Quick, to the rat cave!

A big part of the research in my lab focuses on how rhythmic sensory stimulation can influence brain function, a process known as sensory entrainment. For example, presenting light flashes and pulses of white noise about 5 times a second (a frequency known as theta, which is linked to memory processes) can improve learning in lab studies. Other labs have shown that entrainment at different frequencies can deepen sleep and may even reduce the buildup of toxic proteins in the brain. An exciting thing about this technology is that because it’s so noninvasive, it has the potential for easy use in daily life. Could we perhaps make a wearable that lets you entrain theta oscillations while sitting in class, to help you remember the material better? Or entrains slow oscillations at night to boost your sleep?

The current systems we use in the lab entrain brainwaves through pulses of light and rhythmic white noise. But there’s another potential way to do it: using vibrations applied to the skin, which is already possible with many wearable devices. Because entrainment happens as a result of the brain processing stimuli, there’s every reason to think that a wearable vibration device should provide an easy, non-intrusive way of influencing brain activity. To test this, I recorded some of my brainwaves while using an OpenHaptic stimulator attached to my ankle. I tested entrainment at two frequencies: theta at 5 Hz (associated with memory) and alpha at 10 Hz (associated with relaxation and mind-wandering).  To measure the effects, I recorded EEG using a Muse headset while I listened to audiobooks with my eyes closed. Results To look at how stimulation affected brain activity, I computed the power spectrum of the EEG--a graph which shows how much brain activity is present at each frequency. I compared the EEG from all of the “vibration on” periods to the EEG during the baseline periods, where stimulation was turned off. These graphs are shown below.When alpha stimulation was on (green line), I saw more alpha oscillations than it was off (gray line). The alpha stimulation also appeared to reduce the amount of theta oscillations between 4 and 8 Hz

The Muse has four independent sensors, and breaking down the data by electrode location it’s clear that there’s two things going on: alpha entrainment mostly increased alpha oscillations over the frontal areas, and mostly suppressed theta oscillations over the posterior areas (directly behind the ear)

The data for theta stimulation is similar--stimulation increased power in the broad theta band (4-8 Hz), peaking at around 7 Hz. Unlike alpha, the increases were most pronounced on the rear sensors.

Update 2/26 I also recorded some data in a similar protocol using a Fitbit Sense with 7 H stimulation. Much like with the OpenHaptic, we see a clear peak at the entrainment frequency--which suggests that commercial wearables can also be used for this kind of entrainment!

Conclusions

I think this is a pretty interesting proof of concept for sensory entrainment with a wearable device--suggesting that an unobtrusive, wrist-or ankle worn device can increase brain oscillations at targeted frequencies.

This also suggests that wearable devices might be help to produce some of the behavioral and health benefits associated with brainwave entrainment--such as increasing learning ability with theta stimulation. However, this is something that will still have to be tested, as stimulation with a wearable device is not exactly the same as the stimulation protocols used in labs.

OpenHaptic GitHub: https://github.com/nathanww/OpenHaptic If you get a lot of ads for wearables (like I do), you might have noticed that wearables that work by stimulating your sense of touch are suddenly A Thing. Many are geared to make you more alert or less stressed (like the Apollo Neuro, Cove, or Doppel), but others like Neosensory have other uses, like augmenting hearing or treating tinnitus. I’m always a bit skeptical about claims like this, but there is some interesting science in this area. For example, vibration stimulation has been used to reduce pain, suppress inflammation, and modulate blood pressure and fatigue  Haptic devices can also be used for other interesting purposes like sensory substitution, giving people the ability to experience artificial senses, or biofeedback. So needless to say, it’s an interesting field. But unfortunately there aren’t many good options for experimenting with haptic devices. Commercial haptic devices generally use proprietary patterns, with few options for customization, experimentation, or hacking. Devices like smartwatches and phones can also be used for haptics experiments ( I’ve done this before) but still very limited by their software and low-quality vibration motors.

All of this was the impetus for to design a purpose-built open source haptic device . This device allows me to write Python programs, which run on the internal microcontroller and control three vibration motors. Because the microcontroller is an ESP32, it can connect over Wifi and Bluetooth, and a companion Android app allows me to program and control it.

The OpenHaptic electronics schematic

The assembled OpenHaptic device, with Velcro upper arm/leg strap. The three vibration motors and MOSFETs are attached to the bottom of the ESP32 Feather board, the battery is placed on top, and the whole thing was wrapped in tape and glued to the velcro straps. By using the Android app, it’s possible to write your own haptic applications and download them to the board, where you have full control over the three vibration motors as well as the Bluetooth/wifi connections and internal sensors built into the ESP32. There’s also an API, which simplifies some tasks like synthesizing and combining vibration waveforms.

This system allows one to implement a lot of interesting programs. For example, it can run sensory substitution program that lets you feel Bluetooth signals, or some nerve stimulation programs designed to be (somewhat) similar to the ones the Apollo device uses to make you feel more alert or relaxed.

If you want to try OpenWave on your Wear OS Smartwatch, click here. There’s also a similar application called Neurostrobe which runs on Android phones here

A few months ago I started getting ads for wearable devices that claimed to reduce stress. I get a lot of ads for various kinds of neurostimulators , but these were interesting because they claimed to use vibration to influence the autonomic nervous system, and thus your mood, alertness, and stress levels.

While it might seem odd that something that’s essentially a vibrating watch could do this, there’s a surprising amount of research on it. Much of this is from studying vibration in industrial settings; like how the vibration of driving a truck affects fatigue. Notably, vibration appears to impact alertness, as well as the balance of the sympathetic and parasympathetic nervous system, and even inflammation (when applied to the vagus nerve) Some of these findings need to be interpreted with caution (there’s potentially a big difference between the kinds of vibration you’d experience while driving, and the kind of vibration you get from a thing strapped on your wrist, for instance). But I thought it was interesting enough to try experimenting with. This lead me to make OpenWave, an app for experimenting with vibration stimulation on Wear OS smartwatches. OpenWave lets you deliver stimulation at specific frequencies and lets you modulate one frequency with another frequency, which is a function found in some devices and which may help improve the effect. If you’d like to try OpenWave you can install it from the Play Store here or see the source code on Github here.

