JMIR Publications

ABSTRACT

Background:

Hundred years since the invention of insulin, diabetes management has made amazing progress. But cases and prevalence have steadily increased over the last few decades with 1.5 million deaths reported in 2012 alone. Traditionally analyzing diabetes patients remains a largely invasive approach. Wearable devices (WD) make use of sensors historically reserved for hospital settings. WDs coupled with AI algorithms provide real promise to help understand and conclude meaningful information from the gathered data and provide advanced and clinically meaningful analytics.

Objective:

To provide an overview of AI driven WD features for Diabetes and their usage in monitoring diabetes related parameters

Methods:

We searched seven of the most popular bibliographic databases using 3 groups of search terms related to diabetes, WDs, and AI. A two stage process was followed for study selection: reading abstract and titles, followed by full text screening. Two reviewers' performed study selection and data extraction independently , finally disagreements were resolved by consensus. A narrative approach was used to synthesize the data

Results:

From an initial 3,872 studies, we report the features from 37 studies post-filtering according to our pre-defined inclusion criteria. Most of the studies targeted T1D, T2D or both (57%, 21/37). Many studies reported Blood glucose 41% (15/37) as their main measurement. More than half of the studies (57%, 21/37) had the aim of estimation/prediction of glucose or glucose level monitoring. More than half were wrist worn devices. Only 41% of the study devices were commercially available. We observed the use of multiple sensors with Photoplethysmography (PPG) sensor being most prevalent in 32% (12/37) of studies. Studies reported and compared more than one ML model with high levels of accuracy. Support vector machine (SVM) was the most reported 35% (13/37), followed by Random Forest (RF) 32% (12/37).

Conclusions:

This review has provided the most extensive work to date summarizing WDs that utilize ML for diabetics and will provide research direction to those wanting to further provide meaningful contribution to this emerging field. Given the advancements in WD technologies replacing the need for invasive hospital setting devices, we see great advancement potential in this domain. Further work is needed to validate the ML approaches on clinical data from WDs and provide further meaningful analytics which could serve as data gathering, monitoring, prediction, classification and recommendation devices in the context of diabetes