Artificial pancreas trial

[himtoo]

thanks for the update @diabetic0312

 

[tim2000s]

By the looks of the blue dotted line were you on the standard instruction to have 4 or 5 blocks of rates a day? As it looks like the Artificial Pancreas is doing differently??

This appears to work in a similar way to OpenAPS and Loop. You still need to apply a basal profile, but since the AP knows your IC and ISF values, the Insulin Action curve, does some form of carb absorption calculation and an existing basal rate, it does a calculation that works out IOB, the action of that IOB and predicts a future glucose level based on the listed factors.

This is done with the receipt of every new blood glucose point from the sensor, hence the stepped basal rate that you see.

The Medtronic 670G, from what I understand of the presentations I've seen, works in a slightly different way.

There's a really good round up of all the ongoing APS projects here, put together by the University of Birmingham. Florence is included in that.

 

[donnellysdogs]

This appears to work in a similar way to OpenAPS and Loop. You still need to apply a basal profile, but since the AP knows your IC and ISF values, the Insulin Action curve, does some form of carb absorption calculation and an existing basal rate, it does a calculation that works out IOB, the action of that IOB and predicts a future glucose level based on the listed factors.

This is done with the receipt of every new blood glucose point from the sensor, hence the stepped basal rate that you see.

The Medtronic 670G, from what I understand of the presentations I've seen, works in a slightly different way.

There's a really good round up of all the ongoing APS projects here, put together by the University of Birmingham. Florence is included in that.

Not exactly answering my question....

All I was aaking was whether her profile was originally set up with blocks of hours being at the same rate for a few hours, as this is what it looks like. (I am only viewing a small photo on a phone-can't see in detail).....

Three of my pump hospitals didn't go along with block settings, and were moe than happy with my so calculated individualised profiles that ai had established over 5 years. One hospital, my latest, wanted me to go back to pump again on their block set up of basal profiles.

I just thinking that it looks in the photo that the AP is not agreeing with the mass blocks of original basal rates and making quite a few adjustments.......

 

[tim2000s]

I just thinking that it looks in the photo that the AP is not agreeing with the mass blocks of original basal rates and making quite a few adjustments.......

It rarely does. One of the biggest items that the ongoing artificial pancreas trials have identified is that the 1, 12, 24 or 48 block model is highly inexact and unrealistic. It's very rare for any two days to have the same basal requirements and although a basal profile provides a fall back in the event of something failing, you should really consider the multiple TBRs as 5 minute square wave boluses - effectively a microbolus model.

Because it is reacting to the changes in its predicted glucose level constantly, it's not really a basal profile that you are looking at, as it is also affected by food consumption.

The image below shows two days where I ate different foods and did mostly similar things, and the differences in the adjustment of basal insulin on both days. We all have a variety of factors that affect our insulin sensitivity and these also affect our insulin requirements. Think of it more like never needing a "Weekday" profile, a "weekend" profile and a "nights" profile as the system will always adjust appropriately.

 

[slip]

For those who have a cricked neck!

 

[donnellysdogs]

It rarely does. One of the biggest items that the ongoing artificial pancreas trials have identified is that the 1, 12, 24 or 48 block model is highly inexact and unrealistic. It's very rare for any two days to have the same basal requirements and although a basal profile provides a fall back in the event of something failing, you should really consider the multiple TBRs as 5 minute square wave boluses - effectively a microbolus model.

Because it is reacting to the changes in its predicted glucose level constantly, it's not really a basal profile that you are looking at, as it is also affected by food consumption.

The image below shows two days where I ate different foods and did mostly similar things, and the differences in the adjustment of basal insulin on both days. We all have a variety of factors that affect our insulin sensitivity and these also affect our insulin requirements. Think of it more like never needing a "Weekday" profile, a "weekend" profile and a "nights" profile as the system will always adjust appropriately.

View attachment 21364

Thanks Tim... to be honest that explanation means a lot more to me!! Lol

 

[donnellysdogs]

I find it interesting because I (used) to be able to keep mine between 5 n 7 on a vast majority of readings.

Bith pictures still show peaks occurring.
Ie getting up etc from 5am to midday extra peaks needed of insulin but its not keeping levels as good as I used to do by my own tweaking..

I do have to forget the last two years of no achievement at all on a pump!!

Just finding it interesting.. it would be intersting to see a post from the trial pump for a busy day and a sedentary day. Ie a weekend to a weekday etc...

Would a profile still have to be set up by the user, like I had for AM shifts, PM, weekends etc or would AP recognise and be able to adjust one profile?

 

[tim2000s]

I can only speak for OpenAPS/Loop:

Bith pictures still show peaks occurring.
Ie getting up etc from 5am to midday extra peaks needed of insulin but its not keeping levels as good as I used to do by my own tweaking..

