Description
In this project, we will develop a continuous meal detection method from glucose monitor data of a subject. You will be given CGM sensor outputs for a given person for nearly 6 months. The CGM output is every 5 mins. You will also be given ground truth for every meal instance. You will have to develop a continuous meal detection algorithm.
Task list
- Parse CGM data and synchronize meal ground truth with CGM data
- Write four algorithms for online detection of meal event
- Auto regression based modeling of CGM
- Algorithm development, SARIMA
- Algorithm Instantiation
- Auto regression based modeling of CGM
- Algorithm implementation
- Training and Testing accuracy
- Model based meal detection
- Algorithm development, HYMN based (ask IMPACT Lab for code)
- Algorithm Instantiation
- Algorithm implementation
- Training and Testing accuracy
- Kalman Filter based meal detection
- Algorithm development (Ask IMPACT Lab for algorithm)
- Algorithm Instantiation
- Algorithm implementation
- Training and Testing accuracy
- RNN based meal detection
- Algorithm development (Use Tensor Flow)
- Algorithm Instantiation
- Algorithm implementation
- Training and Testing accuracy
- Apply your algorithms to a new patient
- Provide Execution time analysis





