Question Iso-risk decompression schedules

[LFMarm]

Thanks for that. How is TST defined? Is that a bottom time? Total dive time?
How many dives in the training data set actually ran that long? What was the n/n for DCS on those dives?

Presumably the vast majority of dives in the training set had no DCS, despite un-conservative parameters (due to sanity and ethics). So this model fits are tasked with focusing on a few very rare DCS incidents, imagining some edge case parameters that might eliminate them--unless they don't, due to additional causative or 'random' factors.

Are we sure that the relatively small number of DCS incidences (≤2%) in these models can be explained or predicted on the basis to small differences in GFs and deep stops? How much additional data would you need to statistically control for a random null model or negative hypothesis of DCS occurrence, and other factors such as person-to-person variation, etc?

TST is the sum of all deco stop plus time to move up to next level (except first stop). The training set is thousands of dives from NEDU (Details on the original model / data set is in the article from 2018).

 

[justinthedeeps]

TST is the sum of all deco stop plus time to move up to next level (except first stop). The training set is thousands of dives from NEDU (Details on the original model / data set is in the article from 2018).

Yeah I understand it's a large training set, but what portion of the training set actually covers those very deep depths and very long runtimes? And how many of those dives were associated with DCS? How much data is actually being trained on, for these edge case parameters?

 

[stuartv]

This makes sense if you think about it. GF is a measure of supersaturation. I you spend more time at the same supersaturation, you have a greater chance of an adverse event resulting from that supersaturation.

I think this is an oversimplification that leads to a somewhat incorrect conclusion.

I think it makes sense to base the evaluation of risk of DCS in the context of the algorithm ultimately being used to guide the decompression.

In the case of Buhlmann/GF, each compartment and gas in the model has its own M-value. And each M-value has its own risk level.

If you spend more time at the same supersaturation, you are (potentially) changing which model compartment will control your decompression. As they each have different risk, what that means is that, by staying longer at depth, you might actually reduce your chance of DCS by virtue of changing your controlling compartment(s) to one that has a lower inherent risk.

 

[L13]

I think this is an oversimplification that leads to a somewhat incorrect conclusion.

I think it makes sense to base the evaluation of risk of DCS in the context of the algorithm ultimately being used to guide the decompression.

In the case of Buhlmann/GF, each compartment and gas in the model has its own M-value. And each M-value has its own risk level.

If you spend more time at the same supersaturation, you are (potentially) changing which model compartment will control your decompression. As they each have different risk, what that means is that, by staying longer at depth, you might actually reduce your chance of DCS by virtue of changing your controlling compartment(s) to one that has a lower inherent risk.

I think you totally misunderstood my statement, and maybe misunderstand supersaturation. Maybe because I did not explicitly say "tissue supersaturation."

Supersaturation is when inert gas tension is a tissue is greater than ambient pressure. The M-value is the maximum allowed supersaturation in a tissue. A GF of 100% means that that tissue compartment's inert gas tension is at the M-value of supersaturation. A GF of 0% means that the tissue inert gas tension is at ambient pressure (saturated).

The higher the supersaturation (higher the GF, closer to the M-value) the greater the risk of DCS. The longer the time spent at high supersaturation (longer time at high GF, longer time close to the M-value) the higher the risk.

You seem to be confusing high inhaled partial pressure of inert gas with high supersaturation. They are not the same thing.

While all of your comments are (mostly) correct, they have nothing to do with mine and do not support your premise:

I think this is an oversimplification that leads to a somewhat incorrect conclusion.

It is in no way an oversimplification. it is only a restatement of the definition of GF in different terms. The conclusion is in no way incorrect. Higher GF's (closer to the M-value)are in fact more risky than lower GF's, and more time at high GF's (more time close to the M-value) is in fact more risky. The conclusion and it's basis stand.

In fact all of your statements after your premise support my statement (except where you incorrectly use the the term "supersaturation" in the last paragraph when you probably meant high inhaled inert gas pressure).

 

[inquis]

In the case of Buhlmann/GF, each compartment and gas in the model has its own M-value. And each M-value has its own risk level.

My understanding is that the M-values were originally set with the intent of equalizing risk. Faster tissues (those with greater perfusion) can tolerate higher tensions without issue, for example, and we see this reflected in higher M-values.

However, when the researchers in the field today run a GFlow value that is lower than the GFhigh, it's apparent that they believe the faster tissues warrant greater protection. Presumably, this approximately equalizes risk across compartments at the modified levels of tissue tension.

by staying longer at depth, you might actually reduce your chance of DCS by virtue of changing your controlling compartment(s) to one that has a lower inherent risk

It is here that I think our thinking diverges. As we ascend, a switch in controlling tissue is always to a slower tissue. At the point of switching, the tissue GFs (tissue tensions relative to the critical limits) are equal, meaning comparable risk rather than lower risk. Moreover, the rate of progress (tension reduction) slows, mandating longer time in the water. This increases risk simply from a pragmatic standpoint (e.g., thermal, equipment issues, or even human error). Probability math dictates that for whatever chance of error is present, the probability of *not* having an error is greater the fewer times you roll the dice.

All of which is a long-winded way of saying that I believe switching compartments does not reduce the risk.

