ResMed vs. Respironics: AHI and Variable Breathing

Slinky wrote:Mr Sandman,

I'm rather curious as to why you chose an avatar of Pissing on the USA. Its more than a bit offensive. I've tried to avoid commenting on it - but it does get to me each time I see one of your posts.

Mr.Sandman, mth712, blowfish, Mike, anonymous, etc. got mad (actually I think his baseline is Angry at the Planet) and posted obnoxious pictures, deleted all his posts, stomped his feet and carried on and the like when he didn't get the attention he wanted (somehow he thinks he is "entitled" to free advice, and immediate response at that).

viewtopic.php?f=1&t=35941&st=0&sk=t&sd=a

BTW, as a Veteran of a Foreign War, the poster SAG is equally disturbed by Mike's show of disrespect for this country, and wouldn't give him a mirror even if his head (Mike's) was on fire.

Muffy

Oh! THAT is who he is!!! Thanks for the clarification. It figures. Ignorant is as ignorant does.

I have spent the last few days trying to get into my VB data, to try to “get a feel” for it.

Following Muffy’s lead (and particularly one of Muffy’s patent diagrams), I have taken to thinking of “Variable Breathing” not as an “event” in the sense we might think of an apnea, but rather as a “status”. My VB tends to occur in “runs” rather than in isolated “events”. The “runs” most typically are between about 3 and 6 minutes long, though runs as short as 1 minute and as long as 15 minutes have been recorded.

My runs of VB most commonly simply begin and end spontaneously. Sometimes, however, they begin following an A or an H event. Sometimes, an A or an H event (more commonly an A, though this is not unexpected, since my AI is typically larger than my HI) will “interrupt” a VB run for the duration of one 30-second data window. My data and my analytic tools are not sensitive enough to determine whether A and H events “associated with” VB runs occur at the same rate as A and H events “outside” VB.

Often, a run of VB will have terminated just 30 or 60 seconds prior to my turning off the blower and arising. Not every arising is linked to VB, and not every run of VB is followed by an arising. But the connection occurs often enough to make it appear as a noticeable pattern.

Over about a dozen nights, both my average and my median nightly percent-time-in-VB have both run very close to 30% (range 20% to 56%). With so large a portion of my blower-on time in VB, it is easy to “see” relationships that may not actually be there. Thus far, I have not found a convenient and statistically reliable way to quantify the “patterns” that I have observed. They may not be real.

Slinky --

I agree with your sentiments.

Did you catch that it is the President in that lovely avatar?

jdm2857 wrote:Slinky --

I agree with your sentiments.

Did you catch that it is the President in that lovely avatar?

Even if you don't like or approve of the President,
it's disrespectful.

[quote="cinco777"]Muse-Inc wrote

Velbor and other technically curious folks on this forum are investigating the Respironics Variable Breathing (VB) "mystery". The technical objectives of this investigation include trying to determine: 1) WHY Respironics detects, records, and reports VB events, 2) WHAT is the relationship between VB events and SDB, and 3) HOW are VB events used to monitor and/or modify our CPAP sleep therapy. Of course, VB event detection and reporting may be, as some have suggested, a failed hardware and software implementation that was less costly to leave "as is" than to remove.



cinco777,

For those of us that are trying to follow ... and understand this thread ...

Would you please explain just how the VB chart is actually read (item #3). Plus how it's used.

As I look at it, I haven't got a clue as to what it's telling me ... or how to use it

Thanks,

Jan

Here's a somewhat relevant set of comments I made in another current thread.

While discussing their proprietary statistical measure called Variable Breathing, Respironics wrote: The Auto-CPAP... relies on the ability to trend the steady rhythmic breath patterns associated with certain stages of sleep. When a patient is awake, in REM sleep, or in distress, breathing breathing tends to be more erratic, and the Auto-CPAP trending becomes unstable. It is, therefore, important to interrupt the Auto-CPAP if the patient's breathing becomes too variable...."

