Sleep Science

Can an Earbud Score Your Sleep Like the Lab? What the Research Shows

6 min readThe Wize Sleep Editorial Team

Reviewed by the NextSense sleep science team

The sleep lab is the gold standard. It’s also a single bad night in a strange bed, wired to a machine, maybe once in your life. The real question is whether something you’d wear in your own bed — night after night — can do the job.

For older adults, that question isn’t academic. Sleep quality is tied to cognitive health, and disrupted deep sleep has been linked to the progression of conditions like Alzheimer’s. These are exactly the people for whom a one-time lab visit is least practical and ongoing, at-home monitoring would matter most. A 2024 study in the IEEE Journal of Translational Engineering in Health and Medicine tested whether the ear could fill that gap.

What did the study do?

Seventeen older adults — average age about 72, several with managed health conditions — spent a night under full clinical polysomnography (the lab gold standard) while also wearing a single in-ear EEG sensor. Recording both at once let the researchers ask precisely how well the ear matched the lab.

Then came the clever part. Instead of training a new model from scratch on scarce ear data, they took an existing model trained only on scalp EEG — built on more than 31,000 hours of lab recordings — and pointed it straight at the ear signal.

What happened?

Out of the box, with no adaptation at all, the scalp-trained model scored the ear-EEG sleep stages at 70.1% accuracy. A light fine-tune on ear data lifted that to 73.7% — a statistically significant improvement for 10 of 13 participants, with agreement (Cohen’s kappa) reaching 0.639.

Knowledge learned from a scalp full of electrodes transferred to a single sensor in the ear — without starting over.

The most telling detail: the fine-tuning helped most with N3, deep slow-wave sleep — the restorative stage that clears the brain overnight and is the one you most want to measure accurately. The ear got better at exactly the stage that matters.

What this does — and doesn’t — prove

Honesty first, because that’s the whole point of reading research instead of ads. This was a focused study: 17 participants, a single in-ear electrode, accuracy that is research-grade rather than lab-replacing. Four participants were excluded for poor skin-electrode contact — a real, ongoing challenge in ear-EEG. The headline is not “the lab is obsolete.”

The headline is the direction: that the enormous body of knowledge built from decades of scalp sleep studies can be transferred to the ear, and that the ear can carry the deep-sleep signal that matters most — in the older, real-world population that needs at-home monitoring the most.

Why this is the future of sleep monitoring

A once-a-year lab night tells you about one atypical night. A comfortable in-ear sensor, worn in your own bed, could tell you about your sleep as it actually is — over weeks, months, years. For tracking the slow changes that matter to brain health as we age, that shift from a snapshot to a continuous record is everything.

NextSense Smartbuds are built on this same in-ear EEG lineage — clinical-grade sensing designed to read the brain’s rhythm at home, night after night. The science is published, not promised, so we put it where you can read it.

Frequently asked questions

Can an earbud measure sleep stages like a sleep lab?

Increasingly, yes — though the lab remains the gold standard. A 2024 study took a sleep-scoring model trained only on scalp EEG and applied it to a single in-ear sensor in older adults, achieving 70.1% accuracy out of the box and 73.7% after light fine-tuning, with the biggest gains on deep (N3) sleep. The accuracy is research-grade, and the trajectory points toward reliable at-home sleep monitoring from the ear.

What is transfer learning in sleep scoring?

Transfer learning reuses a model trained on one kind of data for a related task instead of training from scratch. In this 2024 study, a model trained on more than 31,000 hours of scalp EEG was adapted to in-ear EEG with only a light fine-tune — showing that decades of scalp sleep knowledge can transfer to the ear, where large datasets are still scarce.

Why does ear-EEG sleep monitoring matter for older adults?

Sleep quality, and deep slow-wave sleep in particular, is tied to cognitive health, with disrupted deep sleep linked to conditions like Alzheimer’s. Older adults are also the group for whom a one-time lab visit is least practical. A comfortable in-ear sensor enables continuous, at-home monitoring over time — far more useful for tracking brain-health-relevant changes than a single lab night.

How accurate was the ear-EEG in this study?

The scalp-trained model scored ear-EEG sleep at 70.1% accuracy without adaptation and 73.7% after fine-tuning, a significant improvement for 10 of 13 participants (Cohen’s kappa 0.639). This is a proof-of-concept with 17 participants and a single in-ear electrode — research-grade rather than lab-replacing — but it demonstrates that scalp knowledge transfers to the ear and captures deep sleep well.

Sources

Keep reading