Slashdot Mirror
Open-Source Hardware For Neuroscience
the_newsbeagle (2532562) writes "The equipment that neuroscientists use to record brain signals is plenty expensive, with a single system costing upward of $60,000. But it turns out that it's not too complicated to build your own, for the cost of about $3000. Two MIT grad students figured out how to do just that, and are distributing both manufactured systems and their designs through their website, Open Ephys. Their goal is to launch an open-source hardware movement in neuroscience, so researchers can spend less time worrying about the gear they need and more time doing experiments."
Their goal is to launch an open-source hardware movement in neuroscience, so researchers can spend less time worrying about the gear they need and more time doing experiments.
My experiences with lab-built equipment in academia suggest that building your own equipment is not really a good way to "spend less time worrying about the gear". Usually you will spend quite a lot of your time worrying about DIY gear. The advantages are not in time saved, but in two other things: 1) you can build gear that would be prohibitively expensive to purchase; and 2) you can customize it in-house.
10 PRINT CHR$(205.5+RND(1)); : GOTO 10
... if you don't have to jump through all the hoops (quality management, defect tracking, regulatory compliance, manufacturing yield, safety risk assessment/mitigation, customer support, liability, translation into three dozen languages, etc.) that a commercial manufacturer has to.
Is the reason medical devices are so expensive. The actual hardware of medical devices is normally fairly low cost. However the regulatory burden on companies to provide safe and effective medical devices is substantial and rightly so. I am extremely sceptical that these devices have the necessary certifications for use on patients and it would be an irresponsible doctor that allowed their use on their patients without these certifications.. they are there for a reason!
And its not just the devices that need the certifications, the companie that supplies them have to be audited to things like ISO 13485, 14971 as well as 9001/9002. This requires substantial investment, experience and knowledge to achieve, and again I seriously doubt that two grad students have that capability. Finally, these requirements are in place for patient safety.. that is the number one factor in designing these devices. And while it creates a high hurdle for small companies to overcome, it is fully justified and necessary
This Kickstarter project looks promising.
https://www.kickstarter.com/pr...
I don't read your sig. Why are you reading mine?
So how does it differ from OpenEEG project? I read the summary and I read a couple of paragraphs from their site, but it was all some round talk. You can get fully assembled 2 channel (uses smt components - it's small) OpenEEG device from Olimex for 99 euros (+electrodes and shipping costs), if you are not into soldering.
I'm a bio-electrical engineer from MIT, course 6-1b, back when that was the number for the department. I used to design neuroscience equipment: built the highest frequency, fastest pulse stimulators on the planet fand tested on humans in a functional MRI. *MAN*, I miss that work. The pay was horrible, but getting to do genuinely new science was fascinating.
Anyway: there were hard-learned lessons that new designers are unlikely to have learned, so I'd be concerned about amateurish efforts. They include:
1) The ground plain is not actually at ground voltage. It never is: making it bigger will not solve all your problems, capacitive coupling to ground is *localized* and can spread signals via paths you'd never see in a circuit diagram.
2) That line on the diagram that says parts are tied together and the same voltage at the same time is a *lie*. Signals take time to go down wires and can couple to all sorts of things.
3) Really good measurement equipment, low noise, calibrated, used correctly, is worth its weight in gold, and certainly its weight in intern organs. (Which organ depends on the intern.) It can detect subtleties that you never realized were there until you measured them correctly.
4) 60 cycle creeps in *EVERYWHERE*. So do various high frequencies on the power lines from other equipment. I don't care if the "power line filter" vendors say it blocks "95% of all transients", it does not block 60 cycle. Mu metal and magnetic shielding do not block 60 cycle. Running off of clean batteries on an isolated power supply, with a wooden desk and no conductive surfaces blocks 60 cycle. A good motor generator can also block 60 cycles, but make damn sure the coupling between them is wood, not metal.
5) Capacitors leak DC. You cannot completely block DC voltage to nerve electrodes by putting a blocking capacitor on it, and if you do use that as a safety issue, use a *ceramic* capacitor, not a *polarized* capacitor, even if those do come in bigger values for the same size, Make damn sure your stimulators run at a zero DC output *before* the capacitor, or as close as you can get to it, or the leaked DC current can screw up all your results and may cause neurological damage. (This discovery caused a lot of adventures in my lab, I freaked out when I found that some twit had used "polarized" capacitors in a 10 year old design, which do not treat positive voltages the same way they treat negative voltages.)
6) Rebuilding the equipment effectively guarantees somewhat different results, even if you followed the specs exactly.
These kind of hard-won lessons are why I'd be really cautious about student built hardware. Some of those lessons took a long time to learn, and were never really taught in design courses.
This has always bothered me with the current state of neuroscience: The whole point of nerves/brain matter is to communicate/remember/transform information, but we're still relying on crude external cues like heat/bloodflow/electrical activity to tell us "somethings happening around...here", and that's pretty much it. It always bothers me when I hear the term "brain signals".
Nerves should be able to query their neighbors about their state, and the state of other nerves - otherwise, they wouldn't really be able to form something like a mind (as in, "the mind is what the brain does"). Why still can't we find a way to just "ask" the nerves what their state is?
Even in our simulations, we just represent nerves as nodes that grow associations - but those associations are useless, unless they can be traversed in queries by the system, to gather inputs, and send outputs at all levels.
Are we getting anywhere close to a stage where we can communicate with nerves to use that same communication system that logically must exist for it to function? Seems like even with limitations, that would be a LOT more useful than analogously inferring from traffic levels what the function of buildings in a city are, like we're doing now.
Ryan Fenton
I seem to recall that years and years ago Steve Ciarcia wrote a series of article in his Circuit Cellar magazine about making sensors and a home built EEG. If I recall correctly it used relatively inexpensive parts and off the shelf sensors when necessary. All designed for the hobbyist and much lower in cost. Of course I may be wrong. But still wondering, whats the big hoo haw about something this expensive?