Student Post: The BrainGate Neural Interface System

What's a "physiatrist?"

physician specializing in physical medicine and rehabilitation

definition of physiatrist:
http://www.aapmr.org/condtreat/what.htm

For a decade and more, I've been wondering why seeming breakthroughs in treatment of spinal chord injuries stubbornly refuse to translate into real-life benefits. It's been donkeys years since experimental mice could get much enervation back through freezing, precision slicing, and then approximation of a severed spinal chord. It's been a decade and more since a paraplegic college student was taught to walk by a computer, and it seems as though every 18 months or so someone with a spinal-chord defect or injury gets hooked up to a PC and moves the cursor.

But the folks whose spinal chords are injured tomorrow in auto accidents? They'll be para- or quadraplegics (or dead) just as they would have been a generation ago.

Maybe I've just been reading "framed" science all these years. (and I don't mean stryker!)

Cursor moving is pretty cool, and certainly a big step for someone who can't otherwise move anything, but when do humans get the chips that can move robot arms? C'mon! Monkeys got robot arms! Why can't WE have robot arms? (With flamethrowers! And buzz saws! And...)

[Aside: I hope that PZ has said something to his students about ignoring the hell out of us here in comments]

That is simply incorrect. There are a considerable number of people now implanted with muscle implants that allow them to move their arms, hold things, stand up, even walk to an extent. You recall a few years ago that paraplegic walking down the isle for her wedding? What you probably didn't know is that systems like that have been implanted into a large number of people. That was pretty difficult for her to do, but using that implant to just stand up to reach things, move between chairs, get in and out of bed, and other such basic activities can considerably improve the quality of life for a paraplegic. Similarly, just being able to hold and drink from a cup, to grasp a fork or pencil, or other basic tasks, even if somewhat awkward, can also significantly help quadraplegics. These are not universal, but they are far more common than you probably realize.

As for hooking people up to computers, it may seem easy in theory but it is extremely difficult in practice. How much can the motor cortex rewire itself? What sort of signals should you look for and how should they be interpreted? The signals will likely change over time and depending on present conditions, how does the system handle that? How do you filter out all the other stuff happening in the brain? Do you have to put the electrodes in precisely the right spot or can it be anywhere? How many electrodes do you use? How deep should they be? How far apart should they be spaced? Should use one electrode per needle or several? What is the best shape for the electrodes? What should they be made out of? What sort of damage will it do, and what sort of damage is acceptable? Will the electrodes cause the brain to change over time so the system stops working? Will the electrodes stay in place under real-world conditions? Sending signal into the brain is easy, but how do you get a hundred or more channels of data out of the brain without breaking the skin (which will cause infection)? The brain can move around in the skull, how do your prevent the wires from fatiguing and breaking? It is easy enough to do the analysis when you have a large workstation computer and fancy A/D converter, but how do you do the A/D conversion, spike detection, and spike-train analysis calculation in a wearable unit? These are all issues that are easy enough to deal with in a lab but are much harder to deal with in a portable unit that can operate under real-world conditions and must last for decades if not a lifetime.

You don't even need implants. It can be done simply by wearing an EEG cap!

Interesting. Not having any good references available myself, can anybody tell me what the scale required here is? That is to say, what spacing between electrical sensors is required to recover nerve impulses adequately, and how many do you need?

Praise science!

This is the cool stuff we were promised when we were children (aren't I bit young to say that yet).

If I give my monkey army robotic arms, can they be hooked up through a wireless router so I can control them with my brain...

Your little dog stands no chance now!!

oops, link didn't work. here it is.

This is cool.

As a student of electrophysiology, I have to point out that an EEG could never allow for the spatial resolution that an implant does. And I'm fairly sure that you would need an implant to do communication too (signals both in and out of the brain).

To answer Brain Hertz's question - as good as you can get. But, from the looks of it we are nearing the limits of normal electrodes, as the closer the electrodes are (the higher the density of the chip), the more noise you pick up from neighboring electrodes.

It's a trick problem, and one I look forward to throwing myself at as soon as I can!

Could we use nerves that are not "essential" to control devices? For example, take a nerve that goes to some organ that is not needed, like the appendix, put some electrodes in it, and connect those to an apparatus that responds to these signals and does somehting (like playing mp3s)?

I expect that doing that would require the patient to train himself to send the right signals to that nerve, but I do think that people can learn to do that.

Or is this not possible even in principle?

That's an interesting suggestion Valhar. Although your example wouldn't really be valid, because the signals from the appendix don't go to the auditory processing part of the brain.

Perhaps a better idea would be to grow new nerve pathways to hook up to the new device, or use an artificial nerve pathway to implant right into the correct area of the brain.

I wonder if the brain would be able to integrate the signals though. Certainly sound would be possible, but could one, for example, feel realistic tactile sensation from an artificial tail if it was connected to the somatosensory cortex?

My apologies Valhar, I read your comment wrong.

You are proposing that the innervation to the appendix control an mp3 player, rather than an mp3 player send signals to the brain.

Well, the enteric nervous systemon (the nerves controlling the gut) are autonomic. That is, not under conscious control. Of course, while we're on the subject of brain modification, perhaps that could be changed too. But that brings me back to my above comment - how would the brain cope with novel stimuli or novel systems to control?

Not having any good references available myself, can anybody tell me what the scale required here is? That is to say, what spacing between electrical sensors is required to recover nerve impulses adequately, and how many do you need?

The limiting factor is the cellular organization of the brain, not the silicon. You can make silicon structures several orders of magnitude smaller than cells, so that is not a problem. You would most likely want to roughly match the density of cells in whatever structure you are looking at so you have one electrode per cell. One problem is that this is not consistent across the brain, in fact it can vary widely. Second, the specific best way to do this is not necessarily clear. Obviously you would not want to stick the electrode into the cell, this may be okay short-term but wouldn't last. Getting consistent single-cell data from extracellular recording is extremely difficult. You will pick up signal from several neurons in some cases, if not many, cases. And unlike electrphysiological recordings you cannot move it and try again in humans, nor would it likely help if you are using hundreds of electrodes simultaneously.

Roughly speaking, we are probably looking at the range of microns to tens of microns, but getting that exact is not easy.

The depth is a similar problem. There are 6 canonical layers of cells in the neocortex (the are you would most likely be implanting into), but some areas are nearly missing one or more layers, the relative size of the layers vary, and the depth of all but the top layer (which is not the one you want anyway) will vary. Also, the specific layer you want to look at may vary between regions.