This article comes from one of my weekly livestreams, where I work through the neuroscience I am actively refining and take questions from the audience. The night I recorded this, I was setting up to demo HEG training and got defeated by my streaming software. So instead I talked through something I have been chewing on for weeks: why the blood-flow training I have used for 25 years actually works, and what three 2024 papers just told us about it.
What is vasomotion, and why does it matter for your brain?
Your cerebral blood vessels are active oscillators. Small arteries in your cortex contract and relax in a rhythm of roughly one cycle every 10 seconds, a frequency of about 0.1 Hz. That rhythmic squeezing is vasomotion.
This is slow compared to brain waves. Your EEG runs in cycles per second; vasomotion runs in cycles per 10 seconds. That slowness is one reason you can train it with low-resolution tools. You do not need millisecond timing to influence a 10-second wave.
Two things drive the rhythm. The first is the smooth muscle in the vessel walls. These cells run a calcium-dependent oscillation: calcium rises, the vessel constricts; calcium falls, the vessel dilates. The second driver is a small population of pacemaker neurons that coordinate the rhythm regionally. These are the nNOS neurons, less than 1% of cortical cells, releasing nitric oxide to trigger vasodilation. Astrocytes wrap around the vessels and add vasoactive signals through nitric oxide and prostaglandins.
The rhythm matters for two reasons. Fuel delivery is the obvious one. Waste clearance is the second. When vessels oscillate, they create pressure gradients. Blood volume drops, cerebrospinal fluid flows into the space around the vessels, interstitial waste gets pulled out, blood returns, and the CSF drains. That is the glymphatic pumping cycle, and vasomotion is one of its engines (Iliff et al., 2012; van Veluw et al., 2020). This connects to what I covered on biohacking plasticity and the nested timescales the brain runs on.
What did the 2024 vasomotion research find?
Three papers reshaped my model.
Vasomotion travels in waves. Broggini and colleagues (2024, Neuron) used two-photon imaging in awake mice to track arterial diameter in real time across a wide swath of cortex. The vasomotion was not local. Oscillation in one region triggered oscillation in the adjacent region, and the pattern propagated. They saw long-wavelength traveling waves sweeping across the cortex, spanning millimeters in a mouse brain. In a human brain those waves would be larger. The kicker: these baseline traveling waves dominate. They are stronger than the resting perfusion fluctuations tied to active tissue use. The background rhythm carries more signal than the moment-to-moment fuel demand. The NIH writeups described it as waves of blood flow washing across the brain's surface. That is not a metaphor.
This gives a handle on migraine. One model of migraine is spreading cortical depression, a wave of metabolic failure across tissue, the brain effectively cramping up from a shortfall of fuel. When the traveling waves of vasomotion are the infrastructure that prevents that, a failure of those waves becomes a plausible trigger.
Vasomotion is plastic and trainable. A second 2024 paper used mice and rhythmic flashing light. The vasomotion in visual cortex tracked the light, and after the light stopped, the vessels kept oscillating in the entrained pattern. The vasculature sensitized itself to the energy demand. It looked like long-term potentiation, the same wiring-together you see when two neurons are forced to fire together. The entrainment also improved clearance in the perivascular and paravascular spaces. Vascular oscillations can be trained the way neurons are trained. That finding changes how I think about HEG.
After stroke, vessels lose the rhythm. A 2024 group used two-photon imaging of smooth-muscle calcium dynamics. Healthy arterioles ran the 10-second oscillation. After ischemic stroke, even when blood flow was restored, some vessels stayed inert and never picked the rhythm back up. The shock of the stroke changed vessel behavior. This may be one reason stroke recovery stalls even when perfusion returns. More recent work continued to characterize the nNOS neurons as vascular pacemakers.
The thread across all three: rhythmic vascular activity is load-bearing infrastructure, it is plastic, and losing it correlates with disease and aging. Chronic stress blunts or kills those nNOS neurons, and you lose the amplitude and coordination of the whole system.
How does HEG neurofeedback train blood flow?
For 25 years I have worked with passive infrared hemoencephalography, pIR HEG. A sensor measures the heat radiating off the prefrontal cortex as a proxy for blood flow. People train their prefrontal blood flow up, and migraines drop, attention improves, and anxiety eases. The old explanation was vague: you are training prefrontal activation, or improving neurovascular coupling. Accurate, but incomplete.
The 2024 work gives a sharper model. The HEG protocol is repeated, spaced trials. You upregulate prefrontal blood flow, get feedback, rest, and repeat, in short bursts over 10 to 20 sessions. That structure is close to the entrainment protocols that trained vasomotion in the animal models. Over a course of sessions you are likely teaching the prefrontal vascular network to oscillate more coherently and respond more efficiently to neural demand, and possibly improving glymphatic clearance along the way. Label this as theory, not settled fact. It is an inference chain, and the links are plausible.
I run HEG with continuous feedback and voluntary effortful pushes. Dr. Jeff Carman runs it with thresholded feedback and longer baselines. Both work well. I tend to see stronger effects with the continuous, voluntary approach.
I prefer passive infrared over fNIRS devices like the Mendi or similar headbands. Passive infrared measures heat across a broad field, which I think captures vasomotion more reliably. Near-infrared devices use an emitter and receiver looking at red-light absorption as a proxy for blood oxygenation, which reads more like spot oxygenation. That difference may matter for results. Several people I have worked with got little response from fNIRS-style HEG and would likely do better with pIR.
