AI-Generated Inner Ear Simulation: Hair Cells & Hearing Biomechanics

Okay. Hello everyone. We'll be quickly reviewing this web application. It's a simulation of what's happening inside your inner ear. These are neurons called hair cells cuz they kind of look like hair cells. Um the page is interactive. You can go check it out. It's on bionic.com/hair cells. You can interact with the tool yourself. see what it does. There's a description at the bottom. Go check out all the other tools we have on binos.com. And don't forget to provide your feedback. We have a lot of new tools in the works. You can look at an early version of them as well. It's all on the page. It's all on the site. I'll do an overview by notebook of the simulation. You can regenerate this uh yourself. You have this so we can put the tool in full screen so you don't have the the annoying browser bits. And when we hit the start demo, it will actually cycle through the different parameters, different settings, and show you around. So picture this. You are standing in the very center of a massive, perfectly silent cathedral. >> Okay, I can picture that. >> The air around you is entirely still and then someone drops a single metal pin onto the stone floor like 100 ft away. >> Right. So the sound that actually makes it to you is incredibly faint. >> Exactly. It's just this microscopic disturbance in the air. You know, a tiny ripple of pressure. Yet, without even thinking about it, you instantly turn your head and look directly at the exact spot where that pin fell. >> It is pretty wild when you really think about it. >> We take that ability completely for granted. Yeah. >> But how does that actually happen? I mean, it is certainly not magic. >> Far from it. That simple everyday reaction is actually the result of a high-speed microscopic biomechanical ballet. And uh it is happening inside your head in a matter of milliseconds. >> Welcome to today's deep dive. We are taking you on a tour of that exact biological ballet step by step. >> Yeah, we are really getting into the weeds today. >> To figure this all out, we're diving into an incredible digital simulation from bionicos.com that is bionic chaos. >> Right, which allows us to trace a soundwave from the air through the fluid of the ear and directly into the brain circuitry. We are going to use their demo mode as our map today >> to understand exactly how a physical vibration transforms into the electrical signals your brain perceives as sound. >> And this simulation is incredibly comprehensive. To really grasp what we're looking at, you kind of have to shift your perspective. >> Shift it how? >> Well, you need to think of the mamalian cookia not just as a part of your ear, but as a fluidfilled spiral-shaped biomechanical frequency analyzer. >> A biomechanical frequency analyzer. That sounds intense. >> It is. Its entire job is to take acoustic airborne pressure waves and translate them through some incredibly complex and chaotic fluid dynamics into electrical action potentials that your auditory nerve can actually read. >> Okay, let's unpack this. Before we zoom all the way into the microscopic cells, we have to start at the macro level, >> right? We need the big picture first. >> We need to see how a sound wave physically travels through this biological machinery. H. >> So if we jump into the simulation, the first lens we can look through is called the tonotopic wave. >> This gives us a bird's eye view of the main structural component which is the basil or membrane and how a physical wave travels down it. >> So we can throw a curveball at this virtual ear to see what happens. The simulation has an acoustic frequency slider and I can sweep it all the way from a low rumbling 200 Hz, >> which is like a deep bass note. >> Yeah. up through the mid-range of 4,000 hertz all the way to a piercing 16,000 hertz. >> And as I move that slider, you see this continuous physical waves traveling along the basler membrane. >> It is really cool to watch. >> The best way I can describe looking at this membrane is like a rolled up biological xylophone. It literally changes its physical shape and properties from one end to the other. >> And that structural gradient is the absolute key to everything. The basil membrane is not uniform at all. >> How so? Well, at the base, which is right near the oval window where the sound first enters from your eardrum, the membrane is narrow. It is taut and it is very light. >> So, kind of like a guitar string pulled tight. >> Exactly like that. Because of those physical properties, it naturally resonates at very high frequencies up to 20,000 hertz in humans. >> Okay? So, it catches the high pitches first. But what happens as that wave travels further inward deeper into the spiral? As the wave moves toward the apex, the membrane becomes drastically different. It becomes approximately five times wider and significantly more floppy. >> More floppy, right? >> Yeah. Or compliant is the technical term. Because it is looser, heavier, and wider, it resonates at much lower frequencies going all the way down to 20 hertz. >> So the physical geographic location on this membrane determines the exact pitch you hear. >> Yes. And we don't just know this conceptually. I