Man compares his prosthetic hand to a 3D-printed one
3duniverse.org
3duniverse.org
Models like this are quite exciting - assistive devices and prosthetics are incredibly expensive due to low volume and customisation requirements. This is preventing innovation and production of new devices and technologies.
They are also (rightly) highly regulated. A next step is trying to figure out how to maintain quality and safety in these kinds of DIY projects.
Also, as an FDA certified medical device, the rough rule of thumb (that I was taught while on an internship) is ~5-7 times mark up on BOM+assembly cost.
The -real- amazing thing about this is the idea that we can use far cheaper materials with much lower safety margins, since we can then just reprint individual components that break - remember the price of the materials isn't the bulk cost of material, its the price it takes to shape them, which is quite high for traditional methods, for low run parts like prostheses. What's going to be amazing is the combination of newer myoelectric devices, with these new manufacturing/design approaches, that -should- enable intermediately priced prosthetic.
Of course, the biggest cost saving comes from not being an FDA approved device at all.
I think the markup varies quite a bit more widely than that, probably along with device complexity. As someone who has been a member of the development teams for various complex FDA-regulated medical devices (think image-guided surgery), the markups I saw were more in the 10-15x range.
I can see regulation as soon as you are penetrating the body - screwing something in, surgery to make an attachment point, and so on, but otherwise? Don't make people's lives even harder because you think you can think and make decisions better than they can.
I'm making generalisations that may extend beyond this particular device for this particular user because its an entire field that isn't well serving its users at the present time. Implementation far lags technology because business and operations models don't exist to promote the development of these technologies. The kind of approach article is suggesting is promising, but not a complete solution.
The reason why regulation is needed is that the people making/designing/marketing the device have by doing so made an appeal that they are an expert.
Often, failure of these devices have far higher consequences than failure of typical consumer devices (maybe ice axes and crampons are also in this category!). Its the same reason why doctors need to be licensed to practice, whether it be for open heart surgery or to diagnose someone with a runny nose.
Probably assistive devices (and medical technologies in general) are over-regulated now, but that doesn't mean the solution is no regulation.
In the end it doesn't matter much if my prosthetic finger breaks - I'll just print a new one. Could I get a rash if I use PVC plastic vs some bioplastic? Sure,but I can also get a rash from the $5 sneakers from Walmart. Is medical grade stainless going to be better than something I get from Home Depot? I would imagine so, but I buy all my metal from Home Depot and the like, and I have yet to suffer any real consequences from that. I have a stiff neck from the $10 pillow from BB&B - shall I buy a $3800 orthopedic, medical grade, hypoallergenic pillow? Nah.
I agree that there is gray area, but these are people with very real needs that are finally being addressed in a cost effective way ($2000 or less!). There is a world of difference between open heart surgery and uninvasively strapping something on the end of my arm.
How about a simple disclaimer - not to be used in safety critical applications - and let people have their $50 hand? When people start selling hands with bigger promises that have more grave consequences, dial up the regulation on those.
I can see why medical devices like pacemakers or implanted replacement joints need more regulation, but I don't understand why a prosthetic needs more regulation than other consumer products like gloves or eyeglasses.
That said, the myoelectric device at a minimum falls under some level of regulation since it is a biopotential device. The moment you start measuring potentials on the human body, you add the risk of injecting current across the human body - obviously bad. If you wanted to ensure that the myoelectric device was safe, you would at a minimum ensure that a) sufficient isolation between the sensory and actuator circuits b) a safe sensory circuit. Now, while doing that is probably relatively trivial to actually design, you do need to verify it.
So, once again, I agree that there's probably too much regulation piled on, but there is a good reason to regulate/verify in some form or way at least the myoelectric device.
More generally, notice that the examples that you and sister component pointed out are relatively simple, especially compared to a prosthetic hand. Complexity in itself can cause the need for regulation.
Oh, and on the side, you may find this document with regards to impact resistant and safety glasses (http://www.fda.gov/medicaldevices/deviceregulationandguidanc...) a fun read =P.
Agree. It's headline grabbing for sure. But when you really boil something down the actual material cost isn't always that big factor in the final product cost, in low volume quantities at least. [1]
There are overhead costs (some of which you have already highlighted) [2] that are necessary in running an actual company that makes products. These costs are not always visible or obvious.
[1] An example might be comparing the cost of a low volume production car such as the Porsche 911 Turbo S to a high volume car like a Honda Accord.
[2] I would also add legal and insurance costs which, with a device like this, are almost certainly not trivial.
There's R&D for the $50 arm too. They didn't just download the code and print plans from Github.
