These are chat archives for nightscout/intend-to-bolus

26th
Feb 2015
diabeticgonewild
@diabeticgonewild
Feb 26 2015 00:59
Simulink (MATLAB extension) makes life interesting. Obviously I don't like change! :(
diabeticgonewild
@diabeticgonewild
Feb 26 2015 02:39
@bewest , somebody is selling a used Veo pump (754) in the United States. I can give you the link.
Jason Calabrese
@jasoncalabrese
Feb 26 2015 06:31
just merged nightscout/cgm-remote-monitor#436 thanks @mgranberry
we should let everyone know that they can now experiment with different durations
diabeticgonewild
@diabeticgonewild
Feb 26 2015 06:59
+1
Scott Leibrand
@scottleibrand
Feb 26 2015 07:42
Yay! Thx to you both!
@GarthDB has made a basic iPad UI for care portal data entry:
diabeticgonewild
@diabeticgonewild
Feb 26 2015 11:45
So, I stayed up all night and I worked on this stuff for at least 10 hours straight. I got like 8 out of 10 differential equations set up to where they can be solved nicely in Simulink, but of course the last two consist of some crazy math/chemistry equilibrium process using Michaelis–Menten equations (at least the paper mentioned it so I am not freaking out on what to do!). I have to solve for some crazy stuff in the steady state, kind of described here: http://www-jmg.ch.cam.ac.uk/tools/magnus/michmenten.html . I'm not that good at chemistry or thermodynamics, either. When I finish this crazy stuff, I will be at least halfway done with Hovorka mathematical modeling, which is by far the most challenging part of my entire project. The only challenging stuff I have left for Hovorka's work after the Simulink DiffEQ stuff is intrinsic Gaussian conditional autoregression with random walks for 3 probability distributions (not bad) and Markov chain Monte Carlo, with rejection Gibbs sampling for nonlinear hierarchical stochastic differential equations. So, I am almost out of the woods, when it comes to completing the really difficult stuff. After that, unless the nonlinear model predictive controller turns out to be a challenge, the only issues I will face are programming hurdles and logistic issues.
diabeticgonewild
@diabeticgonewild
Feb 26 2015 18:57
Yeah this is absolutely crazy but 3/4 doctors that I see thought it was lupus...and I get joint pain, lymph node pain, and chills, among other thing...guess what!? It's not lupus. The lupus anticoagulant antibody was a false positive due to being on blood thinners and I was likely ANA positive due to receiving antibodies from over 1000 donors.
however chemo as immunosuppression is likely in my future and other more aggressive treatment but rheumatologist gets to manage it instead of my neurologist
I can't believe it!
I was anti-dsDNA antibody negative which is highly specific for lupus if you're positive
Matthias Granberry
@mgranberry
Feb 26 2015 19:00
hmm. That's good news except, of course, for the symptoms you're having.
diabeticgonewild
@diabeticgonewild
Feb 26 2015 19:03
Yeah it's impossible to tell what is going on with me medically because I have so much stuff going on. I have to rely on my doctors and a bunch of tests in order to get diagnosed with things.like a bunch
Like I never really feel well, but I am really happy right now
Jason Calabrese
@jasoncalabrese
Feb 26 2015 19:25
if anyone here isn't on FB you may have missed this: https://www.facebook.com/groups/cgminthecloud/716400681814524
lots of people starting to get IOB on their pebbles
@mgranberry we'll need to update pebble.js to use profile.dia, I may work on that tonight if no one does it first
diabeticgonewild
@diabeticgonewild
Feb 26 2015 19:44
Awesome! @jasoncalabrese
Matthias Granberry
@mgranberry
Feb 26 2015 20:13
@jasoncalabrese That doesn't calculate insulin activity, right? Then it's probably fine. I think when I scaled the IOB curve I didn't consider that the activity calculations weren't operating as a function of IOB and thus they are probably off by a factor dia/3. I'll need to take a look at what they are actually doing when I get home and make sure.
diabeticgonewild
@diabeticgonewild
Feb 26 2015 20:18
You could just take the Medtronic curves for the 530G from the Excel sheet provided on the OpenAPS group, as that has a really nice fit. It's already been scaled, so the normalized variable x, which is to be applied in the polynomial equation x = (time elapsed - mean of DIA time)/(standard deviation of DIA time). Easy peasy.
obviously you would have vectors of DIA time, mean of DIA time, standard deviation of DIA time, and the same indice location of DIA would be the corresponding index of mean of DIA time and standard deviation of DIA, as they are all the same length
Matthias Granberry
@mgranberry
Feb 26 2015 20:26
@diabeticgonewild that's what I've done on my own work. I just hate ripping out code that other people are depending on for insulin and replacing it with something completely different.
diabeticgonewild
@diabeticgonewild
Feb 26 2015 20:28
Oh, got it.
Very understandable.
Jason Calabrese
@jasoncalabrese
Feb 26 2015 20:38
when I tested I only compared to dia: 3 on an other site and saw them match
then changes up and down and saw something different in the right direction
Matthias Granberry
@mgranberry
Feb 26 2015 20:45
It matches my pump pretty well, although they obviously differ in some places. The Tandem curve doesn't go to 0 for starters.