The research literature on HEG is encouraging. Studies report reductions in migraine frequency and severity, gains in attention and executive function, and drops in anxiety and depression (Carmen, 2004; Stokes & Lappin, 2010; Dias et al., 2012). What changed is the why. You are training the rhythmic, plastic dynamics of the cerebrovascular system.
Why is HEG less stable than EEG training for migraine?
In my experience reading these cases, EEG neurofeedback tends to be associated with stable, durable change. HEG holds well for some complaints, including post-concussion, post-COVID brain fog, and brightening with brain aging. For migraine, I see less long-term stability. Migraines triggered by weather pressure changes, which is a vascular sensitivity, or by seasonal allergy and inflammation, tend to recur.
My working theory: once you are training the vasculature, you are into regulatory mechanisms the peripheral nervous system runs. Smooth-muscle control does not stabilize the way neuronal firing does. So I treat HEG as part of a full course of neurofeedback, then often suggest a second round, sometimes a year later, when something trips the migraine tendency back up. For a complaint as debilitating as migraine, a touch-up round is a small price.
There is a useful upside to the peripheral-nervous-system character of this training. In peripheral biofeedback, like heart rate variability work, you get skill transfer. You practice dropping into the parasympathetic zone with a device, then you reach for the same lever in traffic without it. I find HEG behaves similarly. People who have abolished migraines with pIR HEG can often catch one as it ramps up and shut it off. The vascular component seems more amenable to voluntary control than EEG, which fits if peripheral autonomic drive is helping shape vasomotion.
Does migraine risk show up in a QEEG brain map?
Sometimes. There is no single deterministic migraine signature, and you can have a migraine tendency with a clean EEG. When it does appear, I look for bilateral temporal lobe slowing, excess delta and theta, sometimes alpha, with raised amplitude of slow waves. A tired-looking temporal lobe often travels with metabolic complaints: brain fog, vestibular and eighth-cranial-nerve issues, balance problems, tinnitus, and migraine. My read, from looking at many concussion and post-COVID maps, is that this temporal slowing reflects the brain failing to draw up the energy it needs. You can learn more about what these maps show in my QEEG brain mapping guide.
Where HEG fits among metabolic-support tools
I group HEG with hyperbaric oxygen and photobiomodulation (red light) as metabolic-support techniques. They raise vascular and metabolic tone and bring energy to the system. They can also be too stimulating or fatiguing for some people, which is consistent with the idea that they are pushing the same vascular and energy machinery. EEG neurofeedback, by contrast, is more about information flow. Stacking HEG with EEG training tends to be associated with better outcomes on both migraine and vertigo.
Questions from the livestream
I take audience questions every week. A few from this session, with names removed.
ADHD with epileptiform discharges, do they interfere with each other? For neurofeedback, no. The same training addresses both. SMR training builds the inhibitory tone that the research links to the high theta-beta ratio of ADHD and the low-beta SMR that supports the brain's ability to resist seizure events. Subclinical epileptiform features show up in a meaningful minority of some neurodevelopmental, autistic, anxiety, and ADHD populations. Getting the brain in shape generally lets it resist those events. See my neurofeedback for ADHD guide for more on that pattern.
Do benzodiazepines reduce neurofeedback effectiveness? They appear to blunt it, not abolish it. The strongest blunting I have seen was in people on benzos or z-drugs like Ambien for many years, so I suspect the effect is more about duration than dose. At worst maybe half the impact, at best a 20% slowdown. That is my read of the patterns I have seen, not a controlled finding.
Can neurofeedback help developmental delay in kids? It depends what changes and what does not. My rule of thumb centers on language. Some brain tissues stay plastic because their job is to adapt to complex information. Others encode systems during a critical period and then lock down. Vision is one: without visual fusion early in development, you get drifting images for life. Language is another: without exposure by about nine or ten years, productive language does not develop. A six- or eight-year-old with no language is not a pattern I usually see change much. Receptive language, prosody, sarcasm detection, stuttering, executive function, attention, sleep, sensory and social processing, and even bedwetting are tissues that keep adjusting, and I see those move.
Are epileptiform discharges treatable with neurofeedback? Sterman and Egner's review of the seizure literature documents substantial reductions in seizure activity across studies (Sterman & Egner, 2006). My own estimate, from the cases I have read, is that roughly two-thirds of the seizure cases I have worked with reach full control, though it often takes two rounds of neurofeedback six months to a year apart. That last third has not generally returned with more seizures either. This is opinion, not a head-to-head literature comparison, but I think neurofeedback is one of the most impactful tools we have for seizure short of removing the tissue.
The practical takeaway
If you have migraines, brain fog, post-concussion symptoms, or attention complaints, HEG neurofeedback is a low-cost, simple addition to a full neurofeedback course, often paired with EEG training. Plan on a possible second round for migraine. The software stack for pIR HEG has gotten harder to source, and I am building a replacement I expect to release in alpha around the first or second quarter of 2026.
I will keep developing this model and likely do a follow-up tying vasomotion together with perfusion, photobiomodulation, and migraine specifically. If you want a brain map to see your own patterns, or to start remote neurofeedback anywhere in the world, reach out to my team at Peak Brain.