Video illustrating test setup:

The machine is connected through a single tube to a respiratory valve (partially red) – air then goes through the pressure and flow sensor (test setup, not final sensor implementation) – then through the Fluke VT900 to an adjustable test lung. Exhalation goes back to respiratory valve and out through manually adjustable PEEP-valve (green)

( Fluke on loan from Rood, test lung on loan from Technimed)

This is still prototype 7. We will switch the test setups to Beta version 1 as soon as these arrive.

NOTE: we are aware connection between respiratory valve and patient should be as short as possible – just not very practical to do in this test setup.

Video demonstrating interface (non-touchscreen version):

Setup start at RR=20, Ppeak=30cmH20 and a PEEP setting of 10.
Video shows adjustment of RR, Ppeak, ramp settings, IE-rate, and the adjustment of the PEEP-valve.

NOTE:
-alarms have not been set correctly yet in this setup
-IE units should be adjusted in the interface.
IE=0.5 in this version means 50% of total time is inhalation. So IE=0.5 is 1/1, IE=0.33 is 1/2.
-Values in the teal rectangle at the right of the screen should show measured values for P, V, RR, MV – not active in this demo.

Static image, where you can read the small print
(open in new tab to enlarge)

Video demonstrating the use of breathing trigger:

Video demonstrating breathing support through pressure-trigger or flow-trigger.

Endurance tests on the ambu bags showed failure after 160’000 cycles. Bags were fully depressed at 1Hz. This translates to failure after 3 days of RR=35 BPM at full volume (1000 mL).

Failure mode is a tear near the hard, plastic endcap of the ambu bag.

We have updated the plunger design, to a narrower & more rounded version. This still achieves the requires Vt of 800mL , while putting less stress on the edges of the ambu bags.

The test version also did not have the silicone covers.
This updated version has been sent to the test facility.

Documention was moved to this site.

The team is sourcing parts for the Beta1 – series.

The interface has been further developed.
There is still a lot of work to implement multiple alarms, but the PC and the Arduino can now be disconnected and reconnected without restarting.

If sensors are disconnected, the machine now goes to a safe open-loop mode and keeps running.

Quick interface demo from today.

The pc-interface is working.
Not the final graphical design, but we basically control the pump from a touch screen or keyboard now

This is amazing work from Thomas Van Den Dries & Tom Bruyneel, Frank Vanbever, Stein Crispel, Laurent Segers, Branimir & all 

Video: https://photos.app.goo.gl/UnNw3LmDGuY6QVMTA

Photos: https://photos.app.goo.gl/8dyWrT299qnPVCM59

Flow curve is missing on the screen, but we have the data. Flow sensor drift has been solved. User can control Vt, RR, IE, Ppeak, ramp time and can set alarm ranges on the measured values.

Computer gives alarm if microcontroller stops working and vice versa. There are no more push buttons.

It has become clear the machines will not be needed in Belgium. 
We are continuing the project, but there will not be 1000’s built in Belgium .

It also means fast-track certification in Belgium is off the table, since no need.

We have started a run of 50 with the help of Audi Brussels; we are definitely skipping a few steps and expect these beta-models to need fixes, but we will have enough to send out for testing. We expect to have those build by the end of next week
(IF the PCB’s we ordered work without dramatic mistakes,
otherwise at least one week later)

We will be putting a group on getting the documentation organised, so we can get clean files out to teams in other countries.

Ronald has set up a group in Montreal, and we also got a request to share files with Sannio University (Italy).
We also sent them the dropbox archive from the day before.

Everything is changing quickly, so we will have to reorient next weekend, but we have every intention to finish the project and get to a reliable, functional and open-source machine.

Best, Lieven

Hi Ronald & all, 

As our insights in this problem grow, I now think:

-a reliable mechanical pump design is 20% of the solution

-PEEP-valves, overpressure valves,tubing, 3-way valves and filters are relatively simple parts, but not necessary easy to source. If you can find a way to massively source those outside of medical suppliers, that is 15%

-the software and sensors is 65% of the value. The more I discuss this, the larger this number gets.

In essence: if you accurately monitor & control your pressures, volumes, RR, IE and oxygen content & you have a clear interface with visual feedback, you have functionality – no matter what  the air source is.

