrast is sharp. Traditional spike burden still carried information, but the enriched vHFA subset carried far more of the localization signal. In bedside terms, that suggests a resected contact was not merely one that fired often; it was one that generated spikes with a specific high-frequency signature and a tendency to appear in spatially linked clusters. Such a pattern is consistent with a network that is both irritative and structurally organized enough to participate in seizure generation.
Residual vHFA in preserved cortex tracked postoperative recurrence
Postoperative seizure recurrence occurred in 20 patients, and a higher proportion of vHFA-positive co-occurring spikes left outside the resection was associated with recurrence. The odds ratio was 3.22 per 0.01 increase in proportion, with a 95% confidence interval of 1.14 to 9.12. Residual activity in preserved areas appears to have retained prognostic weight, which fits the biological premise of the study: if the most informative spikes are those carrying distributed high-frequency activity, then leaving those regions behind may leave part of the epileptogenic network intact.
This is the kind of signal epilepsy surgeons care about because it links localization to outcome. A biomarker that distinguishes resected from unresected tissue is useful only if it also identifies tissue whose preservation has consequences. Here, the same signal that aligned with resection in seizure-free patients also tracked seizure return when it remained outside the surgical field.
External validation supports generalizability, with limitations still visible
The investigators then tested whether a seizure outcome prediction model incorporating vHFA-positive co-occurring spikes would hold across centers. In bidirectional cross-center validation, the model produced areas under the receiver operating characteristic curve of 0.80 and 0.69. Performance improved when the model was developed using the larger sample, a familiar reminder that model stability still depends on training size and case mix.
Cross-center validation is the most persuasive part of this preprint because it moves the biomarker beyond a single operating room or a single signal-processing pipeline. Even so, the cohort remained modest at 56 patients, and the analysis is retrospective. The study also relies on electrocorticography, so the biomarker’s usefulness will likely depend on the technical quality of intraoperative recordings and on whether similar patterns emerge in other surgical settings, other epilepsy phenotypes, and different electrode configurations.
Still, the signal is coherent. Very high-frequency activity preceding the spike by milliseconds, clustering with co-occurring discharges, aligning with resection in seizure-free patients, and persisting outside the surgical target in those with recurrence all point in the same direction. Intraoperative mapping may be gaining a more selective marker of epileptogenic tissue, but this preprint remains an early step, not a practice-changing verdict.