Call for Excel VBA help: to support effort to clarify/understand FL

[2SleepBetta]

For some reason the Reply with Quote would not work. Black font shows my, 2SleepBetta's, earlier post.

Sheepless asked questions of me, 2SleepBetta, in blue-colored font and I reply in green-colored font. 


Main Conclusion - A FL flag may only indicate that threatening wave shapes have been detected and that pressure   will  be  raised or eased  a  bit  stepwise or a lot and fast . Tidal Volume may not be reduced. (New to me, if true as samples and curve inspections indicate.) 

 Facts (as I understand ) - "FL" flags reflect detection of one to four things: (1)  troublesome inspiration wave-tip shapes (such as "M" tips and "Chairs" among the 7 to 47 clasified tip shapes), (2)  wave-tip flatness, (3) respiratory rate change and, (4) reductions in Tidal Volume. (Facts re FL definition)  

it seems your hypothesis, or initial assumption, is that fl is either a function of or should at least vary with the degree of fl and that you have disproven(?) this. that tv should vary with flow limitations seems intuitively correct [yes, agreed by definitions IMO] to me and I'm surprised restrictions in flow from any or all sources - including (1) - (3) above - aren't necessarily reflected in tv. do you have a sense of why that is? maybe something to do with how tv is calculated (like over a longer period than a flow limitation)?

My point is that careful use of the term "flow limitation", in our context, is understood to mean there is at least a brief reduction of the Tidal Volume (TV) delivered in one or more inspiratory waves. But my analyzed sleep data and graphs show instances where Resmed FL flags are not always accompanied by TV-drop. One has to conclude such FL flagging arises solely from one of, or some combinations of, items 1-3 above as Resmed patents disclose. To me: surprising. (Separately, of course there is the [often hidden] baseline flow limitation of real UARS which may or may not have or show significant superimposed flow limitations.) 

Central pursuit of thread and post: - This thread has dealt only with measuring reductions of actual Tidal Volumes: how those  volume drops are associated with and differ from flow limitations, whether or not the limitations evoke a FL flag. (Fact)

and your conclusion is that volume drops may or may not be associated with flow limitations? Not quite. Vice versa with respect to FL flags, some (how few?) do not have an associated TV drop. IMO and, again, apart from serious UARS with its mostly continuous, steady and latent flow limitation: it's by definition our other more common  actual flow limitations (partial restrictions we may care about)  are reflected in commensurate TV reductions, serious or not. have you a sense of what factors or circumstances influence whether volume drops are associated or not associated with flow limitations? have you determined "whether or not the limitations [do you mean flow limitations?] evoke a FL flag" and if they don't, why not? It is my understanding that Resmed Autosets and above use their on board tables  (in memory) for wave-shape pattern recognition as well as pressure-sensed determinations of fluctuating airflow (indicating TV irregularity) to mark the time, duration and perceived severity of flow limitation sensed. But I've seen that TV reduction is not a necessary condition for flagging a FL according to my samples and work to this point.
 
Content:  - Attachment 1: Its bottom two graphs present a  data based, different and more revealing method to present  and  see  ventilation  (TV) reductions from all flow limitations familiar to us at AB. The bottom graph integrates TV and its drops with the FR graph. The upper (red colored) TV projections that rise upward show TV > 0.5L, The lower graph ignores TV > 0.5L .  Both graphs show nothing, zero TV drop, when TV = 0.5 L, the assumed baseline level. (Interpretive matter)

Impracticality and need for context: The TV-drop  method demonstrated (as covered  above) is labor and time intensive, although conceptually simple:  not practical to perform day by day (with ordinary Excel and skills) without an appropriate data analyzing tool and an OSCAR-like presenter of the FR, FL,  TV, TV-Drop  (and, possibly I, E and Pressure) curves..

Validity of novel (?) TV method:  I have no doubt about validity of the method demonstrated for revealing actual TV and its reductions in detail. The detail product lends itself to simple summary with, for example, valid percentile scores.

I'm not clear where the practical utility of the novel tv method lies if tv doesn't necessarily vary with flow limitations. would this "more revealing method" replace or supplement the current scoring/ranking of flow limitations? (my gut feeling is that wouldn't because it doesn't take the other variables (1) - (3) into account?)

