Connecting Living Neurons to a Computer

this is your brain the size of a grapefruit it's a soft, greasy mass that contains everything that makes you your hopes and dreams your talents all the things you've learned and all your embarrassing memories it also takes care of a lot of things you normally don't have to do consciously like breathing or finding a comfortable place in your mouth for your tongue, both of which you now have to do manually just because I reminded you, you're welcome, the brain is one of the most powerful

computer

s we know, but brains have a problem, well , two problems, the first is that most of a brain's computing power is spent guiding the behavior that keeps the brain alive, after all, every animal has to navigate in three dimensions and deal with a wide variety of factors environmental as we try not to be eaten, find food, and find a mate to reproduce with.

The second problem is, of course, the lack of convenient auxiliary ports. This will feel a little strange, but what if we took some

neurons

out of a brain and grew them separately if we gave them nutrients to keep them alive and a new set of signals through some electrodes, can they learn to do other things and If so, is there a limit to the tasks they can perform if we choose the signals correctly? Discover that inside this strange-looking container is approximately one 220th of a rat's brain or approximately 75,000

living

neurons

. This container is our latest prototype of what is called a multi-electrode array, which is exactly what it sounds like: a special arrangement of electrodes embedded in Plastic for neurons to grow so we can pick up the signals they emit and input signals so that process them.

So what signs are we putting up? Other groups have used neurons like this to navigate mazes, control small robots, or even fly. planes in a simulation, but we have more ambitious goals, we want our arrays to play the classic video game Doom, depending on how you're counting, this video is actually the third part of an ongoing series we're working on to build a system capable By playing Doom with live neurons, we call it The Storm Nexus, named after the fictional science fiction book, you don't build the Torment Nexus and we've made great progress so far. Let's take a look at our tech tree so we can see. where we are and there have been some unexpected side quests that we need to add here first, we had to build a special incubator to keep the cells alive and we tested it by growing skin cells which are cheaper and much more tolerant to the neurons we also expend.

Several videos building a plasma-based metal coating system to try to make our first fixes. Our first attempts at growing neurons were pretty disastrous, so in the last video we focused solely on that aspect so we could perfect it and also tried some alternative methods of doing it. Arrays that were easier than the plasma scrambling method recently also saw our side quest of growing cells in Gatorade and other sports drinks which was actually a pretext to start developing our own cheap growth media to reduce the cost of these. experiments from hundreds of dollars to tens of dollars and although our recipe is not yet perfect, our first attempts to reduce the cost of growing media worked very well using a Japanese sports drink and finally, here is a side quest that you didn't know about part of the The control scheme we have in mind requires that we not only have a live set of neurons, but also a simulated partner, that way we can test the stimulus on the simulated set before sending the command to the real one and try have an idea of ​​what is about to happen. happen before it happens.

I won't go into details now, but it means that we need to be comfortable simulating neurons in the long term. I don't want to simulate just a handful of neurons, although I eventually want to simulate an ensemble. Brain, anyway, here's the first step towards all that. What you are seeing is 1 second of simulation of a single cortical microcolumn. These are the building blocks of a brain and in this visualization you are only seeing a small fraction. of the actual number of connections between neurons, here is what 1% of them look like for context. We will eventually use what we learned from these simulations to improve the connections between the current virtual and real neurons, although we focus on the From the physical side and taking the next step towards the game, we will focus on getting signals from the neurons and getting used to stimulating them. in various ways to see how they respond, and we will do it both electrically. but also visually with the help of an incredible die that we will talk about shortly.

To do this, we first needed very serious hardware, we needed better matrices than last time because our homemade attempts have not been particularly successful, ideally. we needed commercially made neuron arrays so that we at least had something that we knew for sure would work, then we could study them and see if we could improve our homemade ones enough to make them work and if we have commercial arrays it would be great to have a system as well. commercially manufactured recording system, so that we can get some data on what the signals actually look like as the neurons grow, so that we have a real baseline.

Well wouldn't you know it after the last video we got in touch with? the wonderful people at multichannel systems and they offered to help us on both fronts, very generously sending us a bunch of their amazing neuron arrays to use and lending us one of their mea 2100 mini systems so we could take our first real recordings. We'll talk about how cool this amazing machine is in a moment, but huge. Thanks to them, this alone has propelled the project forward practically light years. Let's start by looking at some of these outfits because they send us a variety of the types they carry. and, frankly, each one is prettier than the last.

