Slashdot is powered by your submissions, so send in your scoop

 


Forgot your password?
Close
typodupeerror
×
Medicine Technology

Raspberry Pi-Powered Ventilator To Be Tested in Colombia (bbc.com) 86

Posted by msmash from the how-about-that dept.
A team in Colombia is to test a ventilator made with a Raspberry Pi computer and easy-to-source parts. From a report: The design and computer code were posted online in March by a man in California, who had no prior experience at creating medical equipment. Marco Mascorro, a robotics engineer, said he built the ventilator because knew the machines were in high demand to treat Covid-19. His post prompted a flood of feedback from healthcare workers. He has used the advice to make improvements. "I am a true believer that technology can solve a lot of the problems we have right now specifically in this pandemic," he told the BBC. The Colombian team said the design was important for their South American country because parts for traditional models could be hard to obtain. By contrast, Mr Mascorro's design uses only easy-to-find parts -- for example, the valves it employs can commonly be found at car and plumbing supply stores. The machine is set to be put through a fast-tracked round of tests at two institutions in Bogota -- the University Hospital of the Pontifical Xavierian University and Los Andes University.
This discussion has been archived. No new comments can be posted.

Raspberry Pi-Powered Ventilator To Be Tested in Colombia More

Raspberry Pi-Powered Ventilator To Be Tested in Colombia

Comments Filter:

  • Watchdog (Score:1, Insightful)

    by Anonymous Coward
    Pray that this thing has a working watchdog. Pis are specifically NOT for industrial use, much less medical use. South Americans are NUTS. Why am I even going this far? WHAT is he doing with a Pi that can't be done with an ATMEGA?

    • My guess is it is running Ubuntu and Python.

      • Debian.

    • Re: (Score:3, Insightful)

      by alvinrod ( 889928 )
      I'm sure not having any ventilator is a far superior option. Let's not let perfect be the enemy of good.
      • uhh, it's basically impossible to breathe with a tube in your throat. So no, if a machine stops breathing for them it is not better than labored breathing without intubation.
        • If they're putting someone on a ventilator it's because the labored breathing without intubation isn't possible. So once again, any ventilator is better than none at all. Your choices are basically might die even if this machine does work perfectly vs. will certainly die. Decisions, decisions.

        • it is not better than labored breathing without intubation.

          Laboured breathing is not the reason people are put on ventilators with COVID-19. Without that tube down your through you've effectively signed your own death certificate.

      • I'm sure you're not a doctor.

    • Re:Watchdog (Score:5, Interesting)

      by MadCow42 ( 243108 ) on Monday April 13, 2020 @11:28AM (#59941152) Homepage

      Pi's are a lot more reliable than you may think...

      I have about 600 currently running in the field (for the last ~3 years), and the failure/lockup rate is about 1 system every 3-4 months. That's an MTBF (calculated, of course) of about 200 years. Of course that won't hold up over time, but it's impressive.

      Basically all of the failures I've had are attributable to external events - either electrical storms/surges, or people physically shorting out GPIO pins when installing things or doing other service work.

      I've designed hardware-based watchdog systems into my future control board that these plug into, but based on the data they're not really needed. But, the extra security is always a good thing.

      • When I run python programs like meld or meshroom they always leave behind a trail of exceptions and stack traces! Not exactly tidy. And 80% of the docs about the language/libraries are wrong since each point version does things differently...
        • Admittedly this was on ubuntu not a pi. For a controller I used to like an stm32 on ChibiOs but that has rather less power.

        • Well, either that's poorly written software, or a poorly configured environment for what you're trying to do. Python is solid... but any language can look "bad" if you're using it wrong.

        • And? For example, Erlang leaves a trail of errors - in fact it's basically designed to do so - but always recovers. Whether the whole thing works seems much more important.

      • I'm guessing you're not booting from an SD card? That seems to be by far the leading cause of Pis doing down, cheap flash storage failure.

        • Actually I am using SD. However, the software stack ensures that anything that's being read/written often is done on a Ramdisk instead, and only written to disk when necessary. We haven't had a single SD card failure yet... in ~4 years of systems being out there (oldest ones still running fine).

        • In the industrial PI systems I have built (PanTilt protocol translators) the system disk (SD card) is read only. I have never had a disk failure. I do put a redundant disk taped to the underside of the chassis but it has never been used.