IMO, of the four amalgamating factors the Resmeds are known to consider it is only TV-reduction that is significant for us human users outside the real time machine control loop. I do believe those devices almost always respond appropriately to TV drop in their own patented way, but we have members who have  criticized the Resmed and its algorithms as being unduly aggressive in raising pressure. It would take a lot of work to see how frequently Resmed pressure increases were false alarms caused only by wave shape or RR irregularity rather than TV reduction, if that were even possible to check. (Made me think of old jokes about testing flash bulbs to see if they work or putting spilled toothpaste back into the tube or putting .... back into the horse.)

But, as I've acknowledged, I think, the "novel TV-drop method" outlined is absolutely valid and quite accurate, but it would only be practical/feasible if enabled for regular use by an easy data prep tool to make a file OSCAR could import and present. My  preliminary checks suggest that TV detail and TV-drop detail could be scored the same (e.g.,using percentiles) as any other uniform kind of data--but be done far more meaningfully than the RM FL mixed bags. Nevertheless, my checks from some one-night are suggestive not at all conclusive. It  may be rare that FL flags are not reflected in TV drop.

**************

(Rant:

 I'll spare more about scrolling-daunting, huge Excel files (am catching on to using range names with "go to's"), will just mention these whining details. It's all a fight after my Excel 2010 in Office Pro crashed and had to be replaced at my start here by 2016. It has a whole new/different (idiotic, IMO) scatter chart formatting approach, one that displays options that sometimes cannot be invoked on the spot but must be found, as the MS text I have says (somehow) somewhere else among the now proliferated, shape-shifting, obscure and scattered format menus and must-click places. 

For a different example I wanted to change that crude avatar draft, make its features better and change colors. I found, last night (first time ever) , that certain chart-superimposed text boxes would not copy (using up to date Windows 10) with the rest of the avatar image they clearly overlaid inside Excel. Never ending! It's the usual more macro and cosmetic things in many books or atomized illustrations with little if any context and the now often mute index. In-package Help  is not nearly as good as it once was so you need to Google it up from somewhere or just keep bloodying your head against it. Bah! The unbroken was fixed to make it prettier (?), add a feature or formula or two or five so the earlier versions move toward stranding as legacy versions.)

[2SleepBetta]

I should have posted long ago an illustration like "B" at right in the attachment. It's the curve with yellow-tan fill under it. To its right there is an explanation how that curve illustrates conversion of flow rate data to Tidal Volumes.

It seemed strange for a moment when I saw that the inspiratory part (actuallly all) of a FR curve would be identical--it has to be--in form to the TV curve, but TV only involves the area under the curve but above the axis, which has been determined as shown in this project. 

Additionally, there is an excerpt from the first paper "A" I've seen that offers us a lot of help with understanding and recognizing arousals, the RERA we see. It is based on a PSG controlled assessment of a way to  determine arousals with pressure and snore sensors, not just the EEG. Here is a link to the paper where all is visible: Flow Limitation/Obstruction With Recovery Breath (FLOW) Event For Improved Scoring of Mild Obstructive Sleep Apnea without Electroencephalography (nih.gov)

   

[sheepless]

thanks 2sb. I think I have at least the gist of things now. while the details are important to those immersed in the subject, I just get lost in them, so a simplified & abbreviated view is helpful & appreciated.

[2SleepBetta]

The attachment presents nothing new and, consistent with my most recent posts, it shows that deformations of inspiratory wave tips can give rise to unexpectedly long and moderately  "higher severity" level flow limit flags, FL. 

The second image of post 10 (the highlighted area), and some comment there, and the post 17 image are context for recent images and the image below. In #10 I noticed there seemed to be a disparity between high FL level but little corresponding drop in tidal volume.

Upshot as I see it now: 

1. If FL seem high without obvious and corresponding drops in the FR curve then take a good look (2-minute or shorter zoom) at the inspiratory peaks. They are very likely a large factor contributing to or causing the FL. Also, check whether inspiration time seems large relative to expiration time. "Normal" for many is a ratio of 1:1.8 for I and E times, but it varies for and among individuals.

2. There may not be a significant loss of vital Tidal Volume during periods marked with moderate prolonged FL flags. But it bears checking.

Note re images in #10, if you review them: Oscar was set to a different time zone--a 1  hour difference, so times in recent posts are an hour later.
   

[2SleepBetta]

I'm kicking myself but feeling liberated. After struggles using Excel to deal with huge BRP data files (for flow rate mainly) I used Somnopose import in OSCAR to display analysis results. That OSCAR cannot be beat: one can see the whole night or one second spread out with time-synched context. No pain, no fuss. But, wow, how much time I wasted using Excel charting after getting the analytical result needed for OSCAR import.

