#153- Sleeping in an MRI Machine

0:02Matt Walker: Hello, and welcome back to the podcast. It's Matt here. Today's episode is about something hidden in the drawer in your house somewhere. Foam earplugs. They're tapered, slightly gritty, the color of a highlighter pen. Roll them between finger and thumb until they're thin enough to slip into your ear, and then hold them there until they expand back out again and seal themselves in place. We've all used them before, right? On an airplane, next to a snoring partner, or at a concert we may or may not later regret going to. But what happens when you take that earplug with you somewhere completely different, into a room with a superconducting magnet in it, cooled with liquid helium, doing a brain scan?

0:47Matt Walker: There's a protocol we know about. 51 minutes at a sound level of between 103 and 111 decibels, your head inside the thing you're trying to image. At this volume, the earplug is no longer a luxury. It's a piece of safety equipment. And in this room, supine position, head immobilized in a cradle, with all of that going on around you, people fall asleep. They do it systematically, intentionally, as the basis for an entire scientific discipline. The first systematic review of the literature found 205 studies where people had their brain scanned while they were confirmed to be asleep, spanning 30 years and three different imaging methods.

1:30Matt Walker: So the question I want to talk about today is this, can you sleep at 110 decibels? Yes, you can. And everything interesting about the process is in the how. To get a person to sleep in the world's worst bedroom, you first need to understand which aspects of sleep are fragile and which are shockingly resilient. Once a person is under in there, the machine reveals something we can't see any other way, what falling asleep looks like from the inside. Before the sleep comes, the machine, because it's a lot younger than a lot of people think. The science rests on two foundational papers, both published in the early 1970s.

2:14Matt Walker: The first documented that the magnetic signal from tissue within a tumor was different from the signal of normal tissue. It wasn't an image. It was just a difference in the signal. The first image of a living subject followed. It was a clam. A few years later, scientists were able to run the first whole-body MRI of a living human. Barely a decade took the technology from clam to clinical practice. The first commercial scanner shipped in 1980, and the US FDA approved MRI for clinical use in 1984. MRI, or magnetic resonance imaging, is commonly used to scan bodily functions and structures like blood flow, tissue, and organs.

2:59Matt Walker: Think of an MRI almost like a camera. It can take photos or videos depending on what you want to look at. An MRI does two types of scans, functional and structural. Functional or FMRI scans look for blood flow and oxygenation and act more like a video recording. Structural scans are the photos and look at a specific organ like the brain or part of the body to see the anatomy. Beyond the walls of the clinic, that scanner is our main tool for investigating the brain, and the average study is really quite tiny. Imagine 12 people sitting around your kitchen table. That's the median number of subjects in the thousand or so most cited brain imaging papers published between 1990 and 2012.

3:44Matt Walker: Even in the latest years examined, even in the top tier journals, the median sample size had risen only to 23 or 24. Now we move on to describe the worst bedroom ever constructed. First, let's talk about the noise. It's the aspect that most people struggle to accept. 26 healthy young adults were scanned with that 51-minute protocol. They wore earplugs. They wore padding. Their hearing was assessed before within 20 minutes after, and then again 25 days later. After the scan, their hearing threshold had moved 5 decibels. By the 25th day, it had returned to baseline. There are two interesting follow-up points here.

4:30Matt Walker: First, it's temporary. It falls well below the 40 to 50 decibel shift that's been linked to changes in the auditory nerve. Second, the effect is still significant, and it occurs despite the hearing protection. But be cautious. Those numbers, 103 to 111, refer to just one protocol and just one machine. An acoustic survey of newborn brain protocols on five different machines found an average of 93 decibels, with a range of 81 to 105 over protocols lasting roughly half an hour to nearly 40 minutes. The reality is a broad spectrum. Sound levels between 80 and 110 decibels for anything from half an hour to nearly an hour are routine in brain imaging.

5:21Matt Walker: When you're in an MRI, it's not one of life's more pleasant experiences. You are typically lying on your back, which may or may not be how you normally like to sleep or fall asleep. You must keep your head still, otherwise the image can come out distorted. You are essentially enclosed in a tube with the researcher's voice coming through between sequences. Moreover, there's one of the most reliable findings from sleep studies. The first night in any new environment is noticeably poorer. Combine 53 studies on over 1 ,400 healthy volunteers from kids to the elderly, all studied over two nights. And you see that the first night is different from the second on nearly every variable that matters.

