Scientists Invent a Literal Thinking Cap

Medscape

29-05-2025

This transcript has been edited for clarity.
Welcome to Impact Factor , your weekly dose of commentary on a new medical study. I'm Dr F. Perry Wilson from the Yale School of Medicine.
My job (my real job) as a clinical researcher is complex. It's cognitively challenging; there are multiple studies to keep track of, grants and papers to write, a large group of mentees and trainees and staff in the lab to manage. It's emotionally stressful too — recently more than ever, in fact. But if I'm tired, or I ate a bad burrito for lunch, or I get some bad news on a personal level, it's not a crisis. I'm not making life-or-death decisions in a split second. I can take a break, gather myself, prioritize, and come back when I'm feeling better.
Not every job has that luxury. A surgeon doesn't get to take a break in the middle of an operation if they feel like they are not at 100%. An air traffic controller can't walk away from ensuring that planes land safely because their kid woke them up in the middle of the night. These jobs and others like them have a unique challenge: a constant cognitive workload in a high-stakes environment. And the problem with constant cognitive work is that your brain can't do it all the time. If you force it to, you start to make mistakes. You can literally get tired of thinking.
Think of how the world might change if we knew exactly how overloaded our cognitive processes were. I'm not talking about a subjective rating scale; I'm talking about a way to measure the brain's cognitive output, and to warn us when our ability to keep thinking hard is waning before we make those critical mistakes.
We're closer than you think.
The standard metric for assessing cognitive workload is the NASA Task Load Index. Yes, that NASA. The Task Load Index is a survey designed to assess how hard a task is. It was originally designed to be used in human-machine interactions, like piloting a spaceship.
It's subjective. It asks you to rate how mentally demanding a task is, how frustrating, how much effort it takes, and so on. Cognitive researchers have used this scale to demonstrate how successive mentally stressful tasks degrade task performance. Science has demonstrated that taking breaks is a good thing. I know — news at 11.
The problem with subjective scales, though, is that people have a tough time being objective with them. Astronauts might tell you a task was easier than it really was because they want to be chosen to ride on the rocket. Or a doctor might evaluate a complex surgery as less mentally taxing so they can continue to operate that day.
Bringing objectivity to the brain is hard. Sure, you can do an fMRI scan, but sitting inside a metal tube is not conducive to real-world scenarios. You can measure brain fatigue in the real world with an EEG, though. The problem is that an EEG involves wires everywhere. You're tethered. And the goo, the sticky stuff that they use to put the electrodes on your head, is very sensitive to motion. In anywhere but a dedicated neuroscience lab, this isn't going to work.
I thought the day of real-time monitoring of cognitive load would be pretty far off because of these limitations, and then I saw this study, appearing this week in the journal Device, from CellPress.
It reimagines the EEG in a way that could honestly be transformational. There's a not-too-distant future when you'll be able to recognize people with highly cognitively intense jobs because they will look something like this.
What you're looking at is a completely wireless EEG system. The central tech here is what the researchers call an 'e-tattoo' — but think of it like those temporary tattoos your kids wear. Conductive wires are printed on a thin transparent backing which conforms to the forehead.
Electrodes make contact with the skin via a new type of conductive adhesive. The squiggles in the wires allow you to flex and move without breaking connections. That whole printed setup is made to be disposable; apparently the material cost is something like $20.
The blue square is the ghost in the machine, a processor that receives the signals from the electrodes and transmits them, via low-energy Bluetooth, to whatever device you want. It's got a tiny battery inside and lasts for around 28 hours. In other words, even in this prototype phase, you could wear this thing at your cognitively intense job all day. And yeah, you might get a few looks, but the joke will be on them when the algorithm says your brain is full and you need to take a 15-minute rest.
Of course, cool tech like this is only cool if it actually works, so let's take a look at those metrics.
The first thing to test was whether the device could perform as well as an EEG on a simple task. Six adults were recruited and wore the tattoo at the same time as a conventional EEG. They were then asked to open and close their eyes. There's a standard finding here that with eyes closed, alpha frequencies, mid-range brain oscillations, dominate.
You can see the patterns recorded by the standard EEG and the new tattoo system here. They are basically indistinguishable.
But the tattoo system, with its flexible design, offers some particular advantages. One of the problems with conventional EEGs is how sensitive they are to motion. You turn your head, you get a bunch of noise. Walk around, and the signal becomes useless.
Not so with the tattoo. These graphs show the electronic noise levels when the participant was doing various motions. Broadly speaking, you can see that the tattoo continues providing solid, reliable recordings even when walking or running, while the EEG goes all over the place with noise. The only exception to this was with eyebrow raising — maybe not surprising because the tattoo goes on the forehead.
But I didn't start off telling you we have a new flexible EEG tech. I told you we had tech that could quantify our cognitive load. Here's how they tested this.
In the lab, they had their volunteers do a cognitive task called the N-back test.
It starts at level 0. Basically, they ask you to click a button whenever you see the letter Q or something. Easy.
Level 1 is a bit harder. You have to click the button when the image on the screen matches, in either location or content, the image from one screen ago — one image back. Get it?
Level 2 is even harder. You click when the current image matches, in content or location, the image from two screens ago.
Level 3 gets really stressful. You have to click when you see something that matches three screens ago. And, of course, this keeps going, so you have to keep this information in your memory as the test continues. It's hard. It taxes the brain.
Here are the results on the NASA survey scale. This is what the participants reported as to how mentally taxed they were. As the N gets higher, the cognitive stress gets higher. So the system works.
The participants, you won't be surprised to hear, performed worse as the N increased. At higher N, the detection rate — the rate at which matches were appropriately clicked — declined. The reaction time increased. False alarms went up. All hallmarks of cognitive stress.
And the e-tattoo could tell.
Feeding its wireless output into a machine learning model, the researchers could predict the level of cognitive stress the participant was under.
They show the results for the participant where the system worked the best — a bit of cherry-picking, certainly, but it will illustrate the point.
The blue line indicates what level of the N-back test the participant was actually taking. The red line is what the machine learning model thought the participant was doing, just from reading their brain waves. They match pretty well.
Again, that was just the time the experiment worked best. The overall results aren't quite as good, with a weighted accuracy metric ranging from 65% to 74% depending on the subject. Clearly better than chance, but not perfect.
Still, these are early days. It seems to me that the researchers here have solved a major problem with monitoring people doing cognitively intense tasks — a way to read brain waves that does not completely interfere with the task itself. That's a big hurdle.
As for the accuracy, even an imperfect system may be better than what we have now, since what we have now is nothing. But I have no doubt that with more data and refinement, accuracy will increase here. When it does, the next step will be to test whether using these systems on the job — in air traffic control towers, in operating rooms, in spaceships — will lead to more awareness of cognitive strain, more rest when it is needed, and better decision-making in the heat of the moment.