ReMed Airsense software/firmware

[Ralph436]

1) Does the Airsense 11 Autoset Elite utilize the Resmed S-10 software?

2) Does the Airsense 11 Autoset Elite utilize the following algorithm to detect hypopneas? :

     IF a 12 second RMS average inspiration volume is reduced by 25-50% for 10 seconds or more AND
         at least one breath (in the 10 second segment) has a positive flow limitation index
     
      THEN that 10 second segment is flagged as a hypopnea.

My principle focus is on the duration of the comparison segment (that is the segment prior to the potential hypopnea); is it 12 seconds or some other duration?

Secondarily, is it average volumes that are being compared or average peak flow rates?

[Muggzy]

I don't have any links handy at the moment but I'm reasonably confident it's a 120 second sliding window on the A10 AutoSet, or at the very least using a120 second window allows me to reliably reproduce the exact same results.

Also, I think it's RMS of the absolute flow, though I'm not sure it makes any difference vs. just looking at inspiratory flow. No need to track peaks, the RMS of all samples works.

Not at all familiar with that particular machine though.

[dataq1]

looking into it.

[Muggzy]

Tidal Volume is a derivative of flow, so the flow data is really the ground truth data to look at. When I say absolute flow I mean the absolute value of all of the flow samples, in other words making them all positive before doing the calculations.

Give me a bit and I'll try to come up with some better explanation, kinda hectic at the moment

[Ralph436]

me too, hectic that is, so take your time and thanks

[BoxcarPete]

For what it's worth, before I started taping my mouth, the machine would always register lower TV during the times that I was mouthbreathing on exhale, causing the machine to lose track of some of the air and adjust my leak rate and inspiration/expiration flow rates off the nominal center point to balance that out. On the other hand, this typically caused my mouth to leak like a leafblower, so I take any stats during this time period with a grain of salt.

The point is, that the machine has only one flow sensor, so the volume numbers it shows you are calculated and derived from flow rates already.

[Muggzy]

Actually on my way to a doctor's appointment, so I hope this makes some kind of sense because it'll be a few hours before I can try again lol.

It looks like they actually compare using a 60s window for the baseline and a 2s window for detecting flow reduction:

   

Here is a visual representation of how I interpret that, with the more stable red line being a 60s window used for a baseline, and the other red line being a 2s window used for comparison :

   


What I've done is use a 120 second sliding window for a baseline (worth noting that the recorded data is also half the sampling rate of what they specify), which is fairly resilient to minor/short changes in flow and seems to give roughly identical results in practice (direct comparison of bespoke apnea/hypopnea detection versus ResMed's flagged events just for kicks and grins). A 12 second window doesn't work well as a baseline, I think they probably mean that they use it for the comparison against baseline?

   

[Muggzy]

Regarding my comment earlier about Tidal Volume being derivative of flow...

Here is an example of calculating Tidal Volume using nothing but Flow data, overlaid on the recorded Tidal Volume waveform provided by ResMed.

Not quite identical, probably because I didn't integrate the signal properly, but obviously representative enough to illustrate what I'm talking about.

The Respiration Rate, Inspiration Time, Expiration Time, Tidal Volume, and Minute Ventilation can all be calculated using nothing but the Flow data, and I believe that's what OSCAR does when those signals are not provided by the user's CPAP machine.

[pholynyk]

Since nobody has touched on the software of the AS10 versus AS11, I will say that the file names and content are different, so as least some parts of the software are different. I suspect the algorithms are the basically the same.

I would not be surprised if the microprocessors are different, which would, of course, preclude using the same software in both machines.

[Muggzy]

I suppose that would depend on whether the code could be cross-compiled or not. There are obviously at the very least cosmetic differences.

Somebody somewhere on the forums previously made the point that new algorithms would likely require new patent applications and approval process etc, and that it was most likely that the algorithms used are the same, which I thought sounded reasonable.