Question on apap

[StillAnotherGuest]

Good find there o., those results seem to be pretty similar to Rappaport et al re: centrals.

Is O'Nare an Irish respiratory event index?

Speaking of Rappaport (the reference that was originally used to establish the sensitivity)(BTW, I'm Not Rappaport)(I wonder if you're not from NYC, you would understand that)(did I tell you I have a little ADHD? I'll be going on Oh look there's a butterfly Here Butterfly Can you come play....)

...anyway...

We find that:



A better question might be, will you end up in a group that 420E is identifying all of the events correctly, only some of them, or maybe even none of them?

Looking at the total number of patients in that study:

10.9% of the time, no cardiac oscillations were seen in any of the central apneas (missing the boat entirely)
52.2% of the time, identification was mixed (somewhat helpful)
37.0% of the time, cardiac oscillations were seen in all of the central apneas (completely accurate)

But some of the patients had statistically insignificant numbers of central apneas. I mean, if you only had one or two centrals, and 420E picked them up, I don't think that you can go ahead and say, "Well, 420E has 100% specificity for me."

If you only include those patients who had >5 central apneas (which still ain't a lot), then:

7.4%% of the time, no cardiac oscillations were seen in any of the central apneas
70.4% of the time, identification was mixed
22.2% of the time, cardiac oscillations were seen in all of the central apneas

Taking a simple average of the % of identification in the mixed group of the all-patients total showed a 49.4% ability to correctly identify centrals.

So I think what you have to get from all this is that most of the time, central apnea event identification will be mixed, and in that case, 420E will only pick up about half of the central apneas that are actually occuring. So great, it doesn't respond when it sees a cardiac oscillation-oriented central, but if you're in a less than completely accurate group, then it's reacting to centrals about 50% of the time. And if it's reacting, and you're in a group with a residual central apnea index of >5 (which would probably put you in a CA problem group, like CompSAS), then there's a good chance that that pressure increase will in turn create more centrals.

Course, as -SWS once suggested, you could always run "Command on Apnea", say at 10 cmH2O.

Or just buy a ResMed machine.

SAG

[StillAnotherGuest]

Mouth taped, as usual.

YOU HAD YOUR MOUTH TAPED?!

[ozij]

Mouth taped, as usual.

YOU HAD YOUR MOUTH TAPED?!

And she changed her name too!

No, I'm not from NYC....

And when you add this table:

http://www.chestjournal.org/cgi/content ... 6/3/660/T1


To the cental apneas per patient table, then you can start talking about baaad statistics.

That table (quoted by PB...) treats all apneas as though they were independent. The histogram by patient shows how ridiculous that is. Those apneas are not equivalent to subjects.

Dr. Rapoport and Mr. Norman have a financial interest in the development of automatic CPAP in the form of patent rights (assigned to New York University and licensed to Mallinckrodt Nellcor Puritan Bennett)

So they skip the inter-case variablilty in both the abstract and the conclusions and discuss all apneas as though they were the same.

Takes a sleep pro, (who is not Rapoport...) who analyses patient profiles to remind us of where the important data is hiding.

Even taped profiles. Full face too....

O.

[rested gal]

Mouth taped, as usual.

YOU HAD YOUR MOUTH TAPED?!

Yeah, big guy!!!

AND... I set the pressure myself!!!!!

So, there!

By the way, I meant to tell you that you have the best sleep techs. And/or you warned them well ahead of time to let the crazy lady from down South do whatever.

First night, when I mentioned to the tech to not be alarmed if she noticed tape over my mouth, she tactfully said, "Have you considered a chin strap?"

Explained, "Been there, done that, doesn't work for me. Been taping for two years. I'll be fine. No problem."

She smiled and went on.

Word got around, I guess. The other two never blinked an eye.

[StillAnotherGuest]

AND... I set the pressure myself!!!!!

YOU SET THE PRESSURE YOURSELF?!

By the way, I meant to tell you that you have the best sleep techs.

Thank you, they truly are.

And/or you warned them well ahead of time to let the crazy lady from down South do whatever.

Yeah, pretty much.

Actually, I said that the CLFDS would probably have a lot more to teach them than the reverse.

I've kept forgetting (for two years now!) to ask about something:

If the 420E is 100% accurate that when it says it saw a central, it really was a central, how come there was this...

Because NOT, NOT, NOT!

OK, that's another story.

Let's browse. Again, the 420E output was not being used, there was a separate transducer, so admittedly, I don't know what happened. But what the hey, the Cardiac Oscillation Police aren't around, sooooo...