A few weeks ago, Caputron sent me a new tDCS device to try out and review--the LIFTiD brain stimulator.

So, what does it do?

Interestingly, a lot of Liftid’s marketing focuses on using it as a replacement for caffeine or other performance enhancers, and as a way to boost attention.  There is some research that supports this as a use for tDCS, although weirdly the study that Liftid’s website cites (comparing tDCS with caffeine) used a completely different electrode configuration than the Liftid actually uses.

Still, there is some evidence that the Liftid’s configuration can improve attention, particularly if you’re doing a task that’s monotonous and requires focus for long periods of time (say, driving on the freeway) . There’s also evidence that this kind of stimulation can make people faster (though not better) on a number of cognitive tests.

Still, the bottom line is that this is all still very much experimental . The tasks used in laboratory experiments are much simpler than the things we deal with day-to-day, and often don’t account for the potential negative effects of the stimulation (one example: a study finding that stimulating the prefrontal cortex reduced people’s scores on an intelligence test).

Entry-level tDCS

The experimental nature of tDCS is exactly what makes Liftid’s design stand out. Most tDCS devices are built for experimenting--they let you tweak the intensity, timing, and position of stimulation. Liftid’s interface consists of just one button, which when pressed will deliver 1.3 milliamps of current to the left and right dorsolateral prefrontal cortex for exactly 20 minutes.

This simplicity is appealing--when I was feeling sleepy I opted for the Liftid device because its fixed-in-place electrodes are a lot easier to set up than those of my other devices . But it also didn’t seem like it was working, and this is where the simplicity became limiting--would it work better with higher current, or a different electrode position? There’s evidence that higher current levels might work better--but Liftid doesn’t give you any way to try that. (An unrelated issue: it’s quite easy to put the device on upside down and thus accidentally swap the locations of the anode and cathode) The device also comes with a kit that’s designed to help you get started with tDCS, which includes electrodes, a dropper for salt water, a spoon for measuring salt, and a mirror to help  you stick the Liftid over F3 and F4. To start it up, you mix the salt water, wet two of the white electrode disks, and place them into the black electrode holders on the Liftid. You can then attach the device (a velcro strap across the back holds it in place) and turn it on. Overall, everything seemed comfortable and well-built.  The fact that this device uses sponge electrodes as its default (and only) option, is a nice safety feature that’s likely to reduce the risk of skin burns. An unavoidably tricky part of of using this (or any) device is ensuring the electrodes were positioned properly. The manual and website instructs users to put the electrodes just under the hairline, but this is likely a bit too low to adequately stimulate the frontal cortex. (Most studies and clinics use the F3 and F4 points, which are higher). Luckily, because the device uses sponge electrodes that can work through hair, I didn’t experience any issues using the device higher on my head than recommended.  It’s really important to stress this: the device will not work if the electrodes are not correctly positioned.

The Liftid with its removable sponge electrodes in (top), how the company shows you to position it on the forehead (middle).

Technical stuff

The Liftid’s electrodes and current output fall well within the accepted ranges for current and current density, and I didn’t note any major issues here.

A nice feature of the Liftid is its automatic current shutoff if the resistance becomes too high. This is something that’s found on a handful of consumer devices, and is an important way to prevent skin burns should an electrode be damaged or have a poor connection.

Like other devices that use a digital control, the Liftid’s current regulator was sometimes slow to respond to a drop in electrical resistance, resulting in momentary current spikes of a few milliamps. Current spikes at this level don’t seem to pose a safety risk, and the device was often able to prevent spikes by shutting down when resistance rose too high.  It’s unlikely that you’ll experience any spiking using the device normally, unless you attempt to take it off or put it on while it’s running.

Conclusions--a neat brain stimulator, but not for everyone

A lot of what Liftid wants to be--a one-touch device to enhance your brain--is simply not possible at the moment. There are simply too many remaining questions about the efficacy of tDCS in general, and the Liftid’s electrode placement in particular. For some goals--like boosting learning or athletic perfomance--you might be better off with a different device that allows you to place the electrodes in more optimal locations.

(Want to play the game or see the code? Click here) When I read Ender’s Game, one of the parts that most stuck with me was the delightfully creepy Mind Game--a game designed to probe the player’s subconcious. In the book, we learn that the Mind Game isn’t actually programmed--it’s powered by an AI that makes up the gameplay as it goes, reacting to the player’s decisions and getting progressively more surreal.

Ender’s Game was written in 1985, and since then AI has actually gotten pretty good at creative tasks since then. The real-time graphics of something like the Mind Game are probably still out of reach, but what about something simpler like a text adventure? I decided to try writing a game like this by training GPT-2, a state-of-the-art predictive text model, on some transcripts of classic text advetnures. These look something like this: YOU STAND IN FRONT OF THE DOOR

>open door THE DOOR OPENS TO REVEAL A CLUTTERED OFFICE.

GPT-2 is a predictive text model, which just means that it tries to predict what comes next after some text that you enter. That means if you give it >Open door It will try to predict what happens next, based on it’s training data. Let the user choose their next action based on the response, and you have the makings of a text adventure game. This makes for a playable, if thoroughly weird, text adventure YOU ARE STANDING IN A GRASSY FIELD WITH A PATH ALONG THE STREAM, AND A SMALL CRACK ACROSS THE FLOOR AT THE BOTTOM OF A SMALL PIT

>follow path THE PATH HEADS NORTH-SOUTH HERE, AND APPEARS TO BE A STRAIGHT PATH

>continue ROAD NEAR BRIDGE cross the bridge CROSS THE BRIDGE AND YOU WILL FALL INTO A PIT

