 You can send a lot of light through human bodies. So if you take a flash and stick it inside some of these throat and trigger the flash, you can see light go all the way through their mouth coming out through their legs, also through their throat and all their faces. You can also see that the light is predominantly red, and that turns out to be important, because not only is human flesh, it also has color. Most of the color from human flesh comes from hemoglobin, which is the red mixed inside blood. And hemoglobin has a few different forms. One of the forms is oxyhemoglobin, or oxygenated blood. One of the forms is geoxyhemoglobin. They have different absorption spectra, which is to say that they absorb light at different rates, at different wavelengths. Specifically, if we look at 6669 meters, there's the largest difference between the oxyhemoglobin absorption and the geoxyhemoglobin absorption. And over by 800, the absorption is almost even. So this device that we have here has two LEDs on it. One is at 6669 meters, plus one is from the LEDs, and not single wavelength devices. The other LED is at 850, which is close, but not exactly this whatsoever isospecific point. Using those two wavelengths, we can measure the color of the blood in a person's body, and color is another way of saying oxygenation. So we can measure or approximate the oxygenation of a person's blood. And because light passes through human tissue fairly well, and because photodetectors, especially photodiodes, are very sensitive and linear, we can do this over moderate distances. So this device has a distance between the emitters, the LEDs, and the detector of about three centimeters. So three centimeters from here to here, which means that on average, a photon will penetrate about one and a half centimeters between two persons, which gives us the surface of the cortex. Blah, blah, blah. The oxygenation of blood in the brain is useful because it's a good metric for overall metabolic activity. When energy is used in the brain, that causes more blood flow to come to that area. The mediating factor in this chain is actually carbon dioxide. Carbon dioxide increases blood flow. So when you use oxygen that produces carbon dioxide, which causes more blood flow to the area, which causes osteoporosis in the brain. And this is the principle behind fMRI, which is the most commonly used imaging technique in our science today. So it's a very well understood mechanism. The point of doing this is that when you're doing these measurements, you need them in real time. And given real-time feedback about Cindy's body, most people can learn how to control that signal. And this applies to many different types of signals. This applies for heart rate, this applies for skin conductance, how sweaty your palms are, if you're told in real-time, how sweaty and how largely conductive your skin is. You can learn how to control that. You can also learn how to control your finger temperature. And as it so happens, you can learn how to control brain blood oxygenation as well. It's not too hard. About 60% people can get decent control of the brain blood oxygenation within a couple of weeks. But their 40% should take a few hours of practice if they all see what you're doing. So what we're seeing here is my diagnostic software written in Python, of course. It shows three signals. There are also some other signals down below, but they're not as important. The middle one is the infrared signals. This is the 850 ampere light below we have the red. And the top is the ratio of these two, which gives us an oxygenation metric. And as you can see, there are a lot of spikes in the red and big red signals. And that's because he's not dead. Specifically, his heart's still weak and doesn't have those mechanical hearts that are at appropriate pumps instead of listening pumps. You can also see that there are some long-term changes on the order of blood blood birds. Those are called myer waves or basal waves. And those are variations in blood pressure that are mostly or sort of indirectly influenced by reading in some complicated ways. You also see some large spikes. Those are motion artifacts. So if you move your eyebrows, there'll be some large spikes that will be very short in transit. These large spikes and these myer waves show very well in the raw absorbance or transmittance data, but they don't show very well in the ratio data of the oxygenation data. And that's the main point of taking that ratio. Total blood volume is really interesting and a great thing to measure, if you can measure it, but it's hard to measure because there's a lot of noise sources that influence that. But most of the noise sources that influence each of these raw readings influence them equally, and so don't influence the ratio very much. So one of the cool things about this program is that what it's actually doing is taking the text in this box and evaluating it, bringing eval on that text with a set of local and local variables. So if I do this, then all the values get divided by 20. If I decide that I want to take the square root of it, I can do that. If I want to