          It takes some time to make your system read only and there are lots of little hacks here and there to accommodate logging, dhcp, fake-hwclock,etc I wish there was a more straightforward way of doing this.

      • I know of PI's being used in industrial PLC's so they can be reliable but what they aren't is safe.

        • Interesting... care to share your data to support that opinion?

          • Re: (Score:2)

            by DrXym ( 126579 )
            There are two kinds of PLC used in factories - standard and safe. Standard runs the every day parts of the machine - the motors, heaters or whatnot. The safe PLCs are as the name suggests, safety critical. They're what monitor and stop the machine if somebody pushes the button, or opens a door or something else that is a hazard when a machine is running. They have added functionality and robustness enhancement to ensure their reliability, to ensure they don't receive false negatives/positives from their inp

            • Honestly I'd generalise even more than that. There are safety critical PLCs and standard ones, but the standard ones are not "unsafe". A lot of effort still goes into ensuring that the output state is somewhat defined, that through redundancy the logic keeps getting run (and often checked), and inputs can be set to a pre-defined value on fault.

              There's safety PLCs
              Then there's safe control systems like PLCs.
              And then there's completely unsafe stuff like Raspberry Pis which are not deterministic on any kind of

              • Re: (Score:2)

                by DrXym ( 126579 )
                This article [controleng.com] covers it pretty comprehensively. Whether we're talking medical devices or factory equipment, it is important to consider for and mitigate modes of failure.

                Fundamentally if you're sticking a raspberry pi in a ventilator then the question isn't "what if it fails", the question is more "what if it didn't exist at all". It's all good and fine to question if a device is safe here from our desks, I won't be applying that same level of critique when I'm laying on a hospital bed unable to breath.

                Hence me saying "We could argue the point that in an emergency an RPi powered ventilator is better than nothing but there still needs to be mitigations for possible modes of failure.". It's great to have emergency ventilators but if it overpressures, or the pump fails, or a it fails to complete a cycle, or the power goes off or whatever then

          • Interesting... care to share your data to support that opinion?

            Safe has a definition for industry, that is that for the many MANY failure modes in a device the action on failure is known, defined, and consistent. A Raspberry Pi is none of those things. When the software locks up on the Raspberry Pi there's no setting you can set that automatically sets all outputs to zero. When a fault occurs on an input there's no settings you can set that automatically makes that input have a pre-defined value. Hell the inputs float. The system cannot detect faults anywhere other tha

      • I don't know... my Raspberry Pi 2 had a bad habit of overheating and locking up every few days when I tried to use it as an embedded WordPress server (which it did fairly well). Maybe it would be more reliable if they underclocked it?

      • I have done some work on designing medical equipment, the electrical part that is. These things need to be safe even when electrical stuff goes wrong. I.e. how does the system react when a signal wire breaks due to wear? How does it handle ESD when someone touches it? What if there is a voltage spike/drop on the line. What if there is a burst of these? The list goes on... . In that perspective, having a few breakdowns in three months is a lot. If you would do extensive safety testing, you would find a lot m

      • Other people have gone down the same route https://theface.com/life/oxven... [theface.com] and similar https://www.mercedesamgf1.com/... [mercedesamgf1.com]

    • Not only that, but "made with a Raspberry Pi" and "easy-to-source parts" usually don't go in the same sentence, especially if they're using the Raspberry Pi Zero to lower the cost.

      • If you are mass-producing something you should not be using a prototyping board like a R-Pi. At the least you should go through all the software stack and cut out anything not used. Pull up resistors can guarantee your control outputs are known values during reboot.
  • When regular people are able to replace expensive equipment with off the shelf material and create the same usable product. Then are able to maintain it and enhance it without needing to entirely replace it or pay for high cost service plans (SAS). Gasp... It's like right to repair and right to replicate and then right to replace. It would be horrible if prices begin to go down as a result. I am sure the lawyers will be working overtime trying to sue whoever makes one or if someone dies on one. The legal s
    • Don't worry, the government agencies will just make it illegal because it hasn't been properly certified. The big medical companies don't mind a little regulation because when it keeps competition out.

    • Re: (Score:3, Insightful)

      by Tailhook ( 98486 )

      When regular people are able to replace expensive equipment with off the shelf material and create the same usable product.