mean it has been mapped mathematically to an astonishing degree of precision. >> Right. >> In the simulations documentation, you see this defined by Greenwood's frequency mapping function. >> Yeah, I saw that in the lab notes. There is an actual empirical formula for human hearing. But I'm going to be honest, looking at the math, the function relies on constants like a is 165.4 and little a is 2.1. >> It is a lot of numbers. >> It feels incredibly abstract. Why should anyone listening care about those specific numbers? Well, you don't need to memorize the numbers, but you should absolutely care about what they represent. >> Okay, play it on me. >> That equation proves that your ear is tuned on a logarithmic spiral. The exponential nature of that math is why human hearing is so uniquely disproportionately sensitive to the specific frequencies of human speech. >> Oh wow. Wait, really? So, the physical tissue has literally mathematically evolved to prioritize us talking to each other. >> That is exactly what the math dictates. When a pure tone soundwave enters, it creates a traveling wave in the fluid. Okay? And that wave propagates from the tot base towards the floppy apex. >> As it travels, it slows down and builds an amplitude. It gets bigger and bigger until it hits its exact resonant spot on the membrane. >> The spot that perfectly matches its frequency, >> right? It hits a peak and then immediately beyond that spot, it underos a rapid spatial decay. It just crashes and stops entirely. >> Okay? Because the wave hits this very specific resonant peak based on the frequency. Let's use the simulation to zoom directly into that exact spot on the membrane. >> Let's do it. >> Let's see what happens at the crash site. >> Moving from the macro to the micro. This is where we change our lens in the simulation to look at the organ of cordi. >> Let me set the scene for this. We are looking at a microscopic cross-section of the tissue. I am setting up a specific scenario on the dashboard to see how the system handles it. >> Sounds good. So, the exact spot we are zoomed in on has an active characteristic frequency of 5,913 hertz. But let's send in a sound that is slightly off peak. I'll set the incoming acoustic frequency to 5,850 Hz. >> Okay. A little offc center. >> Yeah. And for the loudness, the sound pressure level, I'm pulling the slider way down to just 15 dB. That is a mere whisper. >> What's fascinating here is what stands out when you look at that crosssection. The organ of cordi sits right on top of that flexible basler membrane, but it is populated by a very specific cast of characters. >> It looks like two distinct groups of cells. On the left side, there is a single solitary row of cells. The simulation labels these as inner hair cells or IHC's. It says there are only about 3,500 of them in the entire human ear. >> Yeah, those 3,500 cells are your primary sensory receptors. Mhm. >> Almost all of the wires going to your brain, over 90% of the aference spiral ganglion and synaptic connections start right there on that single row. >> But then on the right side of the screen, there are three parallel rows of a totally different type of cell. The outer hair cells or OC's. >> Yeah. >> And there are way more of them, about 12,000. >> We will get to the outer cells in a moment because they do something entirely different. >> Okay, so sticking with the left side for now, >> right? Let's focus on that single row of inner hair cells on the left. That is where the physical wave is actually captured. Above these cells is the tectoral membrane and the space between is flooded with a specialized fluid called endolymph. >> Oh, I see it in the visualization as our 15 decel whisper wave rolls through. You can literally see the fluid sheer displacement happening. >> Exactly. >> The fluid movement between the tissue layers physically bends this little staircase bundle that's sitting on top of the inner hair cell. They look like tiny hairs, which I guess is why they're called hair cells, but they are technically actinfilled stereocyia. Notice what happens when those hairs bend toward the tallest row. >> It puts tension on these tiny microscopic connections, stringing them together. I absolutely love this analogy. Think of them as microscopic protein trip wires. >> That is a great way to picture it. >> The simulation identifies them as tip links, specifically caderin proteins like CDH23. When the little hairs bend in the fluid, these protein trip wires get pulled tight and they physically yank open biological trap doors on the surface of the cell. Those trap doors are mechano electrical transduction or meti ion channels. And the simulation models the dynamic probability of these channels opening using a two-state boltsman distribution. >> You're you're losing me a bit with the statistics there. A Boltzman distribution. I mean, I thought a trap door is either open or closed. >> It's it's not just a simple onoff switch because we are dealing with microscopic chaotic fluid dynamics. Those trip wires are constantly vibrating slightly. I see >> the Boltzman distribution math basically just calculates the exact probability of how many of those thousands of trap doors are yanked fully open at any