So above, Respironics tells us that Variable Breathing covers a variety of conditions, including wakefulness. Periodic Breathing may be included in that broad VB statistical metric.

While discussing the clinical event known as Periodic Breathing, Respironics wrote: Periodic breathing is defined as alternating periods of hyperventilation with waxing/waning tidal volume and periods of central hypopneas or apneas. There are many forms of periodic breathing, one of which is Cheyne-Stokes Respiration (CSR). CSR is characterized by a cyclic pattern of waxing and waning during periods of apnea, and deep, rapid breathing.

Periodic Breathing is not that somewhat ordinary wakefulness variability measured by VB. Similarly, Periodic Breathing is not rather ordinary REM associated variability also measured by the proprietary VB statistic.

But the new Respironics algorithm does: 1) employ Forced Oscillation Technique (FOT) to differentiate an open-airway or central apnea, and 2) track and score the highly patterned waxing and waning of Periodic Breathing, to notify clinicians that yet other SDB issues besides ordinary OSA require attention.

What I am very surprised to read is that Respironics claims to be able to somehow differentiate and score Respiratory Effort Related Arousals (RERAS). They can somehow discern brain arousals from a flow signal?

So according to Respironics a variable patient flow pattern is now sufficient to score EEG-associated RERAS???? http://i741.photobucket.com/albums/xx58 ... ure7-2.png

If so (and I can't help but doubt it), that has to be a refinement to their former VB signal processing and differentiation techniques. Wow. Differentiating brain-based RERAS solely from variability in a flow signal? Really?



Regarding Muffy's earlier comment/question about whether Respironics may have exclusionary A & H scoring criteria while the VB controller is at the helm: I suppose it's possible. But I can't find any text describing that. On the contrary, Respironics tries hard to service higher-priority apneas and higher priority hypopneas while the lower-priority VB layer is in control. That requires first detecting those higher-priority A & H events before abandoning the lower-priority VB control layer. ( on edit: if Respironics "normalizes" data toward scoring fewer hypopneas, then I think that can be thought of as exclusionary criteria"---ideally exclusionary of hypopnea false-positives.)


How does Respironics score those A & H events during some rather wildly-varying patient flow circumstances (when a somewhat narrow-window recent flow baseline is really required for that comparison/scoring of A & H)? They algorithmically "normalize" those wild fluctuations:

Respironics Patent Description wrote: The present inventors appreciated that using the standard deviation alone as a measure of the degree of variation in the weighted peak flow data may not produce consistently correct results. This is so, because the standard deviation of the weighted peak flow data when the mean patient flow is relatively low is not exactly comparable to the same standard deviation for a higher mean patient flow. The present invention, therefore, seeks to normalize the standard deviation to the mean patient flow, and then takes the mean flow into consideration when analyzing the variation in the data.

FIG. 11 is a chart illustrating a normalization curve 290 that describes the relationship between the mean patient flow and an adjusted mean patient flow. It can be appreciated from reviewing this figure that there is a linear region 292 in which the adjusted mean flow (vertical axis) has a one-to-one match with the actual mean flow (horizontal axis). If the patient's mean flow for the 4 minute window is within region 292 , no adjustment to this mean flow is made. There is also a first region 294 having a ½ to one relationship between the adjusted mean flow and the actual mean flow. Thus, if the actual mean flow falls within region 294 , which is between 15 and 25 liters per minute (1 pm), then an adjusted mean flow is calculated based on curve 290 . There is also a flat region 296 where the adjusted mean flow is clamped to a baseline value even if the actual mean flow is decreased. Thus, if the actual mean flow is less than 15 lpm, the adjusted mean flow is clamped at 20 lpm.

Above Respironics compensates a highly-variable flow sample by raising or normalizing the current VB flow-comparison value toward a higher mean (during certain skewing circumstances). If Respironics did not clamp some of their low outlying sampled mean flow higher, in an attempt to normalize flow data, then they would score more hypoponeas than they currently are. However, that scoring assumption rests on instantaneous flow being similarly normalized with four-minute mean flow. That scenario would be exclusionary.