First let's try to figure out what events we're talking about. Superimposing the 420E download on the NPSG, it seems pretty clear where they are (the light green respiratory events on NPSG histogram are hypopneas):



Now, the CLFDS's sleep is marked by a heavy concentration of PLMs in the first half of the night. When she sleeps through them (maybe a little autonomic stuff there), they look like this:



But in the area in question looks like this:



Now, the scoring rules say that PLMs should not be scored if there's a respiratory event there (but, c'mon, them are PLMs). So the question is, are those respiratory events, and if so, what type. Obstructive or central, apnea or hypopnea, caused by PLMs or "causing" PLMs (BTW, respiratory events don't cause PLMs).

There is considerable flow reduction, but what apnea and hypopnea definitions do you want to use? Definitions are relative to baseline, what baseline do you want to use?

The abdominal channel shows some increasing effort as the event progresses, so they would seem to have some obstructive component. The reduction in flow is very significant. The new rule has apneas at 90% flow reduction, the old rules had it at 80%. Visually scoring, and relative to the immediate surrounding breaths, these events could easily go either way, but physically measuring them with software shows about a 75% reduction, and relative to the somewhat more stable baseline above they're a little more like hypopnea. The effort channels, particularly the abdominal channel, don't show a huge drop, so I'm going with "obstructive hypopnea."

Continuing the endless discussion points, was the 420E right? More to follow.

Can we debate these events, obstructive or central, apnea or hypopnea? Sure, till the cows come home.

I think a better understanding would be gained by considering the following concept: Sleep instability begets respiratory instability. I would offer to explain these differences (PLMs with and without associated respiratory disturbances) by the PLMs causing a hypereactive respiratory response, giving the event a little central flavor, and the unstable sleep state permitting some airway instability to occur.

But before I went out and dropped 4 grand on an ASV machine I'd be thinking about fixing them PLMs.

SAG

[rested gal]

But before I went out and dropped 4 grand on an ASV machine I'd be thinking about fixing them PLMs.

SAG

Wow, what a thorough analysis.

The first thought never crossed my mind (and still doesn't... ) 'cause the second thought will buy a lotta' Mirapex.

[StillAnotherGuest]

Hey o., get a load of this.

That 30 minutes from 86.5 to 87.0 shows 420E listing 19 events-- 2 apneas, 8 apneaCAs and 9 hypopneas:



The actual area shows 20 distinct events:



Yeah, I know, kink in the tubing and the 420E signal could have been affected somewhat, but overall event identification looks pretty close so perhaps "concurrent ticks" does not mean "double-counting events".

SAG

[-SWS]

Howdy folks. A few thoughts or comments about some of the 420e concepts raised in this thread:

Event Tick Marks Silver Lining produces a field-clinician's data set--not a scientific data set. The hash marks in particular bear the true purpose or nature of this data set. Look at the amount of empty space available for hash marks in any given half-hour segment. Now calculate the maximum number of hash marks that can be fit in that half-hour segment. Now think about how a moderate-to-high AHI could ever be fit in the allotted amount of white space. Do you see how those hash marks are only intended to give the clinician an imprecise albeit high-level view of 1) general efficacy, and 2) the general temporal and ratio-based nature of event distribution? Not even close to a scientific data set--but extremely helpful IMO.

CA Associated Events I don't think unintentional double-counting occurs. However, the tabular portion of the 420e data set does endeavor to report these two event types associated with cardiac-oscillations: 1) frank central apneas, and 2) mixed apneas. How will a detected mixed apnea be represented via the hash marks? Concurrent vertical ticks. Again, any given white-space epoch is capable of being "outstripped" by large numbers of discrete events in real time.

420e Central Apnea Specificity Rating of 100% Ozi, that 100% rating was not made up by us as we went along. Rather, it was published in an older version of SL as a "carry-over" specification from the 418P, which utilizes the same sensor as the 420e. To question the accuracy of that specificity rating is right-on-the-money healthy discussion. There really is no such thing as a 100% specificity or 100% sensitivity instrument in the real world IMO. However, the specificity of cardiac oscillation detection in general should be fairly high (much higher than SAG hints). The signal lends fairly high specificity by virtue of the highly cyclic and "shape-unique" nature of the cardiac-oscillations (that same geometric signal simply tends not to occur in the immediate external environment). The snore signal, by contrast, possesses neither of those signal processing advantages--and undoubtedly entails a much lower specificity of detection.