 

[L13]

... by staying longer at depth, you might actually reduce your chance of DCS by virtue of changing your controlling compartment(s) to one that has a lower inherent risk.

Good catch by @inquis! I did not register his suggestion that staying longer at depth might reduce DCS the first time I read it.

Longer time at depth increases the saturation of ALL tissue compartments, which increases the supersaturation (and therefore DCS risk) of ALL compartments as you ascend to the surface. On ascent the controlling compartment only changes because the supersaturation (and therefore DCS risk) of a longer TC compartment is not dropping as fast as the supersaturation/DCS risk of the faster compartments are, so it becomes the one with the highest risk. "Controlling compartment" is just the compartment with the highest percent of its M-value and therefore highest DCS risk. It is only "controlling" in the sense that it is the one that most constrains your ascent to the next stop because it has the highest DCS risk.

The "controlling compartment" has no direct effect on the other compartments, it only controls when you can ascend to the next stop. However, it does have an indirect effect, by keeping you deeper longer, it might increase the saturation of an even slower compartment which then might become controlling later in the ascent (basically the problem with deep stops/low GF_Lows). Staying deeper longer does let the faster compartment's supersaturation/risk drop further, but by definition they are not "controlling" because they are not the ones with the highest risk and their risk is dropping faster than the risk of the "controlling" compartment is dropping.

In ALL cases, staying deeper longer will always either increase DCS risk, or increase deco time to prevent the increased DCS risk. There is no case where staying deeper longer reduces DCS risk.

 

[LFMarm]

Did some more experimenting with Doolette’s numbers. Shallower dives (max depth) require lower GF High to keep DCS risk equal for an equal TST time. Hence I am going to use the 66m curve as the limit for my GFs. Red dots are the GF High numbers from the XVAL tables from the 2018 article. Black dots are the interpolated ones. The red line is the squared profile that you can program in Shearwater computers (multiples of 5).

And here is the table of GFs for 2.3% DCS risk (corresponding to the red line in the chart above):

TST (min)​	GH Low​	GH High​
45​	70​	80​
50​	65​	75​
55​	65​	75​
60​	60​	70​
65​	55​	65​
70​	55​	60​
75​	50​	60​
80​	50​	55​
85​	45​	55​
90​	45​	50​
95​	45​	50​
100​	45​	50​
105​	40​	45​
110​	40​	45​
115​	40​	45​
120​	40​	45​
125​	40​	45​
130​	35​	40​
135​	35​	40​
140​	35​	40​
145​	35​	40​
150​	35​	40​
155​	35​	40​
160​	35​	40​
165​	35​	40​
170​	35​	40​
175​	35​	40​
180​	35​	40​
185​	35​	40​
190​	35​	40​
195​	35​	40​
200​	35​	40​
205​	35​	40​
210​	35​	40​
215​	35​	40​
220​	35​	40​
225​	35​	40​
230​	35​	40​
235​	35​	40​
240​	35​	40​
245​	35​	40​
250​	35​	40​

 

[inquis]

@LFMarm, I thought the XVAL tables were for heliox. Are you aware of any evidence (either way) about using them on trimix?

 

[LFMarm]

@LFMarm, I thought the XVAL tables were for heliox. Are you aware of any evidence (either way) about using them on trimix?

No, I am just assuming similar behavior.

 

[stuartv]

I think you totally misunderstood my statement, and maybe misunderstand supersaturation. Maybe because I did not explicitly say "tissue supersaturation."

Actually, I think you totally misunderstood my first statement. I wll say, I did not communicate very well.

I will try again.

Your time at depth has zero risk of DCS ... until you ascend. Thus - to ME, anyway - it only makes sense to quantify the risk of DCS in terms of the algorithm being used to control the ascent.

Saying that staying at (for example) 200' of depth for longer increases your risk of DCS is not all that helpful.

Saying that staying at 200' for longer and using Buhlmann w/GF with parameters of 50/80 for ascent is higher risk (than staying a shorter amount of time) is a hypothesis which can be analyzed in a useful and meaningful way.

Your earlier statement to which I replied seemed to be saying that when I am at a certain depth, staying longer increases my chances of DCS. If I interpreted it correctly. For reference, you said:

This makes sense if you think about it. GF is a measure of supersaturation. I you spend more time at the same supersaturation, you have a greater chance of an adverse event resulting from that supersaturation.

And again, I will say this is not exactly correct. You are ignoring that the model (a specific model implied by your use of the term "GF") uses multiple compartments - each of which has its own GF and its own risk level.

If you evaluate the risk based on using Buhlmann w/GF to control the ascent, then you can find that there are instances where the risk is lower by staying at depth longer.

Why? Because staying longer can change what tissue compartment becomes the controlling compartment duing different portions of the ascent.

The M-values for each compartment and gas are not exact science. Ideally, they would all yield the same risk of DCS, but I assert that they do not.

And, as they do not, the actual risk of DCS varies depending on which compartment is controlling the ascent. Which brings us to the inevitable conclusion that a longer bottom time could result in a different controlling compartment, which could then result in a different (including lower) risk of DCS.

Disclaimer: I am not a scientist and these statements are not based on formal scientific research. They are based on my lay understanding of decompression combined with my own personal experience.