normalize it, I can do that too. So this didn't actually change on the shape of the graph, but if you looked at the left side, you'd see changes on the axis. We can also normalize this, and we can see two and three. We can overlay them so that they're both on the same graph. And these graphs are just the WX Python plot widget. So there are a few things I've added to our right clicking and stuff like that, but all the graph is done by WX Python. So the graph is actually, oh, and that WX Python widget uses numpy as a stack, and so it's relatively fast. We can do 100,000 point graphs without too much trouble. So, yeah, if we want, we can also clip, so we clip to 0.95 and do much. But you can do any kind of transformation that you can do in a single line with this program. That's really useful for debugging and for testing out algorithms. So because half of this is a hardware device and half of this is software, it's really annoying to program hardware. It's really annoying to program C++ or C for microcontrollers because the device that I'm using has four kilobytes of RAM and four kilobytes of long-term memory program space, flash space, and it's a 10-minihertz processor. So trying to program for that is a really restrictive environment, and it's much easier to practice my algorithm than to Python. And if I can see real-time feedback, that's quite useful. So, and see, I've got the code up here. Oh, and that program is open source. So, yeah. So this is basically the code that controls those graphs with that value. Is this the one size bigger? Yeah. So we create a dictionary for the local variables using the dictionary from one Python space as the starting dictionary. And then we just basically add a bunch of other variables with values based on that, as to see my convenience and then you can do calculations using all the facilities in NumPy in those single line things. So that's just my debugging program. There's also a lot of other stuff you can do with it. So the main software that I've written is designed to be easy to use and to give feedback in a format that's really easy for humans to interpret. And as computationally simple as simple numbers are, they're actually really hard for humans to repeat and to process code. It's much easier to process motion over a language statement, for example. So right now when your cerebral blood oxygenation is decreasing, it flies forward. And when your cerebral blood oxygenation is decreasing, it flies backward. We would take a massive GPU array with some really good algorithms to get it to process as an internal statement, but as simple as Python for humans. So if I want to think of something that really irritates me, do you see some change? That's a good question. Let me try to read it and read out a lot now. Take a deep breath. Breathing is complicated and it doesn't have the effects that most people expect it to have. It does have effects, but they're not. The frustration thing is more likely to have an effect. Frustration tends to... So right now we're measuring from the dorsal prefrontal cortex. There tends to be an antagonistic or inhibitory relationship between the prefrontal cortex and the limbic system, the lower down deeper emotional areas of the brain. So the limbic system is heavily active. That tends to shut down the prefrontal cortex and vice versa. That's likely to shut down the prefrontal cortex and make his limbic system more active. So that should tend to make a sickle decrease. That said, it's often hard to interpret things on a single trial basis. So you might only be able to see this if you had him in a randomized fashion. I didn't think of puppies, unless puppies would be annoying. I didn't think of something that annoys him. Maybe rabbit puppies. And then average all those trials together. There's just so much variation in what's going on in a person's brain normally that it's hard to pick out a single trial of things. Usually what does activate this or make things go forward and make that graph on top go up is a sort of... Excuse my hand weakness, but a Jedi focus type, mind trick type thing. You have to use the force in a really cheesy way and suspend your disbelief about whether or not you can control your reality just by thinking about things moving and just willing to move. That tends to work more often than not. It seems that most people happen to a distilled concentration focus state that involves a lot of prefrontal activity and that state gets picked up and gets translated into forward motion. So when people concentrate and focus it's really hard for them to move forward. They tend to move forward. When they concentrate and focus, it's really hard for them to move backward and move forward. It doesn't care which direction you're focusing on. Just back your focus. Anyway, so this software took me a while to work this well because that landscape you're seeing is not real-time generated. It's basically an Abbey file and what I'm doing is I'm decoding one frame from the Abbey file and displaying it on the screen. When it moves backward, I see it to a previous frame. When it moves forward, I see it to a