>fine,go into the pit FINE,GO INTO THE PIT

>Look around LOOK AROUND, AND STRANGE SHAPES (APPARENTLY MADE BY AN ARTIST WHO NEVER WAS IN THE ROOM) ARE SEEN ALL OVER

>talk to the shapes ARGUE WITH THE SHAPEK FAN, WHICH APPEARS TO BE HANDS-FREE. THE CHEER GOES UP! THE HEAVENS ARE NOW WIDE ENOUGH TO ALLOW FOR SINGING >I'm scared I'M SCARED THAT YOU HAVE DONE SOMETHING DELIBERATELY

>Get out of the pit YOU ARE IN THE PIT, SO TO SPEAK, BELOW THE RAMPARTS THERE IS A NARROW PASSAGE FROM EAST TO WEST AND A STONE STAIRWAY LEADING DOWN THE ROOM IS DEAFENINGLY LOUD WITH AN UNDETERMINED RUSHING SOUND THE SOUND SEEMS TO REVERBERATE FROM ALL OF THE WALLS, MAKING IT DIFFICULT EVEN TO THINK >go down YOU ARE IN THE PIT

>go down the stairway GO DOWN THE STAIRWAY IS BARRED BY EVIL SPIRITS, WHO BY SOME MIRACLE OF ZORKIAN TECHNOLOGY HAVE STOPPED THE FLOW OF THE RIVER

>kill the spirits THE SPIRITS, SENSING A GREATER POWER, FLEE THROUGH THE WALLS

>go down the stairway THE PIT IS TOO NARROW FOR MOST INSECTS (Even after defeating the spirits, I was still unable to get out of the pit.  I was warned I guess) If you want to play the game you can do it online here! There are some interesting things going on here. First, GPT is lifting elements from some of the games in its training set (i.e. Zork) and sort of mashing them up and putting them in new places. It does this surprisingly well, i,e, the spirits reacted to me in a reasonable way. What’s the goal of this game? Exploring, I guess. Even the game is making it up as it goes along, and that’s what makes it interesting Because none of the logic is explicitly programmed, you can do some other cool things, like making your character start in space for example, instead of a meadow, GPT2 will roll with it, though the training on classic adventure games means that it will occasionally try to fill your spaceship with moss. Technical stuff There were a few interesting trends I noticed while programming the game

One of the coolest phenomena in sensory neuroscience is sensory substitution, or the brain’s ability to make sense out of non-natural sensory information. For instance, when an algorithm is used to convert video into an audio signal, people can learn to use the generated audio to navigate and pick up objects—and even activate the visual cortex while doing so! This ability is likely owed to the brain’s remarkable ability to detect statistical patterns, which allows it to quickly work out how patterns of tones correspond to the spatial features of the environment. Amazingly, this ability can extend even to senses that we don’t have; for instance one team found that people were able to improve their navigation performance when fitted with a device that communicated the direction of magnetic north through vibration. After six weeks of training, people reported not only learning to interpret the vibration signals but also developing a new intuitive sense of spatial orientation; tests of eye movements also showed that some subjects integrated information from the vibration signals into their automatic eye-stabilization reflexes.

The possibility of augmenting our sensory systems has become particularly exciting now that most people carry around a phone constantly connected to the internet and loaded with sensors capable of detecting everything from infrasound to ultraviolet. I recently started experimenting with this technology using a smartphone and a Wear OS smartwatch to actually deliver output (vibration) to my wrist. As a first “sense” I wanted to add, I developed a program that used the phone’s radio to detect the beacons broadcast by Bluetooth devices and translate them into specific vibration patterns that were played on the watch. While using the system initially just felt like weird buzzing on my wrist, after using the system for a few weeks I began to notice something like what the subjects in the magnetic-north trial did. I started to develop an intuitive ability to attribute the signals to specific device; to notice people moving behind me based on their Bluetooth transmissions, and to recognize specific people from specific Bluetooth patterns. Building a navigation app

While this was cool, it wasn't necessarily useful. To put together something more practical, I decided to build a smartwatch-based version of Feelspace’s original navigation app. The app uses the heading derived from GPS measurements to generate a vibration signal which corresponds to the cardinal direction you are moving, where the vibration frequency decreases as the travel direction goes from south to west to north to east.

If you want to try this app, you can download it here! You’ll also need to download Sixth Sense which provides the utilities to actually generate the vibration.

I also decided to convert the code that handles communicating with wearable devices and generating vibrations to a separate app (called Sixth Sense), which provides services that can be used by any app that wants to uses sensory substitution functions. Interfacing with wearable devices can be complicated and annoying, so having a dedicated “output” app allows many “sensing” apps to easily generate and execute vibration patterns without needing to know about the lower-level details. In addition to permitting other developers to easily write apps, this means the Sixth Sense system is platform agnostic—for instance, it can generate vibrations on a Wear OS smartwatch if one is present, but also use the phone’s internal vibration motor.

Building new sensory substitution apps

This section talks about developing apps that work with Sixth Sense. If you’re not interested in that it’s probably not worth it.

Sixth Sense allows any app to generate vibration patterns on a wearable simply by sending a broadcast as shown below

Intent intent = new Intent(); intent.setAction("neurelectrics.sixthsense.SEND_PACKET"); intent.putExtra("sequence", "50,0,50,0"); intent.putExtra("priority", 1); intent.putExtra("duration", 100); intent.putExtra("expires",System.currentTimeMillis() + 1500); sendBroadcast(intent);

This command will cause whatever device the user has selected to output vibrations to vibrate according to the parameters passed in extras described below. All parameters are required. Sequence: A String of arbitrary length, containing the vibration intensity at each time point in a comma-separated list. Vibration intensity can range from 0 (motor off) to 255 (maximum). Not all output devices support variable intensity, and these devices interpret any value above 0 as “on”. Therefore it’s important to make sure that important info is not encoded in sequences like “50,100,50,100”, which will just produce a continuous vibration on devices without intensity control. Priority: An int larger than 0 describing how time-critical this output is. Outputs with a smaller value will preempt outputs with a higher value. Because multiple apps can be sending data to SixthSense at the same time, an output can be preempted by an output from another app.

