Seizure Detection Without EEG: ECG and Accelerometry Study Posted to medRxiv

A preprint posted to medRxiv, titled "Multimodal Seizure Detection with Long-term Ambulatory ECG and Accelerometry Data," examines seizure detection using physiological signals outside the electroencephalographic channel that normally anchors this task.

The study focuses on two modalities recorded in ambulatory settings: long-term electrocardiography and accelerometry. Together, these capture the two most common non-EEG physiological signatures of epileptic seizures — ictal cardiac changes, including tachycardia and other rhythm disturbances, and the motor manifestations that surface during convulsive events. The "multimodal" framing of the title signals that the authors combine both signal streams rather than rely on either one alone.

For clinical neurophysiology readers, the approach touches a persistent practical problem. Scalp video-EEG monitoring remains the diagnostic reference standard for seizure classification, but long-term ambulatory EEG is burdened by electrode maintenance, artifact, and patient tolerability. Wearable ECG patches and wrist- or body-worn accelerometers, by contrast, can record for days to weeks with minimal supervision, which makes them attractive for extended detection windows — particularly for convulsive seizures and events with prominent autonomic features, including most seizures of temporal lobe onset and many focal to bilateral tonic-clonic seizures.

The preprint platform itself matters for how the findings should be read. medRxiv posts are not peer-reviewed, and the paper's methods, cohort size, sensitivity, false-alarm rates, and validation strategy have not yet undergone external scrutiny. Readers evaluating the work for monitoring-unit or ambulatory-practice implications should treat any reported performance figures as provisional until the manuscript clears formal review.

The publication abstract available with the listing does not specify the patient cohort, recording durations, or algorithm architecture, so those details await consultation of the full preprint text on medRxiv. The title's reference to "long-term" recording implies multi-day acquisition rather than the hours-long windows typical of inpatient video-EEG, consistent with the ambulatory context the authors name.

The broader direction is well established in the field. Cardiac-based seizure detection exploits ictal tachycardia, which occurs in a large fraction of seizures and can precede or coincide with electrographic onset, while accelerometry-based detection targets the rhythmic motor patterns of tonic-clonic events. Combining the two channels offers complementary coverage: cardiac signals capture seizures with minimal movement, and accelerometry captures the convulsive events that matter most for sudden unexpected death in epilepsy (SUDEP) risk, where post-ictal respiratory and cardiac dysfunction are the leading mechanisms.

Whether the authors' multimodal system performs well enough to approach the sensitivity and false-positive performance of existing wearable and implantable detection devices — such as FDA-cleared wrist-worn or muscle-based systems already in clinical use — is the central question the full preprint must answer. Those cleared devices set a high bar, with published sensitivities for tonic-clonic seizures typically above 90 percent in controlled evaluations.

The preprint is available on medRxiv, and its eventual peer-reviewed publication will determine whether long-term ambulatory ECG plus accelerometry can serve as a practical, EEG-sparing complement — or alternative — for extended seizure detection in selected patient populations.