Inogen

[Aaviationist]

6 minutes ago, wombat said:

I've already got adapters to go from my industrial tank to my Aerox portable tank as well as the onboard tank for the Mooney. Adapting the output from a homefill machine will be easy.

I do wonder how to test the output from the homefill machine to determine if it's making relatively pure oxygen and if there are any unpleasant contaminants.

It does that itself. The concentrator has a measurement at the output and lights a led. 
 

the compressor has a sensor on the input and also an associated led. The compressor will not compress without meeting the threshold for o2 concentration. 
 

when I tested mine it was 98%. 

[N201MKTurbo]

5 hours ago, Aaviationist said:

It does that itself. The concentrator has a measurement at the output and lights a led. 
 

the compressor has a sensor on the input and also an associated led. The compressor will not compress without meeting the threshold for o2 concentration. 
 

when I tested mine it was 98%. 

I think the alarm and limit is 89%. I don’t think they can do 98% very well.

Actually, there are multiple alarm levels. It looks like the highest is 85%.

https://invacaredocs.com/assets/documents/PIM/Platinum_9_User_Manual.pdf

[Aaviationist]

Mine was outputting 98% when we tested the output of a bottle I had filled. 

[N201MKTurbo]

The homefill inlet oxygen sensor requires 90%. I was close.

https://reliablerespiratory.com/wp-content/uploads/2022/01/Homefill-2-manual.pdf

[N201MKTurbo]

Yours is exceptional if it is doing 98%. Good deal.

I used to work for a place that did a lot of cryo work. They had cryo pumps that used molecular sieve. They would recycle it with this rig they made. It was a canister with electric heaters. They would put the sieve in it and heat it to about 500F under a high vacuum. They would cook it for a few days. It was better than factory fresh when it came out.

Air pollution will eventually foul the sieve. Yours gets so little use, it will probably last a long time.

[Aaviationist]

7 minutes ago, N201MKTurbo said:

Yours is exceptional if it is doing 98%. Good deal.

A lot of it probably depends on environment. I’m at sea level and it’s in the air conditioned basement of my house. 
 

I’ve seen others with units in Colorado only doing 91/92%.   When Inogen claims not to use their unit for 2 people above 14k, I believe them. 

Edited June 26 by Aaviationist

[Yetti]

just thought I would stop by and say check FaceBook Market Place for Oxygen Concentrates.      Professional course on a closed driver.

[toto]

5 hours ago, Yetti said:

just thought I would stop by and say check FaceBook Market Place for Oxygen Concentrates.      Professional course on a closed driver.

I’ve seen a ton of these on Facebook Marketplace, but people should definitely approach with a degree of skepticism…. Some of these deals are way too good to be true. 
 

[N201MKTurbo]

I got a colonoscopy yesterday. They left be alone for 1/2 hour hooked to the patient monitor, so I started playing with the pulse ox. The monitor had ECG on the top trace, SpO2/pleth wave on the second trace and respiration rate on the bottom trace. It would take blood pressure about every 5 minutes. With what was connected to me I'm not sure how it was deriving the respiration, but it was. If I held my breath the trace would stop and it would start beeping. Anyway, back to the SpO2. I tried it on all my fingers with different orientations. It would always eventually get a good pleth wave and give a stable SpO2 reading, unless I didn't insert my finger deep enough into the sensor. It seemed to have an AGC function that would try to optimize the red and IR levels. I'm not sure what the pleth wave was actually showing. It could be the red channel, the IR channel, the sum of them, or the difference of the two. It could also be some derived combination of the two. Whatever it was doing, it was pretty robust and once I stopped messing with it, it would get a steady pleth wave and gave consistent readings. 98 in my case. If I jacked with it and the pleth wave trace went to hell, the reading always went down, never up, which is good. I couldn't get it to read high, but I was close to max as it was, and I didn't have CO poisoning either.

