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
1) Parse CGM data and synchronize meal ground truth with CGM data
2) Write four algorithms for online detection of meal event
a. Auto regression based modeling of CGM
i) Algorithm development, SARIMA
ii) Algorithm Instantiation
iii) Algorithm implementation
iv) Training and Testing accuracy
b. Model based meal detection
i) Algorithm development, HYMN based (ask IMPACT Lab for code)
ii) Algorithm Instantiation
iii) Algorithm implementation
iv) Training and Testing accuracy
c. Kalman Filter based meal detection
i) Algorithm development (Ask IMPACT Lab for algorithm)
ii) Algorithm Instantiation
iii) Algorithm implementation
iv) Training and Testing accuracy
d. RNN based meal detection
i) Algorithm development (Use Tensor Flow)
ii) Algorithm Instantiation
iii) Algorithm implementation
iv) Training and Testing accuracy
3. Apply your algorithms to a new patient
4. Provide Execution time analysis