Yes, they do, and this is because these systems can't predict what you are going to do in future and react. Much of the basis of keeping really flat profiles comes from knowing what you're going to eat and dealing with it ahead of time. My stats using these systems are almost exactly the same as Sugar Surfing. The difference is that I'm spending 12 hours a month less doing the tweaking that gives me that data.

The other thing to bear in mind is your eating patterns. If you eat a lot of carbs (and I've been doing that with the loops to really put them through the test) they give you these numbers in spite of that. When Sugar Surfing, I was eating ~80g carbs a day to get my variation and hba1C really good. Using the hybrid closed loops, I've been able to eat up to 400g (for test purposes!) a day of carbs and still stay within the same numbers and not have to do very much at all to get there.

Only the Bigfoot product is coming close to predicting what will be eaten (it has incorporated a learning algorithm that can learn your daily/weekly pattern and predict what you will do and when, and start to adjust thing accordingly) and even then, these can only react to the inputs they see, and when they react , they have to react with slow insulin. The likes of FasterAspart and BCLispro will make a huge difference to that reaction time, and therefore the results.

With OpenAPS, there is a function that you tell it you will eat in the next 90-120 mins and it will start to seed insulin, then when you do eat, if you forget to bolus, it knows and starts to deliver bigger TBRs to compensate based on the rise rate it sees and the length of time it runs for). So there are ways to deal with it, but the real issue is insulin performance!

 

[donnellysdogs]

So my routine of 4am injecting etc (slightly lower today due to correction bolus hangover from midnight) - is this achievable on an AP?

Can you actually get good flat lines with an AP?

This is MDI which is not so good as my pump results used to be... and only starting tresiba last week getting highs at other times...

Looks like AP to me with current results displayed that its a nice option for T1's but needs progress on data, algorythims etc to be able to contend with someone with randomised oddness!! (Like I was ie active certain days, less others, gym, lounging, and a bad stomach!! )....

I find it all very interesting.. thanks for explaining in "simple" terms...

 

[tim2000s]

So my routine of 4am injecting etc (slightly lower today due to correction bolus hangover from midnight) - is this achievable on an AP?

Completely. My DP is handled by it with no issues, as are gym sessions.

Can you actually get good flat lines with an AP?

Yes, but the biggest driver of flat lines is never background insulin. It's food. 100% all the way. You can see how many carbs I had for breakfast and it has kept me in a range of 5.4-7.3 all morning, with the "spike" being 7.3 an hour after breakfast and back down to normal levels 3 hours later.

If I was to eat 120g carbs a day, the line would be pretty much flat, unless that was all jelly babies. It handles activity exceptionally well, because once you start to see the levels drop it just suspends insulin. You are in control of the target level, so if you want to exercise and are worried, you just have your target slightly higher.

For me, the one thing that drives variability more than anything else is the carb content of food. Lounging or exercising it does a great job with, and varied getting up times are also not an issue. So, I'd disagree. I don't think the algorithms need more work to contend with "randomness". They are designed to do that already.

 

[donnellysdogs]

Thanks Tim I do appteciate the greater explanatiom that you have given.

I suppose in the greater view of things that I spent a long time getting basas to cope with hubbys diff shifts, active days etc.. and this would free up and give some improved quality of life without thinking and actioning so much!!

You are right with food too... and ****** stomachs.. For people with stomach and digestive problems our food rarely gives good results because we do not know how long that food is going to really get in to our system. So, the AP could determine that though probably at us having to second guess?? Even though its not necssarily perfect it would be working on real data and recognising when our difestive systems are releasing glucose..??

Thanks Tim..

 

[tim2000s]

So, the AP could determine that though probably at us having to second guess?? Even though its not necssarily perfect it would be working on real data and recognising when our difestive systems are releasing glucose..??

That's what Advanced Meal Assist in OpenAPS is trying to achieve, yes.

 

[donnellysdogs]

That would be something so hugely so significant with digestive problems!

That would turn some really chaotic lives in to much more quality of live!!

 

[donnellysdogs]

So.. sorry about this Tim!!

With somebody like me that has a problem with skin initially accepting cannulas for first 15 hours and then having TBR hangiver hypo's for the next up to 12 hours...then a good few hours till next set change.

This (by the sounds of it) could actually handle the skin reactions with cannulas and insulin in this scenario?

 

[tim2000s]

That would be something so hugely so significant with digestive problems!

That would turn some really chaotic lives in to much more quality of live!!

There are already quite a number of people with GP using OpenAPS precisely because of those issues.

 

[donnellysdogs]

There are so many around the world though that are seriously struggling. When fed by tube etc and surviving on next to nothing so many more people need these......