Skip to content

aptash/AndroidAPS-DynISF

 
 

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

AndroidAPS

Support Server

Build status Crowdin Documentation Status codecov dev: codecov

BTC 3KawK8aQe48478s6fxJ8Ms6VTWkwjgr9f2

Dynamic ISF

To access Dynamic ISF, please use the dev branch. Dynamic ISF is accessed via a plugin in the config builder that needs to be enabled.

Please note that this is experimental code and it is recommended that you are an experienced AndroidAPS user before embarking on use. Additionally, it is not recommended that this is used with children.

Pre-installation notes

Please note that to use Dynamic ISF effectively, the AndroidAPS database needs a minimum of five days of data. If you are installing version 3 over the top of 2.8 or on a new phone, it's best to run the standard master branch first for those five days. Dynamic ISF will work without the full set of data in the database, as it uses an estimate of TDD based on the profile basal rates, then whatever the database has once data is populated, but it will be more stable with five days of data.

Fresh installations of v3 crashing

As mentioned in the last paragraph, if you are installing v3 over v2.8, or fresh on a new phone, the database will be empty. This creates an issue with the SMB plugin on all versions of AndroidAPS, where attempting to use SMB causes the app to crash and restart. To resolve this, you will need to run on AMA for a few hours so that SMB has some data to use for Autosens. SMB will work once this is populated, however, as recommended above, for best results, the database should be populated with five days of TDD data.

Changes compared to standard AndroidAPS

Dynamic ISF uses Chris WIlson's model to determine ISF instead of a static profile settings. This is applied only in the openAPSSMB plugin. The openAPSAMA plugin continues to use profile settings for ISF.

The equation implemented is: ISF = 277700 / ( BG * TDD )

The implementation here splits the use into two. One to calculate current ISF for use with predictions, the other to calculate future ISF for use with dosing.

Current ISF

This uses a combination of the 7 day average TDD and a linear extrapolation of the pumps current total insulin delivered to calculate a TDD for today.

The total daily dose used in the above equation is generally weighted 40% to the 7 day average and 60% to the extrapolation from the pump data. There are some special cases where this isn't applied. These are:

  • When the time is earlier than 5am and the pump TDD is greater than 7 day TDD
  • When the time is earlier than 7am and the pump TDD is less than 80% of 7 day TDD

In both these cases, a value of 80% of TDD is used in the caluclation.

When the pump TDD is less than 33% of 7 day TDD

In this case, the weighting is changed to be 75% pump TDD, 25% 7 day average TDD.

The dynamic ISF value is still adjusted when setting an activity or eating soon temp target, as per the standard oref model.

Future ISF

Future ISF uses the same TDD value as generated above. It then uses different glucose values dependent on the case:

If delta is +ve or zero, or delta is negative but predicted glucose is above target, the future ISF value used for determining insulin required is the same as the current ISF.

If delta is -ve, and the predicted glucose level is below target, then the future ISF value is used, as this is a less aggressive value.

Dynamic ISF Adjustment Factor

The 1.5 version of the code includes an adjustment factor that allows the user to specify a value between 1% and 300%. This acts as a multiplier on the TDD value and results in the ISF values becoming smaller (ie more insulin required to move glucose levels a small amount) as the value is increased above 100% and larger (ie less insulin required to move glucose levels a small amount) as the value is decreased below 100%.

Autosens and sensitivity ratio

In this version of the code, the autosens value no longer uses the traditional oref1 deviation based model and instead uses rolling 24 hour TDD / 7-day average TDD. This is used to adjust basal and targets when the options are selected in preferences.

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Packages

No packages published

Languages

  • Kotlin 72.4%
  • Java 26.5%
  • JavaScript 1.1%