S9 Flow Limitation Graph -- Questions

[billbolton]

although in practice a "wide-open airway" should have a value of about 0.3

0.3 on what measurement scale?

Cheers,

Bill

[-SWS]

although in practice a "wide-open airway" should have a value of about 0.3

0.3 on what measurement scale?

I kinda suspect that 0.0 to 1.0 number might be a statistical correlation coefficient.... Resmed just might perform a covariance comparison of each inspiratory flow curve against a sinusoid (top half) as an ideal template for comparison. That's what the 420e does as well according to the patent descriptions:

This correlation coefficient is an index ranging from 1 (sinusoidal or not flow limited) to 0 (not sinusoidal).

viewtopic.php?f=1&t=31210&p=269799&hili ... nt#p269799

Similarly, here are four comparison parameters Respironics strings together in algorithmically ascertaining whether flow limitation has occurred:


NotMuffy, does that guess sound plausible to you for Resmed's 0.0 to 1.0 index? Or no?

On Edit: I see that Resmed associates a 0 value with no flow limitation and a 1 value with flow limitation---perhaps a coefficient index inversion of the above toward describing likelihood of flow limitation occurrence---with or without further probability formulation(s) added.

[NotMuffy]

I believe the ResMed algorithm is based on the work of Teschler which is

FLI = (Mean Inspiratory Flow 25-75% - Mean Inspiratory Flow)/Mean Inspiratory Flow, where 25-75% represents the middle half of the inspiratory waveform. If the 25-75% is rounded, a value of ~0.3 would result. In really nasty flow limitation with inspiratory "cupping', a negative FLI could result.

I think the 0 to 1 is simply a grid (you have to start someplace).

That said, "Flow Limitation" actually appears as "FFL Index" in the raw data file, so perhaps the calculation has changed from the days of "Flattening" (although what the heck does the first "F" stand for?).

[LoQ]

(although what the heck does the first "F" stand for?).

Fuzzy Flow Limitation

[-SWS]

Also, SU brought up an excellent question about FL "severity".... I sure could be wrong, but I don't THINK there's a neat answer for that in sleep medicine currently.

On one hand we can talk about "severity of flow wave flatness/frequency". On the other hand, there's the "resulting pathology", which I think is more relevant when discussing severity. With FL (not A or H) there are usually no hypoxemic issues. So I'm thinking FL pathology severity is almost always going to be a matter of how disturbed a patient's sleep becomes thanks to those flow limitations. Anybody else out there with thoughts/guesses/answers to this?

Here's how I endeavored to describe it in robysue's recent thread:

That's a halfway point on a proprietary Resmed scale. That half-scale measurement point says nothing about how "arousal susceptible" you happen to be to half-scale, quarter-scale, eighth-scale, or even sixteenth-scale Flow Limitations given Resmed's proprietary measuring scheme. So the more crucial events---resultant sleep arousals---are unfortunately not in the S9 data set.

Depending on patient-unique threshold of arousal, your sleep architecture might severely deteriorate at frequent eighth-scale FL crossings, while my architecture might stay intact despite frequent quarter-scale or half-scale FL crossings. So threshold of arousal relative to FL severity can be unique from one patient to the next. UARS patients, as an example, are thought to be hyper-arousable to flow limitations.

viewtopic.php?f=1&t=57373&p=540485#p540295

So I'm thinking FL severity is more related to how often/infrequently a patient is disturbed by their FLs rather than simply how frequent or flat the FLs happen to be...

[-SWS]

Similarly, here are four comparison parameters Respironics strings together in algorithmically ascertaining whether flow limitation has occurred:


NotMuffy, does that guess sound plausible to you for Resmed's 0.0 to 1.0 index? Or no?

On Edit: I see that Resmed associates a 0 value with no flow limitation and a 1 value with flow limitation---perhaps a coefficient index inversion of the above toward describing likelihood of flow limitation occurrence---with or without further probability formulation(s) added.

As a side note, I think the 420e ran a correlation coefficient comparing sinusoid against patient flow. That 0 to 1 sinusoid correlation coefficient went toward one sub-component in an overall FL probability equation. They performed a similar coefficient comparison against a square wave. That coefficient went toward a second sub-component in an overall FL probability equation. Measuring peak flow's vertical line (also shown above) left-shifted versus centered yielded a third probability sub-component... and so forth.

The 420e algorithm strung a handful of FL probability sub-components toward that manufacturer's final FL probability index between 0.0 and 1.0. In other words, the 420e's final FL probability index was not a single correlation coefficient.

