What is this?
GlucoTune is a toolkit that provides a standardized way to perform and share Type-1 Diabetes dataset preprocessing as well as baselines benchmarking across datasets, hardware, and model training configurations.
Why?
Preprocessing blood glucose time series data is a critical step in developing data-driven methods for diabetes management. The lack of standardized preprocessing pipelines and experimental baselines hampers reproducibility and makes it difficult to fairly compare and validate approaches in the literature. Moreover, sharing of the preprocessed data is commonly forbidden by the datasets usage agreements. GlucoTune’s goal is to support reproducible experimentation with blood glucose time-series data. It provides configurable preprocessing pipelines that ensure consistent data handling, support reliable experimental setups allowing the sharing of the preprocessing configuration for later replicability, and reduce experimental variability. Beyond preprocessing, GlucoTune also incorporates a model library that allows users to construct baseline models for downstream analysis by selecting from a curated collection of state-of-the-art methods for blood glucose prediction and time-series forecasting, further improving the comparability and replicability of experimental studies.
Included in GlucoTune
- 📊 Interactive results tables, take a look at the leaderboard
- ⚙️ Graphical user interface and command line tools
- 🔬 Reproducible experiments