given millisecond depending on the exact tension of the wave. >> Okay, that makes sense. But why does yanking a trap door open matter? What is actually outside trying to get in? >> This is where the brilliant bioysics comes into play. That endolymph fluid surrounding the cell isn't just water. It is heavily loaded with potassium ions. Okay, >> it acts like a biological battery carrying what we call an endoc clever potential of positive 80 m. >> So it's basically a highly pressurized electrically charged fluid just waiting for a way in. >> Precisely. The moment those met trap doors are yanked open by the physical tension of the soundwave, there is a steep immediate inward rush of potassium and calcium ions into the cell. >> And we can see this happening in real time on the data readout. The screen shows the internal voltage of the inner hair cell is currently shooting up to - 38.7 molts >> which is a massive rapid depolarization. Normally at rest these cells sit around 70 molts. So hitting - 38 is a huge spike >> right? >> This voltage change travels down to the bottom of the cell and activates basilateral voltage gated calcium channels. This sudden influx of calcium induces rapid exocytosis at the synapses. >> Whoa, hold on. Rapid exocytosis sounds like a spell from Harry Potter. You were losing me with the biology terms again. In plain English, what is the cell actually doing? >> Fair enough. Fair enough. Think of exocytosis as the cell vomiting chemical messengers. >> Vomiting. Okay, that's a visual. >> Yeah. The voltage spike causes the bottom of the cell to rapidly dump tiny packets of a chemical called glutamate into the gap between the cell and the nerve fiber. >> Ah, I see. >> That glutamate is the chemical signal that tells the auditory nerve to fire an electrical spike to the brain. Okay, I have to pause you right here because I have a major push back on the physics of this entire mechanism. >> Let's hear it. >> Earlier, I set our sound pressure level slider to only 15 dB. We are talking about a microscopic, barely there whisper of a soundwave, >> right? How does such a weak, pathetic little puff of air create a fluid wave with enough actual physical force to yank open thousands of cellular trap doors across this membrane? It really doesn't seem like there is enough physical energy in a 15 decel sound to move the mechanism we are looking at. >> That is the perfect question. And honestly, your skepticism is completely justified because based on simple fluid mechanics, a 15 decel sound isn't strong enough to open those trap doors. >> Here's where it gets really interesting. This is why the mamalian ear is a marvel of evolutionary engineering. Remember those 12,000 outer hair cells sitting in three parallel rows on the right side of our screen? >> The ones that outnumber the actual sensory receptors by almost 4 to one. >> Yes, those. They aren't just passively sitting there sensing sound. They are active biological po electric motors. >> Wait, motors? >> Yeah. The outer wall of every single one of those 12,000 cells is densely packed with millions of voltage sensitive motor proteins called Preston. >> Motor proteins. They physically move. >> They do. When that faint acoustic vibration initially hits the outer hair cell, that Preston protein underos a sudden dramatic shape change. It physically shortens the cylindrical body of the entire cell. And when the wave passes, it elongates again. >> Wait, let me look at the simulation data here. It shows the outer hair cell electromotility is at 100% meaning it's fully active and it's causing a physical shift of -6.38 nm. Right? >> Are you telling me these entire cells are physically bouncing up and down to match the sound wave? At what speed? Because the sound frequency we said is almost 6,000 hertz. >> They are bouncing at tens of thousands of times per second. >> That is insane. >> They dance in exact cycle by cycle phase with the incoming wave. As they bounce, they exert dynamic mechanical force directly back onto the basil membrane. >> That is completely mind-blowing. They are literally pumping physical energy back into the fluid wave to make it bigger. >> Yes. This feedback loop is formally termed cocklear amplification. For very quiet sounds like your 15 decel whisper, these bouncing motors add 40 to 50 dB of mechanical gain to the membrane. >> Wow. They take a weak whisper, physically amplify the fluid wave locally, and make it strong enough to yank open the trap doors on the inner hair cells. >> That is incredible. >> Not only that, this mechanical feedback sharpens the frequency tuning. It gives you the ability to resolve single distinct tones instead of just a muddy wash of sound. So, if you're listening to this and your child whispers a secret to you from across a quiet room, the only reason you can hear them is because you have 12,000 microscopic cells doing a high-speed trampoline routine inside your ear to turn up the volume. >> Exactly. That is exactly what is happening. >> But what happens if the sound is already loud? Does it amplify a shouting voice until it breaks the internal machinery? >> That's the genius of the presser system. At high sound pressure levels, like anything over 80 dB, the mechanical feedback