Bold claim: a single, unified approach could overhaul how doctors locate seizure-causing brain regions, potentially making presurgical planning safer and faster. But here’s where it gets controversial: can one framework truly replace multiple separate analyses for different epileptic biomarkers?
A team from Carnegie Mellon University led by Bin He has introduced spatial-temporal-spectral imaging (STSI), a machine learning–driven method designed to analyze all major epileptic brain signals within one computational framework. Published in PNAS, STSI represents both a technical breakthrough and a promising new direction for noninvasive presurgical planning, aiming to identify the epileptogenic zone—the brain area where seizures originate.
Currently, many epilepsy centers rely on invasive intracranial EEG recordings to pinpoint seizure onset. Patients may be monitored for days or weeks until a seizure occurs. While accurate, this approach is costly, time-consuming, and physically taxing. Noninvasive scalp EEG offers a safer alternative, but clinicians have struggled to determine which biomarkers—spikes, high-frequency oscillations (HFOs), or actual seizures—are most reliable for locating seizure-generating tissue. Traditionally, each biomarker required its own analysis pipeline, leaving the field without a unified way to compare them.
STSI changes this landscape by jointly analyzing where activity occurs, when it happens, and at which frequencies. This enables imaging of transient events like spikes and oscillatory events like seizures and HFOs. For the first time, a single algorithm can handle all epileptic biomarkers, a feat never before accomplished.
In their multi-year study, He’s group examined 2,081 individual EEG events from 42 patients with drug-resistant epilepsy. This study represents the first rigorous quantitative comparison of all major epileptic biomarkers for source localization and was conducted in collaboration with clinicians at the Mayo Clinic, who provided the patient data.
A key finding is that pathological HFOs—those that occur when HFOs overlap with spikes—are the most accurate interictal biomarker for locating epileptogenic tissue. Pathological HFOs localized the epileptogenic zone within roughly nine millimeters of the results from invasive seizure mapping, approaching the seven-millimeter precision achieved when seizures are recorded.
This has important implications. He notes that pathological HFOs can be recorded in under an hour, compared with the days often required to capture seizures. In contrast, general HFOs, which were previously heralded as a promising biomarker, performed poorly in this analysis. The study helps resolve longstanding inconsistencies across clinical research by clarifying which HFOs matter most.
Beyond epilepsy, STSI represents a conceptual shift in electrophysiological source imaging. It offers a noninvasive, faster tool to support presurgical planning and has potential applications across a wide range of brain signals, including EEG and magnetoencephalography (MEG). This could enable new investigations into memory, attention, pain, psychiatric disorders, and normal brain function.
Looking forward, the researchers aim to obtain additional funding to validate STSI in larger patient cohorts and move toward clinical adoption.
“The whole point is to help others,” He emphasizes. “If a noninvasive, precise alternative can spare patients from days of invasive monitoring, that would be a major impact. The commitment is to improve the patient experience through our expertise.”
Source:
Jiang, X., et al. (2025). Mapping epileptogenic brain using a unified spatial–temporal–spectral source imaging framework. Proceedings of the National Academy of Sciences. doi: 10.1073/pnas.2510015122.