Justifying a fourty grand pricetag based on R&D costs is absurd in the face of this competing arm (which we should note is not actually $50, just $50 to replace plastic printed components).
The point is that the price difference is several orders of magnitude, and R&D was independently conducted by both groups, therefore R&D cannot justify the price difference alone.
http://openprosthetics.wikispot.org/Federal_Regulation_of_Pr...
"...regulated as Class I (exempt) medical devices."
As long as the new device, "...is intended for the same use and the same user type as existing products, and the device operates on the same fundamental scientific technology."
This means that most prosthetic devices are not explicitly regulated and are free from regulatory overhead in nearly all cases, especially the 3d-printed ones. While one can speculate on why they chose not to regulate these I think we can assume it's because ostensibly there's no danger to the end-user.
Video demo of the updated, 3rd-generation DEKA arm. http://youtu.be/auetIxOaSe8
I could speculate that because it relies on new tech and new research into prosthetic devices that the didn't just rubber stamp it.
Part of what the FDA does is safety, and the other part is validating the claims that a device or drug manufacturer makes. It's probably for this reason that they imposed additional regulatory validation. Because a cyborg arm? That HAS to be SCI-FI, right? ;)
Both required engineering, fitting, assembly, but one was done using volunteers time and a new technique, while the other was built part of a lucrative activity using a different technique.
They will serve the same purpose to the end-user, and the end-user will either pay 42k$ or 50$.
Consider this; at what point was the first person hired to work on the kernel? At what point were there a majority of patches submitted by paid workers?
^And then consider that up until that point, it was all unpaid. Just... think for a while, just think please.
Publicly funded research may also contribute to open source, 3d printed prosthetics.
If you read Robohand's FAQ, to build a proper medical grade hand will cost $2,000.
"To print the fingers with the knuckle block and hinges for a Robohand costs around US$5.00. To make a proper Robohand, which includes medical Orthoplastic, stainless steel hardware, tools, Velcro, plus the time it takes, costs US$2000.00."
anecdotal, a friend with an artificial leg needs it replaced every few years because of wear and tear on it and him. The fitting, maintenance, as well as physical rehab, is an ongoing costs. Hands don't experience the wear and tear that legs would so I wonder how 3d printed legs would hold up?
All: when you know of a better, more substantive version of a story, please alert us to it in the comments. There's too much material for moderators to check, so community effort makes a big difference. Let's optimize for quality.
If anybody wants to track their dog: http://www.shapeways.com/model/1805713/dog-collar-garmin-mou...
One of the things a mechanical engineer friend of mine suggested is finding a cool SolidWorks file online and learning from it. Basically drag the bar of the feature tree all the way back to the top and you can "play back" the person's design step by step.
Any ideas?
http://youtube.com/watch?v=Yvev6shNvSg
http://www.ncbi.nlm.nih.gov/m/pubmed/19367513/
You take an old soda bottle, cut the bottom off, and use heat to shape it to the stump. You can fix lightweight tools (forks, combs, button hooks, pens) to the screw top lid.
Obviously there is a lot of data missing here, but the fact that it is even approximated is amazing. Imagine the possibilities for people w/o insurance, veterans, inhabitants of poor countries.
Imagine if one day amputees etc. have a standardized interface to their nervous system that controls their hands. Then all 3d-printed hands need to do is adhere to that spec and people can put their health/abilities in their own "hands".
Very cool and the more geeky/creative amputees could actually build tools into themselves to use for fun sometimes. Granted I'm speaking for people when I have no idea of their struggles. But as Kramer said once, I'd like to have "shoehorn hands".
Something about that phrase just sounds really, really awesome.
It seems like a relatively rudimentary neural net with a lot more electrical sensors than today (maybe 16 to 256, something on that order) combined with some of the new sensors like light transmission or even maybe ultrasound to "see" the muscles under the skin, could be mapped to a graph with roughly the same degrees of freedom as the human hand.
Then maybe something like KNN could sift the data and map it to the motors. I’m not very familiar with data mining yet, but surely a lot could be done with a 700 MHz Raspberry Pie and a few GB of SSD storage. I’m thinking that giving the algorithm some notion of time by recording 1/60 of a second slices or something in that range might really improve accuracy. Just throw data and processing at it rather than trying to fully understand how nerves work.
Doing the reverse, to stimulate nerves based on how much power each motor is using, should be simpler. It could be sloppier because the human mind would learn to interpret the signals better than a computer.
The software should be available for free online and work with commodity hardware, and then the prosthetic designer could customize the limb for each user. Plus it could have the ability to upload random samples to help train the next generation AI.