I agree that highly variable breathing is a likely place to account for at least some of the differences between Resmed's proprietary scoring of hypopneas and Respironics' proprietary scoring of the same. When breathing becomes highly variable, then establishing a flow baseline for comparison/scoring of SDB events can become a dicey proposition. Resmed and Respironics are guaranteed to employ different algorithmic flow-baselining methods when breathing becomes highly variable.

I agree that the differing algorithmic methods of trying to cope with that highly variable breathing---while trying to establish a baseline for scoring/comparison---is a likely place to expect different hypopnea-scoring outcomes.

SWS, thanks for providing very pertinent extracts from Respironics patents that describe Variable Breathing, and for your thoughtful and insightful explanations and comments on what Respironics says, and may mean, about their implementation of VB in their current machines, and Periodic Breathing (PB) in their recently announced machine. Your postings and those of Muffy have been extremely helpful in my attempts to better understand VB and pursue my own VB investigation. The two of you, and a few others, provide the Graduate Level "courses" that have become so helpful to us as we continue climbing the steep learning curve of Sleep Apnea Therapy

Muffy wrote:

Slinky wrote:Mr Sandman,

I'm rather curious as to why you chose an avatar of Pissing on the USA. Its more than a bit offensive. I've tried to avoid commenting on it - but it does get to me each time I see one of your posts.

Mr.Sandman, mth712, blowfish, Mike, anonymous, etc. got mad (actually I think his baseline is Angry at the Planet) and posted obnoxious pictures, deleted all his posts, stomped his feet and carried on and the like when he didn't get the attention he wanted (somehow he thinks he is "entitled" to free advice, and immediate response at that).

viewtopic.php?f=1&t=35941&st=0&sk=t&sd=a

BTW, as a Veteran of a Foreign War, the poster SAG is equally disturbed by Mike's show of disrespect for this country, and wouldn't give him a mirror even if his head (Mike's) was on fire.

Muffy

You know this is old news and I think in the end I was the bigger man for TOTALLY coming clean on this here viewtopic.php?f=1&t=38524&p=342435&hili ... 12#p342446 . Isn't it time you let it go? I don't like your online persona and that is all. I do not really know you and you do not really know me.

The avatar is a symbol of what I feel Obama is doing to this country and it is not meant as disrespect from me. As a Marine and Desert Storm Vet. I have earned the freedom of speech for myself and for you to continue to attack me.

Get over it already... To Slinky - we have had PM's and discussions so why do you now act as if you didn't know who I am?

I haven't posted here in many months. I come back to a subject that interests me and then here you come.

Who really has the problem here?

Mike

I will change my avatar out of respect for people offended by it - that is all.

cinco777 wrote:SWS, thanks for providing very pertinent extracts from Respironics patents that describe Variable Breathing, and for your thoughtful and insightful explanations and comments on what Respironics says, and may mean, about their implementation of VB in their current machines, and Periodic Breathing (PB) in their recently announced machine. Your postings and those of Muffy have been extremely helpful in my attempts to better understand VB and pursue my own VB investigation. The two of you, and a few others, provide the Graduate Level "courses" that have become so helpful to us as we continue climbing the steep learning curve of Sleep Apnea Therapy

Thanks for that comment, cinco777.

But I think of it more as collective head-scratching... At least on my part.

I have personally found Velbor's and Muffy's comments in this thread particularly helpful. So thanks, Velbor, for launching this thread.



P.S. More RERA discussion regarding that new PR System One model/algorithm in Rooster's thread here:
viewtopic/t44463/PR-System-One-REMstar- ... chine.html

MrSandman wrote:The avatar is a symbol of what I feel Obama is doing to this country ....

Who really has the problem here?

There is no doubt that you have the problem here!

