From the Glucosym Github repository at:
GlucoSym is an open-source simulator aimed to help in the testing and development of insulin delivery automation systems.
I’ve been migrating all of the simpancreas code to have quick access to all of the initial user settings making the algorithm constants and variables tweaking much easier.
While doing this process, I’ve been trying to get to the set of variables that can work out-of-the-box for all patient profiles on Glucosym.
I think I’ve gotten to that set of variables and I think that most of them can be set to be learnt from the analysis of the blood glucose data fed to the algorithm through time.
I used the following settings for all of the patient profiles on Glucosym:
— bg_objective: 100
— base_insulin_sensitivity: 20
— body_weight: 80 ( Kg )
— insulin_duration: 4 ( Hours )
— max_bolus_limit: 3.0 ( Units )
— max_iob_limit: 15 ( Units )
— Started with NO historical data.
— Ran simulation for 5 days ( 1440 hours ).
— Ran simulation assuming no food was given.
What about carbs?
Here’s a run for Patient A ( In my opinion one of the hardest ones ), with the following simulation characteristics:
— base_insulin_sensitivity: 40
— Ran starting with NO data.
— Ran for 5 days ( 1440 hours ).
— Ran assuming 20gr meals were given 3 times a day. Meals were configured for a 3 hour digestion time.
The simulation was restarted afterwards incrementing the carb amount to 40gr per meal and ran for an extra 5 days: