5 ms·

There have been a lot of these 'open source ventilators', and they kind of miss the point. It's easy to make something that performs the basic functionality. I

There have been a lot of these 'open source ventilators', and they kind of miss the point. It's easy to make something that performs the basic functionality. I have spent basically every waking hour of the last month and a bit working on a ventilator for my employer (part of Ventilator Challenge UK - project recently suspended). Making something that delivers air in the right duty cycle for breathing is easy. There's a few more things you want to do: Blend O2 and Air ("FiO2"). Ventilators need to deliver Oxygen rich air to patients, but not usually 100%. Blenders are harder than you'd think to make. Note also that having pure O2 in your system means that there's a load of stuff you have to get right to prevent fire. Alarms. This is the big one. most of the implementation (even taking account the next feature), was about checks and associated alarms, for blockages, failures etc. Making sure that we beeped loudly if something looked wrong. Note that 'not beeping' is a much WORSE failure than 'not ventilating'. If something goes wrong (including complete hardware failure of the ventilator), as long as a healthcare practitioner is aware and responds, then there isn't an issue. If the ventilator is happily ventilating away but something is obstructed then the patient might die (there are backup checks like pulse oximetry on the patient, but they're slow). Assisted breathing. Driving a solenoid on and off at a particular rate and duty cycle gets you 'mandatory breathing'. That is great for keeping someone alive. If you're not unconscious though, it is both incredibly unpleasant and doesn't help you get weaned off and get better. What you want is a device capable of sensing attempted breaths and using the ventilation to 'help' them. This feature isn't 100% necessary - the early plans in the UK were for a massive shortage and for the simplest things we could get in a hurry. It quickly became clear that actually we needed ventilators to help people recover.

jschwartzi 7y ago
Yes. We spent as much time testing our alarm and monitoring system as we did doing gas delivery changes. And even on the gas delivery side, simple things like the patient changing positions while the ventilator delivered a breath could be damaging without following the standards explicitly. I think there are about 2 or 3 thousand different little functional requirements for building a ventilator. It took us about a year to get to first prototypes and then another year to get to the first passing of our system-level tests. Even just our alarm system was very complicated, with cross-checking alarms and even cross-checking that the audio was indeed playing back. Things like playback tones and frequencies and alarm indicator colors and flash frequencies are highly regulated with relevant standards. And nuisance alarms are almost more important to avoid than having enough actual alarms. No alarm is as useless as the one the clinician turns off because it's the 4th time this hour it's gone off and the patient has been fine every time. These are not easy devices to build, and selling something that doesn't meet the applicable standards and isn't substantially equivalent to something marketed today is unlikely to be particularly safe. You can easily do more harm than good by giving someone an untested ventilator.
- Justsignedup 7y ago
  Thank you for pointing it out. In addition to false-positives... it would really suck if a LED died, or a speaker died, and suddenly a critical alarm isn't going off. Redundancies upon redundancies are critical for these systems. The not cheap for a reason.
carbocation 7y ago
As someone working with these devices and patients relying on them, I agree with each of your points.
Justsignedup 7y ago
Just to highlight it more. Alarms are critical, if you don't know that everything is working PERFECTLY then you don't know if one of your 20 patients is dying because of a failing $400 machine. Also so is lifespam. It would suck to get all these ventilators that break down in a few days. I've seen medical machine designs, and they have to check by the second that everything is working, and redundant systems so that if one system thinks all is well and the other doesn't big alarms start going on. And redundancy so that if part of the machine is failing the patient is kept alive while another machine is gathered. The quality is just as important as the function.
- khazhoux 7y ago
  > Also so is lifespam You just invented an awesome word
tinco 7y ago
Man I felt bad for feeling this while reading the article. I've no doubt that Bill Dally is an awesome guy, especially given Nvidias recent excellent performance. But this project seems almost embarrassing. Why tout something on your corporate blog that's just a personal project of an employee he built in a couple of weekends. It looks simple enough for the final year project of a highschool student. I hate to be so negative, but I just don't see why nivida would associate with it.
- ventilated 7y ago
  Don't feel bad, you are likely a "glass is half empty" type of person, you can't help it. It's hard to believe you actually read the article, they framed the accomplishment in a very accurate light. They are not claiming it revolutionary: "Dally’s aim was to build the “simplest possible” ventilator." He accomplished this goal. He wanted to build a ventilator useful in emergency situations. Would you have us believe that on your death bed you would tell the doctor to give a ventilator of this design to someone else because it looks like a high school student could build it? This is a quality design because it meets the stated requirements.
  - tinco 7y ago
    If I were a glass half empty kind of person I wouldn't feel bad about it. Those guys are always gleefully pointing out flaws in things. I would not turn away anything a doctor would recommend I do, especially not on my death bed. In any case, there's been dozens of these solenoid based ventilators, and this one looks especially pretty, but if they were actually needed surely you'd hear all about hospitals using these diy things to save lives? From what I've seen doctors actually talk about, is that there's ways to hook up even four people on one of those expensive type ventilators, and doctors still don't do it, because just giving people air isn't good enough. The thing that makes me feel bad is this feeling it's some sort of off key joke. Maybe it has a use, but to me it seems like why is a big serious company like nvidia, making noise about constructing a ventilator that no doctor would use unless in an absolute emergency where they couldn't use a serious ventilator?
  - ckdarby 7y ago
WheelsAtLarge 7y ago
One of the items, of many, that worries me about this wild west season of ventilator innovation is that many of these new ventilators lack a common interface. Users are trained a long time on how best to use a ventilator using a very limited variety of user interfaces. I think it's very dangerous to expect a respiratory therapist to handle all the varieties of respirator interfaces that I'm sure are being created. New interfaces always mean mistakes, it's part of the learning process. We see it in software all the time. A user gets a brand new interface so it takes time for the user to acclimate to it. When you are dealing with a wordprocessor it's no big deal to make a mistake but a machine that can save or kill a patient depending on how it's being used is a whole different situation. I wonder, what is being done about that?
- rkangel 7y ago
  I can only speak for the UK regulatory environment here, but actually the situation is better than you might imagine. The best example is alarming - there are some common alarm patterns (the pattern you beep and flash your lights) for different priorities of alarm. My understanding is that these aren't standrards you must adhere to, but because they're 'accepted' using them makes your life much easier with the regulators so everyone does unless they've got a good reason not to.

jschwartzi 7y ago

Yes. We spent as much time testing our alarm and monitoring system as we did doing gas delivery changes. And even on the gas delivery side, simple things like the patient changing positions while the ventilator delivered a breath could be damaging without following the standards explicitly. I think there are about 2 or 3 thousand different little functional requirements for building a ventilator. It took us about a year to get to first prototypes and then another year to get to the first passing of our system-level tests. Even just our alarm system was very complicated, with cross-checking alarms and even cross-checking that the audio was indeed playing back. Things like playback tones and frequencies and alarm indicator colors and flash frequencies are highly regulated with relevant standards. And nuisance alarms are almost more important to avoid than having enough actual alarms. No alarm is as useless as the one the clinician turns off because it's the 4th time this hour it's gone off and the patient has been fine every time. These are not easy devices to build, and selling something that doesn't meet the applicable standards and isn't substantially equivalent to something marketed today is unlikely to be particularly safe. You can easily do more harm than good by giving someone an untested ventilator.

Justsignedup 7y ago
Thank you for pointing it out. In addition to false-positives... it would really suck if a LED died, or a speaker died, and suddenly a critical alarm isn't going off. Redundancies upon redundancies are critical for these systems. The not cheap for a reason.

carbocation 7y ago

As someone working with these devices and patients relying on them, I agree with each of your points.

Justsignedup 7y ago

Just to highlight it more. Alarms are critical, if you don't know that everything is working PERFECTLY then you don't know if one of your 20 patients is dying because of a failing $400 machine. Also so is lifespam. It would suck to get all these ventilators that break down in a few days. I've seen medical machine designs, and they have to check by the second that everything is working, and redundant systems so that if one system thinks all is well and the other doesn't big alarms start going on. And redundancy so that if part of the machine is failing the patient is kept alive while another machine is gathered. The quality is just as important as the function.

khazhoux 7y ago
> Also so is lifespam You just invented an awesome word

tinco 7y ago

Man I felt bad for feeling this while reading the article. I've no doubt that Bill Dally is an awesome guy, especially given Nvidias recent excellent performance. But this project seems almost embarrassing. Why tout something on your corporate blog that's just a personal project of an employee he built in a couple of weekends. It looks simple enough for the final year project of a highschool student. I hate to be so negative, but I just don't see why nivida would associate with it.

ventilated 7y ago
Don't feel bad, you are likely a "glass is half empty" type of person, you can't help it. It's hard to believe you actually read the article, they framed the accomplishment in a very accurate light. They are not claiming it revolutionary: "Dally’s aim was to build the “simplest possible” ventilator." He accomplished this goal. He wanted to build a ventilator useful in emergency situations. Would you have us believe that on your death bed you would tell the doctor to give a ventilator of this design to someone else because it looks like a high school student could build it? This is a quality design because it meets the stated requirements.
- tinco 7y ago
  If I were a glass half empty kind of person I wouldn't feel bad about it. Those guys are always gleefully pointing out flaws in things. I would not turn away anything a doctor would recommend I do, especially not on my death bed. In any case, there's been dozens of these solenoid based ventilators, and this one looks especially pretty, but if they were actually needed surely you'd hear all about hospitals using these diy things to save lives? From what I've seen doctors actually talk about, is that there's ways to hook up even four people on one of those expensive type ventilators, and doctors still don't do it, because just giving people air isn't good enough. The thing that makes me feel bad is this feeling it's some sort of off key joke. Maybe it has a use, but to me it seems like why is a big serious company like nvidia, making noise about constructing a ventilator that no doctor would use unless in an absolute emergency where they couldn't use a serious ventilator?
- ckdarby 7y ago

WheelsAtLarge 7y ago

One of the items, of many, that worries me about this wild west season of ventilator innovation is that many of these new ventilators lack a common interface. Users are trained a long time on how best to use a ventilator using a very limited variety of user interfaces. I think it's very dangerous to expect a respiratory therapist to handle all the varieties of respirator interfaces that I'm sure are being created. New interfaces always mean mistakes, it's part of the learning process. We see it in software all the time. A user gets a brand new interface so it takes time for the user to acclimate to it. When you are dealing with a wordprocessor it's no big deal to make a mistake but a machine that can save or kill a patient depending on how it's being used is a whole different situation. I wonder, what is being done about that?

rkangel 7y ago
I can only speak for the UK regulatory environment here, but actually the situation is better than you might imagine. The best example is alarming - there are some common alarm patterns (the pattern you beep and flash your lights) for different priorities of alarm. My understanding is that these aren't standrards you must adhere to, but because they're 'accepted' using them makes your life much easier with the regulators so everyone does unless they've got a good reason not to.