All of these sets were manufactured using plasma deposition and are all made of glass. These sets have electrodes made of Titanium Nitride or, if you want to see every little detail, they also have electrodes made of Indium Tin Oxide, in either case the electrodes are only a few micrometers wide. We had pretty much stopped making arrays like this because of how difficult it was, so I'm really excited to try them out, something that's been a big deal for us. In the past it was the way we sealed our arrays, these cultures are very sensitive to contamination so they need a lid but they are also

living

organisms and need to breathe, the last part is actually the hardest to deal with because the Most things that give a good seal also prevent air from getting in, so our crops kept suffocating.

These commercial dies use a very clever system of a Teflon ring with a thin sheet of fet film stretched across it that is held in place with an O-ring. The same thing used in 3D resin printers, it is a fluorinated plastic that prevents water from diffusing through it, but has small enough pores to allow oxygen and CO2 to pass through relatively easily and the film Fet is super cheap. I bought a roll of 25 micron thick film. for $25 now before we head into Weeds about our new batch of homemade arrays and our new control scheme for making neurons play a game.

I want to show you how all our new tools work for that we need some neurons like As usual, we will use these prepared tubes that we bought from the manufacturer and they contain about 1 million neurons ready to go. They are hugely expensive for what they really are, so we are working on a cheaper alternative, but for today they will work fine. We went over the process of how to grow neurons in detail like this in the last video, but after publishing it, we were contacted by Dr. Steve Potter, who is a legend in the world of neuroscience.

He had an entire laboratory dedicated specifically to the use of neuron cultures. neurons to interact with a

computer

and had a lot of experience using these exact arrays, so on his advice we made some changes to the protocol so that the neurons would attach to the arrays or just to the bottom of a normal 12-well plate that you normally have . To apply a chemical coating last time we used something called polydeysine which worked, but at higher densities the neurons broke off and started to form these strange brain orbs. This time we are using a double coating at Dr.

Potter's suggestion. polyethylene imine also known as Pei followed by a layer of laminin Pi is a synthetic polymer, while laminin is a mixture of proteins to apply the coatings we put a portion of Pei in each well or set and then let it sit for half an hour and then we draw Take it out again, rinse everything well with sterile water and let it dry under the clean air of our flow hood, then when it is dry we apply the laminin and then let everything sit in the incubator for about 20 minutes, after That we can remove the excess laminin and add our culture media back in, this is where we are trying a few different things.

Dr. Potter pointed us to some literature that showed that, contrary to what is commonly done, ordinary, cheaper DM provided electrophysiology that was more similar to that obtained in a real brain. he suggested that neurobasal media, the material we used last time, was optimized to make the cells look good on the images rather than electrophysiology. Well, we've got a fancy recording system and a bunch of cool tints, so we can try it out and find out which one we prefer. We set up a matrix with neurobasal and a matrix with dm supplemented with horse serum and insulin.

I have no idea why those ingredients, but there will be plenty of links in the description to articles for those interested in the exact recipe we set up as well. eight wells of a 12-well plate with DM or neurobasal so we can take a look using our fancy new dies and see if we can visually tell the difference in their activity after everything has had time to warm up and equalize on the four-hole. meat. let's add our neurons now, these will need a few days to cook before we can really do anything with them; It will take them about a week and a half to two weeks to really establish the network and be really usable, so it is very important that nothing happens to them at that time, otherwise you would have to start over Hey, how are the neurons growing? ?

They all died yes, I don't know what happened, but the first batch died, we have no idea why, but luckily these arrays are reusable to do that. They undergo prolonged soaking with a trypon enzyme solution followed by thorough rinsing to clean dead cells. They can then be re-sterilized by soaking them in 70% alcohol and letting them dry in the hood anyway, skipping them to a few weeks. later when a new tube of cells arrived and had time to grow here, they are already a week into the culture period, as you can see a dense network is already forming, but let's take a look at what they are thinking and try our new one today we're going to use a really cool one called calbright 520 a.m. which I'm just going to call calbright.

What makes this D amazing is that it has two special properties: the first is that in its basic form it doesn't actually have them. It looks like anything, no fluorescence, no color, just a clear solution, but when it enters the cells, the enzymes in the cells react with the dye to turn it into a fluorescent form, which makes it much easier to see the cells and only living cells should glow, although the second property is that the fluorescence is not uniform, in fact the brightness of the glow should be proportional to the level of calcium present.

