Machine Learning Techniques for Hypoglycemia Prediction: Trends and Challenges

Abstract

(1) Background: the use of machine learning techniques for the purpose of anticipating hypoglycemia has increased considerably in the past few years. Hypoglycemia is the drop in blood glucose below critical levels in diabetic patients. This may cause loss of cognitive ability, seizures, and in extreme cases, death. In almost half of all the severe cases, hypoglycemia arrives unannounced and is essentially asymptomatic. The inability of a diabetic patient to anticipate and intervene the occurrence of a hypoglycemic event often results in crisis. Hence, the prediction of hypoglycemia is a vital step in improving the life quality of a diabetic patient. The objective of this paper is to review work performed in the domain of hypoglycemia prediction by using machine learning and also to explore the latest trends and challenges that the researchers face in this area; (2) Methods: literature obtained from PubMed and Google Scholar was reviewed. Manuscripts from the last five years were searched for this purpose. A total of 903 papers were initially selected of which 57 papers were eventually shortlisted for detailed review; (3) Results: a thorough dissection of the shortlisted manuscripts provided an interesting split between the works based on two categories: hypoglycemia prediction and hypoglycemia detection. The entire review was carried out keeping this categorical distinction in perspective while providing a thorough overview of the machine learning approaches used to anticipate hypoglycemia, the type of training data, and the prediction horizon.

A Function Approximation Approach to the Prediction of Blood Glucose Levels

“…The majority of machine learning models (30% of those studied in the survey paper [9]) are based on artificial neural networks such as convolutional neural networks, recurrent neural networks and deep learning techniques. For example, in [17], the authors develop a deep learning model based on a dilated recurrent neural network to achieve 30-min blood glucose prediction.…”

“…It should be noted that the majority of BG prediction models (63.64% of those analyzed in [9]), including many of the references described above, are dependent on additional input features (such as meal intake, insulin dosage and physical activity) in addition to historical BG data. Since such data could be inaccurate and hard to obtain, we have intentionally made the design decision in our work to base our method on historical BG data only.…”

“…, s p (t j−d+1 ) for appropriate values of h and d. We took d 7 (a sampling horizon t j − t j−d+1 of 30 min has been suggested as the optimal one for BG prediction in [4,6]). We tested prediction horizons of 30 min (h 6) (the most common prediction horizon according to the analysis in [9]), 60 min (h 12) and 90 min (h 18).…”

“…A clinically reliable prediction of this nature is extremely useful in conjunction with other communication devices in avoiding unnecessary costs and dangers. For example, hypoglycemia may lead to loss of consciousness, confusion or seizures [9]. However, since the onset of hypoglycemic periods are often silent (without indicating symptoms), it can be very hard for a human to predict in advance.…”

“…In particular, machine learning has been used extensively for this purpose. We refer to [9] for a recent review of BG prediction methods based on machine learning, with specific focus on predicting and detecting hypoglycemia.…”