The attachments are further development of the the same sleep session, as above, after beginning with it in Post #10. That session had a remarkable display, which an attachment shows, of a long period without FL and then another with FL flags after a cluster of obstructive apneas. I omitted as irrelevant the early part of the session which had no FL, the key topic here.

For the most part the charts should be familiar to anyone following this thread. The same matters have been dealt with time after time. Accordingly, I'll limit remarks here and hope charts speak for themselves as displayed now in OSCAR.

A 3-minute view shows that only subtle indications of flow limitations seen in slightly unrounded inspiratory tips are causing FL flags which seem out proportion. I raised a question about that in Post 10 where and when I wasn't sure what was causing the disparity between significantly high FL flagging vs. such minor indications of tidal volume loss.

Speculation: It has occurred to me that mere unrounded tips are a signal of flow limitations at a  low level and should be looked at if sleep is not restful. IMO, those may signal an overshadowing constant flow limitation akin to sucking on an inflexible  large straw all night long. That would cause arousals we are unaware of which impair sleep  . Against that background from firm tissue constriction (I assume), I'm guessing that ones more sleep-relaxed soft tissue at other airway narrowings might cause the tip unroundedness and other flow limitations, FL flags or obstructive apnea we see. Our devices won't flag an underlying constant flow limitation of the kind a rigid tube could/would cause.

               

An added comment after checking Previews of the images. Tidal Volume shown by OSCAR, as I understand, is sampled only at 0.5 Hz, every 2 seconds. I've not examined that or why there is such a difference between the curve OSCAR draws from Resmed data and the TV I calculate from numerical integration of the area between the inspiratory wave and the flow rate axis. The TV I show is the volume enclosed by each such wave for which there are 25 data points per second.

[2SleepBetta]

Here are additional notes I owe readers who take time to follow--bear with--this wordy thread, although they are likely aware of these points. These are important to keep in mind relative to the approach I've used here and what remains to be done, if anything practical, to fix its limitations.

--Natural breathing will have breaths with TV dropping below 0.5 L, and the difference is represented as  TV drop. The fixed 0.5 L TV baseline is an approximation for varying TV that is practical to use at this state of the method, but it would be desirable to find a way to determine the actual varying values. Use those along with an index of TV normality (if one can be found) as remedy for the first point above.

--A thought for those with low AHI, but unrestful sleep to consider. The whole session view has Inspiration and Expiration time range set to show the most detail for the later flow limited period in the least OSCAR real estate, that to the exclusion of showing them in the pre-FL period. A look at those ranges, pre-FL vs FL period, accentuates what I had noted and may or may not have mentioned, but have done so and newly elaborated as below. 

In the pre-FL period I and E times converged just  "off screen"; I-time grew shorter, E-time longer. That shows my overall work of Inspiration was significantly higher in the whole FL flag dense later period when the opposite happened. That thought  just now suggests to me the value in FL flags which seem to exaggerate Tidal Volume drops (which I  have considered to be most important, to the point of being skeptical about Resmed "overstatement" of flow limitations in their "FL" flags ). The FL flags are caused by mere unrounded tips of inspiratory waves, not by TV drops. Those may/do suggest greater physical work of breathing that may contribute to unrestful sleep. (sheepless may have had this in mind, earlier, upon seeing that I was essentially dismissing the Resmed FL shape-sensing value as a machine control matter, not otherwise important relative to more vital TV drop matters we can attempt to mitigate.)   

The just out-of-channel-range of convergence of I and E , in the pre-FL period,  illustrates an aspect of flow limitation that applies generally and can sometimes help us explain unrestful sleep. When there is low AHI and/or little indication of flow limitation --by unrounded wave tips and/or FL flag--look at the relationship and absolute level and varying differences between I time and E time and the I/E value. Lengthening of I time is a first  clue (SleepRider's comments elsewhere drew my attention to this indicator). 

Suggestion: If you want to drill deeper, visually, situate your I time-channel view directly above the E-time channel view and scan their convergences and divergences to see the larger differences, I up and E down, or note increases in I with no decrease in E and vice versa. The points of time with the largest divergent differences will have highest I/E values, giving a clue to increased work of drawing a breath. Those are more likely to cause or be associated with mini arousals.

