Artificial Intelligence and Diabetes

[jwilkes1]

I’ve noticed we’ve took a step to relying on AI for our type one needs, being closed loop. However, we need to set the conditions for the AI: the sensitivity factor. Is there any way in the future for AI to micro analyze our bloods and make basal adjustment recommendations on that. So if there is a trend of going high at x time, it will recommend an increase in basal at x time. Or similarly, if there is a low after eating breakfast, reduce the insulin:carb ratio for breakfast?

Just curious, probably only a matter of time before some big corp jumps onto its back and capitalizes 🙂

 

[SB2015]

It is exciting to see how many improvements in what is available in the last few years.

The 780G closed loop bases it’s decisions hour by hour based on the data from the previous 72 hours. So there is an element of what you are looking for. None of the systems we have are perfect and, like when we manage it ourselves there are times when it doesn’t quite match what we need. Still a vast improvement on earlier systems.

 

[littlevoice359]

Personally I would be much more interested in research that seeks to find other measurable biological indicators related to actions by other bodily organs and/or the immune system that affect blood sugar. We need better predictors so we are not constantly reacting to what our body is doing due to factors into which we have no insight. Currently we are working with an incomplete information set, in my view.

 

[Bruce Stephens]

We need better predictors so we are not constantly reacting to what our body is doing due to factors into which we have no insight. Currently we are working with an incomplete information set, in my view.

Perhaps, but I'd worry that you'd spend lots of time and money trying to measure things which end up having much less influence than food and physical activity. I guess it's worth some effort in seeing if there's something useful there (perhaps there are signals some sensor could pick up indicating anticipation of eating, for example).

And anyway, OpenAPS was developed on the basis of connecting two things that already existed.

OpenAPS Reference Design – OpenAPS.org

 

[everydayupsanddowns]

I think the algorithms and sensors are just going to get better and more finely tuned to each individual as time goes on.

Sensor augmented pumps have already come a long way since the only interaction on the Medtronic Veo was to suspend insulin once you had already gone low.

I think future versions will make the current crop of hybrid closed loop pumps look very crude. 🙂