#simPancreas updates March 2015

Account of changes to the project that led to today’s great test results.


I am going to start this post saying how happy I am with today’s test results. I would like to break down the post in 2 parts: The detail explanation of the test based on the above graph & the improvements I made to simPancreas in order to achieve the above graph.

I. Explanation of test based on graph

( a ) The test started around 12:30pm today. I was at 103, just had recovered from a small morning meal.

( b ) Just after starting the test, I decided to do something I had never done before. I ate a pack of Oreos, because what other reason would I have for wearing an artificial pancreas??? … 6 chocolate cookies filled with vanilla cream ( Hey, it really is something for a T1D to do this and see what happened next ). As expected, BG started going up and got up to 114 ( peak ) then back down to 81.

( c ) I had a meal at about 3:00pm which was loaded with about 110 grams of carbs + fat. At 4:11 I peaked at 105 and stayed there until about 5:45pm. simPancreas shutdown after that.

( d ) I started to descend until 68. I am almost 100% sure it was due to a higher set basal rate at that time in the afternoon. Need to test more and adjust.

( e ) I started going up again at about 7:45pm and simPancreas started kicking in again until I peaked at 113 to go back down to around 80, the point where I am right now 🙂

II. Improvements to simPancreas

Now to the interesting part, how the hell was simPancreas able to manage that?

The first important change I did on simPancreas was changing temp basals to boluses for dosing. With temp basals, you get 2 risks: Not knowing exactly how much insulin was given and getting stuck at a high basal rate and not being able to cancel it.

With boluses, you know exactly how much insulin was given and if the system connection is lost, no other boluses are given.

The second important change I did was to change aggressiveness of recommended rate based on acceleration. The simplest way to show this is with an example:


If blood glucose is 86 and acceleration is positive small,
increase recommended rate by 1.5
== If recommended rate was 0.2, increase to 0.3

If blood glucose is 106 and acceleration is positive normal,
increase recommended rate by 2.0
== If recommended rate was 0.4, increase to 0.8

If blood glucose is 126 and acceleration is positive normal,
increase recommended rate by 3.0
== If recommended rate was 0.5, increase to 1.5


This adjustment basically increments boluses so the resulting effect on blood glucose is seen much earlier, helping me maintain blood glucose at an adequate level.

Here are graphs for part of the treatment today.

See how in the first part, small increments in bg generate adjustments with micro boluses.

Also note how in the second part, “large” increases in bg generate larger boluses that completely stop when leveling off.

Screen Shot 2015-03-15 at 23.15.58

Screen Shot 2015-03-15 at 23.16.23

Screen Shot 2015-03-15 at 23.16.15

Next, I will continue testing and doing micro adjustments until I feel simPancreas works exactly as I want it to.