Home

autoIFS 3.0 Quick User Guide

DISCLAMER

I am not a medically trained professional, and the methods I present here are derived solely from trial and numerical experimentation. For instance, I did not utilize a mathematical model to assess the reaction kinetics of free fatty acids, a factor contributing to temporary insulin resistance. Should you choose to experiment with these methods on your own, please be aware that you do so at your own risk and assume full liability for any consequences.

PREFACE

This document serves to delineate the features of autoISF along with their corresponding settings. Specific tuning guidance is provided in the "How-to-..." guides currently in development or planning stages. These tuning recommendations are individualized and contingent upon your intended usage scenario for autoISF, such as testing, fine-tuning AAPS, adhering to a low-carb diet, engaging in extensive endurance exercise, or implementing a full loop with no carb announcements and manually triggered boluses.

In consolidating previous Quick Guides from earlier autoISF versions, this comprehensive document also introduces the latest features. Noteworthy enhancements in comparison to version 2.2.8.2 include:

• Internal handling of the IOB threshold method, minimizing overshoots, and automatically adapting to sensitivity changes, especially during exercise mode.
• Incorporation of specific automations featuring actions and triggers tailored to autoISF.
• Introduction of an activity monitor utilizing step counting as a more subdued form of exercise mode.

In version 2.2.8.2, a potential deficiency was identified in situations where glucose levels were decreasing, and glucose acceleration was already positive. This circumstance allowed for the extrapolation of a minimum glucose level, which, if projected to be below the target and expected within 15 minutes, warranted caution against strengthening ISF. Consequently, the parameter bgBrake_ISF_weight is now employed instead of bgAccel_ISF_weight in such instances. It is essential to note that these situations were infrequent and, upon closer examination, less critical than initially anticipated. This is attributed to the fact that, in most cases, predictions fell below their threshold, leading to the disabling of SMB functionalities.

Hint for users with Libre 1-minute CGM

In AAPS version 3.2.0.2, the glucose history utilizes data at 5-minute intervals, disregarding the values in between. This approach can lead to awkward consequences for both the graph display and the exponential smoothing method. With each new reading received every minute, the 5-minute pattern shifts forward by 1 minute, incorporating an entirely fresh subset of glucose values and times.

Currently, alternative methods for determining glucose acceleration based on the 1-minute data are still under evaluation. Once a promising method is identified and validated, an interim release will be made available.

As a Libre user, you have three alternatives:

1. AAPS can retrieve values from xDrip+, where you can smooth and reduce the 1-minute data to 5-minute data before sending it to AAPS.
2. Stick to the 5-minute mode in your Libre setup, if possible.
3. AAPS can obtain values from Juggluco every minute, but it will use only the 5-minute subset for looping and parabola fitting.