Distorted or misrepresented source data OSCAR vs. SleepHQ

[Tec5]

The info i provided was accurate. I'm not here for a debate. Cheers

I'm not interested in debating (although what is to be debated is debatable),
However, it would be helpful to know if you are speaking in an official capacity for SleepHQ or if you are just some random internet guy that is spinning an explanation, albeit a plausible one.

[lazarus]

Obviously a lo-res rep, which should be good enough for the average Joe who doesn't want to blow up his browser with unnecessary details.

For hi-res reps, please see OSCAR.

[Tec5]

Prolly just a low-res rep. The hi-res reps are in OSCAR.

'Prolly' correct. I've always assummed that the SD card contains the record of each and every sensor reading, ergo highest resolution possible. I've also assumed that OSCAR acquires ALL that data stream and plots each record.
So now to get a lower resolution (for transmission purposes) one has to translate the data to a lower frequency. That involves some decision process.
Shall we just transmit every third data point? Every 6th data point? Should we average 10 data points, should we use a moving average? Just how do we get to a less granular representation (lower resolution) of the data?

These are questions that I suppose the "Average Joe" is not interested in.

But in a community that seems to be highly analytical ---- "see that little bump there" ----- is mightily important.

'Prolly' Wanderlustralia is correct ----- the degree of resolution , it depends on the intended audience.

[lazarus]

I meant Wanderlustralia must be a low-res representative.

[chunkyfrog]

The uncertainty principle--measurement influencing results.
Of course, we have no way of quantifying how much, or even if . . .
Most of the time, it doesn't matter.
Like I don't invest in expensive audio because I know my hearing is not capable
of perceiving enough difference.

[Wanderlustralia]

Honestly, i love all of you apnea nerds geeking out on flow rates and yawns disguised as centrals etc and helping everyone out. There's no bigger sleep geek than yours truly, but I'm also a realist. You're the 1% (me included), and I'm stoked you have OSCAR for all your needs because SleepHQ was not made for you. It was made for the other 99%. Anyway, In 6 months, all these posts will be completely irrelevant. SleepHQ will be charting sleep data from any wearable that can write data to Apple or google health. It won't just be cpap. It will be a complete sleep health app. Millions of Apple watch, fitbit, garmin users etc tracking their Sleep Cycle and other health data alongside their cpap data on SleepHQ and it will all be completely automatic. Have a good one!

[Rubicon]

There's no bigger sleep geek than yours truly, but I'm also a realist.

With delusions of grandeur.

Anyway, In 6 months, all these posts will be completely irrelevant. SleepHQ will be charting sleep data from any wearable that can write data to Apple or google health. It won't just be cpap. It will be a complete sleep health app. Millions of Apple watch, fitbit, garmin users etc tracking their Sleep Cycle and other health data alongside their cpap data on SleepHQ and it will all be completely automatic.

May 18th, huh?

Holding my breath till then, Nostradamus.

[lazarus]

One App to rule them all,
One App to find them,
One App to bring them all
And in the darkness bind them.

Or for any who may happen to work in Mordor:

Ash app durbatulûk,
Ash app gimbatul,
Ash app thrakatulûk
Agh burzum-ishi krimpatul.

[zonker]

Honestly, i love all of you apnea nerds geeking out on flow rates and yawns disguised as centrals etc and helping everyone out.

i sort of look at this another way. i don't have the faith that you have that it will all be automatic. i think folk will still come to cpaptalk.com for advice on how to adjust their therapy for a better night of sleep. even with numbers on sleephq now, people are here looking for advice.

MY hope is that the more trusted members of the forum, those that continue to give very good advice, will absorb data represented on sleephq and then be able to keep giving great advice.

i DO think more and more people will be going to sleephq. i don't think it will become a replacement for oscar; it'll just be another available tool in the drawer.

