Multimodal Wearable Nelli Tested for Seizure Classification in Dravet, LGS

A non-randomized, single-center feasibility study has tested whether a multimodal seizure monitoring system called Nelli can classify seizures in patients with Dravet syndrome and Lennox–Gastaut syndrome (LGS), two of the most drug-resistant developmental and epileptic encephalopathies. The study appears in a journal hosted on Wiley Online Library.

The choice of population matters. Dravet syndrome and LGS both feature frequent, motorically heterogeneous seizures — tonic, atonic, myoclonic, generalized tonic-clonic and focal events with impaired awareness often coexist in the same patient. That heterogeneity is exactly where single-channel wearable devices have historically struggled. A device that detects convulsive events but cannot separate a tonic seizure from a myoclonic jerk offers limited value to epileptologists adjusting treatment, since seizure types respond differently to antiseizure medications and carry different risks of injury and status epilepticus.

Nelli approaches the problem as a multimodal system rather than a single-signal wearable. The published feasibility study evaluated its ability to classify seizure types — not merely count them — in a defined cohort of Dravet and LGS patients monitored at a single center. The non-randomized, single-center design places this work firmly in the early clinical evidence tier: it is a feasibility study, not a pivotal validation trial, and readers should treat the results as hypothesis-generating rather than practice-changing.

This distinction is worth spelling out for the monitoring unit. The gold standard for seizure classification remains video-scalp EEG with full montage review by a clinical neurophysiologist, typically during inpatient long-term monitoring. Wearable and multimodal systems occupy a different niche: they extend observation beyond the epilepsy monitoring unit, into the home, where most seizures in Dravet and LGS actually occur and where caregiver reports systematically undercount and misclassify events. Families of children with Dravet syndrome frequently report nocturnal generalized tonic-clonic seizures as their dominant fear, and SUDEP risk in this population keeps accurate nocturnal detection a clinical priority rather than a convenience feature.

For the reading room, the practical questions raised by this study are the ones any new classification device must answer. How does the system define its seizure-type labels, and against what reference standard were they adjudicated? What sensitivity and specificity did it achieve per seizure type, not just per event? How did it handle the non-motor and subtle motor events that dominate LGS? And how burdensome was the device for a population that includes children with intellectual disability and behavioral comorbidities, where tolerability often decides whether a monitoring technology survives in daily use?

The study's feasibility framing signals that the investigators prioritized exactly these questions — whether the system works at all in this demanding cohort, whether patients and caregivers can use it, and whether its outputs align with clinical seizure classification — before larger comparative trials. Single-center feasibility studies typically enroll modest numbers, and their results set the stage for multicenter validation rather than regulatory claims. Any device intended for seizure classification as a clinical function would need to clear regulatory review; a feasibility study alone does not establish FDA clearance or CE marking for that indication.

The context is also competitive. The ambulatory seizure-detection field already includes FDA-cleared wrist-worn devices for generalized tonic-clonic detection in epilepsy. Multimodal systems such as Nelli aim one step higher in difficulty: classification across seizure types, in syndromes where seizure semiology overlaps and evolves. If the feasibility results hold up, the clinical payoff would be longitudinal, at-home seizure-type data to guide medication decisions, evaluate fenfluramine, cannabidiol and other Dravet- and LGS-specific therapies, and quantify the seizure burden that trial sponsors increasingly want measured outside the clinic.

For now, clinicians should read this study as an early but targeted data point: a multimodal monitoring system tested specifically in Dravet and LGS, two populations that most consumer-adjacent seizure tech has avoided. The findings will need replication in multicenter cohorts with standardized EEG- or video-EEG-anchored reference classification before Nelli or similar systems can join routine care. The next step the field will watch for is whether the investigators move from feasibility to a powered validation study — and whether the device can distinguish seizure types, not just detect them, where patients actually live.