[2SleepBetta]

-----Again, it's time to step back, review and think about where or whether to go on from here. My conclusions to this point are based on only one person's single but richly illustrative SDB sleep session (mine). Nevertheless I believe the following observations are correct and I see FL more clearly as I hope others do too after wandering along with me down this tortuous thread and, now, as I attempt to put it all together.

Here is my hokey attempt to boil the thread  down to one sentence, better yet, to one  metaphor: The  Resmed flow limitation flag we see designated as "FL" is a multi-warning SDB  road signal that calls for alertness at the crossroads and/or detours around or through, Normalcy, Abnormal Tidal Volume (TV) drops, UARS, Algorithmic Washout and Machine Limitville. In general a FL flag does not tell us our TV or TV shortfall. It and other signs OSCAR presents can give us a good indication of the significance of the TV shortfall or drops .

The FL flag is a broader indicator than mere TV drops which, at the start of this thread,  I mistakenly saw, as many do, as the one most important function of the FL flag. As one astute member commented, no one needs to "do math" to see TV drops which can be inferred, as we all knew, from a glance at dips in the inspiratory half of our FR curve.  But consistent with the EMTs' "ABC rule" an open Airway and Breathing come before Cardio for the stricken; life sustaining  regular ventilation, TV, is first priority, as it is for fully restful sleep, our goal. So what does a FL flag say about TV? What factors cause confusion? 

 Some concluding observations, some more rambling restatements:

1. Resmed flow limit flags, FL, may be based mainly, if not entirely, on either misshapen inspiratory waves, including flattened waves or on  reduction of tidal volumes (TV) and all variations of proportions of those two manifestations. But that was known at the outset from published information. No surprise arose in this thread; only some specific evidence of same is presented. Nevertheless, critics of FL flagging appear to wish, as I do,  to have FL flags tell me when and to what extent my FL flagged breathing is inadequate and needs attention. 

2. The Tidal Volume drop (TVd) method shown in the thread is simply the computation of individual I-wave volume (TV) from flow rate (FR) curve values and deducting that tidal volume (TV) from the published base (normal) human TV of 6-8 ml./kg for one adult body. By that method one usefully estimates the TV drop, (TVd), if any, for each I-wave seen in the FR curve. (My avatar illustrates a TV indicator [green color] and a TVd indicator [red color] at the end of the the inspiratory wave. Added together, the red and green indications total 500 ml, the assumed normal TV for my weight.)

TVd's can be seen in drops of Flow Rate curve which is identical in form or profile to the curve of positive and negative TV values. Note that the TV curve, which has long been provided by OSCAR from Resmed data, smooths TV data and tends to cause the curve to lag FL effects. 

3. TVd could/can be scored and summarized meaningfully--far more so than FL--if in no better way, using percentiles. (Note I assumed and presented my TV inconsistently here as  480 ml and 500 ml. It is 476 ml for my weight at the 6ml. lower rate. My VAuto shows it varies, night to night, as either  480 ml or 500 ml.)

4. The TVd method of looking at flow limitation is new, if at all, only in the sense no one else may have applied it, doubtful as that seems. Most persons who know high school math (or advanced placement math) and a touch of physics likely have thought of the approach, but have had better sense than to spend time to do the tedious chore. I simply felt need to do it because of confusion about the FL flag meaning, including my own and that of others. I wanted concrete illustration for all. 

5. It is not practical to do the required data prep work repeatedly; all the data prep must be streamlined by a computer program to turn Resmed data into a form that can be imported into OSCAR, as has been demonstrated in this thread.

-----All said, what's the value of the method if it cannot be put into practical use? I'm not sure, but here are some matters that seem relevant after being better informed by the method. Pursuit of the following, or of what else they might suggest, could lead to something helpful for those troubled by unrestful sleep, OSCAR indications of dense FL occurrences and/or  deformed inspiratory wave tips. Again there is nothing new to senior moderators and analysts here at AB. 

1. If a FL flag can be based on the shape of an I-wave alone (no TVd), it is, nevertheless, a valuable indicator of flow limitation existing at or above the blower device's FL detection level. The value is in its calls for, (a), inspection of the inspiratory peaks of the nearby FR curve and, (b), a similar  look at whether inspiration time is or has been high at the approximate FL time. Regarding the latter, determine whether the quotient " I-time/E-time" increased at the FL time (allowing for machine algorithm processing time). Suggestion: Place the I-time curve just above E-time curve and look at the times when the curves are farthest apart--that occurs when the quotient I/E is largest.