Now you might be thinking, why does that matter? When neurons fire, a couple of different things happen, the first is an avalanche of sodium and potassium ions in and out of the cell, which is where the electrical current comes from when a neuron is activated, but that avalanche of ions in Movement also triggers a large amount of calcium to enter the cell, so cells stained with calbrite should be very bright. brighter immediately after they fire, these cultures already show a lot of spontaneous activity where random cells will pulse, although we noticed that the Dy seems to saturate the longer you look at it, if you look at any area for too long eventually all the cells get brighter .

It's only in the shiny state and you can't see the pulses anymore, but I think this is just a quirk of this specific die. It must take some time to resetto the off state or something related to the lighting causes it to get stuck in the on state. There are alternatives to this die that can be genetically encoded and in fact we have been testing them in the background, for example a specially designed protein called gamp 6f can be delivered to cells as a DNA fragment and has similar properties to calbrite, however, it reboots. timing is much better and it doesn't get stuck as much in the on state and since it is delivered as DNA you can also choose the exact cell type you want to express the protein in, for example muscle cells also use the flow of calcium as part of its activation. mechanism for muscles that express gcamp to light up when activated, but we're not here for the muscles, we're here for the brain.

Several years ago, a group of researchers produced a variety of zebrafish that was transparent enough that they could see through its skull and then used gamp specifically expressed only in its neurons. The result is this incredible footage of a fish thinking while swimming. You can literally see where the abstract thoughts are versus where the muscle control and navigation bits are which is part of the reason I found calbright was because I was looking for ways to replicate this experiment, the idea of ​​being able to see electrical activity was so Fascinating to me and when I found calbright as a cheaper and theoretically easier alternative I thought I'd give it a try, but Now that I've tried it, I can see why gcamp is preferred for many applications, even if genetically modifying neurons is harder than adding a drop of dye.

One last thing to try with dyes is chemical stimulation. We tried a few different things but by far the most interesting was capin also known as the spicy molecule, as you would expect adding some caps to the cultures causes the neurons to fire more and they also start releasing bursts occasionally where a lot of neurons fire all at once, but to increase the heat we simply drop a few specks of cayenne pepper directly into some of the wells, you can literally see the capin wave pass through the liquid as the wave propagates, the neurons get stuck in the peak state, but I don't know if this means they're not firing, it could just be an interaction with the die.

Okay, we're comfortable with the tints, so let's check out the new recording system and give it a try. Theoretically, this amazing system can not only perform live recordings, but it also has a hole in the bottom for a microscope objective, so in theory we could observe the neurons firing and see the electrical activity at the same time, but when we tried this we couldn't get close enough to see the flickering effect and only Seeing it when you zoom in at this magnification you just see the flat glow but that's okay because the reading from the software package that comes with the system is Amazing, not only can you see the electrical activity of each electrode individually, but there are all kinds of built-in tools, such as filters, peak detectors and analysis tools to perform any experiment with Breeze.

The machine itself has three parts, the main stage where the array is placed, a signal collector unit that can drive up to four main stages at a time, and the interface board that the collector unit connects to is connected to a computer. with a USB cable, although make sure you have a large hard drive installed because this produces data like it's a full time job, you know what else. Data like that, without realizing it, every time you use the Internet, you leave a trail of crumbs that Hoover data brokers package and sell to the lowest bidder, search histories, names, addresses and much more, all available to whoever wants it between data breaches.

From your favorite websites, honest mistakes and unknown vulnerabilities, there is a lot of your data available and while you may not think that matters, those with nefarious intentions would beg to differ, for example if you are in the health landscape of the United States United, your ability to get medical treatment is not controlled by your doctor but by health insurance providers and they are always looking for any excuse to increase their rates and are big customers of data brokers, as are scammers, scammers and other Nero Wells who seek to make money at your expense. enter the sponsor of this video incog incog fights against data brokers and uses an endless barrage of deletion requests and other tools to force them to remove your data from their collection and continually monitors in case your data reappears so they can start to work immediately to remove it if you use the torio code or go to incog docomo imporium, you can get protection now for the incredibly low price of about $62 per year if you opt for the annual plan which costs just over $5 per month from Monday to Friday. just get the peace of mind that your data is safe, but also know that it will stay that way, so head on over to incog docomo Emporium or click the link in the description after this video and get protection today, now back to the machine whose data You all came here to see how to load a die, just press the release button to open the clamshell, just drop the die carefully and close it, then just turn on the control software.