6:07Matt Walker: What you find is that people take longer to fall asleep, and there's more waking time in the middle of the night. Additionally, total sleep is down. Sleep efficiency is down. and there's more of the lightest, flimsiest stage of sleep during the first night compared to the second night. Supine position in a tube in an unfamiliar building at 100 decibels on night one. Pretty much every single variable stacked against you and yet it works. People reach confirmed properly staged sleep inside the magnet. 37 volunteers, all men, were wired up for a full sleep recording a generation ago, and between them held early night wakefulness, deep sleep, dream sleep, and morning wakefulness across a single night in a scanner.

6:55Matt Walker: Each person compared against themselves in groups of 7 to 20 for each pair of states. Before we go further, I want to land the running theme of this episode because it matters more than any single finding. Across those 205 studies I mentioned at the start, the median number of participants is 18, and around three quarters of them used healthy adults only, 18 people. Now set that against the 12 from earlier. Sleep imaging is actually running slightly above the historical norm of 12 for the field. It sits inside, and a little below the 23 or 24 of recent top journal papers. Either way, the norm is small.

7:40Matt Walker: The field knows this perfectly well. It says so in print, in its own review, and lists it as one of the central methodological challenges it faces. So, how do you actually get someone to sleep in an MRI? Also, how do researchers know when someone is asleep? Four problems, four solutions, and not one of the solutions is make the person comfortable. Problem one, on its own, the scanner can't confirm sleep. The machine measures changes in blood flow and oxygenation. And a relaxed waking brain and a lightly sleeping brain look similar enough on that signal that sleep research insists on the brainwaves as well.

8:22Matt Walker: So the electrical recording has to go into the magnet with you. A cap of electrodes reading your brainwaves while the scanner reads your blood. Of those 205 sleep and imaging studies, 141 paired brainwaves with magnetic imaging, 34 with an older tracer-based technique, 26 with a light-based method, and 4 that combined more than one. Anything that recorded the brainwaves and the pictures at different times was excluded because the whole point is that they come from the same moment. Problem 2 is the one I find genuinely lovely because on paper it shouldn't work at all. Picture trying to record a whisper next to a pneumatic drill.

9:07Matt Walker: Hopeless, except for one thing. The drill is identical on every cycle. So you record a few hundred cycles, average them into a template of exactly what the drill sounds like, subtract that template from every cycle, and what's left is the whisper. The fix was basically the drill trick. Average the artifact, subtract it, clean up what's left. After testing it in five people, the corrected recording was within 10-18 % of a recording made in the same scanner with the imaging switched off, and 99 % of test signals they injected into the data afterwards were recoverable, versus just 12 % without the correction.

9:50Matt Walker: Before then, you could get brainwaves inside the magnet, but not clean ones while the scanner was actually taking pictures. Even now, 25 years later, researchers still report significant differences between competing approaches. Simply put, the artifact isn't overcome, it's circumvented. The third issue is getting the subject to fall asleep. Here, the variable to manipulate is neither the environment nor the subject itself, but their sleep pressure. About 28 % of studies in the literature sent subjects in having had intentionally restricted sleep for the prior night, about half used an overnight design, and 20 % used a nap-only design.

10:33Matt Walker: The second half of the manipulation is just boredom. Look at 1 ,147 healthy control subjects whose resting state scans were pooled and analyzed, the basic scans where people are told to keep their eyes closed, stay awake, and let their minds wander. For a third of the population, they fell asleep in under three minutes. These weren't sleep studies. Nobody was told to fall asleep. Simply staying still and not being given anything to do is in itself a sleep induction method, and a pretty robust one. The final issue is one where this is all routine clinical practice, not research. And the answer is the oldest technique in human history.

11:19Matt Walker: The way to get a newborn to stay still while being scanned is to get a newborn who's sleeping. One option is to sedate the baby. Another is to feed the baby, wrap them up, and let nature take its course. A single center review of every scan of a baby under 4 months old over 13 years, 432 consecutive patients roughly split between the two methods found that feed and sleep worked in 95 percent of cases and produced images only marginally inferior in quality

11:57so for oh gosh it must be at least a decade now i have been an avid probably more like rabid consumer of protein shakes. But the hard thing, honestly, in the supplement space, if you really understand and do your diligence, is this critical thing called trust. It is very hard to find a brand that you can truly trust. That's when I came across the company Puri, and the really impressive founder, even though I'm sure it makes them less profit versus other quick and loose companies. Puri takes quality more seriously than any company I've ever seen before. They have so many offerings, gosh, protein powders galore, but lots of other additions, but that's what I was focusing on here.

