The MeDeT approach focuses on building, adapting, and operating high-fidelity digital twins (DTs) of medical devices, employing few-shot meta-learning techniques. These medical devices DTs are designed to streamline testing automation for healthcare IoT applications.
MeDeT works in six phases: (i) Data Generation - generates raw data for medical devices, (ii) Data Preparation - preprocesses raw data for training, (iii) Meta-learning - creates meta dataset & taskset, determines model architecture, and trains/fine-tunes with MAML algorithm, (iv) Build DTs - creates model clones, storage, APIs, and JSON objects, (v) DT Request Handler - processes requests from a healthcare IoT application during testing, and (vi) DTs to Device Communication - establishes DTs communication with medical devices.
This work is a part of the Welfare Technology Solution (WTT4Oslo) project (#309175) funded by the Research Council of Norway.
- Machine: minimum 16GB RAM and 8-core processor
- OS: MacOS or Windows 10
- IDE: PyCharm
- Python: 3.8 or higher
- PyTorch: 2.0.1
- learn2learn: 0.2.0
- scikit-learn: 1.3.0
- Pandas: 2.0.3
- Flask: 2.2.3
- Flask-RESTful: 0.3.9