As long as I am dreaming, we also need a moonshot to standardize nerve implants. It would go inside the limb somewhere and then use something like 802.11 powered by inductance coils above and beneath the skin, with a faraday cage built into the arm to prevent interference. Actually now that I say that, I wonder if the noise produced by the body’s movements would be statistically significant enough for another input into a neural net.
Anyway, chemical markers on the terminals could urge the nerves to grow into the terminals. Or use a bit of foreign nerve tissue that’s been grown on a scaffolding in the device in the lab, and then implant it and put the patient on immune suppression drugs until the nerves grow together, and then take them off gradually until the body dissolves the foreign nerves and has a direct connection to the sensor. Another idea would be to press a comb-shaped device with hundreds of microscopic channels through the nerve bundle and read the voltage through each channel. It should be manufactured for longevity and to read the most amount of data possible, and just let the programmers work out how to sift through it.
Once we had good data, training a neural net to do the mapping to actuators seems like a basic problem. Like maybe sample a few hours’ worth of movements and put the whole thing online like the Netflix contest until someone has programmed a near-perfect AI. Or a series of contests, each aiming to raise accuracy by 10% or something.
And you will basically have brain-controlled port. Then emulating normal(-ish) human hand becomes kinda boring thing to do. Because you can use a “different” hand. Or plug in into music synthesiser. Endless possibilities.
> Once we had good data, training a neural net to do the mapping to actuators seems like a basic problem
You don't strictly need it. Human brain is pretty ok with learning all kind of weird stuff.
Sure brains are plastic, but learning takes time so if you can map nerve outputs to hand in a way that generalizes across people, they could start using it faster.
Lets go through your post and comb it out. I really love the enthusiasm and ideas. My biggest caution going forward though: The body really dislikes foreign objects, like really.
>"...it seems like the complexity in prosthetic is not the mechanics, but interpreting the signals properly to control the device."
You couldn't be more correct! Dr. Richard Weir at CU Denver[0] has some great work in 3-D printed hands. The lack of controls into the hands is a major concern. The issue is you only have so many muscles in the arm to control all the complex motions of a hand. His group is working on the eigen-states of hand movement, trying to decompose them into the number of muscles in your arm.
Why the arm? Well, it makes sense to focus there, as that is where patients' minds are focused. The method is myocardial sensors that sense the nerve firings in the peripheral nervous system and then can send out that data. You basically implant them in a muscle fiber and they last a long time.
>... a lot more electrical sensors than today (maybe 16 to 256, something on that order)...
This would be awesome. But, focusing on a prosthetic hand, we don't have that many muscles in the arm to control that many[1]. The hand is much more complicated than the arm. Decomposing gestures, as Dr. Weir is doing now is a good step in that direction.
>...Just throw data and processing at it rather than trying to fully understand how nerves work.
How true this is! Much is being done in the Big Data era of neuroscience that we live in. Many groups are trying this tactic[2]. The thought is to just describe what the system of a brain does, not how it works. Think Gaussians, not Markov chains.
>...Doing the reverse, to stimulate nerves based on how much power each motor is using, should be simpler.
Feedback is very important to prostheses and helps improve care and quality of life. However, directly stimulating nerves is difficult. Your fingertips have much better 'resolution' than your elbow does[3]. Also, your skin also becomes accustomed to the input and you will ignore that stimulus pretty quickly. Like how you notice the smell of the food when you walk in the door, but don't within 10 minutes of standing next to the stove.
>...It could be sloppier because the human mind would learn to interpret the signals better than a computer.
The body does acclimate to the stimuli, but as I said before, it also starts to ignore as well. Also, people are all different. Genes can account for a lot of the differences, as well as environment, but also even culture and gender[4]
>...As long as I am dreaming, we also need a moonshot to standardize nerve implants.
I'm not sure what you mean here. Each person's situation is very different and personal. Standardizing is difficult with medicine.. Skin color, size, weight, culture, gender, job, etc all greatly vary. For example, a scuba instructor's situation is a lot different than a 1st grade teacher's. The scuba guy has to deal with greatly varying pressure and water. The teacher has to have the new hand be fit for children. The nerve implants will have to follow their situations.
>...a faraday cage built into the arm to prevent interference.
Your blood is mainly salt water, which is very conductive, so there is not real need to cage things. In fact, you want to encase everything in silicone to prevent leaks and make it all non-bioreactive.
>...the noise produced by the body’s movements would be statistically significant enough for another input into a neural net.