There is a calibration tester included that creates bikes with simulated neurons. and other signs that come with the machine we are supposed to use. Firstly with the calibrator installed, when we turn on the Dack we can see the rhythmic pulses it emits on each channel, this means everything is working as intended, however when we switched to a real matrix we discovered that there was a lot of electrical noise leaking in , which made the signals really hard to see, so we made a little scientific neuron burrito and wrapped the main stage tightly in aluminum foil. The main stage box is grounded.

So this turns everything into a Faraday cage and makes taking readings much easier, although we're going to build a better Faraday cage for future experiments to reduce the noise level even more anyway, with a bit of Senti inila in hand we could take some appropriate measures. Readings are that special sensitive Bingle sandwich. We started with the matrix that was grown in neurobasal media under the microscope. They looked like nice neurons and we saw them firing very well before with the machine, once again, we saw nice big spikes and what is immediately obvious is that some electrodes pick up a lot more signal than others.

When we inspected the array under the microscope, we found that the electrodes showing a strong signal had a neuron stationed with its end directly above it. I can see why the instructions say to try to get as much as possible. neurons as you can in the center of the array because the ones in the pads are picked up much more clearly now with that baseline set, let's take a look at what the DM array looks like once again, there are still a lot of peaks, but its magnitude is dramatically smaller than the neurobasal, so maybe this is the difference that Dr.

Potter was talking about, maybe this is more like what you get with an actual piece of brain tissue, but until Let's not test the system with a portion of the brain, we won't know. Surely when we look at these cultures under the microscope, although they definitely look strange, the cell morphology is different and the cells look much more like fibers that burst neurons, but that could simply be a peculiarity of them growing in this medium, however , as with the electrical results with Dy seeing any kind of flash of activity were really difficult, so it's actually hard to say at this point which is better: the big spikes in morphology that we saw with neurobasal, an artifact of growth Media, or are the neurons happy.

They are supposed to look and behave like, and DMs are the underperformers. Let us know what you think below, especially if this is your field. Now we've covered everything other than our home fixes so we can finally talk about it. them and then we will play with the stimulator function of the machine and see how the neurons respond. So how did we make our new neuron arrays and what's different this time? Well, as a quick reminder in the last video, we started exploring how to have our arrays. Made of P by a normal PCB manufacturer, but one problem we ran into was that there was a limit to the size of the hole they could cut in the casing.

These pet circuits are actually two layers of plastic sandwiched around a delicious metal medium and to expose the metal for the electrode, you need to physically make a hole in the top plastic before gluing it in the final manufacturing step last time. . This meant that instead of many small holes like we ordered, the company simply cut one large hole around everything. of our electrodes makes them useless, as we talked about last time, if your electrodes are too big, you actually weaken your signal because you are averaging the activity of many neurons over a large area to solve that problem, this time we try to be smart, we arrange holes small and thin and the electrodes barely stuck out from under the plastic so we could get our little electrodes while still working within the manufacturer's limitations and the large empty spaces leave a flat surface for the neurons to grow on. and be observed through we made and ordered three different versions, each designed to test different ideas on how to implement a game matrix.

The first uses square-shaped holes with multiple electrodes inserted into each one. The idea is that they simulate what we call falling action. groups or collections of electrodes, all associated with a particular action the array might perform while we play, we try different sizes and spacing to see how that affects things. The second test uses arc shaped cuts and we tried different spacing and radii to see how. Those affect the signals. The idea with this is that they could be different inputs to the matrix, such as the positions of different objectives, walls or elements in the game world, although these pads could easily be configured as groups of Acts in the same way and finally there.