Respironics gives an example of Variable Breathing in its patent description, and highlighted is an area where I believe ResMed could score a hypopnea:



In all fairness, neither machine would score a hypopnea because this area is only about 6 seconds long (boy, I'm glad somebody proof-read that histogram!). This patient looks to be breathing at a rate of 15, which could be a little bit high for sleep. But breathing at a rate of say 10, you miss one breath and BANG! there's your event, and certainly 2 or more breaths like that would tip the scales.

Now again, I don't think Muffpothesis applies to apneas based on the anecdotal reports of the Resmed/Respironics comparisons, but on the other hand, why couldn't it?

On the other, other hand (back to the original hand), why couldn't there be a similar filter in A10 to suspend event scoring/machine response in order to address artifact (which, out of necessity, would really have to be there)?

Indeed, there is a screen to take out rogue breaths, based on pre-determined upper and lower limits:



as noted in ResMed US Patent 7013893, which states

As discussed above, in step 200 of FIG. 7, the curve F is checked to insure that it corresponds to a valid inspiration curve. The flow curve F is checked against an upper and lower bound to prevent processing of an inspiratory curve corrupted by a cough, sigh, etc. For example, as shown in FIG. 12, the curve F may be rejected if it exceeds at any time an upper limit curve UL or falls below a lower limit curve LL. UL may be selected at about 150% of the mean inspiratory flow and LL may be selected at about 50% of the mean inspiratory flow.

Muffy

I would offer tot his conversation my data from last night, in that it strikes me as containing some elements which might have bearing on the Variable Breathing discussion.

Over a blower-on total of 9h:24m, my AHI was 2.8, comprised of AI (OA) of 2.1 (20 apnea events) and an HI of 0.6 (6 events). No flow limitations were reported. Variable breathing was present for 34.9% of the therapy time.



During this night, as for all nights I have examined, during the 30-second data-reporting time blocks, there is NEVER a concurrent report of apnea, hypopnea, flow limitation or variable breathing; these “events” or “states” appear to be mutually exclusive in terms of data reporting. (Vibratory Snore does NOT fall into this category; it can be reported concurrently with A, H, or VB).

I would particularly call attention to the first block of VB (top line on the graph), running from 0:11:00 through 0:57:00, an atypically long duration of over 45 minutes. During this “run” of VB, there were only five 30-second time blocks interrupting the otherwise unbroken string of VB entries: four Hypopneas at 0:21:30, 0:31:00, 0:40:30 and 0:53:30, and one Apnea at 0:45:30. I find these interesting in that each of these five respiratory events is distinct. Also, fully 2/3 of the hypopneas recorded for the night are concentrated - and reported - in this time period of otherwise continuous VB.

I would also point out that during this VB "run" period, and during most other periods characterized by VB, there appears to be a distinct LACK of algorithmic pressure change.

In summary, this data suggests that 1) respiratory events, including hypopneas, do appear to be reported while the VB “state” is “on” (which does not prove that there were not additional A or H events which met the definitions but which were not reported as such), and 2) algorithmic responses to a variety of other events, and pressure changes which appear to otherwise occur even in the absence of reported events, appear suppressed during periods of VB.

Velbor

I think that graph is fabulous, Velbor!

In order to try to determine the clinical correlation of the data, do you have the histograms of your sleep architecture from any sleep studies?

Also, how would you rate the quality of your sleep, including periods of wake, relative to events on the graph?

Finally, if

Velbor wrote:During this night, as for all nights I have examined, during the 30-second data-reporting time blocks, there is NEVER a concurrent report of apnea, hypopnea, flow limitation or variable breathing; these “events” or “states” appear to be mutually exclusive in terms of data reporting.

although you say

Velbor wrote:In summary, this data suggests that 1) respiratory events, including hypopneas, do appear to be reported while the VB “state” is “on”

if VB and A/H/FL never appear concurrently, would that not suggest that A/H/FL are NOT reported when VB Mode is active?

Muffy