 

https://en.wikipedia.org/wiki/Photoplethysmogram

[Marc_B]

I love the story Rich.  I can picture one nurse to the other "what's he doing in there??"  "Don't know, must be the drugs!" "Let's just hope he doesn't try the pulse ox on more than just his fingers!"   But perfect description of waveform showing that you have accurate information...waveform acts up = probably not giving accurate information.  Also why most quality pulse ox's have some type of graphical output of waveform (bar going up and down, or a wave tracing).

Respiratory rate is either calculated from the pleth if just using a pulse ox.  All the calculations are complicated, but it can use the data from the pulse ox to estimate respiratory rate and they are pretty accurate most of the time.

https://pmc.ncbi.nlm.nih.gov/articles/PMC4309914/

But respiratory rate in a hospital setting is sometimes based on end tidal capnography (ETCO2).  For you it might have been a a small mouth scoop on the nasal cannula to capture exhaled CO2 via the tube and measure that numerically / show wave form (breathe in CO2 drops, breath out CO2 goes up).  ETCO2 is also connected inline to ventilators and advanced airways.  Respiratory rate & ETCO2 are both great for catching early apnea (not breathing) or ineffective breathing before you begin to desaturate.

Graph showing a basic idea of what happens when you paralyze a patient, they aren't breathing, and how long it takes their oxygen saturation (SaO2) to drop...usually it's somewhere around 3-6 min.  Versus if you stop breathing your end-tidal CO2 (exhaled amount of carbon dioxide) drops immediately.  So for procedures, respiratory rate and ETCO2 are super helpful to catch issues before they result in drop in oxygen levels.

This is somewhat pertinent to pilots as well, as everyone has different baseline physiology that may or may not result in desaturation more quickly with an oxygen issue in flight.  We can't always predict who will, or will not, desaturate more rapidly, but critical ill patients, obese patients, and children usually desaturate quicker than a normal, healthy adult.

[N201MKTurbo]

1 hour ago, Marc_B said:

I love the story Rich.  I can picture one nurse to the other "what's he doing in there??"  "Don't know, must be the drugs!" "Let's just hope he doesn't try the pulse ox on more than just his fingers!"   But perfect description of waveform showing that you have accurate information...waveform acts up = probably not giving accurate information.  Also why most quality pulse ox's have some type of graphical output of waveform (bar going up and down, or a wave tracing).

Respiratory rate is either calculated from the pleth if just using a pulse ox.  All the calculations are complicated, but it can use the data from the pulse ox to estimate respiratory rate and they are pretty accurate most of the time.

https://pmc.ncbi.nlm.nih.gov/articles/PMC4309914/

But respiratory rate in a hospital setting is sometimes based on end tidal capnography (ETCO2).  For you it might have been a a small mouth scoop on the nasal cannula to capture exhaled CO2 via the tube and measure that numerically / show wave form (breathe in CO2 drops, breath out CO2 goes up).  ETCO2 is also connected inline to ventilators and advanced airways.  Respiratory rate & ETCO2 are both great for catching early apnea (not breathing) or ineffective breathing before you begin to desaturate.

Graph showing a basic idea of what happens when you paralyze a patient, they aren't breathing, and how long it takes their oxygen saturation (SaO2) to drop...usually it's somewhere around 3-6 min.  Versus if you stop breathing your end-tidal CO2 (exhaled amount of carbon dioxide) drops immediately.  So for procedures, respiratory rate and ETCO2 are super helpful to catch issues before they result in drop in oxygen levels.

This is somewhat pertinent to pilots as well, as everyone has different baseline physiology that may or may not result in desaturation more quickly with an oxygen issue in flight.  We can't always predict who will, or will not, desaturate more rapidly, but critical ill patients, obese patients, and children usually desaturate quicker than a normal, healthy adult.

They didn’t have a cannula on at the time. They didn’t put one on until they wheeled me into the procedure room. On further reading, it looks like you can get respiration from either the ECG or the Pleth. It also seems the pleth display isn’t well defined as to what exactly it is showing. I assume the engineers at the monitor company have access to all the internal data and have their own algorithms to derive the displayed data. Being an FDA certified patient monitor, it was compleatly validated.

