A soft patch that sticks to the head and beams focused ultrasound into the brain shortened the time it took sleepers to reach REM sleep by 43 minutes and added about 16 minutes of REM, according to a 28-person study from a University of Texas at Austin team published June 4, 2026 in Nature Communications. Each volunteer slept two nights in a sleep lab wearing the device, called NEUSLeeP, with a fake stimulation night first and the real stimulation night second.
Key takeaways
- Time to first REM fell from an average of 177 minutes on the sham night to 135 on the stimulation night, and REM rose from 16.3% to 20.9% of total sleep.
- The order was fixed for everyone, sham first and real second, which the authors say leaves placebo and order effects on the table.
- The device aims at the subthalamic nucleus, a movement structure, not at the brainstem where REM sleep is actually generated.
What the study found
The 103-gram patch combines a ring-shaped ultrasound array with hydrogel EEG, EOG and EMG electrodes, so it stimulates and records at once. Ultrasound ran in 10 blocks, 30 seconds on and 30 seconds off, about 10 minutes of stimulation in all.
Twenty-eight people enrolled, 16 healthy sleepers and 12 with insomnia. Two recordings were unusable because of noise, leaving 26 for the sleep analysis. Across those 26, REM went from 16.3% of total sleep to 20.9%, a 4.6 percentage point rise, or about 16 minutes. The chance that increase was a fluke was under 1 in 1,000 (p = 0.0007).
Time to first REM dropped from 177 minutes to 135. That was clear in the healthy group, with roughly a 1 in 60 chance of coincidence (p = 0.0165), but it did not reach statistical significance in the insomnia group.
Wake-ups did not change and the other sleep stages were not disturbed. Heart rate variability rose the morning after stimulation in healthy participants, and brain scans the next day showed shifts in emotion-related circuits.
“This is the first time we’ve been able to noninvasively target deep brain regions involved in REM sleep, while simultaneously monitoring brain activity,” said Kai Wing “Kevin” Tang, the UT biomedical engineering PhD graduate who led the work. Participants rated the patch as comfortable and safe, with minimal adverse effects, and the team has filed a patent application.
Dr. Kumar’s take
I have placed electrodes in the subthalamic nucleus. It is a lens-shaped structure a few millimeters across, buried near the top of the brainstem, ringed by the substantia nigra, the red nucleus and the fibers of the internal capsule. Getting a wire there takes imaging, a stereotactic frame and intraoperative recording, and surgeons still miss sometimes. “Noninvasive deep brain stimulation” here means a sound beam through the skull, and a focused beam at that depth is not a dot. It is an elongated cigar of energy, which is why the authors list “specificity in STN-targeting and off-targeting effects” among their limitations.
The mechanism deserves scrutiny for a second reason. REM sleep is generated in the pons, in the brainstem, not in the basal ganglia, and the subthalamic nucleus is a motor relay. The paper’s own imaging showed changes in the substantia nigra, the ventral tegmental area and the red nucleus, all neighbors of the target. That fits a beam reaching more than one structure.
The design issue is what would keep this out of practice. Everyone got sham on night one and stimulation on night two. In a sleep lab the first night is usually the worse night: people sleep badly in a strange bed with wires on, and REM in particular is suppressed. So the second night looking better is close to what you would predict with the device switched off. The authors are direct about it. Without randomization and blinding, they “cannot rule out placebo or context effects,” and they call for a double-blind, randomized, sham-controlled trial next.
More REM is a sleep architecture number, not a proven health outcome. Nobody showed less depression, less anxiety or better memory here. And 28 people over two nights says nothing about what nightly ultrasound to the brain does over five years.
What it means for you
Nothing to buy here yet. The device is a research prototype with a patent application, and the larger trials in PTSD, depression and insomnia are still being planned.
If your REM sleep is poor, the levers with evidence behind them are unglamorous: a consistent wake time, less alcohol in the evening, and dim light in the last hours before bed, since room light before bedtime suppresses melatonin. Heart rate variability also moves with plenty of things other than a brain beam, oxytocin among them.
Watch for the randomized trial. If a properly blinded study with the order shuffled reproduces a 43-minute drop in REM latency, this becomes interesting. Until then it is an engineering achievement with a sleep result attached.