[lynninnj]

Honestly, i love all of you apnea nerds geeking out on flow rates and yawns disguised as centrals etc and helping everyone out. There's no bigger sleep geek than yours truly, but I'm also a realist. You're the 1% (me included), and I'm stoked you have OSCAR for all your needs because SleepHQ was not made for you. It was made for the other 99%. Anyway, In 6 months, all these posts will be completely irrelevant. SleepHQ will be charting sleep data from any wearable that can write data to Apple or google health. It won't just be cpap. It will be a complete sleep health app. Millions of Apple watch, fitbit, garmin users etc tracking their Sleep Cycle and other health data alongside their cpap data on SleepHQ and it will all be completely automatic. Have a good one!

My Apple watch tracking my sleep data has zero to do with Sleep HQ.

But yeah, if it can overlap the timing of the sleep data and O2 data from the watch with the sleep data from the SD card, that might be a worthy goal.

Not sure how much I would be willing to pay for it. Maybe a one time fee of $5 or so but thats about it.

[lazarus]

Honestly, i love all of you apnea nerds geeking out on flow rates and yawns disguised as centrals etc and helping everyone out.

i sort of look at this another way. i don't have the faith that you have that it will all be automatic. i think folk will still come to cpaptalk.com for advice on how to adjust their therapy for a better night of sleep. even with numbers on sleephq now, people are here looking for advice.

MY hope is that the more trusted members of the forum, those that continue to give very good advice, will absorb data represented on sleephq and then be able to keep giving great advice.

i DO think more and more people will be going to sleephq. i don't think it will become a replacement for oscar; it'll just be another available tool in the drawer.

Well, if yer gonna get all serious about it like that, then I'll be forced to admit that I agree with you completely.

As a former long-time user of MyAir, I think that anything that gives quick easy access to some sleep-breathing data for the masses is part of the good fight. So more power to the SleepHQ dudes and the MyAir dudes and all the rest.

And there's nothing wrong with making a buck or two at it, as far as that goes, from where I sit, if some users find it worth it. The very existence of OSCAR as a free product should keep other services in line and should keep those services from gouging--which is just one more way the hard-working people of OSCAR are doing a service on multiple levels for the entire planet--round or flat.

Oops, wrong thread again.

[Dog Slobber]

The browsers would crash in an instant if they were hit with super high res data with hundreds of thousands of data points. It would cause all sorts of performance issues. The data is the same, just lower res. It's the same thing youtube does when you're on a slow internet speed. Instead of showing you the 4k stuff and buffering etc, they show you the 720p. For 99.9% of viewers, no one cares! The same goes for SleepHQ. For 99.9% of the users. They're happy with the level of detail provided. SleepHQ is not OSCAR! It was made for the average Joe.

Sorry, but I can't accept that the data loss is simply because of less granularity.

Here is an example of approximately 40 seconds of data.

oscar_vs_sleepHQ.jpg (443.75 KiB) Viewed 1427 times

The top is OSCAR, immediately below it is SleepHQ. At the bottom is the OSCAR graph represented with a granularity of 1 data point per second. The data points have intentional been left enlarged and then connected with straight lines. Making it easy to visualize how the graph was derived.

While there is *some* data loss, there just isn't the same data loss and mischaracterization of the trace that exists in the SleepHQ data. The trimming of the high and low points shows averaging, curve smoothing are also at play, or there are bugs.

The info i provided was accurate. I'm not here for a debate. Cheers

This is probably one of the most disturbing things you've written.

The data you presented clearly demonstrates there is more going on than simple loss of granularity.

And your response is You and your data are correct, and that's not debatable.

How many other bugs exist in your software, that isn't worth investigating, nor will you debate it.

Let's consider your YouTube analogy.

Youtube reduces data, by properly sampling the 4k video and presents the less granular images without changing it's essence. It doesn't simply truncate the top and bottom third and averaging the remain pixels.

The fact that most of the higher resolution traces are represented without serious changes, and not crashing the clients or server is pretty compelling evidence that the same is possible with the mask pressure trace.

