One internship and one *balmy* summer…

In the summer before college started, I did an internship with the MIT Biomechatronics group, which is at the forefront of robotic prosthetic design. It turns out that the prosthesis industry is relatively small, so companies have little competition, and ultimately, incentive to improve their products in technologically meaningful ways (despite continuing to charge… a lot). Instead, Biomech, led by Hugh Herr, designs the new technologies for companies to be able to easily “upgrade” their old devices. Ottobock, for example, sells the Empower Prosthetic Ankle, which was designed in the lab and is now the most advanced prosthetic ankle in the industry.

The point is that the engineering that Biomechatronics does is important and has the capacity to improve the lives of thousands of people - and that’s why I wanted to be an intern!

First, you need a bit of an overview of the process by which amputations need to be performed in order to enable the cool stuff the lab is currently working on. The next post will be about the major projct that I’ve been working on.

Amputations

The AMI

The Agonist-Antagonist Myoneural Interface (AMI) is a surgery designed by the lab to allow amputees to retain some of the ordinary biological functions of their muscles. Practically all muscles have an agonist and antagonist - your bicep and tricep, for example - to enable the the movement of your limbs. The overall mechanical reason for this is obvious, since muscles compress much more powerfully than they extend, but the more nuanced one is that each muscle can then feel the other, allowing one to effectively serve as a sensor for the actions of the other. This relationship, a result of both muscles working together, is something that your spinal cord leverages to accurately allow you to accurately control your muscles and have proprioception.

A good example of this is stiffening your arm in a certain pose - why doesn’t one muscle stiffen more than the other?

The problem with this beautiful biological structure is that it isn’t at all compatible with invasive human surgeries like amputations (sarcasm inteded). Traditional amputations usually connect residual muscle to the residual bone - something which is practically opposite to an opposing muscle (it’s not at all elastic and has no nerve endings to ‘sense’). In consequence, not only do the muscles in amputated limbs atrophy quickly (resulting in smaller and smaller residual limbs - this is expensive considering that one will need new sockets), but the muscle is fundamentally compromised such that it no longer produces similar electromyographic signals (the voltages your muscles receive from your spinal cord) as ordinary muscles.

The solution to this is the AAMI: a specialized surgery which retains the connection between muscles by tying the ends of their tendons together. This knot then rests on the bone, which is grinded into the shape of a pulley so that the muscles are still stretched. The result is that after surgery, amputees can still move their muscles while maintaining their physiological relationship, not only preventing them from atrophying as quickly, but also allowing them to preserve their characteristic EMG signal outputs.

Here’s a picture from the paper, which contains much more (graphic) pictures.

That’s the important part. The lab has also come up with a way to place electrodes for sensing and stimulation in the residual limb through an exposed connector at the end of the leg, but I think that’s less important for now.

My project

My supervisor’s project builds on the work of the AMI and EMG sensing for a novel robotic leg controller that works under neural control - amputees can move the prosthetic simply by… thinking about it, since it just picks up on the muscle signals from the muscles preserved by the AMI!

Although this is super powerful, it has a seemingly trivial limitation - it cannot provide feedback to the user. The human-robot interaction is open loop because the muscles which the robot is sensing aren’t being stimulated in response to a resistance on the robot. As a result, the person can only feel a resistance against the robot through the limbs around that which is being resisted.

That’s where my project comes in - building an interface which allows for simultaneous sensing and stimulating on multiple muscles. This has added difficulty in that stimulators are large, and even one is too large to be strapped to a patient (even for testing), so the interface needs to be able to ‘bit bang’ the stimulation.

So far, work on this has been steady (just got the boards made!) so I’ll make another post soon regarding how the project is actually going.