[N201MKTurbo]

2 hours ago, Marc_B said:

I love the story Rich.  I can picture one nurse to the other "what's he doing in there??"  "Don't know, must be the drugs!" "Let's just hope he doesn't try the pulse ox on more than just his fingers!"   But perfect description of waveform showing that you have accurate information...waveform acts up = probably not giving accurate information.  Also why most quality pulse ox's have some type of graphical output of waveform (bar going up and down, or a wave tracing).

Respiratory rate is either calculated from the pleth if just using a pulse ox.  All the calculations are complicated, but it can use the data from the pulse ox to estimate respiratory rate and they are pretty accurate most of the time.

https://pmc.ncbi.nlm.nih.gov/articles/PMC4309914/

But respiratory rate in a hospital setting is sometimes based on end tidal capnography (ETCO2).  For you it might have been a a small mouth scoop on the nasal cannula to capture exhaled CO2 via the tube and measure that numerically / show wave form (breathe in CO2 drops, breath out CO2 goes up).  ETCO2 is also connected inline to ventilators and advanced airways.  Respiratory rate & ETCO2 are both great for catching early apnea (not breathing) or ineffective breathing before you begin to desaturate.

Graph showing a basic idea of what happens when you paralyze a patient, they aren't breathing, and how long it takes their oxygen saturation (SaO2) to drop...usually it's somewhere around 3-6 min.  Versus if you stop breathing your end-tidal CO2 (exhaled amount of carbon dioxide) drops immediately.  So for procedures, respiratory rate and ETCO2 are super helpful to catch issues before they result in drop in oxygen levels.

This is somewhat pertinent to pilots as well, as everyone has different baseline physiology that may or may not result in desaturation more quickly with an oxygen issue in flight.  We can't always predict who will, or will not, desaturate more rapidly, but critical ill patients, obese patients, and children usually desaturate quicker than a normal, healthy adult.

Cool article you linked. It is a bit tricky to extract a ~16 breaths per minute signal from a ~70 beats per minute signal. Especially when you don’t actually know the frequency of either to start with.

[201Mooniac]

55 minutes ago, N201MKTurbo said:

Cool article you linked. It is a bit tricky to extract a ~16 breaths per minute signal from a ~70 beats per minute signal. Especially when you don’t actually know the frequency of either to start with.

Actually this is a well studied part of signal processing called signal estimation.  The technique shown in the linked article has been a standard approach for almost 40 years now.  There are many other approaches as well but the one presented (which was at least partially developed for a military application of analyzing the modulation and beat frequencies of unknown communications signals) works well considering the non-stationary time varying nature of the signal.

[EricJ]

2 hours ago, 201Mooniac said:

Actually this is a well studied part of signal processing called signal estimation.  The technique shown in the linked article has been a standard approach for almost 40 years now.  There are many other approaches as well but the one presented (which was at least partially developed for a military application of analyzing the modulation and beat frequencies of unknown communications signals) works well considering the non-stationary time varying nature of the signal.

Was just gonna say, they're far enough apart in frequency to be generally separable if not orthogonal.  The characterizations shown in the article demonstrate that it's fairly practical.  Probably don't even need a wavelet, but if that approach works, great.   Sometimes I think wavelets get used just because they make good buzzwords in the presentations or even in the marketing materials.

Pretty cool, though!   

[201Mooniac]

19 minutes ago, EricJ said:

Was just gonna say, they're far enough apart in frequency to be generally separable if not orthogonal.  The characterizations shown in the article demonstrate that it's fairly practical.  Probably don't even need a wavelet, but if that approach works, great.   Sometimes I think wavelets get used just because they make good buzzwords in the presentations or even in the marketing materials.

Pretty cool, though!   

You are correct that in this case a heuristic approach would likely work however the wavelet approach gives less to tune though is possibly more computationally intensive.

