APSDigitalTwin (Artificial Pancreas System Digital Twin) uses a predictive model to simulate a person with diabetes with the intervention of OpenAPS's oref0 algorithm. The system learns a model of the user from past obervational data and allows for this model to be run against different scenarios both with and without OpenAPS intervention.
This project was executed with oref0 version 0.7.1, python version 3.9.7 on the operating system Ubuntu 20.04.5 LTS.
This system requires OpenAPS oref0 to be installed on the commandline. Please check out the following link for instruction on how to do this: click here
Install a conda environment for APSDigitalTwin:
conda create --force -n aps-digital-twin
conda activate aps-digital-twin
pip install -r requirements.txt
This model requires a blood glucose (mmol/L), insulin on board (U), carbohydrates on board (g) and pump output rate (U/h) timeseries at 5 minute intervals to learn the model. This data should be presented in a csv with the following layout:
| bg | iob | cob | rate |
|---|---|---|---|
| 103 | 0.34 | 5.2 | 1.4 |
| 104 | 0.32 | 5.1 | 0 |
In scripts/main.py, modify training_data with a path your own training dataset. You should also update .env with the correct path for profile_path and basal_profile_path.
For windows users, COMSPEC should also be updated to point to the exe of the bash command line which has oref0 installed (eg: \User\GitBash.exe).
You may also update any scenarios as required. To then run the code:
python ./scripts/main.py
In each research question python file, the variable figure_save_path should be set to a path to save figures.
To execute the research questions, run the following commands:
These scripts can be run with the following commands:
python scripts/rq1_model_correctness.py
python scripts/rq2_person_glucose_dynamics.py
python scripts/rq3_openaps_scenarios.py
For main.py, after each scenario, the program will display two graphs representing the scenario with no OpenAPS intervention and the scenario with OpenAPS intervention every 5 minutes. Each scenario run will also return True or False depending if the scenario has less or more deviations outside of a safe blood glucose level.
For each research question script, figures wil be saved to the path represented by figure_save_path. Figures generated are specific to the research question in question.