Our sensors are going to end up to be plastic parts (3D-printed) , with mass-produced electronics. That is 100g of hardware that fits in a shoe box and can be mailed off. Add a laptop for the graphs and control (dump the physical buttons)

I was discussing the italian diving-mask solution (the Decathlon snorkel mask) with Manu Malbrain today, he is in favor of this as he sees a lot of advantages in a solution where he can filter the return air, establish PEEP & high oxygen content, but can avoid or postpone all the complications / nasty side effects of intubating patients.

Our pump would not work with that, not enough Minute Volume for the higher flow needed as you breathe out into the freshly added air mixture, but there is no reason our software with good sensors cannot control a turbine that can deliver the required 60L/min of oxygen/air mixture.

There is also no reason this same control software could not be adapted to fit a proportional valve on a pressurized air/oxygen source.

Please let me know if my view is wrong. This is a project where me and my whole team are learning on-the-go, there was no time for a literature study, so we are potentially making stupid choices.

Here is a link to a demo-video from yesterday. 

https://www.dropbox.com/s/366p6gesj0yt3od/Demo%20video%2028%20Maart%20VUB%20Ventilator.mp4?dl=0

Don’t get too enthusiastic, that Prototype 5 was cursed and died of electrical problems after two nights of useless work-around attempts.

Turns out windscreen wiper motors have their ground on the casing. If you reverse the motor (which you don’t do in a car) you put 12V on the frame of the motor.  We knew that, we thought we could ignore that for now.

If you do a plexi machine, you are just fine, it’s low voltage. If you build an aluminium version, well…  

We think fondly of Prototypes 1 to 4. We will not speak of 5 again.

  6 & 7 have rubber feet on the motor mounts and a plastic piece in the pull rod.  

Also, Audi is coming tomorrow to help with a 10-man team.

SKF did overnight shipping on Sunday, after some of their people spent a day disinfecting  2000 ball-bearings, then re-greasing them. We think we asked for 100 for the first 20 Beta-models, but I am not sure at this point.

best, Lieven

It is hard to describe how hectic the past days have been. To people whose phone calls, messages or emails we have not answered, we apologise. We have been organising the project, and set up team leads for the relevant subsections. This is new terrain for me & Jonathan and it is a very steep learning curve. 

We have grown from 8 people trying a far-out concept in 4 days, to 60 people after another 4. 

It is now only 12 days after the moment when some idiot suggested we should just build & test this thing and see if it could work instead of talking about it.

We have been working with 30+ people in the lab the past days, with another 30 working from home on Slack.

I have been trying to take some video and snapshots on quieter moments:

If you want to get a better idea what it is like here, here are some images & video’s:

https://photos.app.goo.gl/5rjZHk3ZTEPccZYk7

Sensor-based motor control is working, data monitoring is working, breathing detection is working, control panel is finished. PCB designs are at the manufacturer for a small prototyping run. 

We have been receiving massive support from industry and university.
Four days ago we decided to switch from stepper motors to DC windscreen wiper motors, after testing different motor concepts. Three days ago Jonathan sent one email asking where we could get some the same day.

This is yesterday (Nemo, team lead component sourcing)

We got motors donated for testing by Mazda, Ford, Audi, Volvo, Renault and Mercedes.

Volvo Trucks let us know “We ‘ll send 2 motors over with a taxi” and 3 hours later this package got dropped off:

Volvo Trucks for the win…

Eurocircuits is rushing our prototype PCB’s for free. 

We ordered 150 kg aluminium from Dejond, with the request if we could pick it up the same day and explained why we needed them urgently and got an invoice for 0 euro in our mailbox….

Five people from Flanders Make are working full-time to help us roll out the first series as soon as possible. Technically, sooner than possible, since we will de facto be beta-testing while production starts.

FAGG published a shortened certification document for (temporary) use of emergency ventilators in Belgium. So there is now regulation that allows for the use of the kind of machine we are developing.

Lastly, a big thank you to all the friends and colleagues who are helping out from home and sacrificing some of their time, without seeing what’s going on and being in the middle of it. Thank you

Be safe, Lieven

 We have switched to a DC-motor design, based on a windscreen wiper motor.

After testing we concluded the stepper-based design was underpowered to reach 40 cmH20 at 35 BMP at I/E=½

We don’t believe any of the stepper-based designs out there are capable of achieving these specs.

PCB’s are designed

Software is taking shape.