Meal Prediction using CGM Data

We investigate the use of blood glucose levels captured by continuous glucose monitoring (CGM) systems to automatically detect when a diabetic patient intakes a meal in daily life. This is important because it finds application in injecting patients with insulin autonomously prior to the meal. If this prediction is wrong, it can lead to either low or high glucose levels leading to health complications. Hence, we explore and implement two machine learning based approaches to infer future CGM values and predict meals taken by a patient. In this report, we demonstrate our meal detection algorithms and evaluate our results.