Duration: An int specifying how long each time point in the sequence lasts. This value must be less than 255. Longer point durations can be achieved simply by duplicating each point (i.e. a 500 ms vibration could be generated with a sequence of “100,100” and a duration of 250)

Expires: long value describing at what time this output should be discarded if it has not been transmitted, with the time specified in the format of System.currentTimeMillis(). Because outputs can be delayed due to system lag or preemption by higher priority outputs, the expires value allows you to control when the data is considered “stale” and should not be transmitted.

There are also a few other things to keep in mind when building sensory substitution apps.

Minimize memory load: While it’s possible to encode information in complex ways (i.e. intensity ratio between two pulses), these are difficult for users to learning because they require the user to remember the length of the first pulse while waiting for the second). Simple manipulations (like changing the frequency) are easier to learn.

Provide an option to control the speed of vibration sequences. Sequences can be easier to learn when they are initially presented slower.

Hey! So I haven’t posted on this blog for a while (grad school more or less took over my life), but I’d thought I’d write a little about some of the experiments I’ve been doing looking at whether smartphones can measure brain activity based on blood flow and optical plethysmography. Optical plethysmography is a term for measuring blood volume in the body by looking at light transmitted through tissue: it relies on the fact that blood is pretty good at absorbing light so when blood vessels are “fuller” more light is absorbed. Plethysmography has actually found a lot of application in smartphones already, primarily to measure heart rate--a number of apps use the phone’s camera and flashlight LED to provide pulse measurements based on oscillations in light absorption. This method is potentially useful for looking at brain function because when areas of the brain are activated the local blood vessels dilate (through a process called neurovascular coupling) to deliver more oxygen to the area. Dilation increases the amount of blood present in the tissue, which is a signal that could theoretically be detected by photoplethysmography. In fact, researchers have reported using photoplethysmography to measure blood volume in the brain however the experiment involved artificially manipulating blood pressure rather than measuring the natural hemodynamic response during a cognitive task. To test whether phone-based photoplethysmography could be used to detect changes in blood flow due to brain activation, I wrote an Android app for my Galaxy S7  which determines the average brightness of red light in the camera’s field of view. When pressed against the skin with the camera LED on, this yields a fairly uniform image.

An example of what an image from this setup looks like.

The app averages red values over a period of 250 milliseconds (so the final measurement rate is four measurements per second) and transmits them to a server. This arrangement makes it easy to stream data from the phone: any app that knows the phone’s “stream id” can connect to the server and retrieve realtime readings. With the phone and streaming system set up, I wanted to know if the data being streamed could actually be used to measure cognitive brain functions, As a first test, I decided to use an oddball protocol: a simple cognitive task where a subject hears two beeps, one common and one rare, and hearing the rare beep requires the subject to make a response. The oddball protocol is a good test of a brain imaging method because the rare stimulus reliably evokes a larger response over much of the brain (particularly the frontal cortex) than the common stimulus. To determine whether the rare beeps produced a larger response than the common beeps, I used an event related design (a common method in EEG and fMRI studies) to look at changes in brain activity immediately after the beep occurred. I placed the phone  with its LED and camera approximately over the Fp1 point.

Using EEGLAB to compute event-related changes, I found that the rare beeps produced a larger brain response than the common beeps!

Right click and select "view" to get a higher resolution image Response to the “typical” sound (top) and the “oddball” sound (bottom) in terms of the change in intensity of red light detected by the camera’s sensor. The bottom part of each plot shows the average response in milliseconds relative to stimulus onset, the top part shows the response in each trial. All scales are the same between plots. Note that the oddball stimulus evokes a stronger and more consistent response around 2000 milliseconds.

Average responses superimposed. The green trace is the oddball response, blue is the typical stimulus response. Polarity here is inverted relative to the previous two plots. Note that  the absolute magnitude of this change is very small; a camera sensor measures redness value on a scale from 1-256, while the change in value is only 0.8 points. This is to be expected; most of the light emitted from the LED is absorbed or reflected before it reaches the brain and passes back to the sensor. Still, I was kind of suspicious that other factors (like movement of the sensor on the head) might have contributed to this signal. As a control, I did another experiment which  presented both tones with the same chance (each presented 50% of the time) and where I responded to every tone.. This provided a way to isolate the signal due to the oddball effect from other signals like the brain’s recognition that “something happened” or artifacts due to movement. In these data, the spike that occurred 2 seconds following the oddball stimulus was not present, although a decrease in the red value occurred at longer latencies. This suggests that the early positive spike may represent a true oddball response while the later decreases represent a more general effect (i.e. of planning or executing movement).

Response to all tones in the control experiment, with the same scale and units as the plots above.

Another thing I was curious about was how well this method could measure brain function over longer periods of time. To test this, I positioned the camera approximately over the left dorsolateral prefrontal cortex (an area involved in memory) while I performed working memory tasks. By varying the difficulty of the working memory task (switching between a dual 1-back and dual-3-back task), I was able to vary the working memory demand while keeping other factors (i.e. movement) fairly constant. Theoretically, this should result in increased engagement for the higher working memory load, and therefore a stronger signal compared to a no-task baseline. However, a lot of factors can interfere when recording over a period of several minutes—things like a decrease of signal quality or change in heart rate potentially create “artifactual” differences between conditions. To reduce the impact of these variables, I exploited the fact that photoplesymogram signals contain a “high frequency” oscillating component (due to the pulse) as well as (hopefully) a low frequency component corresponding to the brain cerebrovascular response. In theory, differences between conditions due to changes in signal quality or other non-brain factors (like movement and muscle tension) should affect both the high and low frequency signal. By using a technique called multiple regression, we can determine how much the overall photoplethysmography signal differs between conditions. Because we assume that the high frequency signal is related mostly to processes we are not interested in, we can also ask “what amount of difference between conditions is NOT attributable to difference in the high frequency signal?”. The differences that can’t be attributed to this signal are the ones we are interested in because they more likely represent true brain activity. I also applied the same technique to eliminate differences between conditions that were due to motion or posture (by streaming accelerometer readings from the phone and including them as covariates in the same way as the high frequency signal power was included.)

