Ninety minutes before their usual bedtime, 55 adults in Suzhou sat upright in a small shielded room, eyes closed, and did nothing at all for ten minutes. They came back the next morning, ninety minutes after waking, and did it again. Between the two sessions they went home and slept in their own beds. What the recordings caught, according to Feng Cui, Kejia Hu and colleagues at Suzhou Hospital and collaborating institutes, is that the brains of people with insomnia looked most unlike everyone else's in the evening, precisely when a brain is supposed to be winding down.
The tool was a 64-channel OPM-MEG, a magnetoencephalography system built from optically pumped magnetometers: small sensors that sit directly against the scalp and pick up the faint magnetic fields produced by neural currents. Conventional MEG needs a bath of liquid helium and a rigid helmet you must hold still inside. These sensors are worn, tolerate head movement, and sit closer to the head, which improves the signal. The team says this is the first time the technology has been used to track insomnia across an evening-to-morning cycle.
Twenty-six participants met DSM-5 criteria for insomnia disorder, with a mean Pittsburgh Sleep Quality Index score of 16.1 out of a possible 21. Twenty-nine controls averaged 3.7. The groups matched closely on age, sex, body mass index, education and cognitive screening scores, and everyone abstained from caffeine, alcohol and nicotine for at least a day before each session. An Apple Watch and a seven-day sleep diary checked that people actually slept when they said they did.
A spectrum that should tilt, and doesn't
The first measure the researchers looked at was the aperiodic exponent, and it is worth a moment. Brain recordings contain rhythms (alpha waves, beta waves) sitting on top of a broad background that falls away smoothly as frequency rises. For decades that background was treated as noise. Researchers now read the steepness of its slope as an indirect index of how excitable the cortex is: a steeper slope suggests more inhibition, a flatter one more excitation and more scattered, asynchronous firing.
In the evening, the insomnia group's slope was markedly flatter: an exponent of 1.54 against 2.19 in controls, a gap the authors describe as a large effect. By morning the two groups were statistically indistinguishable (1.78 versus 1.99). The direction of travel differed too. Controls' exponents drifted slightly upward toward the evening, consistent with a cortex settling down. The insomnia group's went the other way, lower in the evening than in their own morning readings. The authors read this not as a permanent trait of the disorder but as a failure to apply the brakes at the right hour. They are careful to add that the exponent's link to excitation and inhibition is indirect and can also reflect firing rates or synaptic timing, so they treat it as a correlate of arousal rather than a direct readout.
The second measure was microstates: brief, quasi-stable patterns of activity across the whole head that persist for roughly a tenth of a second before flipping to another. The team pooled every recording from both groups and both sessions, clustered them, and settled on five recurring patterns, which they then mapped approximately onto cortical regions. Two behaved oddly in insomnia. A state loosely corresponding to right temporal-limbic areas, tied to emotional processing, took up more of the evening in patients and lingered longer per visit. A state resembling sensorimotor cortex was elevated in patients at both sessions, covering about 30 percent of the evening recording against 9 percent in controls, an unusually large gap. The authors suggest it may reflect impaired sensory gating, difficulty tuning out the body, which fits clinical reports of patients lying awake acutely aware of their own heartbeat and limbs.
An exploratory mediation analysis tied the threads together: within the insomnia group, a flatter evening slope predicted longer dwell times in the limbic state, which in turn predicted worse sleep quality scores, accounting for about 39 percent of the association. The authors label this strictly exploratory and disclaim any causal reading.
Why it matters
Insomnia is diagnosed almost entirely by asking people how they slept. That is an awkward footing for a condition affecting roughly one in six adults worldwide, and it makes it hard to tell which patients might respond to which treatment. An objective marker measurable in ten minutes of sitting still would change that arithmetic, and the evening specificity here is the interesting part: it suggests the thing to measure is not a fixed brain property but a transition that fails.
The caveats are real and the authors list them plainly. Fifty-five people is a small sample. There was no polysomnography, so wakefulness during recording was confirmed only by video and by asking, though a flattened spectrum is the opposite of what drowsiness produces, which argues against participants dozing off. Without individual MRI scans, the anatomical labels on the microstates are approximate, and the authors say so repeatedly. Three co-authors work for the company that makes the scanner, though the paper states they contributed only technical support during recording. Nothing here is a diagnostic test yet. It is a fairly clean demonstration that a wearable scanner can see something in the pre-sleep hours that questionnaires cannot.