Bubble model vs. Gradient Factors redux

[scubadada]

Hi @Brett Hatch

Bubble detection requires transthoracic echocardiography. This cannot be done underwater during ascent. Doing it every 15 min starting at 30 min is pretty intensive. I might have liked to see the 15 min grade. In this study, 50/75 peaked at the 1st point, 30 min. Interesting, it's often said bubbling peaks at about 45 min, as it did for 20/85. The average bubble grade, not sure if that's valid, is lower for 50/75 than for 20/85. If you did a one hour SI, you'd be starting with a slightly lower bubble grade with 50/75 than with 20/85.

Bubbling may be a marker for risk of DCS but that is all. As there were no episodes of DCS, we can't tell. Take NEDU for example, there were episodes of DCS and an excessive number in the bubble model group. As bubbling is only one marker, it would be smart to check other valid markers, perhaps like the chemokines measured in the Spisni study, see Bubble model vs. Gradient Factors redux

 

[Storker]

I don't believe there are any. The testing would be done on the assumptions of what the models are supposed to do..

The model is pure ZHL-16C. Although I hate to admit it, Ross were onto something when he dissed the GFs, because they are fudge factors added on to the original model to make the deco less aggressive. Without knowing very much about VPM, I more than suspect that its conservatism levels are similar fudge factors, but if I'm wrong I'd like to be corrected on that.

Parameters was perhaps a poor choice of words. I should have said the output of the algorithm. If we're going to test the veracity of the algorithms we need to follow the output.

Well. Given that we know that as long as you enter the realm of mandatory staged deco, no decompression whatsoever carries an unacceptable risk of getting bent, and that a more aggressive ascent profile carries a higher risk of DCS than a less aggressive ascent profile, an initial assumption that more deco = less risk of getting bent is reasonable.

Now look at what we have here. We have two fundamentally different approaches to deco. One says that we should start stopping early to protect the fast tissues, the other says we should go to the border condition for the limiting tissue before we stop. One approach favors deep stops, the other favors shallow stops.

For a simple comparison of the approaches, we should keep as many parameters as possible constant. Perhaps the simplest parameter to keep constant is total deco time¹, because we know that total deco time affects decompression stress. Most of the other parameters can only be handled by having enough data points ("man-dives", sorry for not being able to cook up a gender-neutral term here). That way, the only thing we vary deliberately is the deep vs shallow stop distribution.

So, to repeat myself, I'd use whatever accepted fudge factor I'd have to use to keep the total deco time constant. Whether I'm comparing VPM-B+4 to ZHL-16 73/73 or I'm comparing VPM-B+2 to ZHL-16 85/85 doesn't matter to me. What matters to me is that I'm comparing two fundamentally different ascent strategies at (as) constant conditions (as practically possible).

¹ Which, incidentally, was what those really smart dudes at NEDU chose to do. Still, it's just experimental science 101.

 

[EFX]

The model is pure ZHL-16C. Although I hate to admit it, Ross were onto something when he dissed the GFs, because they are fudge factors added on to the original model to make the deco less aggressive. Without knowing very much about VPM, I more than suspect that its conservatism levels are similar fudge factors, but if I'm wrong I'd like to be corrected on that.

But what has GF's or +n factors have to do with the underlying assumptions I listed in my previous post. We need to use these factors in a test for two reasons: (1) to come up with profiles that reflect what divers are actually using and, (2) to bring the two profiles deco ascent schedules as close as possible notwithstanding the deeper stops of VPM (which is what we're testing) and the longer shallow stops of ZHL (which is what we're not testing).

 

[Storker]

notwithstanding the deeper stops (which is what we're testing) and longer shallow stops (which is what we're not testing).

I don't get you. If you have a fixed total deco time, if you forego the deep stops you have to increase the shallow stops (and vice versa)

 

[EFX]

I edited post #63 for better clarity. I'm not advocating a fixed total deco time (TDT). The TDT is an output of the algorithm (model) and we should not change that. If you change an output you are not testing that model but some bastardized version of it. That was the mistake of the NEDU study. Thanks for mentioning it [NEDU study]. I wasn't going to bring it up.

If I remember correctly the NEDU came up with two profiles based on two models, their dissolved gas (DG) model and their bubble model for a 170 ft 30 minute dive on air. I believe the bubble model gave a longer TDT. They changed the TDT of the bubble profile to match the TDT of the DG model by shortening the shallower stops and adding that time to the deeper stops. That's the redistribution of deco time they talk about. So, here's the problem. The NEDU study reported a higher incidence of DCS with the deeper stops (no surprise there). What conclusion can we reach? Was the shallow stop model safer or was the deeper stop model more dangerous because we made it so? If they made the shallow stop model's TDT shorter and left the deeper stop profile alone would there be more DCS reported with the shallow stop model?