Across 6 trials (3 in each condition) the difference between the levels of memory load was statistically significant (p=0.049), and highly significant when looking specifically at a signal smoothed with a moving average (p < 0.001).  There was a consistent trend toward decreased red light detection (indicating higher absorption and greater blood volume) in the high-load condition compared to the low-load condition.

Representative example of low pass filtered red value over time in dual-1-back and dual-3-back conditions. The data is low pass filtered because a large amount of the signal consists of high frequency oscillations due to heartbeat.

Discussion

These results suggest to me that there may be something going on here. Is it conclusive? No. There are still potential sources of error to be ruled out, and I’ll be working on some experiments over the next few months to do so. However I do think it is interesting enough to be worth discussing.

There is also a lot of room to improve this technology in the areas of data acquisition, processing, and statistical analysis. To facilitate this, I’ve released several components as public-domain software, including:

*Physphone, an Android app for acquiring and transmitting data

*Biostream, the software that facilitates streaming data

*Python data acquisition and statistics utilities that connect to Biostream To use the app, you can download it from here. Data is streamed to the server, and can be retrieved using the stream ID generated by the app. To retrieve data, simply send a request to http://biostream-1024.appspot.com/get?stream=<stream id>. You can also view the source code for all these components on GitHub.

I’ve been intrigued by the Thync device ever since I heard one of its scientific leaders (William Tyler) speak at a conference on brain stimulation, but between its rather expensive price point and a lack of Android compatibility I’d had a lot of trouble getting my hands on one--until John Humphrey at DIY tDCS sent me a unit to try out! Over the last few weeks I’ve been putting the thing through its paces and found some interesting things about how (and how effectively) it works.

How does it work?

The Thync device actually consists of two parts--the Thync module and an adhesive ‘strip” that both passes current through your head and physically holds the module on. The device is attached at two points--the module sits over your left forehead, while the other end of the strip attaches to either the back of your neck (for the “calm” mode) or behind your ear (for the “energy” mode). These two connections complete a circuit that allows the device to send its “vibes” through your head.

Thync headset in the “energy” configuration

An interesting consequence of this design is that Thync is very spatially non-focal. Although the company claims to target specific nerves, there is probably quite a lot of “off target” current interacting with various other systems, including current passing through the brain directly. All this means that, it’s wise to take Thync’s claims about how their device works by stimulating specific nerves with a grain of salt. The actual current that the device is sending through the head is a current-regulated set of rapid pulses. It’s a high frequency alternating current signal that is in many ways similar to the one generated by a TENS device used to treat pain (although the pulse frequency of the Thync device is much higher, typically somewhere between 1 and 10 kilohertz depending on the “vibe” being used.)

The other difference between the Thync and other nerve stimulators is that the “vibe” signals are extremely complex. While most nerve stimulators use a single, repeating pattern the Thync vibes are not only all different from each other but also vary continuously throughout each session.

One way to deal with this complexity and figure out how the vibes actually work is to figure out what’s different between vibes with supposedly opposite effects--the “awake” and “sleep” vibes, for instance. One difference here is anatomical--the vibes that are supposed to make you calm, like the sleep vibe, use an electrode placed over the back of the neck, while the “energy” vibes place the electrode behind the ear. There’s also an interesting difference in the signals the device uses though:  the calm vibes seem to be significantly “burstier” (their power varies more at high frequencies) than the energy vibes. This property is interesting in light of Thync’s (largely hypotheticall) interpretation of how how input to sensory nerves might affect arousal level, which focuses on how these inputs might cause norepinephrine-containing neurons in the locus ceruleus to switch between constantly-firing states (associated with high arousal and energy) and intermittent firing (associated with lower arousal). It’s therefore possible that the “bursty” input from the Calm vibes is designed to encourage intermittent firing, while the more constant input from Energy vibes is intended to increase constant firing. (Keep in mind that while I think this is plausible, it’s just my interpretation and not necessarily correct)

Bursty signal (top) vs continuous signal (bottom)

Vibes also have a number of other interesting parameters however. The power and rate of nerve-stimulating pulses delivered by the vibe fluctuates throughout the session (these changes are slow enough that I suspect that they’re related mostly to comfort and reducing habituation rather than having a direct functional role) and different vibes use differently shaped waveforms, something that might be useful in tuning stimulation to affect specific subpopulations of nerve fibers (you could even, theoretically, use this property to make the Thync only activate the nerve fibers under one of the electrodes and leave the ones under the other electrode alone).

These vibes also have another property--they  sometimes include what is essentially tDCS, or more specifically an AC waveform superimposed on a constant DC current. Unlike most nerve stimulators (where the flow of current one direction through the circuit is balanced by flow the other way), the Thync device uses unbalanced waveforms where the total current flowing one way is higher than the total current flowing the other--creating a DC current which can be quite strong (up to 5 milliamps at the device’s maximum power!)

An example Thync waveform with a DC offset. The positive current (arbitary units) is greater than the negative component, creating a net flow of charge.

Oddly, the tDCS doesn’t seem to have a clear functional role. In all cases, the forehead-mounted part of the device acts as the anode of the circuit--something that would make sense if, for instance, Thync was trying to replicate this protocol which showed caffeine-like effects from frontal stimulation . But the device delivers the same kind of tDCS when you turn on the “sleep” vibe, a somewhat strange design decision if the tDCS was what was supposed to keep you awake. Instead, I suspect that the use of tDCS is sort of incidental--the device uses a waveform that is optimized for something else, and just happens to have a net flow of charge.