12:50Why did I choose them? Well, it's really for four specific reasons. Number one, they are free from hormones. Number two, they are free from GMOs. Number three, they are free of any pesticides. And number four, and this is the critical confidence part, every single batch is third party tested for over 200 harmful contaminants, things like heavy metals, etc. All you have to do if you are a neurotic nerd like me, you can even scan the QRI code on the back of the product to get the third-party independent laboratory test results. That's how much trust and confidence they place in their product. By the way, if you want a recommendation, go for the dark chocolate-flavored protein powder.

13:41It's just exceptional. In fact, it's wonderfully addictive. I didn't just say that. Anyway, it delivers 21 grams of protein per serving. And for folks of my podcast right now, you can save 20%. So go over to puori.com forward slash Matt Walker, and that is spelt P-U-O-R-I dot com Matt Walker. So again, that is P-U-O-R-I dot com forward slash Matt Walker, and you will get 20 % off your product.

14:21Matt Walker: Next, we'll move on to talk about the switch that isn't really a switch. When we talk about falling asleep, we often talk about it like it's a threshold, awake, then asleep, with a firm and distinct line in between. The best direct evidence we have says the brain doesn't do anything like that. To see it, you need to watch two structures separately at the moment of sleep onset. One is the thalamus, which is situated deep in the middle of the brain, and acts as the gateway for almost everything your senses send up there. The other is the cortex, the wrinkled outer surface, where thought and experience occur.

15:02Matt Walker: No scanner has yet caught the timing of those two structures cleanly enough at the moment of sleep onset. For that, you need an electrode actually in the brain. And science has been able to do this. Patients who were evaluated prior to epilepsy surgery had electrodes inserted for medical reasons. They were not healthy volunteers. And this was not a scanning study at all. Think of a big house at the end of the night. The front door is locked and barred first. The lights upstairs remain on for a long time after that. Nothing else is coming in, but the rooms are still occupied and still active. That is what those recordings showed.

15:42Matt Walker: The thalamus, the gate, most frequently falls silent before the cortex, and by several minutes. On awakening, the asymmetry disappears entirely and the two return together in harmony. And it gets better. The cortex does not drop as one. And that same paper also found that changes in cortical activity during the descent into sleep were more extensive than previously assumed, with some areas still active while others had shut down. Imaging studies see the same thing from the outside. People drift between wakefulness and sleep rather than flipping a switch. All agree on the identity of the gatekeeper.

16:25Matt Walker: The deep central nervous system that drives alertness falls dramatically asleep in deep sleep and is reawakened in REM sleep. The authors hypothesize that this desynchronized fall from wakefulness may explain what we see in hypnagogic hallucinations, the images and that weird sensation of dropping you might experience in the moments before sleep. Part of the cortex still working, but no longer connected to the world. The MRI is one of the main instruments of human brain research. Its workhorse is what we call a resting state scan, and a third of people are falling asleep within just a few minutes of being in one.

17:07Matt Walker: That's a measurement problem sitting underneath a very large body of literature, which is precisely how the researchers themselves frame it. They call for wakefulness to be monitored as standard and for a serious reassessment of what might have been missed or invented in the resting state field. Once you are fully asleep, though, the scanner begins to show some interesting sleep structure. Sleep spindles, short bursts of electrical activity found in non-REM sleep, have a discernible blood flow signature. In a small group of people, spindles were accompanied by activity in the thalamus, in brain regions associated with emotions and self-awareness, and in the auditory areas of the temporal lobes.

17:52Matt Walker: Slow spindles also recruited areas in the front of the brain, fast spindles pulled in motor areas and the hippocampus, the brain's memory center. Two types of spindles, two different neural networks. Going even further, people have been able to get into deep non-REM sleep while in an MRI scanner. Everyone thinks they know about deep sleep. The brain calms down, activity levels decline, the lights go out. Only this last bit is wrong and it's been tested directly. 14 healthy volunteers were scanned while the analysis was aligned to individual slow brainwave rhythms, the big slow rollers of deep sleep.