      This is incorrect and illustrates common naiveté. The product of the commercial medical equipment industry isn't equipment. It's regulatory compliance. And current aberration aside "regular people" can't deliver regulatory compliance. Regular people don't employ the sons and daughters and sons-in-laws and daughters-in-laws of the regulatory regime in no-show non-profit chairpersonship jobs and, therefore, are incapable of obtaining the approvals necessary to sell into lucrative markets.

      • Oh please, as much as regulatory compliance sounding like some under the table conspiracy requiring an insider setup to prevent competition, the reality is quite simple: Regulator compliance costs money. Nothing more, nothing less. No need to get your family or some mythical chairperson involved.

        • Re: (Score:2)

          by cusco ( 717999 )

          Need? Almost certainly not. Does it make the regulatory process **much** easier, cheaper and faster? Almost certainly yes. The FDA amply demonstrates this again every few years.

    • It's one thing to make an off-the-shelf substitute ventilator (or whatever medical product) for an emergency, but the real price of medical equipment isn't in the materials, but in the testing and certification process. Has anyone tested Home Depot's pipe fittings for pathogen retention? Will they maintain structural integrity if you put them through an autoclave to sterilize them? Will they start to outgas into patients' lungs? Will they still perform adequately after a year of heavy use? After five y

      • Re: (Score:2)

        by cusco ( 717999 )

        Keep in mind they're talking about Colombia, where it's likely that many (most?) hospitals don't have **any** ventilators to start with. Is a $$$ ventilator with certifications better? Sure. Is no ventilator at all worse? Hell no.

      • Insulin is almost 100 years old, it's well out of patent protection. So how come it's $800 in the USA and $30 in Canada?

  • what it takes to make a ventilator (Score:5, Informative)

    by arielCo ( 995647 ) on Monday April 13, 2020 @11:32AM (#59941170)

    A short video on why most amateur attempts are terribad, and what a decent ventilator has to do:

    A Guide To Designing Low-Cost Ventilators for COVID-19 – Real Engineering [youtube.com]

    • This.

    • Re: (Score:2)

      by PPH ( 736903 )

      Some good advice in that video. In looking at some of the DIY designs that claim auto supply parts* were used, I suspect that pressure feedback to the control program is not an included feature.

      However, simple u-tube manometers with optical position sensors would probably suffice. And even a maximum pressure limiting valve can be fashioned from a tube immersed in a beaker of water.

      *I'm assuming that they aren't using turbo blow-off valves.

    • /Thread won.

  • I expect a ventilator could be run from Arduino or any other programmable board providing you're not too concerned about failsafes or redundancy.

    • An Arduino has lots of input and output pins, along with a watchdog timer. If the machine is designed with failsafes and the microcontroller is programmed properly, it should be safe.

    • Why would you be concerned about those? It's not like if it malfunctions it could kill somebody.

  • i knew techies liked to shove Raspberry Pi (Score:5, Funny)

    by hatman9 ( 6732920 ) on Monday April 13, 2020 @11:41AM (#59941214)
    i knew techies liked to shove Raspberry Pi down my throat, but this is going too far!

    • Hospital PA system: Code Green! Code Green!
      Doctor 1: What the hell is a code green?
      Doctor 2: A Raspberry Pi Zero trapped in someone's trachea!

  • Bah (Score:4, Interesting)

    by ArchieBunker ( 132337 ) on Monday April 13, 2020 @11:53AM (#59941262)

    I could do better with a few 555 timers and some potentiometers. How did they operate for decades before being microprocessor controlled?

    • You'd need bunch of 4000 series chips for the alarms and such as well, by which point it's easier to put in a microcontroller. A Pi seems like overkill, and more complexity means more that can go wrong.

    • Re: (Score:2)

      by ediron2 ( 246908 )

      Does your 555 config live on a circuitboard, so everything is rigid and not likely to break due to movement? Those pots are nice... are you sure you understand the problem well enough to know which variables should be adjusted and which shouldn't? How about user-friendliness: are you available to teach people which pots to adjust for what result? What about complex adjustments, such as when X goes up and Y goes down 20% of that, or whatever? Maybe a UI could improve on your wires-chips-and-knobs plan...

    • I could do better with a few 555 timers and some potentiometers.

      You could not. Unless your goal is to make something even less accurate than the already horrible designs out there (including this one). Your shitty 555 timer is a relic of the past for a reason. I mean shit it costs more than a 32bit microcontroller and requires far more external parts, will be less configurable, less accurate, less repeatable, and I'm beginning to wonder if you haven't practiced electronics since the invention of the computer.