naturally saturates. >> Oh, it caps out. >> Yeah. It stops amplifying and instead acts as a biological shock absorber. It provides compressive protection to the delicate sensory structures so they don't literally tear themselves apart from the force of loud noises. >> Okay, that makes perfect sense. And the simulation actually has bethology controls on the dashboard to demonstrate how critical these dancing motors are. >> He does. You can toggle on conditions like noise trauma from going to too many loud concerts or ototoxic loss. The source text specifically mentions otoxic drugs, certain heavy duty antibiotics or chemotherapy drugs like cyplatin. >> Right? If you select those pathologies in the visualizer, you can literally watch the outer hair cells degrade and die. You permanently lose those preston motors >> and the result on the data readout is immediate. Without those biological amplifiers, your baseline hearing threshold is instantly elevated by 40 to 60 dB, >> which means that whisper completely disappears. >> Plus, it ruins your ability to distinguish specific pitches because you lose that sharp single tone resolution. >> You are left with a system that can only detect loud, muddy, blended sounds because the active localized amplification is just gone forever. >> So, let's trace the signal. The acoustic signal was mechanically captured by the inner hair cells, heavily amplified by the bouncing poelectric outer hair cells. The trap door is yanked open, the potassium rushed in, and the glutamate chemical is dumped. >> Right, that's the sequence. >> But how does the brain actually read this chemical dump? Let's switch the simulation to the final perspective mode, the neural PST, which stands for post stimulus time histogram. >> This is where we leave the ear itself. We are now looking at the auditory nerve action potentials, the actual electrical spikes traveling up your cranial nerve directly into the brain stem. >> So, what does this all mean? When I look at this screen and I turn on the simulation sonification feature, I don't hear the pure acoustic tones anymore. >> No, you won't. >> The audio guide toggles on this synthesized auditory nerve spike crackle. It sounds like intense static, like popcorn popping really fast. How does our brain take that chaotic static and hear a beautiful symphony or you know understand a language? >> Well, the auditory nerve fibers transmit sound information to the brain using two very distinct coding strategies. The first is called the rate place curd >> rate and place. Okay, break that down for me. >> The place refers to tenoty which we talked about at the very beginning with the xylophone analogy. Because the basil membrane is physically mapped to different frequencies, the brain simply looks at which wire is firing. >> Okay? >> If a nerve spike comes from a wire connected to the base of the cookia, the brain instantly knows it's a high pitch. If it comes from the apex, it knows it's a low pitch. The physical location of the wire tells the brain the frequency >> and the rate. >> The overall firing rate of the nerve tells the brain the loudness or intensity. These nerves can fire anywhere from 5 to 300 spikes per second. >> That is fast. >> Yeah. So the faster the popcorn popping sound, the louder the brain perceives the volume to be. >> Okay. So location equals pitch and speed equals loudness. That seems like a pretty straightforward code, but you said there were two strategies. >> The second is the temporal or phase locking code. And this is vital for frequencies below roughly 4,000 to 5,000 hertz, which includes almost all human stach. >> Okay, how does that one work? >> In this range, the nerve spikes don't just fire rapidly at random intervals. They discharge preferentially at the exact same phase of the incoming acoustic cycle. >> Wait, let me make sure I had this right. If a physical sound wave is at its absolute peak crest in the air, the nerve fires its electrical spike at that exact corresponding microcond >> precisely. The electrical spike is locked to the physical wave. If we connect this to the bigger picture, this phase locking is mathematically quantified in the simulation by a metric called vector strength or R. >> I see that on the dashboard. >> Yeah, >> it calculates the stimulus phase angle at each spike arrival. If the value of R is near 1.0, it indicates perfect flawless phase synchrony between the sound in the air and the electricity in your brain. Looking at our data readout, our baseline vector strength is sitting at 0.15 right now. But that's because we are looking at a much higher frequency near 6,000 hertz. >> Right? Phase locking degrades at those high pitches. But when you are listening to lower frequencies and R approaches 1.0, the fidelity is so mathematically perfect that your brain stem can do something incredible. >> What's that? >> It compares the phase locked spike arrivals from your left ear against the spike arrivals from your right ear. It uses this timing to compute interal time differences or ITDs. >> Okay, I know sound takes time to travel, but your ears are literally inches apart. What kind of time differences are we talking about here? >> We are talking about microcond level