subsequent frame. I like to do things in a cross-platform manner. Cross-platform toolkits or libraries for decoding videos are either only C++ or C or lane. So I struggled a lot with different things to get that to work. There was a media control widget in WX Python which worked pretty well. It worked really well on OS X. It worked okay on Windows XP. Didn't work at all on Linux. And then Windows 7 came out and that widget for Windows used the Windows Media Player ActiveX control along with some a little bit of DirectShow stuff. And the ActiveX control for Windows Media Player version 10 and earlier would display frames if you had a view pause and you see it, which was perfect for you. And in Windows Media Player version 11 and later, it didn't display frames. So that stopped working with Windows 7. What I'm doing now is I'm actually using MPlayer as a separate process. You can communicate with MPlayer using a pipe. So I'm just telling MPlayer to get a pipe. Seek space X% whenever I want to seek to that position. And MPlayer can also, on Windows and I think on Linux or on Linux, they can display it onto any window handle they can display if they give to it. So I just give MPlayer a window handle on a startup. And MPlayer handles all the rendering and all the decoding for that new file. So this saves me from having to do any Direct3D or OpenGL code for display if MPlayer handles all the accelerations itself. So that works pretty well. But I can't do that on OS X because MPlayer doesn't know how to dump things into handles on OS X so I'll just platform like this. Anyway, blah blah, software, blah blah, brain stuff, having fun? Yes, yes. Yeah, so the idea behind this is that if you do biofeedback like this for a while you can increase your brain's ability to either do cognitive metabolic work or to have what's applied to that again. So on the left is before doing sessions with this instrument on the right is after, I think, 30 sessions. These are spec images or a single foot on computer photography. It's kind of like PET. It's an imaging method that uses a radioactive tracer. But instead of, well, that's complicated. What this does is it just measures where blood is flowing and how much blood is flowing to each area. So blue areas indicate lower than average for that age and sex blood flow. Yellow areas indicate average and red areas indicate red and white indicate above average. On the left you see the legend that did the standard mutations. So this is like a little bit over two standard mutations above normal. And as you can see, blood flow below normal, before roughly normal, afterwards all over, so there's still some areas especially in the anterior or single area that still have high blood activity and another person with a different issue. Similar things. A lot of the high blood activity, lower activity in the temporal regions was resolved after doing, I think this is more like 20 sessions. So whether this actually helps people in long run with whatever they want to be helped with is still technically up for debate. There's been a lot of data that's been collected on this question. All the data sucks. Most people who've been doing these with my grandfather have not been like to live in control groups because doing so is expensive and complicated. So it's hard to say whether it works from a purely scientific perspective, but it's promising. So that's what we are now. How does it come back to the traditional EMR techniques? So this is what we call near infrared imaging. Near infrared spectroscopy is for functional near infrared spectroscopy FNERS. So MRI has a few different gradients. There's the standard structural MRI technique, which just gives you an image of the shape of the brain, where the cortex is, where the white matter is, how large the vegetables are, whether there are any novelties like scars or other things like that. There's also the T2 star wing imaging which gives you an image that's mostly structural and has a small functional component to overlay on it. So the T2 star wing signal gives you about 2% or so that the total signal has a functionally dependent component to that signal. So with that T2 star thing, if you take two images at different times and then subtract them, what you're left with is all this exclusively functional information. And that's the imaging method of the brain with a heat map on top of it saying area X of the brain is involved to the task Y. Now, fMRI requires a superconducting magnet and requires a few liters of liquid helium and many liters of liquid nitrogen. The instrument itself costs about $3 million. And what it does is it measures essentially cerebral blood oxygenation. The functional aspect of the T2 star wing signal is actually hemoglobin oxygenation. Oxo-hemoglobin has different magnetic properties compared to teoxygenation. So we're measuring the same thing. We're also measuring blood oxygenation but we're doing it through an optical method instead of a magnetic method. And that means that we don't need to go back to magnets which drops the price by, you know, a small amount. It also means that light's a lot harder