[robysue1]

Dog Slobber,

I want to thank you for taking the time to draw that graph that shows so clearly that sampling at a lower resolution (1/second) still shows a good approximation of the correct peaks and troughs of the hi res mask pressure graph. Clearly SleepHQ is either doing something weird with the data to create their mask pressure graph OR there's a bug in their software.

In either case, Wanderlustralia refuses to admit there's a problem with a graph that is quite obviously wrong in the SleepHQ presentation of the data. And that does call into question what other stuff might be buggy.

When I look at my Resmed AirCurve 10 VAuto data, all of the other graphs, on a cursory look in my data seem to be OK, although they only draw the IPAP pressure graph for the pressure graph instead of both the EPAP and IPAP curves.

But when I look at my PR DreamStation BiPAP Auto data, however, I have a serious problem with the way SleepHQ presents the pressure data: SleepHQ draws one pressure graph that is neither the EPAP graph nor the IPAP graph that is shown in both Encore Pro (the PR software) and SleepyHead. Oscar puts in a third graph that looks remarkably like the Pressure graph in SleepHQ. Personally, I would call the way SleepHQ draws the pressure curve for PR BiPAPs a bug that ought to be fixed.

Here are the IPAP and EPAP pressure curves for Oct.29, a night that I used the DreamStation, as shown in SleepyHead. (The graph in Encore Pro is similar.)


Here are the IPAP and EPAP pressure curves, along with a mystery "pressure" curve that is not present in Encore, as shown in Oscar:

I asked about that mystery curve in this thread. Nobody gave a definitive answer, but the consensus was that it must be some kind of a moving average mask pressure value. (The Oscar documentation at ApneaBoard says this is just lo res sampling of the mask pressure data, but they haven't really really explained where that data comes from; when I look at my husband's PR System One APAP data, Oscar shows the pressure set curve and an EPAP curve that appears to be derived from the Pressure set curve and some kind of mystery data in the same way SleepyHead does.)

And here is the so-called pressure curve as drawn by SleepHQ:

Given that this is BiPAP data and the PS is allowed to from 3 to 5 during the night, only drawing the mystery "pressure" curve is badly distorting the pressure data. Looking at this curve, you'd think that my pressure never reached 7cm, which is my minimum IPAP pressure and so the pressure went to (at least) 7cm on each and every inhalation that I took over the course of the night.

Looking at the Statistics part of the SleepHQ data for this day is also misleading:

It correctly identifies the median, 95%, and maximum EPAP pressures. But missing are the median, 95%, and maximum IPAP pressures, which were 7cm, 9cm, and 9cm respectively on this night. The median, 95%, and maximum pressure levels shown in SleepHQ have nothing to do with my machine's therapeutic settings.

I could see the case where a newbie or average PR BiPAP Auto user might look at the pressure information in SleepHQ and start to wonder whether there is something wrong with their BiPAP since the pressure is never reaching the typical IPAP levels that would be expected given the clinical settings.

Finally, it should be clear that the PR pressure curves are most definitely not hi-res. They represent the pressure settings for IPAP and EPAP, not the actual pressure at the mask. And PR machines take their own sweet time in adjusting the pressure settings when they decide to increase either the IPAP or the EPAP or both.

[Wondering1]

Out of curiosity, what is the sampling rate for Resmed Airset 11 machines for flowrates and mask pressures , and is each of those datapoints recorded independently on the SD card.
I assume that OSCAR plots all available datapoints from the SD card - is this not so?

[palerider]

Out of curiosity, what is the sampling rate for Resmed Airset 11 machines for flowrates and mask pressures , and is each of those datapoints recorded independently on the SD card.
I assume that OSCAR plots all available datapoints from the SD card - is this not so?

Oscar plots the high rate 'flow rate' and 'mask pressure' at 25 data points per second, which is what is recorded on the SD card.