The noise in a muscle is hard to figure out. Again, it depends on the person, the muscles, the salt content in the blood, etc. In a nerve, the picture is much the same. I will quote without citation that the firing rates of nerves increase with temperature to the 4th power (I'd love a citation, if anyone can find one, my brain is just borking now)
>...Anyway, chemical markers on the terminals could urge the nerves to grow into the terminals.
This is non-trivial and may not be possible. Those markers often are only present in a real soup, and the behavior of nerve growth is not well understood (at least by me, haha). I'll admit it's possible, but a lot of engineering is needed to do it.
>...put the patient on immune suppression drugs until the nerves grow together,
Wow, that is not a good idea. Going on immunosuppressants is not taken lightly, and you'd have a heck of a time convincing people to take that big of a risk just for a hand or something. Also, they won't be necessary in the peripheral nervous system, just the central nervous system.
>...Another idea would be to press a comb-shaped device with hundreds of microscopic channels through the nerve bundle and read the voltage through each channel.
Dr. Theodore Berger at USC is working on a device like this already![5]
>...It should be manufactured for longevity and to read the most amount of data possible, and just let the programmers work out how to sift through it.
If you'd like to join up, many neuroscience departments really favor physics, CS, and EE majors for grad study these days, at least it seems to me like they are.
The peripheral nervous system is a real good area for these neural ideas. But the central nervous system (CNS) is not. The reason is glial scarring[6]. You body really really does not like things in the CNS. This is why paralysis happens mainly. The scar prevents the nerves in your spine from regrowing and making the legs and whatnot work again (sort-of). This takes place in your brain to a huge degree. I used to work with place cells[7](very cool cells) and everyday we would have to 'burn off' the scarring by passing current through the electrode bundle. It turns out, you are your own worst enemy here. Ideas like opto-genetics[8] can bypass the scarring issues, but they also have a host of others.
Again, great ides and enthusiasm! I love it! If you see a future with neuroscience, the field is wide open these days and there seems to be a bit of money floating around. Apply to a progam or 12 and you'll have a great time.
[0]http://www.ucdenver.edu/about/newsroom/newsreleases/Pages/Bi...
[1]http://www.innerbody.com/anatomy/muscular/arm-hand
[2]It's easter, I'm not counting how many! You do it.
[3]http://visalakshiramani.files.wordpress.com/2010/07/homuncul...
[4]http://ccr.sagepub.com/content/34/2/135.abstract
[5]http://www.usc.edu/programs/neuroscience/faculty/profile.php...
[6]https://en.wikipedia.org/wiki/Glial_scar
What then to make of the widely observed phenomenon that passing small DC currents through electrodes can temporarily rejuvenate them? It could be that this effect is a result of an electrochemical change to the electrode surface itself, rather than to the surrounding tissue.
1. http://www.ncbi.nlm.nih.gov/pubmed/24550823 2. http://www.ncbi.nlm.nih.gov/pubmed/23562053 3. http://www.ncbi.nlm.nih.gov/pubmed/23891081
http://science.howstuffworks.com/prosthetic-limb5.htm
I wonder if eventually "they" can shrink this down to the point where individual nerves are controlling another type of cell, like bioluminescent, and then the flashes could be picked up by a photodiode. I guess the reverse would be if they figured out how to get implanted light sensing/retina cells to react to diodes the other direction.
I'm glad that people are working on that stuff. I guess what I was trying to get at is that a lot of the problems that seem hard, like say self driving cars or voice recognition, are getting to the point where they are well understood and error rates are comparable to what humans achieve. So maybe some of that expertise could be open sourced and used to solve problems that seem intractable like mind-machine interfaces.
I would dearly love to work on real problems like that, but there is such a separation between grants and tinkerers like me that I can honestly say hackers are 90% underutilized. I wish there was a kickstarter for such things. I graduated college in 1999 and have only spent about 2 years out of the last 15 doing what I would consider "real work". The rest of the time I've been scurrying about trying to make ends meet, constantly choosing between the pain of being under someone's beck and call or hunger. I can totally relate to researchers constantly fretting about money. Sometimes I think we could use a Walk for Breakthroughs, the same way that people have fundraisers for AIDS or cancer.
It was founded by a couple who had a disabled child who needed a back brace, and were horrified by the current process for getting one. The measuring process is upsetting for the child (it involves taking a slow-setting wet plaster cast), it takes weeks to make the brace, it's expensive, and when it finally arrives, it fits badly. They think they can turn round a perfectly-fitting brace from laser scan to 3D print in a couple of hours.
It's a totally brilliant project. It's very rare that an idea is so obviously right.