differences. >> Wait, microsconds as in a millionth of a second. >> Yes. Your brain stem is constantly comparing the arrival time of a sound at your left ear versus your right ear, calculating delays down to fractions of a millisecond based on the phase locking code. That is completely insane. My brain is crunching microscond math right now just listening to you talk. >> It is that incredibly precise temporal coding is exactly what allows you to pinpoint where a sound is coming from in three-dimensional space. >> Which brings us right back to that massive cathedral we started in. >> Yeah. >> The next time you are standing in a quiet room and you hear a tiny pin drop and you instantly turn your head directly to the source of the sound, take a second to picture what just happened to make that possible. >> It's a lot to take in. A faint pressure wave entered your ear. It navigated a mathematically perfect, exponentially tuned spiral of fluid. It was actively amplified by tens of thousands of microscopic dancing cells bouncing at khertz speeds. It yanked open protein trap doors, flooded a biological battery, and fired a stream of electrical spikes so precisely timed that your brain stem calculated a microcond difference between your left and right ear to build a spatial map of the room. All of that in a fraction of a second. It really is a flawless high-speed biomechanical ballet. >> It truly is a marvel. And looking at this simulation raises an incredibly important question to leave you with. >> What's that? >> If you look at the related biomedical simulations linked on bionicos.com, this open access educational dashboard is really just the beginning. They are building commercial enterprise tools. things like a caullear implant insertion simulator that models surgical path mechanics and a live neural mapping simulator for cortical dynamics. >> Yeah, I saw those tabs. Enterprise tools for custom feature engineering and hardware telemetry. >> Think about the implications of that. If we have mathematically mapped the exact bioysics, the fluid dynamics and the precise electrical spike code of human hearing well enough to put it in a fully interactive browser dashboard. How close are we to perfectly synthesizing this process in the real world? Oh, I see where you're going with this. >> If we understand the exact microcond code the auditory nerve uses to speak to the brain, how long until we can entirely bypass damaged biological hardware? I mean, bypassing the hair cells and the fluid completely and using an implant to stream perfectly encoded highfidelity sound directly into the human cortex. >> That is a wild thought to end on. If we already have the mathematical blueprint for the code, it might just be a matter of time before we build the direct USB connection to the brain. Something to ponder the next time you hear a pin drop. >> Okay. So, that was notbook lamb. Yeah. I'm not sure. So, it's first of all, it's amazing what what it can do. The whole the whole tool was generated by Gemini, one of the Gemini code agents. Yeah. The demo was generated by Google whatever the Gemini notebook used to be called notebook. I think they changed the name. So that kind of works. I'm not sure that it actually changes the sliders, the controls correctly throughout the audio. Kind of could have done a better job there. Yeah. when we turn. Oops. Yeah, that's too loud. That's way too loud. Yeah, we can reduce the pressure level to like 10 dibels. Yeah, that's what happens when you turn the sonification on. Yeah. So, it changes uh so it changes frequency and amplitude as expected. Okay. You can select your pathology, healthy organ of quarti, healthy coia, noise trauma. So you have this ones that are not active. Autotoxic. So this is drug induced hearing loss. That's what it would look like roughly in the in the coia. And this is synoptotopy whatever that is can read about it some more there's a so if you come out of the full screen the tools available on bkaos.comsl you can read more about it description at the bottom of the page have some equations and things and also go out. This is all available under creative comments license. And we have other implant simulations. This one is uh when you turn your microphone, it will show you how a implant converts audio sound into the stimulation of each of in this case 22 electrodes. Yeah. So you can play with the parameters there. Go check out this tool as well. bios.com/cia scene for cockia simulation. We also have a implant insertion tool. So we had a recent update to those. Yeah, this one you can play surgeon and try and insert a an electrode array into the inner ear into the without damaging it. Because if you yank it or something, you get this red indication that you did some damage. So you probably will get your surgeons license V failed. You did something like that. So you can try these tools. Yeah. And we have many more. They're all on the landing page. Go check out the website bony.com. We have biology, cardiology, data science, some games, healthcare, hearing and speech imaging, math, music, neuroscience, physics, and vision. We have some really interesting optical illusions there. Let me know if you want me to cover any of those. And I'll see you next time. Bye.