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

2nd
Feb 2015
Darrell Wright
@beached
Feb 02 2015 01:37
If there is anything I can code up, I'm willing. I've been tackling a different problem with the aggregate data to get enough info to create basal test like data from the periods of time where a bolus/food/exercise/temp basal event has not happened for more than 4hrs. It's not perfect but allowed me, code is broken right now, to make much more educated basal changes without consciously doing a basal test. I'm a c++ programmer, but python/js isn't really a far stretch either.
Scott Leibrand
@scottleibrand
Feb 02 2015 02:30
@beached: would love to see you do a parameter estimation to determine ISF, IC ratio, and carb absorption rate based on data from people running wip/iob-cob and entering insulin and carb treatments.
We've done it manually, but would love to help someone code up an automated version.
Dana Lewis
@danamlewis
Feb 02 2015 02:31
Also need someone to fix WIP/IOB-cob so it supports a DIA different than 3 :)
Ben West
@bewest
Feb 02 2015 02:32
also some auth and UI to control it
Darrell Wright
@beached
Feb 02 2015 02:32
One thing I have been thinking about, excuse if I am ignorant of the reality, but have not codified is that a lot of decisions seem to be made based on IoB and not an estimated active insulin(i.e. Relate the peak of a basal and/or bolus to glucose clearing). Not sure how to approach it yet, but it seems to be closer to what I would want
diabeticgonewild
@diabeticgonewild
Feb 02 2015 02:34
Will try to @danamlewis . I sure hope the Seahawks win, especially for my 90 year old grandma, who lives there who is in rehab right now. The 12th man flag is flying on top of the Space Needle, and my grandma has a view of it from her room in rehab. Here is a pic (pretty!) http://imgur.com/fsb8aEj&VHEehCw#0
Darrell Wright
@beached
Feb 02 2015 02:34

It could probably be estimated by a pure correction bolus and tracking the rate of change. Aggregate that and it might be meaningful.

But all of this is offline, not online processing

Dana Lewis
@danamlewis
Feb 02 2015 02:34
@beached Have you looked at IOB-cob? Purple prediction lines show inflection points of where insulin and carbs peak. That's where active insulin is visualized, not just IOB which is stacked
Scott Leibrand
@scottleibrand
Feb 02 2015 02:40
Yeah, sounds like wip/iob-cob does the kind of prediction you're talking about. And calculates retrospective predictions. you could optimize parameters to minimize RMS of the divergence of 1h retrospective predictions from actuals.
diabeticgonewild
@diabeticgonewild
Feb 02 2015 02:43
interesting @scottliebrand
diabeticgonewild
@diabeticgonewild
Feb 02 2015 03:00
We lost, no!
Darrell Wright
@beached
Feb 02 2015 03:03
@scottleibrand did you notice any differences regarding time of day with carb absorption.
Dana Lewis
@danamlewis
Feb 02 2015 03:06
@beached no. But this is just me as n=1. So hoping more ppl test and validate. @scottleibrand
Been using carb absorption rate consistently for over a year, haven't noticed a difference for time of day or meal makeup.
Jason Calabrese
@jasoncalabrese
Feb 02 2015 03:15
I want to get IOB into the mainline as an enableable option, all the refactoring and cleanup I want to do is going to be hell if we can't get the mainline and iob-cob the be more in sync
seems like it can all be options
Ben West
@bewest
Feb 02 2015 03:17
:+1:
Jason Calabrese
@jasoncalabrese
Feb 02 2015 03:19
after I get raw wrapped up I'm going to focus on IOB and plan to use https://www.npmjs.org/package/browserify-express to share code between client/server
Ben West
@bewest
Feb 02 2015 03:20
double :+1: :+1:
Jason Calabrese
@jasoncalabrese
Feb 02 2015 03:20
with that we can start pulling client.js apart and get it tested
Ben West
@bewest
Feb 02 2015 03:20
:+1: * :100:
that's how I've been hoping to do it
Jason Calabrese
@jasoncalabrese
Feb 02 2015 03:21
did you see the coverage on nightscout/cgm-remote-monitor#386
over 70%
Ben West
@bewest
Feb 02 2015 03:21
there's node-trumpet and node-harmonize/harmony or osmething with do interesting things
that substack guy has several super-effecient ways for re-using stuff between server/client
testling can run tests in browser, very nifty
Jason Calabrese
@jasoncalabrese
Feb 02 2015 03:22
got merged early, don't want to make a mess with undo/redo, etc
Ben West
@bewest
Feb 02 2015 03:22
oh wow
last time i looked you had just gotten frustrated
hehe
Jason Calabrese
@jasoncalabrese
Feb 02 2015 03:22
still frustrated
I want it all now
Ben West
@bewest
Feb 02 2015 03:25
yeah
crap, I was just low and binged on half box of honety nut cheerios
Toby Canning
@TC2013
Feb 02 2015 12:39
I have a paper that will help with the finding insulin sensitivity from CGM readings, but it requires insulin basal bolus data from ones pump
Toby Canning
@TC2013
Feb 02 2015 13:18
@beached Could you take a look at this? http://care.diabetesjournals.org/content/37/5/1216.full.pdf+html I have hired a mathematician programmer to begin work on this, but she has no knowledge of diabetes, so our first work has all been to get her up to speed. Right now she is just going to work on programming in C/C++ a program that can calculate insulin sensitivity straight from the Nightscout database (mongo), but as I mentioned before I have to get pump history data on insulin delivery into mongo first. Jesus (and others) may have already solved this problem, but I haven't look through their work yet.
Darrell Wright
@beached
Feb 02 2015 13:55
I'll start reading that, I wanted to look into an uploader that takes my csv files from Medtronic and uses that as current data so I can better test.
one thing I noticed is that the bower install will silently fail if you happen to run the npm install as root. Bad thing to do anyway, but might warrant a note somewhere
Darrell Wright
@beached
Feb 02 2015 14:54
Where was this paper a year ago. Wicked. It wasn't around yet, but this is awesome
Tyler Rhodes
@trhodeos
Feb 02 2015 16:15
@TC2013 stay tuned, medtronic support via carelink is coming up for the nightscout uploader :)
Dana Lewis
@danamlewis
Feb 02 2015 17:54
@tc2013 cool! Have you looked at the retrospective predictions in iob-cob? Could be a potential starting point and/or helpful for her to model the changed sensitivities to see if they work better than the ones currently set.
Toby Canning
@TC2013
Feb 02 2015 18:31
@danamlewis Yes, the iob-cob stuff is great. I hope to be able to modify it to support a dynamic profile, like adjusting Duration of Insulin Activity and time of day issues (like insulin resistance in the am and changing basal rates).
@trhodeos That is awesome! Rough ETA? 1 month? 2 months?...
Tyler Rhodes
@trhodeos
Feb 02 2015 18:41
Not sure. Depends on how busy @ktind is :)
Ross Naylor
@rnpenguin
Feb 02 2015 19:22
:)
Toby Canning
@TC2013
Feb 02 2015 19:49
That is exciting!
I will go edit that password now :) LOL
diabeticgonewild
@diabeticgonewild
Feb 02 2015 21:51
Finally was able to fix a programming issue that has been bothering me for days. It's hard stuff though. And I got R and WinBUGS to interface nicely with MATLAB!