JMIR Publications

ABSTRACT

Background:

Diabetes mellitus is a chronic metabolic disorder that results in abnormal blood glucose regulations. Blood glucose level is preferably maintained close to normality through self-management practices, which involves actively tracking blood glucose levels and taking proper actions including adjusting diet and insulin medications. Blood glucose anomalies could be defined as any undesirable reading either due to a precisely known reason (normal cause variation) or unknown reason (special cause variation) to the patient. Recently, machine learning applications have been widely introduced within the diabetes research in general and blood glucose anomalies detection in particular. However, irrespective of their expanding and increasing popularity, there is lack of updated reviews that materialize the current trends in modelling options and strategies for blood glucose anomalies classification and detection in people with diabetes.

Objective:

The objective of this review is to identify, assess and analyze the state-of-the-art machine learning strategies and its hybrid systems focusing on blood glucose anomalies classification and detection including glycemic variability, hyperglycemia, and hypoglycemia in type 1 diabetes within the context of personalized decision support systems and blood glucose alarm events applications, which are important constituents for optimal diabetes self-management.

Methods:

A rigorous literature was conducted between September 1 and October 1, 2017, through various online databases including Google scholar, PubMed, ScienceDirect and others. Peer reviewed journals and articles were considered. Relevant articles were first identified by reviewing the title, keywords, and abstracts as a preliminary filter with our selection criteria, and then reviewed the full text articles that fulfilled the inclusion criteria. Information from the selected literature was extracted based on some predefined categories, which were based on previous research and further elaborated through brainstorming.

Results:

The initial hit was vetted using the title, abstract, and keywords, and retrieved a total of 496 papers. After a thorough assessment and screening 47 articles were left, which were critically analyzed. The inter-rater agreement was measured using Cohen Kappa test, and disagreements were resolved through discussion. The state-of-the-art classes of machine learning have been developed and tested up to the task and achieved promising performance including artificial neural network, support vector machine, decision tree, genetic algorithm, Gaussian process regression, Bayesian neural network, deep belief network, and others.

Conclusions:

Despite the complexity of blood glucose dynamics, there are many attempts to capture hypoglycemia, hyperglycemia incidences and the extent of an individual’s glycemic variability using different approaches. Recently, advancement of diabetes technologies and continuous accumulation of self-collected health data has paved the way for popularity of machine learning in these tasks. According to the review, most of the identified studies used a theoretical threshold, which suffers from inter and intra-patient variation. Therefore, future studies should consider the difference among patients and also track its temporal change overtime. Moreover, studies should also give more emphasis on the types of inputs used and its associated time lag. Generally, we foresee these developments might encourage researchers to further develop and test these systems on a large-scale basis.