The takeaway

One of the big takeaways here should be that the Thync’s mechanism of action is very broad. In addition to acting on the targeted cranial nerves, the fact that current is allowed to diffuse through the head means that it is potentially acting on a number of other mechanism, such as direct current stimulating and high frequency alternating current stimulation/random noise stimulation acting directly on the brain. That is not bad per se--lots of interventions,particularly those targeted at the brain, have complex and poorly understood mechanisms of action. But the complexity of Thync’s mechanism of action also leaves a lot of unanswered questions--could it produce subtle cognitive deficits, as some forms of tDCS seem to do? Does it affect inflammatory processes in the brain?  Can it alter brain plasticity? Answers to these questions seem to be critical for informed use of the device.

But does it work?

All these potential mechanisms of action are, of course, contingent on the assertion that the Thync device actually does what it says it does--makes you calmer or more energetic, depending on which vibe you’re using. So for that, it’s worth taking a look at the evidence behind the device. The idea that pulsed electric current can influence energy and mood is actually not a new concept--its modern form, which bears some interesting similarities with the techniques used by the Thync device, was first developed by European and Soviet scientists as a way to reduce the need for chemical anesthesia during medical procedures. This technology was later adapted to forms known as “electrosleep” and “cranial electrotherapy stimulation”, which found some success in treating arousal disorders like anxiety and insomnia. More recently, sedative effects have been reported from the Cefaly nerve stimulator,  an experiment that is interesting in that it is one of the only cases where the sedation could be attributed to stimulation of a particular nerve (in this case the trigeminal).

Early "electrotherapy” protocol, somewhat reminiscent of what the Thync does.

Source:https://www.alleviahealth.com/wp-content/uploads/2014/06/Zaghi-et-al.-2009.pdf This means that the claim that the Thync device can induce calmness or help sleep is quite plausible. However, because of the variations in protocols used by these devices, it’s important to look at efficacy of the Thync device specifically, something that Thync did in two blinded studies.

The first of these studies , looking at the immediate subjective and physiological effect of Thync’s calm vibes,  found a number of changes indicating that it decreased physiological arousal. Most importantly for users, the researchers reported that the Thync reduced reported anxiety on the Profile of Mood States. The second study found that using Calm vibes before bed resulted in a significant increase in sleep quality, as well as an improvement of several other aspects of mental health likely due to the better sleep.

While these results suggest that the Thync calm vibes can reduce physiological arousal, for a consumer device it’s important to take into account the practical significance of an effect, not just the statistical significance.  The reduction in stress seen on the POMS, while statistically better than the placebo, is still quite modest--a reduction of 0.37 points on a 36 point scale. In practical terms that’s a very small effect--for instance it’s slightly smaller than the baseline difference between relaxed men and relaxed women. 

While the effects on sleep quality are more dramatic (including an increase of about 20 minutes in the nightly  time asleep that likely drove several other improvements in mental health )  I’m not convinced that using the Calm vibes for an immediate effect (the way they’re advertised) would produce any real noticeable effect. As for the Energy or Fitness vibes, Thync has produced no research no them whatsoever.

Impressions

There are a lot of things about Thync that made me go “wow”--the technology, the vision of affective and cognitive enhancement through electrical signalling, and the prospects for what technology like this could tell us about how the brain works.

What didn’t particularly wow me was Thync as a consumer product.  While it definitely has a sci-fi coolness to it, there’s a question here that’s important to address: Is the Thync more effective in any way than breathing exercises, listening to music, taking a walk, meditation, or any of the dozens of other methods for modulating arousal that don’t require sticking a $200 machine to your face? Given the modest effects of the calm vibes and the unknown effectiveness of the energy vibes, I’m not sure it is.

Another aspect of the “is it worth it” calculation is risk, and I think this is one of the biggest ways that Thync’s entry as a consumer device is premature. There is of yet no universally-accepted standard for showing that a device like this is safe--for instance,should there be screening just for obvious problems like skin burns, or also for more subtle ones like depressive symptoms?--and our incomplete understanding of how the Thync actually affects the brain makes it even harder to tell what adverse effects we should look for. That’s not to say that the Thync device is dangerous (so far there have been no reports of serious adverse effects), but the lack of good safety standards and a solid mechanism of action means that adverse effects we simply haven’t detected yet are a very real possibility.

None of these are intractable problems of course. Thync could always do more research, or publish research that they’ve already done, and they have an opportunity to play a major role in the project of figuring out how we should think about the safety and effectiveness requirements for these devices. But for now, the Thync device reminds me a lot of what Dr. Tyler said about neurostimulation technology at NYC neuromodulation--that much of it was a “neat lab trick looking for a way out”. So far, I don’t think they’ve found it.

Other observations

In addition to its nerve-stimulating effects, the Thync app makes heavy use of psychological suggestion. Notice the number of times that the voiceover says something like “when I use this vibe, I feel <x>”--this is a classic technique that’s used (by hypnotists, for example) to actually induce specific sensations. That’s not to say that the Thync relies entirely on this kind of suggestion ( the placebo-controlled studies show that there is some effect directly attributable to stimulation), but the suggestion might be just as powerful (or even more so) than the effects of the stimulation.

Thync has two “generations” of Calm vibes: the first uses higher frequency stimulation, while the newer generation uses a lower frequency. This is likely due to research showing that the low frequency stimulation was more effective.

Unlike, say, tDCS, you’re supposed to be able to feel a sensation from Thync--in the target range it feels like tingling, and if the power is turned up too high you get a dull burning sensation. One implication of this is that placebo stimulation can feel different from real stimulation and studies need to be very careful to make sure their blinding is actually effective.

It’s entirely possible to use the Thync with custom or third-party electrodes so long as they can snap into the two snap connectors on the bottom of the Thync module. The Thync won’t detect it as a calm or energy strip, but you can use any vibe with the custom electrodes.

The actual effect of the Thync vibes on POMS scores is surprisingly small compared to the glowing anecdotal reviews that people give of them. This is mostly likely because people are reporting placebo effects, but it could also be related to flaws in the experimental design. One interesting thing is that  the subjects who got a placebo were also quite relaxed, suggesting that they might have been at their individual “floor” of anxiety before the treatment or that the placebo might have actually had some active effect.

