Mind-Controlled Prosthetic Arm

Losing a limb often means getting fitted for a prosthetic. Although there have been some scientific and engineering advances (compare a pirate’s peg leg to “blade runner” Oscar Pistorius’ legs), they still are just inert attachments to your body. Researchers at Johns Hopkins hope to change all that. In the Journal of Neural Engineering, they announced a proof of concept design that allowed a person to control prosthetic fingers using mind control.

The test subject did not lose any limbs but was the subject of brain mapping due to epilepsy. The electrodes used to map the brain also allowed the subject to control a modular prosthetic limb. The mapping told researchers what parts of the brain activated for movement of each finger. The researchers then set up the prosthetic to operate from the same regions of the brain.

The process required 128 electrodes on a small rectangular sheet of film. Each electrode measures about a 1 mm area of the brain. Initial tests showed the subject could correctly control the limb 76% of the time. Further refinements drove the accuracy up to 88%. This was without special training. The entire experiment took about two hours. The video below shows the arm in action.

20 thoughts on “Mind-Controlled Prosthetic Arm

    1. It kind of is, The arm itself runs on about 5 uPics and an arm processor (cortex I think? don’t really remember). This particular work though wasn’t the point of the study. They were working on this poor girl to treat her epilepsy and since they had electrodes there they just happened to add on a third arm. I’d call that hacking the human body.

      1. The arm doesn’t interpret thoughts you have into desires and actions, you actually have to learn to use it like you would a T.V. remote or a game controller. The arm responds to specific brain patterns in certain areas of the brain and you train yourself to activate those patterns to produce the response. Once you become good enough at it, the act becomes more or less transparent to the user and it in effect becomes a new limb. As I understand it, it typically takes between a few days to a week in order to become reasonably proficient at using such a limb interface but I would expect that if someone could use one for months to a few years, a lot of the apparent clumsiness you see in videos like this would go away.

        1. The article that I read asserted that is not the case for this system, it’s actually reading the movement from the motor cortex directly. Thus it has “no learning curve” (I wonder).

          1. Shannon,
            You’re right. I jumped the gun a bit because I tend to jump on and off here in between bits of work. It really goes to show that I should slow down a bit and be more thoughtful. I tried to get my hands on the journal paper today but found that I don’t have access. You can find it here, with at least the very interesting abstract publicly available
            http://iopscience.iop.org/article/10.1088/1741-2560/13/2/026017/meta;jsessionid=91B10173F7632E0FD3E16057FA4D65DA.c4.iopscience.cld.iop.org

            Ultimately, it looks like they were specifically mapping motion of an existing hand to that of the robotic hand. The paper details success rates for their classification scheme. Which hit around 70-80% depending on constraints.

    1. Actually the skin-barrier can be sufficiently easily be held intact through using implantable wireless transmitters as in cochlea-implants etc – power for these implants is currently supplied via induction, the signal can be modulated on top of this…
      The problem i see though is that required open-brain surgery is the most severe and risky procedure i can imagine and probably something to avoid in any possible way…
      I admit, recording from residual peripheral nerves is a less sexy research topic but can be made in a more specific and less invasive form…

      Still the involved signal-processing in this work is impressive!

      1. Actually, the real problem is the surprisingly aggressive accretion of scar tissue and immune response to electrodes implanted in the brain region. Because of this, these sorts of implants can only be left in the brain for short periods of time before permanent damage results. This particular work is notable because while thought controlled arms are nothing new, (especially this arm which has been the poster boy for thought controlled prosthetics for damn near 10 years now) this arm is the first that allows granular control of individual fingers rather than cycling through preprogrammed motions and positions. They also used a flexible surface electrode sheet to do the brain mapping rather than a deep implant such as a Utah Array, so there is hope that this will enable the implants to stay in the body for longer periods of time at only the cost of higher processing power to interpret the signals.

        1. You seem quite knowledgeable concerning this installation. Regarding the flexible surface electrode sheet, do you happen to know the dimensions of it. I’m curious if the effect of mapping too closely the signals that move one finger to the finger adjacent to it is problematic like it is to musicians.
          There are “rewiring” techniques employed to segregate/separate the spacial mapping of the fingers so that they don’t move as one. I myself have experienced this problem as a guitar player. The techniques range from placing one finger in hot water and an adjacent finger in cold while one is moving in circles and the other moving back and forth to other similar methods employing electrical shocks or textures and such. The different sensations combined with the motor movement signals reinforces the mapping out of the very narrow Everything-Is-Basically-The-Same mode into a larger more distinct set of areas. I guess my point is that similar methods may improve the function of the BCI whatever it’s used for.

          1. Zenzeddmore,
            I do not know the exact sensor they used, but it was listed as a high density ECoG array. That tells me it is probably about 4 mm^2 in size, with somewhere between 60 and 120 separate electrodes. I would guess that the signal bandwidth goes up to about 75-150kHz, and that they get around 80dB of dynamic range. From a quick literature search, this will give you a brain mapping resolution of a few hundred umeters. From there, there is some cool math and signal processing you can do to separate the mapping of signals that overlap or co-fire. Otherwise, yes it is possible to train yourself to a degree to operate them separately and techniques to train your hands would likely show up as changes in brain mapping, but I personally haven’t seen anyone actually try and map that so I don’t really know. I do know that some of the authors of this paper used to use a modified guitar hero controller to train people to use the limbs and interfaces.

  1. I expect that if they can get the electrodes and surrounding neurons to last long enough (yeah that is a big problem) you will eventually see 100% accuracy due to https://en.wikipedia.org/wiki/Cortical_remapping

    Researchers in Australia have got around the die-back problem by putting the electrodes on a stent and guiding it up to the right location, from the inside via blood vessels. It always stays outside of the blood brain barrier so there is no chemical interaction, but the received signals are still usable.

    And you don’t need to hack a hole in the persons skull….

Leave a Reply

Please be kind and respectful to help make the comments section excellent. (Comment Policy)

This site uses Akismet to reduce spam. Learn how your comment data is processed.