[idamtnboy]

But when I run my mouse cursor over those scribbles, it could say 1.00 whether I'm above the top line at the "beehive" or 3 lines below that top line... In fact, I don't see a spot that makes my cursor say 0.

When I say mouse over I'm referring to the graph line, not the symbols. With the window set to 8 or 10 hours what you'll see on the graph line are lines that drop straight down from the level line. That's because, in my case anyway, all of the FLs are only a few seconds long. If you expand the window to 10 minutes you'll see the graph line go up and down in steps. Put your cursor right on a level portion of the line and it'll show the value at that point. In fact, you can do that with all the graphs in Resscan to see what the numerical value is at any point on a graph line.

I haven't actually seen a 0 point on my graph, and you probably won't either. But that is the value range as defined in the BLP file which has the Flow Limitation data in it. The clinicians data information brochure gives the min and max values as flat to round, no numbers. I think that may be because the number is kind of arbitrary, and has no units like liters/sec or cm or # of breaths, etc., associated with it. It's kind of like saying that sure is a bright rainbow, or that's a faint rainbow. What is bright, and what is faint? How dark does a cloud need to be before it means dangerous weather? I don't think there is an answer. But we know a dark cloud can bring more of a storm than a white billowy cloud.

[idamtnboy]

On Edit: I see that Resmed associates a 0 value with no flow limitation and a 1 value with flow limitation

I see the exact opposite on the Flow Limitation graph. The graph line at the round symbol level has a value of 1.0. Graph lines moving toward the flat symbol end of the Y axis move toward 0. Resmed defines the round symbol to be no flow limitation.

[idamtnboy]

(although what the heck does the first "F" stand for?).

Fuzzy Flow Limitation

Hey SleepingUgly, that link LoQ gives is your answer! If you can understand what they're saying you automatically qualify for a Master's degree in Respiratory Technology from the University of Australia! I sure don't!

[-SWS]

On Edit: I see that Resmed associates a 0 value with no flow limitation and a 1 value with flow limitation

I see the exact opposite on the Flow Limitation graph. The graph line at the round symbol level has a value of 1.0. Graph lines moving toward the flat symbol end of the Y axis move toward 0. Resmed defines the round symbol to be no flow limitation.

Thanks for straightening that out, idamtnboy ---I typed it the exact opposite of what I meant to say. Unintentional reversals are my forte.

My point still stands: if Resmed compares several waveshape characteristics toward ascertaining FL, as the other manufacturers do, then that 0.0 to 1.0 index might be a multi-factorial index rather than uni-dimensional.

[idamtnboy]

then that 0.0 to 1.0 index might be a multi-factorial index rather than uni-dimensional.

The complexity of this issue as expounded by those of you who are more immersed in background knowledge directly related to the instant matter of discussion is approaching perilously close to creating a multi-FRACtorial cranial condition within me.

[-SWS]

Supercalifragilisticexpialidocious!

[NotMuffy]

So I'm thinking FL severity is more related to how often/infrequently a patient is disturbed by their FLs rather than simply how frequent or flat the FLs happen to be...

I agree. A RERA (Respiratory Effort-Related Arousal)(which is really what we're talking about here when we talk about FLs) is only a "RERA" if it has an "A".

This reminds me of the groundbreaking thread

Back To The Mountain(woman)

where the concept of anatomical fixed-flow restriction was raised as well as the use of IFL1 as a diagnostic tool (albeit inadvertently).

[NotMuffy]

Supercalifragilisticexpialidocious!

-SWS, you know I'm a hippopotomonstrosesquipedaliophobiac!!

Aaaaaaaaaaaarrrrrrrrrrrrrrrggggggggghhhhhhhhhhhhhhhhhh.........!!!!!!!!

[NotMuffy]

(although what the heck does the first "F" stand for?).

Fuzzy Flow Limitation

Son of a gun!

That's IT!

Never would've guessed "Fuzzy"!

Certainly a number of additions to Flattening, including more timely analysis (breath-by-breath, although in practice the value looks to be updated every 2 seconds) and the addition of ventilation anaysis.

1) as flattening severity increases (e.g., the shape index decreases) the FFL measure increases, 2) where flattening is occurring and ventilation is decreasing, the FFL measure is even more severe, and 3) where ventilation is high (e.g., recovery (big) breaths etc. are occurring) the response to flattening shape index is tempered.

Consequently, it appears the "flip" was done simply to keep the graph on the bottom:



Good find, LoQ!