The Sensor Science Behind Detection
Not every wearable can tell a seizure from a jog. Here's what our sensor research found — and why one signal in particular changes the accuracy picture.
Detecting a seizure reliably isn't about piling on sensors — it's about picking the ones that actually carry a signal, and being honest about what the rest of the market leaves out. Here's what our research into wearable sensors turned up.
Six Signals, One Goal
Each sensor on the wrist tells a different part of the story. Together, they're what a detection model has to work with.
Why EDA Changes the Picture
Electrodermal activity — a measure of skin conductance — spikes with sympathetic nervous system arousal. Many seizures, particularly tonic-clonic ones, trigger exactly that response. Motion and heart rate alone can be noisy or ambiguous; EDA tends to hold up.
Ranges from our sensor research phase — informed by the published literature and the hardware we evaluated, not yet a completed clinical trial of our own. See how we're validating this properly.
Why Most Smartwatches Skip It
True EDA sensing needs skin-contact electrodes, adds cost and battery drain, and is niche enough that most consumer platforms skip it in favor of a cheaper proxy: heart-rate variability. That trade-off shows up clearly once you compare what's actually on the market.
| Wearable | True EDA sensor | Notes |
|---|---|---|
| Fitbit Sense / Charge | Present, SDK-restricted | No real-time or continuous access for developers |
| Empatica EmbracePlus | Yes — clinical grade | FDA-cleared for seizure monitoring; research/clinical use only |
| Samsung Galaxy Watch | Not available | "Stress" readings are HRV-based, not true EDA |
| Apple Watch / Pixel Watch / Garmin | Not available | Same HRV-based proxy across the board |
No mainstream Wear OS, Apple, or Garmin device exposes true skin-conductance data today. We picked our current hardware with that trade-off in full view — read how we chose it →
Two Models, Not One
Without EDA on the wrist today, motion and heart rate have to work harder — and that's exactly why detection doesn't stop at the watch. A fast on-device model flags a pattern in real time; a deeper cloud model then reviews the fuller picture before anyone gets alerted, catching the everyday movements — exercise, driving, tremor, shivering — that motion data alone can confuse for a seizure.
Curious how that handoff actually works end to end? Read how your watch talks to the cloud →
The sensors are one half of the story.
The other half is validating it properly.
Read About Our Clinical Approach