18:33Matt Walker: These waves were associated with sizable increases in the frontal regions, in the midline of the brain, and in the deep structures below. There were no decreases at all. The author's interpretation is that deep sleep is not a passive state, but an active one. Another study using a different technology and involving many more subjects reaches the same conclusion. The tracer experiments showed that deep sleep leads to a selective deactivation of higher order integrative brain areas, while the lower order sensory areas maintain their activity levels, which caused these investigators to argue that deep sleep might not be a global reduction in brain activity after all.

19:16Matt Walker: This suggests an obvious question. What happens to sound? If you play tones to someone in non-REM sleep, there is still a response in the thalamus and the auditory cortex, just like in wakefulness. The brain is still listening. The only exception is the spindle wave. During a spindle, those responses get more inconsistent, almost as if the switchboard had paused every incoming call. When a sound instead elicits one of the giant single waves the sleeping brain throws up in response to a stimulus, the signal is enhanced and reaches far-flung frontal areas. Taken together, these findings suggest that sensory input is filtered before it reaches the cortex, especially during spindles, effectively disconnecting the cortex from the outside world.

20:09Matt Walker: That's the mechanism in a handful of people inside a scanner. The population scale evidence runs hard the other way, and both halves need saying together. Pool 11 newer studies covering more than 100 ,000 responses, with 25 earlier ones covering 64 ,000. And for every 10 decibel rise in nighttime noise, the odds of being badly sleep disturbed rise by a factor of around 2 for aircraft, 2.5 for road traffic, and 3 for railways. The internal gait works, but the external volume dose still matters. And then there's REM sleep, which flips the entire scenario upside down. In one study with seven subjects with sustained REM sleep, blood flow went up in the brainstem, the thalamus, and both emotional processing centers, but decreased in a large swath of the executive cortex above the forehead and in the areas involved in spatial processing and self-reflection.

21:06Matt Walker: Higher order and self-regulation circuits go down, emotion and memory circuits go up. The tracer data show the same picture, and the overall review shows that the pattern holds true across three decades. This is, at least roughly speaking, why the dream feels so vivid and immersive when you're in it, yet completely illogical once you're awake. The latest advance in imaging doesn't measure electricity, but fluid. Your brain sits in a bath of clear liquid, and that liquid moves. Using scanning fast enough to capture the body's own rhythms, it's now possible to simultaneously record electrical activity, blood flow, and that fluid in the same sleeping brain.

21:48Matt Walker: During non-REM sleep, a repeating pattern emerges, a sort of tide with three phases. Slow brain waves, followed by an oscillating change in blood oxygenation, then large-scale slow waves of fluid flow. A separate group using three simultaneous methods has shown that the relationship between them changes with state. When awake, electrical and fluid changes both predict blood flow changes. Asleep, the unidirectional flow breaks down and changes can be predicted by each other. Fluids and brainwaves become bidirectional and rhythms from the blood vessels themselves have more sway. Now, I want to be careful about that last claim.

22:29Matt Walker: We've talked about the brain's waste clearance system in depth on this podcast and portrayed it as a genuine question. Nothing here reopens the discussion. This is a claim about what we're able to see, not what it's for, and whether that's significant or useful is another matter. Overall, everything in this episode is based on small studies. 18 participants in a cohort that's mostly healthy adults engaged in an activity most of us don't do at home is the average study. But that's not an idiosyncrasy unique to sleep researchers. The most cited brain imaging papers of the 90s and 2000s had a median size of 12.

23:12Matt Walker: The MRI captures sleep, but only the kind that survives being observed. But here, at the end, is the takeaway. That earplug in the drawer. You already know what it does. It blocks noise, which is why it's useful on planes. But in the scanner, it does something different. It reduces 111 decibels to a level tolerable for an hour, and then it ceases to function. It doesn't make you sleep. Sleep comes from a debt you brought with you, from an hour tailored to your individual needs. And if those recordings hold for the rest of us, from a thalamus that locks the door before the rest of your brain realizes it's locked.

23:53Matt Walker: The plug controls the volume. The gate controls the rest. In the people we've been able to record, that gate usually closes minutes before the cortex follows. That may be why it's nearly impossible to catch yourself falling asleep. Thank you so much for listening. I'm grateful for your time and your attention. Take care. Enjoy your waking moments. and when the night comes, may you sleep soundly. Hopefully somewhere more comfortable than an MRI scanner. And until next time, bye for now.