False Detection of Central Apneas When false detection of a central apnea occurs with the 420e, then cardiac-based acoustic reflection is the likely candidate in my opinion. Laura and SAG are both interested in why Laura tested false-positive for central apneas the night an argyle sensor tube was substituted. The argyle tube is perfectly fine for conduction and signal processing of a fluid gas moving at the slow speed of a beverage being drawn up a straw. That is close to the slow speed at which respiration moves air up and down the 420e signal line. However, cardiac oscillations travel up and down the sensor line at the speed of sound (see camels & monkeys analogy below). Here basic considerations of flow impedance give way to transmission line theory. In physics, transmission line theory considers a wave guide or tube's "characteristic acoustic impedance". That acoustic impedance is derived by, among other acoustic factors, the tube's interior surface reflectivity. The pressure transducer's acoustic impedance must match the acoustic impedance of the signal tube.

If an argyle tube is selected that does not match the acoustic impedance of the signal transducer, then there will be subtle or greater energy wave reflection up and down that tube (regarding speed-of-sound signals only). This can result in two signal-processing skewing caveats for the 420e: 1) the resultant oscillation signal will be amplitude reduced (making the signal harder to algorithmically normalize and baseline), and 2) there will be a residual echo signal that the 420e can conceivably detect during a closed-airway obstructive event. End result: false positives for central apneas. Again, transmission-line acoustic impedances are a separate issue in physics than low-velocity gas-flow impedance.

Camels & Monkeys in Real Time When we see a cardiac oscillation graphically multiplexed/juxtaposed on top of a flow signal, many people tend not to realize their relative speeds. We see a snap shot of that "cardiac oscillation hump" sitting on top of the larger inspiratory flow-signal waveform. It looks as if those two signals or occurrences traveled up the human airway at the same rate--because we see a snapshot. In reality the respiratory wave moves up the airway rather slowly, and the cardiac oscillations "ride the back" of that respiratory wave at the speed of sound.

To give you an idea of how these two events travel in real time, imagine an endless caravan of camels, slowly meandering along the dessert. Each camel in that succession represents one respiratory cycle. Now imagine an endless succession of monkeys skipping along the backs of the camels--at the speed of sound. Those speed-of-sound-traveling monkeys are the cardiac oscillation events/signals. The monkeys travel along the tops of the camels at a drastically different rate of speed than their host carriers (as do respiratory-flow and cardiac-oscillation waveforms travel juxtaposed together at drastically different rates of speed).

Regarding flow signals multiplexed with cardiac oscillation signals along the same tube (or "transmission signal line"): the vast discrepancies in the rate of travel of these two energy waveforms make for entirely different signal processing and tubing compatibility considerations. And yet those two sleep-related signals--among many others--are multiplexed on but one waveform.

I would like to comment next on loonlvr's saga...

[ozij]

Hi there, -SWS!

Howdy folks. A few thoughts or comments about some of the 420e concepts raised in this thread:

Event Tick Marks Silver Lining produces a field-clinician's data set--not a scientific data set.

Agreed.

The hash marks in particular bear the true purpose or nature of this data set. Look at the amount of empty space available for hash marks in any given half-hour segment. Now calculate the maximum number of hash marks that can be fit in that half-hour segment. Now think about how a moderate-to-high AHI could ever be fit in the allotted amount of white space. Do you see how those hash marks are only intended to give the clinician an imprecise albeit high-level view of 1) general efficacy, and 2) the general temporal and ratio-based nature of event distribution? Not even close to a scientific data set--but extremely helpful IMO.

I was thinking of that too - and when they changed from a 30 second sampling (418P) to a 60 second sample (420E)- it became an even rougher guide. But again, I agree - extremely helpful.

CA Associated Events I don't think unintentional double-counting occurs. However, the tabular portion of the 420e data set does endeavor to report these two event types associated with cardiac-oscillations: 1) frank central apneas, and 2) mixed apneas. How will a detected mixed apnea be represented via the hash marks? Concurrent vertical ticks. Again, any given white-space epoch is capable of being "outstripped" by large numbers of discrete events in real time.

On the 418P's details screen they had separate hash marks for mixed apneas. On the 420E the they no longer do. When it comes to reporting those mixed apnea on the numerical data lists, they count them once among the apneas. And I haven't figured out yet when they drop them from the centerals and when the don't.

And then, on the export, those concurrent tics, which had been added to the to the apnea score are dropped, gone, disppear without trace..

