DARPA's Mind-Controlled Prosthetic Arm Could Be on the Market in Four Years | Hacker News Reader

DARPA's Mind-Controlled Prosthetic Arm Could Be on the Market in Four Years

fastcompany.com

64 pointsraphar21 comments

So I totally want this to actually happen. However, they gloss over all of the actually hard parts. Cool, DEKA's built a sweet anthropomorphic robotic arm. I have no doubt that part works. But you don't just jump to running a human study from that: first, you do it in non-human primates and show that they can effective control the arm with the implant. Then you show that they can interpret feedback from the arm stimulated into their brain. Then you show that the implant is stable and immunologically accepted over a long period of time (warning: deep, hard unsolved problem). Then you show that they continuously integrate it into their body representation (not necessarily hinted by the prior control study). Then you cautiously do a human surgery and see if they can control it, and the brain doesn't reject the implant after a year. Then five patients. Then apply for FDA approval, expedited or otherwise.

AFAIK there's never been an implant study done using the Luke arm. Therefore we've never tried to build control models from cortical activity for it. That's a hard problem in and of itself. Jumping straight to a 5-person human study is insane. Furthermore, the idea that human patients are going to get intracortical feedback from the arm is crazytalk. It's just not happening. Electrical stimulation cooks the brain and causes seizures, and optical stimulation requires gene therapy (!) and has all kinds of long-term problems right now.

TL;DR: awesome I hope it works. They'll bring a needed device to market and solve about 10-15 of the hard problems of brain-machine interfacing for us at the same time!

Indeed. I have no idea why they would want to include solving the "don't cook the neurons" problem in their road-map to a usable product; there are other feedback methods available.

A haptic display of some flavor would certainly be plenty to close the loop in controlling the important variables like "are any of the motors encountering resistance" and "how much pressure am I putting on this thing I'm squeezing in my robot hand." A more advanced haptic display should allow enough feedback signals to hold much more than that. Heck, if one is going to go electrically stimulating neurons, better to do that to sensory neurons in the user's skin than in their skull.

Electrical stimulation cooks the brain and causes seizures

From Wikipedia: deep brain stimulation has been FDA-approved for tremor since 1997. More treatments have been approved since, like Parkinson's, dystonia, major depression, and chronic pain. https://secure.wikimedia.org/wikipedia/en/wiki/Deep_brain_st...

Then you show that they continuously integrate it into their body representation

This part is easy, isn't it? People integrate tools into their body image innately and easily. I don't think that will be very difficult.

Deep-brain stimulation is different than cortical microstimulation. DBS is injected into the basal ganglia and excites an area of the brain that's effectively atrophied. This can be done relatively safely for things like Parkinsons, even though it's definitely not your first choice of treatment. It's still somewhat cooking the brain, however; in the cases where DBS is used that just doesn't outweigh the benefits. Cortical stimulation into primary sensory cortex is a totally different animal, and tends to cause seizures much more frequently. The brain is kind of wired to seize, when you look at it, and it's amazing it doesn't more often. Also, you can't map receptive fields -- which specific areas of cortex correspond to what sensation in what part of the body -- if your patient is an amputee, and there's no "master map" that's precise enough since it varies between individuals widely. Further, DBS is just designed to grossly "drive" an area of the brain; in ICMS you'd want to precisely excite a few cells to create a sensation. ICMS in practice just floods its surroundings with current in a rather blunt way. Optogenetics are a solution but have their own problems (i.e., requires genetic transfection) that make them unacceptable for use in humans today.

This part is easy, isn't it? People integrate tools into their body image innately and easily.

Yes -- but as a patient lives with a prosthetic full time, as opposed for a few hours at a time like in most monkey studies, the brain will adapt to the prosthetic and the neural representations shift. You need to be able to deal with this and understand what that learning looks like, which we can sort of do right now, but no one's really done a long-term-continuously-living-with-a-prosthetic study. You generally don't try to solve complete unknowns like that for the first time in humans. Humans would be a poor platform, anyway, since you can't observe and record from them 24/7 in a controlled environment.

  > Also, you can't map receptive fields -- which specific
  > areas of cortex correspond to what sensation in what part
  > of the body -- if your patient is an amputee, and there's
  > no "master map" that's precise enough since it varies
  > between individuals widely.

This is a very good point but I don't think it's an insurmountable problem. After amputation or deafferentation the "unused" parts of the cortical map get taken over by adjacent representation areas. However, after hand transplantation cortex recovers some of it's normal organization[1]. So it's at least possible that cortical stimulation could lead to the same result.