[EricJ]

8 minutes ago, 201Mooniac said:

You are correct that in this case a heuristic approach would likely work however the wavelet approach gives less to tune though is possibly more computationally intensive.

Yeah, I was wondering how much computational bandwidth a pulse ox has to have to be able to do that, unless it's sent offline to a separate display or something.

[201Mooniac]

5 minutes ago, EricJ said:

Yeah, I was wondering how much computational bandwidth a pulse ox has to have to be able to do that, unless it's sent offline to a separate display or something.

Given the power of the MCUs I use these days, there is way more than enough to do this calculation.  When I was in grad school 40+ years ago, I worried about every cycle and every byte, these days a $1 MCU runs my deep learning models which was completely unthinkable back then.

[N201MKTurbo]

38 minutes ago, EricJ said:

Yeah, I was wondering how much computational bandwidth a pulse ox has to have to be able to do that, unless it's sent offline to a separate display or something.

Getting the SpO2 doesn’t take any signal processing. Getting the respiration from it does.

[EricJ]

28 minutes ago, 201Mooniac said:

Given the power of the MCUs I use these days, there is way more than enough to do this calculation.  When I was in grad school 40+ years ago, I worried about every cycle and every byte, these days a $1 MCU runs my deep learning models which was completely unthinkable back then.

Yeah, I've done a bunch of deployed projects with software radios running on RPis over the last ten years, and they're doing a lot more processing and wind up not using much of the total available resources.   It's crazy what you can get done for a few bucks these days.   And now I see people are writing libraries to offload typical signal processing stuff into the GPU in the RPi (e.g., FFTs), so it's getting interesting.   I have a hard time keeping up with the actual state-of-the-art any more since I'm no longer in it full time.

 

[EricJ]

Just now, N201MKTurbo said:

Getting the SpO2 doesn’t take any signal processing. Getting the respiration from it does.

Yeah, I was wondering where they were computing the wavelet transforms for the respiration, or however they're actually doing it.

[MikeOH]

1 hour ago, EricJ said:

Yeah, I was wondering where they were computing the wavelet transforms for the respiration, or however they're actually doing it.

Why is a wavelet transform needed vs. just an FFT?

[EricJ]

6 minutes ago, MikeOH said:

Why is a wavelet transform needed vs. just an FFT?

That's what the article said they were using, presumably to better handle the time variance and non-stationarity, which is something that wavelets can be good at.   I suspect it could be done with more traditional means, and the article mentioned short-time FFTs, etc.    I think if the signal always looked like it does in Figure 1 it'd be pretty straightforward, but the corner cases and time variance, etc., probably led to their alternative approach into the buzzword algorithms. 

But if it works, more power to them.

[MikeOH]

1 minute ago, EricJ said:

That's what the article said they were using, presumably to better handle the time variance and non-stationarity, which is something that wavelets can be good at.   I suspect it could be done with more traditional means, and the article mentioned short-time FFTs, etc.    I think if the signal always looked like it does in Figure 1 it'd be pretty straightforward, but the corner cases and time variance, etc., probably led to their alternative approach into the buzzword algorithms. 

But if it works, more power to them.

Thanks.

Although, I think your 'marketing' explanation is at least as likely

[201Mooniac]

3 hours ago, MikeOH said:

Why is a wavelet transform needed vs. just an FFT?

An FFT could work for this signal but the analysis relies on some assumptions that don't hold for this use case including the signal and noise both being harmonic and stationary throughout the window interval.  These assumptions might be okay if you can maintain sufficient precision throughout however I'm not clear on the analysis requirements to know whether that is true or not for this use case.

[dkkim73]

13 hours ago, N201MKTurbo said:

got a colonoscopy yesterday. They left be alone for 1/2 hour hooked to the patient monitor, so I started playing with the pulse ox

And I thought actual science was dead in the centralized funding + corporate media era. 

Great thread. 

This is an area begging for some inexpensive NORSEE safety equipment. 

I think the guy who mentioned capnography has some ideas...