to direct in a brain than our magnetic fields because human tissue scatters like tremendously. So it's a lot harder for an instrument like this to get good spatial resolution. It's quite easy for fMRI to get good spatial resolution to get spatial resolution between 2 milliliters or so. Sometimes as low as a quarter of a milliliter per box. And because of the light, I don't know, 4 centimeters or 4 cubic centimeters per box. Oh, and this also has much better temporal resolution. You can take as many samples per second as you want. Currently I'm taking 10 because I'm having to make it faster than ever. But fMRI is looking to sampling rates around. I've seen some newer stuff get down to about 200 milliseconds but traditionally it's about 2 seconds per sample. So when you have 2 second samples you have interesting issues with heart or pulse aliasing your sampling rates much slower than pulse. So we're going to take the sample at a different part of the cardiac cycle which is going to be you'll have some emotional compensation to choose and so on. So there's a lot of noise in fMRI data measurements. Not enough noise to make it not work. It's just an omnipresent issue back of fMRI. Anyway there are a bunch of other differences too but you don't have to worry about having a lot of limitations like this. Other questions? I'm basically done. So I used the NumPy main space as the initialization for the local variables that I used for the evaluations. It's just to make all the NumPy functions easily available like side or potential. Is it different from NumPy and normal? No, it's not a NumPy dict. It's a regular Python dict. The contents of that dict are the NumPy dot underscore underscore dict very well. So it's just that's where I'm going to find my values for sitting into that dict. The underscore underscore dict underscore underscore property is the namespace for whatever module you're working with. So you can use that with anything else. If you just want to copy namespace or just copy that dict. It's also I think is returned by DIR. DIR space NumPy that will give you that dict or I guess actually it will just give you the keys for that dict but it's almost the same here. Other questions? Why are you doing this? Because I think it will help people and because it's cool. Yeah, that's the best way. So I purchased another headset called the Mind Wave from a company in Neurosky. It lets you go through a whole lot of fun things like you know you look at animals on screen and playing to shoot and things like that but I got out of all of that maybe within days I think it was interesting and now what I do is when I start to read I start off a journal over there and I am training myself to learn to focus better because I don't know how to quantify this but I just am able to teach myself to focus better. I look at the graphs occasionally or I set an alarm or something and what I have found is some 10 minutes after I have started to read is when I am able to really focus and what also happens is in case someone disturbs me it can take me a good 10 to 15 minutes to focus it. I'm sure you have already researched this. Yeah, so there's a lot of behavioral data on focus and so it's rich with interruptions and it also tends to show that after an interruption it takes usually 10 to 15 minutes or so before a person gets back into the task there's a lot of data that's specifically on programming too so one of the issues with programming is that it's a high working memory load task so you have to have a lot of your mind in order to program effectively you have to keep in mind several different local variables you have to keep in mind functions and a few different modules and it takes about a fair amount of time and as soon as you're interrupted all that stuff leaves your working memory and is replaced by, I don't know, cricket or whatever so then it takes another 10 minutes to get back in and if you're uninterrupted you can keep those things in working memory and actually get stuff done so yeah the neuro sky system is measuring brain activity patterns it's measuring the electrical activity and the electrical measurements that we make are actually a lot more indirect than metabolic measurements even though neurons are primarily electrical and blood flow is more of a secondary trait it seems to me that the blood flow correlates better with the overall the overall activity level of neurons whereas the electrical activity just shows patterns in the information flow so the reason for this just comes from the way neurons are shaped within the cortex so when a neuron receives information from another neuron you get an electrical dipole you get a current flowing in one direction at that neuron synapse and that produces a dipole that you can sense on any line parallel to that dipole so if you have one electrode on one side of that line and another electrode on the other side of that line you can see a voltage difference if under the hand your electrodes are oriented that perpendicularly to that line you'll see nothing you might also have other neurons oriented in the opposite direction so when they receive