1. Frey et al. Chronically deafferented sensory cortex recovers a grossly typical organization after allogenic hand transplantation. Curr. Biol. (2008) vol. 18 (19) pp. 1530-4

the brain will adapt to the prosthetic and the neural representations shift. You need to be able to deal with this

Why? Sounds like the brain is dealing with it pretty well on its own.

If the brain is good at one thing, it's taking note of, and adapting to, useful relationships in the structures it finds itself embedded in (affordances).

I think if you can get some kind of probe or grid of probes into an area that has even a vague chance of getting some activation during physical therapy sessions, then there's a pretty good chance the patient will be able to get from there to skilled motor control through practice.

>> first, you do it in non-human primates and show that they can effective control the arm with the implant... Then you show that the implant is stable

Uncited, I'm not sure my reply is worth much - but such studies have already taken place.

Non-human primates have shown dexterity with such devices - but the 'interface' has a life of around 6 months due to prolonged , subtle vibrations in the 'brain probe' and scarring.

EEG inputs, while non-invasive, are susceptible to interference and work best with additional triggers such as speech to extend ones 'vocabulary' of movements. The reason being it's hard to to calibrate.

Lastly, electrocorticography (ECoG), seems a happy medium: picking up high frequency signals within the skull and atop the brain. So no issues with scarring and the signals are easier to map to brain regions.

'Google Tech Talk' on the subject (1hr)[1]: http://www.youtube.com/watch?v=mFWnTONOvVo

[1] Includes footage of monkeys + robotic arms @ 9m30s

> Finally, laypeople will benefit from the Defense Advanced Research Projects Agency's (DARPA) mad scientist projects

Yeah, because that whole Internet thing and GPS didn't really count for much...

If this turns out to be a real FDA approved implant, the potential to improve the lives of paralyzed people is staggering.

This looks like the DEKA arm. Here is Dean Kamen talking about it:

http://www.youtube.com/watch?v=AoY1cItRiHA

Even if it isn't the same thing, this is a pretty interesting video.

I worked at DEKA for a while. Although I didn't work on the Luke arm, I got to see a few demos. It's my understanding that the control system is very loosely coupled to the user inputs. I think it's an expansion of the control mechanism for the iBot, which took input from hundreds of sensors but could be controlled with just a joystick and a single button. The Luke arm currently is controlled by pressure sensors in the shoes and on the chest, so you wiggle your toes and flex your muscles to command the arm. It also has very impressive force feedback. But other input mechanisms and feedback systems could be implemented, including direct mental manipulation.

Edit: here's a good demo video http://www.youtube.com/watch?v=R4X_l2XOz8g

At first, I thought that it would be very cool to have such an arm, for a second entertaining the thought of someone chopping their arm off just so they could be a cyborg. However, a moment later, I thought "three arms", and immediately I saw the improvement.

I bet some rich guy will soon turn into some robo-Shiva.

Cue the Dr. Strangelove jokes.

More like the Six Million Dollar Man theme...

Watch the arm: http://www.youtube.com/watch?v=59YKlP--PhU

Did Arnold Schwarzenegger's face flash through anyone else's mind while reading this?

Sigh, why does this have to come out of the military industrial machine?

Little known fact: over 1,000 soldiers in Iraq and Afghanistan have lost a limb (out of 41,000 wounded that survived because of more advanced medicine).

I hope they don't use this development to justify that it's okay because we certainly don't talk about it much otherwise.

Isn't it possible that this is why it comes out of the military industrial machine? Surely a lot of advances in trauma treatment, bandages, surgery, and all kinds of other medical advances have come from the military industrial machine as well.

I don't see the need to politicize this, and I don't think the numbers associated with the wars would be little known to people here.

Indeed. If you watch Dean Kamen's TED Talk, he says that's part of why he created it. It's not to justify sending soldiers to lose their limbs, it's to help those that already have. He showed a lot of respect and ...sadness, I guess for the soldiers who had lost their limbs, and genuinely seemed to want to help.

Why does it matter where this comes from? It's great technology that will improve people's lives.

Historically speaking, the majority of innovations take place as a consequence of war. Advancements in aircraft, electronics and industrial processes, medicine are all prime examples. Read up on the technological breakthroughs that were expedited by world war II that might have other otherwise not taken place. Think about the mass production of penicillin and skin grafts. Necessity breeds innovation.

Reply on news.ycombinator.com