Help Interpreting Data From SleepyHead

[richart]

Here is the data for last night from SleepyHead. A little help please on understanding what the flow rate data is telling me.
1) The positive flow rate seems to average around 30. The negative flow rate seemed to change in cycles from about -30 to about -45 to about -60. What is the significance of that?
2) It appears that there were a couple of episodes of mask leak but to my untrained eye they do not appear significant. Correct?
3) To me, it looks like I had a very good night with the sleep apnea well under control. Am I missing anything?

[archangle]

Here is the data for last night from SleepyHead. A little help please on understanding what the flow rate data is telling me.
1) The positive flow rate seems to average around 30. The negative flow rate seemed to change in cycles from about -30 to about -45 to about -60. What is the significance of that?
2) It appears that there were a couple of episodes of mask leak but to my untrained eye they do not appear significant. Correct?
3) To me, it looks like I had a very good night with the sleep apnea well under control. Am I missing anything?

1) Zoom in and look at the "weird" areas and see what's happening.

2) Leaks look OK, but the bumps down in leak rate probably mean you're blocking your exhaust vent. Since it's not dropping below 20, you're probably not going to suffocate, but watch it. Maybe you're rolling over and blocking it with your pillow or arm.

3) Zoom in and look at the duration and severity of the apneas, but it doesn't look that bad on the zoomed out view.

[jedimark]

The first issue might be a simple plot error.. I'll look into it to make sure it's not my screw-up... (I doubt it, as this appears to be one consecutive session)

It's weird your leak drops like that at 5am and 7am, considering your pressure is constant.. It's kinda screwing with my current unintentional leaks theory. :-/

The new 0.8.8 version adds another Leaks plot that subtracts the 0th percentile (ie, the minimum) from the leak figure, effectively giving a display of unintentional leaks.. It's not a perfect solution yet, as it works on the assumptions that for part of a session the mask is fit right, and the pressure is constant (which is not the case on APAP mode)

It could be for those 10 or so minutes you were laying in a position that resolved mask leaks, or a blanket over your exhaust? An evil cat/child/partner trying to smother you?

Perhaps it needs to raise to a higher than 0 percentile to counter for this stuff.

Hmm.. If this does work, maybe if I followed the pressure graph I could calculate a better one for APAP users. You've given me an idea!

[archangle]

It could be for those 10 or so minutes you were laying in a position that resolved mask leaks, or a blanket over your exhaust? An evil cat/child/partner trying to smother you?

I'm assuming the "real" intentional leak rate is around 30 because I find it hard to believe that any of the sources of unintentional leak would be that constant.

[jedimark]

If I took for instance the 10th percentile for all leaks at each particular pressure, and used these values as a running baseline to subtract from total leak values, I could probably do away with needing the mask leak profile business, and having to know about the umpteen thousands of mask variants..

The percentile figure would be adjustable in preferences..

I wonder if this would be enough to keep people happy as far as unintentional leaks go?

[archangle]

If I took for instance the 10th percentile for all leaks at each particular pressure, and used these values as a running baseline to subtract from total leak values, I could probably do away with needing the mask leak profile business, and having to know about the umpteen thousands of mask variants..

The percentile figure would be adjustable in preferences..

I wonder if this would be enough to keep people happy as far as unintentional leaks go?

There is a perception problem with autoscaling of many of the graphs. If you have a night where there's very little variation in the value, the autoscaling blows up the variation and it looks "bad" because it's visually very choppy and looks like it varies a lot. If you have another night where most of the night is the same, but there's one or two "blips" where the value gets very high, it looks like a very smooth line for most of the night, when in fact, it's actually the same level of variation.

Us geeks are somewhat familiar with the problem of autoscaling graphs, but it's probably more confusing to normal people.

I'm not sure I have "THE" answer, though.

I know color choice on graphs is planned in the future, but please change the yellow for total leak to something else. It's nearly invisible on my display. Since it's not on the same graph as other data, there's no reason to not make it a darker color by default.


By the way, WOW! You've really made a lot of nice changes. Thanks again.

I've also said thank you the right way. Forget flowers, say it with CASH. Hint! Hint! to other satisfied users.

[jedimark]

If I took for instance the 10th percentile for all leaks at each particular pressure, and used these values as a running baseline to subtract from total leak values, I could probably do away with needing the mask leak profile business, and having to know about the umpteen thousands of mask variants..

The percentile figure would be adjustable in preferences..

I wonder if this would be enough to keep people happy as far as unintentional leaks go?

There is a perception problem with autoscaling of many of the graphs. If you have a night where there's very little variation in the value, the autoscaling blows up the variation and it looks "bad" because it's visually very choppy and looks like it varies a lot. If you have another night where most of the night is the same, but there's one or two "blips" where the value gets very high, it looks like a very smooth line for most of the night, when in fact, it's actually the same level of variation.

Us geeks are somewhat familiar with the problem of autoscaling graphs, but it's probably more confusing to normal people.

I'm not sure I have "THE" answer, though.

I know color choice on graphs is planned in the future, but please change the yellow for total leak to something else. It's nearly invisible on my display. Since it's not on the same graph as other data, there's no reason to not make it a darker color by default.


By the way, WOW! You've really made a lot of nice changes. Thanks again.

I've also said thank you the right way. Forget flowers, say it with CASH. Hint! Hint! to other satisfied users.

You can set the max value in the Graphs tab in preferences for leaks to 120, which will stop it from auto-scaling under this point.. (double click on the field allows you to edit it)

All graphs can have a recommended min/max value now.. It will only scale over these limits when it has to..

Later, when I've fully finished this bit, it will probably be the default for leaks..

I know the yellows bad.. I can hardly see it myself.. sometimes I just pick a colour by name out of thin air.. just to be different from the others.. I should be using RGB here..

Once again, when I'm finished the bit that saves and loads the XML file full of graph defaults, someone else with visual skills can come up with better default colors for us..

[archangle]

You can set the max value in the Graphs tab in preferences for leaks to 120, which will stop it from auto-scaling under this point.. (double click on the field allows you to edit it)

STUPID!!! STUPID!!!! STUPID!!!

I looked at that menu several times and didn't notice that.

Must be dain bramage.

[jedimark]

You can set the max value in the Graphs tab in preferences for leaks to 120, which will stop it from auto-scaling under this point.. (double click on the field allows you to edit it)

STUPID!!! STUPID!!!! STUPID!!!

I looked at that menu several times and didn't notice that.

Must be dain bramage.

It doesn't help when I haven't really documented how any of this new stuff works.. :-}

The Users Guide on the wiki needs updating..
If anyone's feeling linguistic they are welcome to improve on stuff there..

[Pugsy]

The Users Guide on the wiki needs updating..
If anyone's feeling linguistic they are welcome to improve on stuff there..

What do you think about adding definitions to the user guide? You know...sort of a mini tutorial on what all this stuff means and what to look for? I have been mulling it over in my head but composition is something that I totally suck at (despite some of my novels )
Don't want it to be a long elaborate explanation but just definitions and what to look for...thinking newbies here.
Used to we just told people to look at AHI...less than 5 good...leak line...more or less stable...good. Now with all this new stuff available I suspect newbies eyes will glaze over.