2. Notice whether increases in FL could be explained by more pronounced cardiogenic waves (cardio ballistic artifacts "CBA) at the FL time and whether deformations of the I-wave tip seem to be "in step with" the CBA pattern just before inspiration starts at end of expiration. In graphics presented here, it was necessary to filter out CBA and those in the range--I was feeling my way here--of volume 0.015 to 0.020 ml were eliminated.

-----Some highly speculative and dubious thoughts from studying FR, FL, TV, I and E curves and I-tip deformations of mine and of one or more other persons. My sense is that these anomalies arise when extreme I-wave irregularities--in form, density on the time line or/and duration-- impinge algorithm design limits of the Resmed blower. 

1. FL flags are as described above and elsewhere, but can show, I believe, two other confusing attributes that are evident in FL curves of UARS affected apneic persons. On  one hand confusion arises from FL being or becoming "sticky" upward in a FL-dense or tip-deformation-dense context. On the other hand, FL are or become  sticky downward mutatis mutandis.  Upon seeing a persistent upward tendency I believe I mentioned in another person's thread that there seemed to be a "bridging effect" where a clean drop of FL for some unspecified period was needed for flagging to drop to zero. See attachment http://www.apneaboard.com/forums/attachm...?aid=34845

2. Along the same line, one can see flagging of certain deformed wave tips in one sleep session (or time span within a session) and not see flagging of what appear to be the same tips in another sleep session or segment of the same session.


3. Adding more confusion, I believe that under certain conditions there is, or appears to be, an erroneous kind of multiple counting--as in "double counting"--in the stacking up of FL from each member of a series of similarly deformed I-wave tips. A series of, say, M-tips that seem to be alike (visually) can cause an upward stairway form of FL curve, having a new step upward in severity value for each succeeding I-wave, all waves being seen as superficially  alike (that is to say that there was only one kind of FL wave deformity and it had no severity greater at any later point than it had had at the beginning. Severity never got worse although indication of it did. See attachment at item #1 immediately above where there is little I-wave tip deformity variation but there is stacking up of severity. Note that  "severity" is the vertical dimension of a FL  as opposed to its duration dimension.)

[sheepless]

"Note that "severity" is the vertical dimension of a FL as opposed to its duration dimension."

unrelated, or at least tangential, perhaps, to your excellent summation, I wonder if you have any better understanding of the duration dimension as a result of your inquiries.

as I brought up in another thread, intuitively it seems to me that duration of flow limitations, episode by episode and in the aggregate per night, is at least as significant as severity. e.g., high severity for brief periods may have less impact on sleep quality than longer periods of lower severity flow limitations. but as far as I know duration isn't a factor in our current 'score' / 'rank' / 'grade' scale of 0-1.

I didn't get the impression from the other thread that there was much interest in this. or maybe folks don't agree with the premise. idk. but until I read a reasonable argument against it, I still wonder if we are ignoring duration of flow limitations to our detriment.

so my questions are:

does 'severity' include duration as a variable in the resmed flow limitation algorithm?

is the 'stacking' you refer to in #3 above how duration is factored in? if so, is that an adequate characterization of the duration dimension?

if not, is there a way to capture and report the horizontal duration dimension separately from or integrated with the vertical dimension?

[2SleepBetta]

sheepless  you always raise good questions that help the project.

I think taking a close look at the attachment along with a couple comments here will clear these matters up better for you and others as your question did for me just moments ago.

I do not ever recall seeing prolonged FL of any severity in one of the ranges of 0.01 to, say, .05 for more than about 12 seconds duration--if that long. 

Accordingly, I believe, the continued stacking, as can be seen in the attachment can be a means of expressing duration for an extended period of repeated FL flaggable occurrences of one and the same FL severity. 

More and more I see the FL as a warning sign and that both its width and height are significant. Size, the FL flag size, matters as to degree of warning but  not necessarily as to TV drop. Again, it suggests to me that summing up the areas could be helpful.

   

Edit: The graphic states a range as .015 to 0.20. The 0.20 should have been 0.020.

[sheepless]

"More and more I see the FL as a warning sign and that both its width and height are significant."

I read that to support my contention that duration is important but I'm still not clear whether duration is included in the machine determination of the flow limitation 'grade'. your suggestion about summing these areas implies duration is not currently a factor in the assignment of severity on the 0 to 1 scale(?). is that correct?

switching gears, do you think the 4 second stair steps correspond to the time between the pointed waveform troughs below the zero line in the flow rate or is that coincidence?