Is this one different than the other two? This one is designed specifically for playing a specific game. While we were thinking about how to get a matrix to play Doom, we thought it might be good to start with a game that has the fewest options for the matrix. make as much data as possible for it to process, but only a small number of actions it can perform, that way we can develop the most complex case of Doom. When we were thinking about a good test case, what came to mind was a uniform case. older game called asteroids on asteroids you control a ship and you have to shoot at the flying asteroids while trying not to get hit, but if we immobilize the ship in the middle and prevent it from leaving that place, it becomes more like a defense game towers, so in theory this should be simple enough that a matrix can take it on from the start, all you have to do is turn and shoot, so just three actions from left to right and shoot, but We can input a ton of target data giving the Asm and distance of the targets, although while designing this set with this special version of asteroids in mind, we realized that the same control scheme could be used to play Space Invaders , Likewise, left, right, and shoot are your only options and there are many. of targets to shoot now before you get your hopes up no we're not going to test that today, while this system is awesome it's designed for researchers not a bunch of crazy nerds so it has a recording system and amazing stimulation, but it can only do two stimulation pulses at a time, while to play a game we actually need mainly stimulation electrodes and only a very small number of recording electrodes comparatively, that means we need to build our own system, but it's okay, we already have it designed and I think we know what parts we need.

It will be built around Intan Technologies' amplifier and stimulator chips. They are specifically designed to work with neurons in the way we have in mind. The only downside is the $10,000 price tag for all. The bits we need for today are just going to use the current system's built-in stimulators and see how our arrays handle stimulation if the neurons grew properly on them. The P arrays we ordered are great but they are super flexible and bendy, which is not useful when we want them to be a solid array that we can handle, soThey were mounted on a glass plate with UV resin and then to turn them into actual culture wells we used marine grade silicone and some sections of polycarbonate tubing, since none of this can be autoclaved the plan is to sterilize everything chemically with alcohol and all of these materials are inert to that, then, like commercially made dies, we turned some Teflon retaining rings on the lathe, although we chose a larger O-ring not because we thought it would be better, but because I'm an idiot and I forgot to press order the day before the experiment when I was supposed to buy them, so we used what we had on hand, oops, which said when the new O-rings were the right size.

We did a little Indiana Jones swap, but in the end the cover and head still couldn't fit over the o-rings or caps, so we moved the head towards the hood of the flo and very carefully removed the cap to close. the cover and then putting the cover back through the hole in the main stage and we did the same in reverse later when we needed to take it out again, not ideal so next time we will choose a different mounting option. or to make the wells less tall and thinner, one last note is that since DM was suggested to be the best, that is what we used to grow the neurons in our home-made arrays, so their activity was similar to the other DM test, there was still a lot of clear evidence. peaks, but none of them were as large and forceful as we got with the neurobasal media.

Finally, let's play with the functions of the stimulator. The software definitely allows you to find an arbitrary waveform so you can stimulate the neurons however you want and you can choose which electrodes to stimulate. We were primarily interested in how signals propagate and how neurons respond, so we tried a variety of different pulses and electrode arrangements. What we found was that, as expected, the neurons closest to the pulse tend to fire very strongly, but what was interesting was the increase in activity after a pulse even far from the stimulated electrode and we saw the same behavior in all the arrangements, including homemade ones.

This is pretty crude evidence, but this video was already getting very long the next time we watch it. much more extensive experiments to see what exact stimulus patterns we should use and how they affect the dynamics of the neurons' activity, but the most important thing is that the r-rays seemed to work as well as the commercial ones, which is a big improvement and means that next time we can focus on fine-tuning its exact design and getting the neurons to actually play a game, and while of course Doom is the goal, we might just be handling asteroids in the next video, but even that would be a huge achievement. no matter what we are tantalizingly close to achieving. neurons to play a game to commemorate this momentous occasion.

We are doing something very special. If you have been following this series and ever thought you would like to have one of our arrangements for yourself, now you can get one and the best one. Above all, it's mounted on what might be the coolest poster we've ever made, not only is it hand-printed here in Quebec, but we incorporated Secrets upon Secrets with white lighting, you can enjoy this amazing design, but turn off the lights . And the magic begins in the dark, the brilliant phosphorescent layer lights up to reveal additional details and the glow of stained neurons, but if you use the ultraviolet light that comes with the poster.

You can reveal one more secret in the form of a hidden message. If you want one of these amazing posters, head to the link in the description or just visit the thought emporium.com we have. There are stocks between sponsors, posters and product sales. Your support goes a long way in helping us keep this project going, so thank you very much in advance. We will make as many of these signs as we receive orders, but they are handmade, so it will be a few weeks before the first bat is shipped. As an added bonus, the first 50 people to use the code torment Nexus will get 5% off their order so be sure to check out the link below and get your poster today and for those interested there will be plenty of links to items and other things in the description.

That's all for now and see you next time.