Recordings

To download recordings of the Thync vibes,click here

I’ve made recordings of the complete Thync vibes for anyone who’s interested in taking a look at them. Because of the frequencies that the Thync device operates at, audio recording and analysis software is well suited to recording and analyzing the vibes; in this case I used an external USB microphone input to capture the voltage drop across a resistor. The values in this file represent the current output by the Thync device at 2% power. Due to the limitations of the audio card, high-amplitude vibes had to be recorded with a different resistor than low-amplitude vibes--this means that for files prefaced with “100” the scale is -4.35 mA to 4.35 mA, while for files prefaced with “33” it is -13.18 ma to 13.18 mA. For files prefaced with “473” the scale is -0.92 mA to 0.92 mA.  In all these recordings, the Thync output closest to its USB port is connected to the “terminal” input of the microphone TRS connector, the other output is connected to the sleeve.

A while ago, I was asked by the makers of TheBrainDriver to write a review of the device. Being curious about this device myself, I got a review unit...and prombptly fogot about it for a while while I flew around the country for interviews. But now that that's mostly done, I'd like to share my impressions on this device!

Opening the box

The first impression that I got of the BrainDriver was that it looks sort of like a prototype iPod. Unlike most consumer tDCS devices, there are no physical controls for current intensity, etc. on the device. Rather, everything is controlled through menus displayed on the device's screen. While the device feels cheap when you pick it up (which, objectively, it is), it also seems sturdy enough that it's not going to crack apart if dropped a few times.

One of the most pleasantly surprising parts of the design is the sponge electrode and cable assembly that comes with the device. Many “entry-level” tDCS devices either don't ship with electrodes or use self-adhesive electrodes that only stick to bare skin and are therefore useless for most tDCS montages (unless you're bald or willing to shave your head). TheBrainDriver instead ships with several round, ~13 square centimeter electrodes , a pair of silicone electrode holders, and a stretchy head strap.

BrainDriver included electrodes. The red and black leads can actually be unplugged and used with other electrodes that accept approximately the same size of pin.

The electrodes connect to the stimulator via an included cable, which plugs into a jack at the top of the stimulator and provides color-coded pins for the electrodes (current flows out of the stimulator and into the head through the red electrode, which is conventionally considered the anode). It's overall a very high-quality electrode system, although the use of the DC jack means that to use electrodes other than the included ones you may need to either hack the cable (by default, it seems to be compatible with my Caputron rubber-carbon electrodes but not Amrex-type ones which need a larger pin)

It’s actually hard to overemphasize how much I liked the electrodes on this device--the sponge+headband arrangement makes for a really quick and easy setup. One thing to keep in mind however, is that because the headband is stretchy it might cause the electrodes to “drift” when placed in certain positions.At 13 square centimeters, the electrodes (while not extremely small) may also be samller than optimal for some protocols.

One final thing worth noting is the documentation that comes with the device. While the manual that comes in the box is excellent in some ways (for instance, it provides a very good step-by-step guide to running a session that will be helpful for new users), I found myself quite leery of the manufacturer's online montage documentation, where I noticed a few errors (for instance, in describing the location of the dorsolateral prefrontal cortex) and vague claims (what does “improved socialization” or “savant learning” actually mean?). As in any case, it’s probably best to use this site as a starting points but review the actual studies before using a montage.

Turning it on

The device is started up by holding the power button for a couple of seconds. Once it wakes up, the backlight turns on and it displays the simulation options. You can set the power between 0.5 and 2 milliamps, as well as the stimulation duration (20 or 30 minutes), and there's a display at the bottom left that shows the battery status. There’s also a start/pause button which lets you start and interrupt stimulation.

While the display is definitely easy to use, the silkscreened LCD is quite limited. What you see when it turns on is what you get—there's no way to set the timer to a duration other than 20 or 30 minutes, for instance, or to see a meter of the actual current output. While the UI works well for what it’s designed to do, it definitely doesn’t offer as much control as the foc.us v2, the other popular digital tDCS device

Stimulation and safety

The Brain Driver is advertised as being “A New Generation of Safer tDCS device”. With that, it's worth noting that it's included some good safety features.Unfortunately it also comes with a number of quirks.

One feature that might be more accurately called a comfort feature is a built-in capability to smoothly “ramp” the current level over a few seconds in order to reduce the feeling of an electric “zap” when the current level changes. While this works fine when starting a session, for some reason the ramping isn’t applied when you press the pause button, meaning you’ll get a definite jolt of electricity. It probably isn’t dangerous (a lot of simpler stimulators do this simply because they have no ramping whatsoever) but after doing this a couple of times I felt a little dizzy.

Another interesting advertised capability is the ability to cut off stimulation if the resistance rises too high. This function can serve two functions on a tDCS device: it prevents the device from ramping up voltage when an electrode loses contact with the skin (thereby preventing the stimulator from generating spikes of current) and it can protect against burns by cutting off current when electrodes are not contacting the skin well (i.e. poorly attached or drying out).

Unfortunately, on the BrainDriver it doesn’t work exactly the way you’d expect--the device continued to insist that the connection was fine when the resistance reached one megaohm! (about a hundred times the resistance you would expect from a typical connection across the head). This is particularly weird given that the device’s maximum output is 24 volts, which means there is a large region where it can’t deliver the target current but also doesn’t give any indication that there’s anything wrong with the connection. The shutoff does work if one if the electrodes is dangling in the air, but that’s about it--it’s not sensitive enough to detect more subtle issues like poor electrode contact or  resistance that is too high to deliver the target current.

Moving on to the device’s current regulator...