information they produce another dipole in the opposite direction if you put those two next to each other and you put your electrodes at some distance away on the skin you'll see nothing on average you'll see a very very small signal because they'll cancel out so when you do get stuff from neurons making it to the scalp in EEG what's actually happening is you have a large number of parallel neurons in one area all receiving information at the same time so it's the synchronicity or the synchronicity of information flow that primarily drives EEG rhythms and one of the most synchronous or one of the most synchronous sets of rhythms in brain activity are the idle rhythms so the largest rhythm that you see in EEG by far is the alpha rhythm the amount of time it takes for signals to go from the visual cortex to the switch yard of the brain the foulness and back is roughly 100 milliseconds incidentally the alpha rhythm is about 10 hertz so 100 milliseconds when the visual system is not doing anything and it's not processing information there's still a tendency for neurons to fire and the neurons will fire and send information to the foulness and then those neurons will fire and send information back to the cortex and this will go on and on for a while there's some positive feedback there so each time it goes through this loop the signal gets amplified until eventually you have a big sinusoid waveform going on between the visual cortex and the foulness and that only stops if you interrupt it with some chaotic high entropy data like looking around so if you close your eyes so that big rhythm is actually just pure brain doing nothing and the actual activity is what shuts down that rhythm so that synchrony measure though is still quite useful it tells you something about what the rhythms are in the brain and what kind of mental states going on which is why I can show some information about focus for attention so the neurons guys try to tap into that and try to distill some metrics that attract that that relate to that however those measures are or most of the other measures you can get with ED are hard to control because they don't map on to the overall activity of the brain it doesn't map on to our subjective experience of actually thinking we're doing some tasks as well and that's generalization but all generalizations are false but it also implies because I'm competing with NASA but this is not so with neuroscience and other EDG systems it's more of a brain state kind of thing that's getting into sort of a mindset or getting into sort of a mood with this with EDG or your input spectroscopy by feedback it's much more of an exercise it's much more of a pushing yourself to use your brain harder and to think it will harder than you normally do so they're very different systems Any other questions before I appear with you guys? You can ask questions if you want Thank you Thank you Thank you Thank you Thank you Thank you Thank you Thank you Thank you Thank you Those images cost about $2,000 each They were not taken with us That technology is a public spec it's commonly used for doing well it's not to back home movies but it's somewhat commonly used for diagnosing brain problems of race sorts So some people who have come to my grandfather's clinic for therapy have the money to give them $4,500 which is that sort of thing Is this a radioactive course here or an MRI? Are there radioactive places that are more costly than MRI? MRI doesn't use radioactive tracers MRI also doesn't because fMRI doesn't tell you overall activation levels they only make a difference in activation between two times between two times So a standard fMRI experiment will be take this subject in have them sit down have them look at words that are consistent with double letters so not actually real words but still consistent letters do that for 30 minutes and then just send them to a different state where they're looking at actual words that have letters in the right shapes and can be pronounced and been allowed and have meanings You take images from both of those to lots you send them together to take an average of them you subtract it to do some statistics eventually you'll get an image that shows you which areas of the brain are active in the experimental conditions for real words So that doesn't tell you about baseline activation and you just can't get that information with fMRI because you have to do that subtraction but it's actually you don't have to do that subtraction you get baseline activation levels so baseline blood flow levels and compare that to a population we can see what's hyperactive compared to what's hyperactive here and unfortunately it involves radioactive people very small dose it's very comfortable to and see it's comfortable to about like 500 nannies or something like that even 500 nannies and that's what we got that's the point there's the has the thing this is Geeks of RFS Geeks of RFS Yes, it's the password Req Geeks of RFS Yes, RFS Hello everybody My name is Laos Ma I'm a security researcher I'm an operation tester I'm also a developer of the security tool I'm a little intermediate