Wearable physiology adds a second channel to epilepsy monitoring
The PreEpiSeizures project set out to answer a practical question with immediate bedside relevance: can wearable sensors capture physiologic signals during seizures well enough to support epilepsy research beyond the EEG trace? Video-EEG remains the clinical reference standard for seizure characterization, but it is still confined to the hospital, where sleep disruption, limited movement, and a short observation window can blunt the value of what is being measured. A wearable platform that preserves signal quality during real-world activity would extend epilepsy monitoring into daily life and could make seizure detection, burden assessment, and event characterization more usable outside the ward.
The dataset assembled so far is substantial by epilepsy-wearable standards. Investigators enrolled 59 patients across 2 hospitals and collected 2,721 hours of wearable data, alongside 348 seizures. Hospital-based electrocardiographic data added another 5,838 hours. Those numbers matter because wearable development in epilepsy has repeatedly run into a basic bottleneck: algorithms can look impressive in isolated testing, but they cannot be validated without a sufficiently large corpus of synchronized physiologic recordings tied to clinically verified seizures.
ECG signal quality matched the hospital reference
The strongest technical result in the cohort involved ECG. The quality of ECG signals captured through the proposed wearable system was comparable to the hospital electrocardiography setup. In a field where motion artifact, imperfect skin contact, and variable wear time can degrade traces quickly, that is a meaningful engineering achievement. ECG is not merely an accessory channel; it can reflect autonomic changes during seizures, including tachycardia and other rhythm shifts that may help define event timing or identify seizure-related stress responses. When wearable ECG preserves the fidelity of the hospital reference, it becomes much easier to imagine using the same signal for prolonged monitoring at home, in transit, or during sleep.
Other biosignals also reached state-of-the-art quality. The project used three different wearable form factors, and each was deployed in the context of video-EEG monitoring. That design choice is important because form factor influences both signal stability and patient acceptance. A wearable that performs well but is too intrusive will fail in routine care; a comfortable device that loses data during seizure activity is equally unusable. By testing multiple configurations, the project generated both performance data and implementation experience, which is exactly the combination needed for the next stage of epilepsy monitoring research.
Seizure-adjacent physiology is the real prize
Seizures are electrical events, but their clinical footprint extends well beyond the scalp EEG trace. Changes in heart rate, movement, and other physiologic signals can precede, accompany, or follow the ictal period, and those changes may vary by seizure type and patient. A wearable system that captures those signals during confirmed events can reveal patterns that EEG alone cannot provide. In practical terms, that means the same seizure might be seen not just as a discharge on a screen, but as a coordinated physiologic storm involving autonomic and motor features. Such data are especially valuable for algorithm development, because models trained on a single biosignal often underperform when asked to generalize to messy outpatient life.
The study also documented 18 challenges encountered during data acquisition, with possible solutions presented alongside them. Those operational obstacles are not trivial side notes. Wearable research fails most often in the handoff between prototype and routine use, where electrode placement, battery life, synchronization, data loss, patient comfort, and workflow interruptions can undermine an otherwise elegant design. A dataset built inside a hospital can still be fragile if these issues are not anticipated. Here, the accumulated lessons may be as useful as the physiologic recordings themselves, because they map the points where real-world monitoring breaks down.
A single seizure trace is easy to admire. Sustained acquisition across dozens of patients is what makes the work clinically useful.
Hospital monitoring remains the anchor, but wearables widen the frame
Video-EEG remains indispensable because it links observed behavior to electrical seizure onset. Yet hospital monitoring is inherently selective. It captures a patient in a controlled environment and for a limited duration, which means it can miss the physiologic variability that defines epilepsy over days, weeks, and ordinary routines. Wearable data can widen that frame by filling in the gaps between admissions and by providing a substrate for future home-based studies. The implication is not that wearables should replace EEG; it is that they may eventually complement it by adding context that the clinic cannot continuously observe.
This is also why datasets matter as much as devices. In the absence of large, well-annotated recordings, even a technically promising sensor cannot be judged fairly. The project directly addresses that limitation by contributing original wearable data linked to seizures captured during video-EEG monitoring. For researchers building algorithms to detect seizures or characterize physiologic signatures, a dataset with confirmed events and synchronized biosignals is the difference between speculative modeling and credible validation.
What the current dataset can and cannot answer
The cohort already demonstrates feasibility: wearable collection during video-EEG was possible, ECG quality was comparable to the hospital system, and additional biosignals reached strong quality standards. It also demonstrates scope, with 2,721 hours of wearable data and 348 seizures now available for analysis. Those are concrete outputs, not theoretical promises. At the same time, the work remains ongoing, and the current report does not establish clinical performance for seizure detection in everyday home settings, nor does it claim that any wearable configuration is ready for routine substitution of hospital EEG.
The next clinical question is narrower and more demanding: which signal combinations, device placements, and acquisition workflows can survive the variability of real life without sacrificing interpretability? Answering that will determine whether wearable physiology becomes a tool for epilepsy care or stays confined to research settings. For now, the study supplies something the field has needed for years—original physiologic data, confirmed seizure events, and a detailed account of the practical problems that shape wearable monitoring before it reaches the bedside.