Agonist-Antagonist Myoneural Interface

https://www.media.mit.edu/projects/agonist-antagonist-myoneural-interface-ami/overview/

by kaycebasques • 6 months ago

73 7 6 months ago

media.mit.edu

w10-1• 6 months ago

This is a very significant result for a high-traffic real situations -- work to be proud of.

To summarize:

When a prosthetic lower leg was attached, they connected antagonist muscles to the leg, with sensors from those muscles in a control loop to the leg (ankle), mimicking how proprioception works. (The sensors are the new interface technology.)

The patient knew and could move the position of the foot when he couldn't see it. He walked up stairs with the usual natural coordinated movements. And he felt like the leg was part of him.

It's one thing to (cortically) plan and execute and track prosthetics visually; it's another for the cerebellum to autonomously monitor and control them, and not to feel cut off.

This seems workable as a standard of care for arms and hands as well. And in this case, it was installed years after the leg was lost, so it works for retrofits (granting this is N=1 young patient in otherwise excellent condition).

kaycebasques• 6 months ago

I stumbled on this work while researching cyborgs. There's quite a bit of jargon on this official page. The main idea (as I understand it) is that previously, when you got an amputation above the elbow (for example) the bicep and tricep muscles became dead ends. They would just attach the ends of the muscles wherever feasible. Apparently the act of one muscle like the bicep contracting (agonist) while another related one like the tricep extends (antagonist) is a really important feedback loop in our brains. AAMI essentially restores this feedback loop, making the prosthetic feel like part of the body. The lead researcher is apparently himself a double amputee.

Osseointegration was another example of interesting real-life cyborg technology that I stumbled upon.

observationist• 6 months ago

The real magic is the plasticity of the brain, the cerebellum as well as the cortex. In this case, they're tapping in to existing neural structures by correctly aligning the prosthetic with the configuration of the original.

One of the ways the brain works is to construct and predict, or assume, the way things will be, and use proxy signals to confirm the success or failure of a prediction or assumption. By aligning the prosthetic, the proxy signals result in not only successful manipulation of prosthetic orientation and placement, as if it were a foot, but the patient feeling feedback from the prosthetic, as the brain reconstructs some pieces of the sense of actually having a foot.

Phantom limb and phantom pain taps into some of the same phenomena, and advanced prosthetics with myo/neural electrodes actually tap into the nerves and nerve endings to create new input/output pathways, with some experiments actually succeeding at reproducing touch, hot/cold, pressure sensation, and control of the cybernetic limb.

Input is hard, but with nonintrusive ultrasound techniques seemingly working for certain deep brain stimulation, maybe it won't be too long before we see highly precise ultrasound phased arrays able to stimulate precise neurons when installed over a patch of skin, and we'll get wearable third limbs, prehensile tails, and other cyborg augments with full neural io integration for non medical purposes.

Very cool paper!

whymauri• 6 months ago

Hi, it was a long time ago but I worked on this and can answer high level questions.

https://www.nature.com/articles/s41598-018-38096-z

uoaei• 6 months ago

For what percent of trial participants did the interface function well enough for normal function?

Did you find any evidence, even anecdotal, about alleviation of phantom limb symptoms? I imagine it would be complete and instantaneous but I'm not an amputee with any experience.

PetitPrince• 6 months ago

Man, is this still sexy science !

In a parrallel universe, I am still be working in that domain (I was in Silvestro Micera's lab (he did similar kind of feedback for the hand) for my Master's thesis - also a long time ago; it didn't go so well due to an expectation mismatch from both myself and my supervisor)(I now work as a software engineer... pay and oppotunities are better).

If I understand correctly (I only skimmed your paper), the method you used is to take a muscle, cut it in two lengthwise, use those as a pair of muscle to graft, then put two nerves close to it and pray for re-inervation. Then you use EMG as a basis for your signals.

- Help my brush up my EMG knowledge: what's the tradeoff in choosing the muscle ? For a human case such as the one provided in the link, do you have the same signal quality choosing a smaller or bigger muscle ?

- I assume you're using intramuscular EMG (you're doing surgery anyway, so you might as well put some electrodes). How does this behave over time ? I had some experience in brain-computer interface, and I know scar tissues and the like is a real issue that can come up over time.

thomasm6m6• 6 months ago

post title should be dated. earliest archive.org snapshot is from 2018:

https://web.archive.org/web/20180501000000*/https://www.medi...

Agonist-Antagonist Myoneural Interface

Related Stories