Artificial Pancreas

Looks great but pretty sure it's gonna be way too pricey for NHS. They won't offer CGM or Freestyle Libre now to the vast majority. Prices will eventually drop but I doubt we'll see this in UK for NHS patients for many, many years.

It's interesting. This is in competition to similar work being done out of MIT that uses two Tandem pumps and CGM to deliver both insulin and glucagon.

All very interesting stuff.

noblehead said:

Never say never, aren't the NHS looking into funding the Libre sometime soon?

Click to expand...

Hope so but I wouldn't hold my breath. Looking into something is a long way from dishing them out. We'll see.

The algorithm part of this interests me. I'm sure it can't be that hard to retro fit a predictive algorithm using cgm data to back test. Hmmm.

tim2000s said:

The algorithm part of this interests me. I'm sure it can't be that hard to retro fit a predictive algorithm using cgm data to back test. Hmmm.

Click to expand...

How well would these things work in real like though with different living?
How could a person with sedentary living in an office have the same algorithims in a machine as a person that is a lumber jack working in the cold?

To me, we are all individuals. Algorithims will not be able to reprogramme the differences in our individual lives..

Pump users all have differnt capabilitys of setting up different profiles for their own needs. Even me swapping from AM to a PM shift is programmable on my pump.

How are algorithims able to cope with individuality?

The Algo doesn't care about you. It looks at the reference dataset, which is the cgm data. If it is going higher than the required limits it acts to bring it down, and vice versa.

It probably requires some calibration, ie, insulin carb ratios, but in theory it's quite straightforward.

The people I work with do similar things, and with a single data stream and the correct parameters, you could also incorporate machine learning.

I can see it being approved sooner rather than later.

tim2000s said:

The Algo doesn't care about you. It looks at the reference dataset, which is the cgm data. If it is going higher than the required limits it acts to bring it down, and vice versa.

It probably requires some calibration, ie, insulin carb ratios, but in theory it's quite straightforward.

The people I work with do similar things, and with a single data stream and the correct parameters, you could also incorporate machine learning.

I can see it being approved sooner rather than later.

Click to expand...

So how would the algorithms know from the CGM data for example... I've sat around for 2 hours.... ( so needing more insulin)... Which will affect me for up to 5 hours.... But then I'm going to do 4 hours hard labour digging? I eat the moment I can reduce my basal 2 hours before I go out to keep me perfect.....

An algorithm surely can't fine tune actions in advance?? And fast acting insulin can act up to 5 hours later. I could see that I would be needing extra glucose constantly being fed into me from the CGM readings.... Which currently I don't have to have.

That extra glucose would actually mean more calories daily being pumped in rather than just reducing my Basals in advance?

Which is why the MIT one uses glucagon. When you are heading low, suspend insulin. If you go low, increment glucagon to cause an appropriate liver dump.

You have to be able to go up and down, via hormones. Only using insulin doesn't give you enough control or a back stop.

You need to be able to force the body to bring glucose levels up in the event of an overdose of insulin as the alternative doesn't bear thinking about.

The other point is that the Algo is predictive, so based on vast amounts of data it knows what the reactions of a percent of the population are and acts accordingly. The rate of change of bg level is one of the key indicators in predicting what happens next and what it needs to do. It's really fascinating stuff, but the algos that do this in other fields are advanced and very effective. It's not a huge step to make it work in an artificial pancreas.

But then if the MIT one pumps glucagon in, then that's just adding in more stuff (so to speak) that I wouldn't have needed to pump in if I had altered my own basal in advance? So it would then be just adding in more calories from the glucose that I wouldn't have needed if I was managing it by myself??

Or are you saying that these algorithm things could basically learn that 5 days a week I am a lunatic physically, 7 days a week during summer and January I like to take the month off(ish) from gardening?? Can algorithms fine tune like that?

Personally I wouldn't want to be reliant upon a machine pumping glucagon in me as well!!

Algorithms can learn patterns. They are more likely to be able to learn a 7 day one than which months of the year you do and don't have physical lunacy. The thing is that your body naturally would establish a balance using both, so why shouldn't a pump?

Wow!!! Didn't realise algorithims could do all that.

My fear would be of having to have that extra glucagon being pumped in wheras I don't need to now....

It's all too confusing to someone in my "new age". I would rather count cabs take the appriate insulin, I know when I am going a bit low or a bit high and can alter that. I would not like to trust my life to a machine designed by a man made by a machine, and as we all know machines or computers neverfailneverfailneverfail. ...