    • How did they operate for decades before being microprocessor controlled?

      Originally they were all-mechanical. Everything was done with pneumatic valves and such. The Bird Mk8 is a pretty good example of that tech: https://en.wikipedia.org/wiki/... [wikipedia.org]

      Of course, they had very limited adjustability and no monitoring or alarms. They would not be considered adequate for any modern hospital use. A 555 and a few pots would essentially be an electronic reimplementation of the same thing.

      The early electronic era was all analog control systems. For example, the Servo 900: https://centu [centurionservice.com]

  • This:
    https://www.woodlibrarymuseum.... [woodlibrarymuseum.org]
    Is what we used back in the day, no computers needed, any patents have way expired, you can probably use a 3d printer on most of it and it's been around for years.

    We get so caught up in having to throw electronics at everything that we forget that we don't always need them.

    This is a fun read: http://rc.rcjournal.com/conten... [rcjournal.com]

    • Given the domain name I was hooping to see one made out of wood ;)

    • People create with what they know. I saw a ventilator prototype made by someone who builds 3D printer.

      As you can imagine, his version of ventilator uses a NEMA 17 stepper motor, a GT2 belt and 8mm rails with linear bearings.

  • Not only these. 80% of patients put on them die.

    https://www.businessinsider.co... [businessinsider.com]

    • That means 20% recover. Without the ventilators 100% of those patients would have died.

    • Re: (Score:2)

      by Compuser ( 14899 )

      Saving 20% is a big deal. Losing even one life because we did not have equipment is crazy.
      That said, I wish we paid more attention to ECMO machines. They can help much more but they are much harder to build and validate so their shortage is being swept under the rug. We are not even talking about the devices we actually need the most. Instead every hillbilly from the dimmest of backwoods to the brightest bulbs at MIT are going for the less important but lower hanging fruit. It is frustrating.

    • Saving 20% of the most critical patients is not overrated.

  • From personal experience, I'd go beaglebone. I've got 4 BBone's that have been running for years as controllers with no problems. I've got 2 pi's I use for AP's. The originals were pi-2's but now both have fried and replaced with pi-3's. The pi's in my experience have had a much higher failure rate (2 out of 4 in about 4 years) versus 0 out of 4 for around 5 years for beaglebones. Oh and 3 of those BBones are outside or in the attic. The pi's are inside. Frankly, I've been very surprised the BBones have su
    • What failed on the Pis ? Cheap SD Cards are known to be weakspots. The guy above you has 600 Pis in the field with a calculated MTBF of about 200 years.
      • Was not the SD. I tried replacing them but still failed to boot. So no idea what the problem was. They are not worth investigating for my application. Maybe the 3 is better than the 2 was. I've not had a 3 fail yet (well except for the one I fried by accident) so maybe better than the 2's.
        • I've lost count at how many Pi's I've deployed at home and at work. I've had exactly one failure and it was my oldest home Pi, an original B+. The SDcard failed. It has run 24 hours a day for five plus years as a media streaming NAS. The failing SDcard was a junk card from an Amazon Pi kit.
  • Doctor: "Time of death?" Nurse: "10:55pm." Doctor: "Prepare the corpse for the morgue. I'll inform the family." Nurse starts rolling out the corpse... Doctor: "Stop! I'm missing some of my LEGOs!!"

  • After having seen the Arduino code on github:

    Being attached to this piece of shit is the last resort. My death/permanent brain damage must be near certain before you do it.

    Reason for this estimation:
    * protocol between PI and Arduino is complex
    * timing functions are done by a loop, not timers
    * loop functions contain serial communication and synchronous reads from ADC
    * i have thee feeling than some variables miss modifiers
    * should use interrupts to do the separate monitoring and control functions
    * should make

    • Don't take this too serious. These projects keep popping up all over the world and are often a form of virtue signalling as well are a show of helplessness. Many of us feel helpless, but taking these projects too serious means we give more importance to our helplessness than to the people who are actually suffering. Some people will die regardless of how much help we give them and we need to allow for them to have some dignity. Give them the best we have to offer and not make this about our helpless. So tak

Slashdot Top Deals

Real programmers don't bring brown-bag lunches. If the vending machine doesn't sell it, they don't eat it. Vending machines don't sell quiche.

Close