The device seems to deliver on average the target current in a “steady state” configuration (resistance held  constant), but there's a small (about 0.1 mA at 2 mA output) low-frequency oscillation present in the signal which is likely not biologically relevant (given that 0.1 mA current levels are sometimes used as a sham control). Edit: Someone on reddit correctly pointed out that this may not be entirely true, since this is an AC ripple current and about AC is actually used in studies at very low current levels approaching this. Therefore it seems reasonable to think that this oscillation might actually change the effects of stimulation compared to “pure” tDCS though what the effects would be exactly is not quite clear. (Keep in mind that oscillatory tDCS is not neccesarily throught to be more dangerous than pure tDCS, it’s just a different thing)

Steady state voltage drop measured over a 100 ohm resistor. A change of 100 millivolts represents a change of 1 milliamp current,

But the resistance in a tDCS circuit doesn’t stay constant--if it did, we wouldn’t need a regulator in the first place. Therefore, another important aspect of a tDCS device is its ability to adapt to rapid changes in resistance.

Here the BrainDriver falls short-- its regulator seemed to be very slow to adapt to changes in resistance caused by varying circuit conditions. A consequence of this is that when resistance varies, rather than a smooth DC signal the BrainDriver tends to produce a series of odd-looking pulses.

The most significant consequence of this is that under certain conditions the device can produce quite large current spikes. For instance, a drop from a steady-state resistance of 1 megohm (which significantly exceeds the capability of the device’s 24-volt maximum output to deliver the target current) to 100 ohms produces a current spike over 40 milliamps!

Current spike generated during sharp transition from 1 megohm resistance to 100 ohm resistance (voltage measured over a 100 ohm resistor). The spike ends when the device powers off (possibly due to tripping some protection mechanism)

Interestingly, spikes of this size seem to trip some sort of fuse in the device’s power supply that cause it to completely shut down after about a quarter of a second.

It’s worth noting that while this test highlights some of the limitations of the BrainDriver’s regulator, the changes in resistance are by design far larger and more rapid than you would expect to encounter during a typical tDCS session. To get a better picture of the device’s “typical” behavior, I spliced a current sensor into the electrode cable and ran a session with electrodes on F3 and F4 which were subjected to vigorous head movement movement and readjustment while the stimulation was running.

The device’s performance during a real session (even one where I was deliberately trying to mess with it) was a lot better than in the previous tests. Even during vigorous shaking and pressing on the electrodes produced only very small current excursions, suggesting that despite its relative sluggishness the regulator was able to keep up with these demands.

Voltage over 100 ohm resistor during vigorous head motion test. 100 millivolts represents one milliamp of current.

Conclusions

I really liked most aspects of the Braindriver. Probably the best aspect of the BrainDriver is that for is entry-level price it’s a very capable system, and it seems to cover the basic demands that people often have of a tDCS system such as an output voltage high enough to easily overcome electrode resistance, sponge electrodes, and ramping capability, though the lack of an onboard current meter is a little disappointing.

To try BrainPrint, click here!

Recently I've been working on a new project: an app for interpreting data from consumer EEG devices. The goal is ultimately to build a system that can take recordings from devices like the Neurosky Mindwave , automatically clean and interpret the data using some sophisticated neuroscience measures, and compare your data to a population as well as to your history in order to gain insights into your brain function.

Why build this? I've always been interested in individual differences in brain function and electrophysiology—it's what I did my senior thesis on!---but a major inspiration was getting genotyped through Genes for Good a few months ago. While it's amazing to be able to look through a data file on your genes for various brain functions, when it comes to brain function genes are very far from the whole picture—your genotype doesn't capture how your brain physiology changes over time, for instance. This lead me to think about building a system that would operate like 23AndMe or Promethease—highlighting useful and interesting biomarkers in a user's brain activity.

It turns out there are a lot of reasons that this is a bad idea. The neuroimaging and EEG field as a whole isn't quite at the point of having huge public standardized datasets (particularly for EEG data) comparable to the ones that enabled these genomics products.

But there's still a surprising amount that you can do with the hardware capabilities of consumer EEG devices combined with publicly available data. The easiest thing to do is compute statistics from spontaneous brain activity and this is currently the basis of BrainPrint—we record data while you look at your phone's screen for about a minute and a half. From this data,the  BrainPrint server can  compute some neurologically interesting statistics on the EEG (peak alpha frequency, eyeblink rate, etc.) and then compute how “unusual” each measure from an individual is compared to the population. This is a pretty standard method for highlighting interesting individual differences in quantitative EEG work, and in BrainPrint it's accomplished by comparing each person to a collection of about a hundred recordings previously collected to test a brain-computer interface system. (Data posted on physionet). If you've recorded at least ten sessions, it also lets you look at how any particular recording compares to your past history.

To download LearnBoost for Android, click here! Recently, one of the things I've been interested in is how manipulating sleep can be used to affect cognitive performance. In the process of developing Sleep Boost (an app for enhancing slow waves during sleep), another interesting technique that I stumbled on to was targeted memory reactivation (TMR).  Much like the slow-wave enhancement methods I previously talked about that have been used to improve memory, TMR uses sensory stimulation activated during deep sleep. The mechanism of action is rather different though: it works by “cuing” specific memories (typically by sound or smell), which enhances their ability to be recalled the next morning.

Why does cuing improve recall? The best answer is likely because one of the main functions of deep sleep is optimization of memory and strengthening of important memories. This function means that memories seem to be unusually “plastic” during this period—their strength can be altered to a significant degree by activating them, an effect that's been investigated in improving vocabulary learning and enhancing the effects of counter-stereotype training.

In most cases, detection of deep sleep in studies is carried out with EEG. However, like in my previous app it's theoretically possible to do the same thing using measures of body movement (more or less; it's very difficult to distinguish REM from deep non-REM sleep using movement since both states involve little movement).

Therefore, I decided to write an app for implementing targeted memory reactivation, called Learn Boost. It's available for free on the Google Play store and uses the verbal cues like the ones described in this study  to reactivate memories.

Using LearnBoost requires first setting up a list of cues you want to use (ideally single words or phrases related to something you've recently learned) and placing the phone in your bed near your pillow. Unlike the previous Sleep Boost app, there is no manual calibration of the deep-sleep detection required—the algorithm will automatically calibrate itself during the first sleep cycle. Like Sleep Boost, it probably doesn't work terribly well on memory foam mattresses or if there's more than one person in the bed.