 

[scubadada]

I don't get you. If you have a fixed total deco time, if you forego the deep stops you have to increase the shallow stops (and vice versa)

Exactly. Again, the NEDU profile, 170 feet, 30 min, air. MultiDeco, 1 min stops:

It is pretty easy to get run times to match perfectly with enough tweaking

 

[Storker]

I'm not advocating a fixed total deco time (TDT).

Depends on your target. If comparing different ascent schedules is your target, it seems to me that a fixed total deco time is rather vital.

I dont know what your target is, so if you'd care to elaborate on that it'd be appreciated.

 

[EFX]

Yes, I agree. My target is not ascent schedules but the underlying assumptions (methodology if you will) of the two algo's. Whether we choose VPMB+0 or VPMB+5 it makes no difference to the method. Regardless of the conservatism the bubble phase of VPM will still try to limit the size and quantity of bubbles from exceeding a critical bubble size (BS). All you're changing is one parameter of VPM, the critical bubble size.

The same goes for ZHL16C. GF's only change the limit of supersaturation during the ascent phase. The model will still use the amount of supersaturation as the limiting factor. ZHL does not care about BS so we need to find out if keeping the level of supersat low regardless of BS is enough to keep us safe.

I would also like to know if bubble size or its combination with bubble quantity (BQ) is an important factor in safe decompression. One underlying assumption of bubble models and VPMB in particular is that smaller bubble size coupled with a somewhat larger bubble quantity is OK. Now, I say higher BQ because that is what we see happening with deep stops compared with shallower stop models in ultrasound VGE studies. As a generalization I think we can all agree that large bubbles are inherently more dangerous than smaller bubbles because they can lodge in capillaries blocking blood flow or increase pressure on nerves among other issues.

Why should we care about deep stops?

We want to prevent very small bubbles at nucleation sites that are both created during the descent and level phases of the dive and preexisting bubbles pre-dive, from growing on ascent beyond a critical size. Bubbles will grow anyway because of the reduction in ambient pressure on ascent. This poses a problem though in that there is a somewhat positive feedback mechanism in bubble growth. In terms of pressure there are three components that affect bubble growth: (1) ambient pressure (AP), i.e. outside the bubble, (2) the pressure caused by surface tension (ST), and (3) the pressure inside the bubble (BP). Both ambient and ST act to oppose the inside pressure. When BP = AP + ST, growth or shrinkage stop.

The problem is on ascent when AP drops at some point the BP exceeds AP + ST and the bubble grows. This growth increase the volume of the bubble which causes the BP to decrease which enables gas from a higher AP to flow into the bubble. The reduction in AP also creates a greater volume of gas available to diffuse into the bubble which contributes to bubble growth. So, VPM creates deep stops to prevent preexisting bubbles from growing on ascents to shallower depths. The hope is that some of the dissolved gas in the tissues will offgas at the deep stops to reduce the resevoir of available gas on successively shallow stops that would contribute to bubble growth perhaps in a runaway situation. You also stabilize bubble growth at the deep stops and give time for existing bubbles to either dissolve or get passed out of the body via the lungs . Sorry for the long-winded explanation. Also I'm just giving basics here. There are more complicated nuances that are part of the model. See the attached file for more information.

 

[Storker]

Thanks for elaborating. However:

regardless of the conservatism the bubble phase of VPM will still try to limit the size and quantity of bubbles from exceeding a critical bubble size (BS). All you're changing is one parameter of VPM, the critical bubble size.

If bubble size can't be measured, how can we consider it? It's all just speculation. Bubble quantity can be measured, but as long we can't measure bubble size it's just yet another confounding factor in the analysis. So we can't afford to bother with it.

The bottom line is P(DCS). And if we can't get those numbers, we have to make do with bubble count as a proxy.

If the bubble models and their associated algorithms prove to provide a lower P(DCS), then it's time to find the reason why. But until then, the burden of proof is on the bubble models.

 

[EFX]

Thanks for elaborating. However:

If bubble size can't be measured, how can we consider it? It's all just speculation. Bubble quantity can be measured, but as long we can't measure bubble size it's just yet another confounding factor in the analysis. So we can't afford to bother with it.

Being able to measure bubble size and correlate that with DCS incidences would help but it's not necessary in proving the validity of the VPM model. I want to test the algorithm and showing it's safer will also validate (in an indirect way) bubble mechanics since that is a strong proponent of the method.

And we shouldn't forget that the strength of ZHL was based on numerous animal and human tests long before VGE studies (showing bubble quantity) became available. So, the lack of bubble size tests should not deter us from testing.