Another problem is "fat tics". A "fat tic" in apneas and hyponeas is counted twice - or more. I've got an example of 4 tics in hypopnes, two regular, one very fat and one fat that together make 9 (nine) hypopneas for that session. Or a concurrent (thin) tic in apneas and a fat one (which to my eye looks double at most), which comes to 4 apnea numerically.

Which is why I don't count tics. And hich is why I think that if we had the pat file for that famous Argyle night, we would see SL3 counting far more events than 19.

Thanks for that monkeys and donkeys explanation - it makes things very clear.

O.

_________________

CPAPopedia Keywords Contained In This Post (Click For Definition): 420E, AHI

[StillAnotherGuest]

...which is why I think that if we had the pat file for that famous Argyle night, we would see SL3 counting far more events than 19.

Hey o., that is the Argyle night file there. The only thing I didn't do was run the actual 420E flow signal into the NPSG. So again, prefacing these comments with "I don't know what the 420E saw in the RG waveform", I am frankly beginning to think it was right, at least with what limited information there is available from this data collection and the Evolution patent '133. In that patent, it is noted:

If on the other hand the breath amplitude is determined to be less than the first limit at 52, it is again examined at step 60 to determine whether a second amplitude limit such as 10% is exceeded. If so, hypopnea is indicated 62 and the waveform is again checked at 64 to distinguish between obstructive and central hypopnea wherein a flow-limited shape is indicative of obstructive hypopnea. If the determination is made at step 60 that the breath amplitude is below the second amplitude limit, the thermistor amplitude is compared to a preselected limit such as 10% of the average amplitude sensed in a preselected time period such as for example in the previous 5 minutes. If such limit is exceeded, an unspecified type of hypopnea is indicated 72. If the thermistor amplitude fails to achieve the preselected limit, apnea is indicated 74 and the pressure waveform is examined for cardiac oscillation at step 76, i.e. regular, small-amplitude flow pulsations with a frequency in the range of the cardiac frequency. These pulsations can be detected from the flow signal after it is appropriately filtered and transformed to magnify their amplitude. The signal transformation function (which preferentially magnifies the amplitude of the signal near its average value) may include, but not be limited to, nonlinear mathematical functions (e.g. square root) and look-up tables. These periodic fluctuations are then detected in the transformed signal with valiance and/or period amplitude techniques which identify fluctuations at a frequency similar to that of a cardiac cycle (e.g. 40-120/min). If such oscillations are not present, obstructive apnea is indicated 78, if on the other hand such oscillations are identified, central apnea is indicated 80.

What this suggests to me is that if 420E sees an event, and that event has respiratory frequency-based waveform activity that is <10% of the previous 2 minute analysis window, and the event is at least 10 seconds long plus five-eighths of the average of the preceding three breaths, and then it finds cardiac frequency (40-120 cycles/minute) waveform activity, then it will call that an ApneaCA. Consequently, the speed of waveform transmission becomes academic.

So if somebody were able to produce a waveform with 40-120 cycle/minute activity during a central (defined as >90% flow reduction from baseline) event...

SAG

[ozij]

SAG, I mean the *.pat file, SilverLining's data file.

If you have it, what are the indices for that night ?

Because RG was showing the *.jpg of the "detailed record" screen - and that doesn't show us how SilveLining counts those tics.

So you're saying that according to the patent the machines looks air flow oscillations (same old camels but dancing or limping) and not sound (monkeys)?

O.

[StillAnotherGuest]

SAG, I mean the *.pat file, SilverLining's data file.

If you have it, what are the indices for that night ?

OK, you mean like a fat tick represents multiple events. Let me dig that up...

SAG

[StillAnotherGuest]

Hey, look at that!! By my calculations, the ApneaCA aren't really counted, and hypopneas can be ApneaCAs!

Hmmm, gots to think about that...

SAG

[StillAnotherGuest]

OK, in that specific 30 minute time frame, assigning 1, 2 and 3 events based on the width of the tick ("fat tick"-- that's funny), I get 3 Apneas and 17 hypopneas, 12 of which have associated cardiac oscillations. That would seem to explain the PSG results.
SAG

[Guest]

OK, in that specific 30 minute time frame, assigning 1, 2 and 3 events based on the width of the tick ("fat tick"-- that's funny), I get 3 Apneas and 17 hypopneas, 12 of which have associated cardiac oscillations. That would seem to explain the PSG results.
SAG

WAIT! THAT'S NOT MY FINAL ANSWER!!

The BIG CA one is a 4, so that's 15 CAs.