This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in the Journal of Medical Internet Research, is properly cited. The complete bibliographic information, a link to the original publication on https://www.jmir.org/, as well as this copyright and license information must be included.
Prevalence of diabetes has steadily increased over the last few decades with 1.5 million deaths reported in 2012 alone. Traditionally, analyzing patients with diabetes has remained a largely invasive approach. Wearable devices (WDs) make use of sensors historically reserved for hospital settings. WDs coupled with artificial intelligence (AI) algorithms show promise to help understand and conclude meaningful information from the gathered data and provide advanced and clinically meaningful analytics.
This review aimed to provide an overview of AI-driven WD features for diabetes and their use in monitoring diabetes-related parameters.
We searched 7 of the most popular bibliographic databases using 3 groups of search terms related to diabetes, WDs, and AI. A 2-stage process was followed for study selection: reading abstracts and titles followed by full-text screening. Two reviewers independently performed study selection and data extraction, and disagreements were resolved by consensus. A narrative approach was used to synthesize the data.
From an initial 3872 studies, we report the features from 37 studies post filtering according to our predefined inclusion criteria. Most of the studies targeted type 1 diabetes, type 2 diabetes, or both (21/37, 57%). Many studies (15/37, 41%) reported blood glucose as their main measurement. More than half of the studies (21/37, 57%) had the aim of estimation and prediction of glucose or glucose level monitoring. Over half of the reviewed studies looked at wrist-worn devices. Only 41% of the study devices were commercially available. We observed the use of multiple sensors with photoplethysmography sensors being most prevalent in 32% (12/37) of studies. Studies reported and compared >1 machine learning (ML) model with high levels of accuracy. Support vector machine was the most reported (13/37, 35%), followed by random forest (12/37, 32%).
This review is the most extensive work, to date, summarizing WDs that use ML for people with diabetes, and provides research direction to those wanting to further contribute 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 meaningful analytics that could serve as data gathering, monitoring, prediction, classification, and recommendation devices in the context of diabetes.
Diabetes, also known as diabetes mellitus, is a metabolic disease characterized by elevated blood glucose levels, which can ultimately result in many complications such as heart attack, stroke, kidney failure, leg amputation, vision loss, and nerve damage [
Despite the advancements in blood glucose monitoring techniques, the mainstream detection technology remains largely invasive. The commonly used home electronic glucometers involve people with diabetes invasively self-pricking to draw blood from fingertips, opening them up to infections as well as stress and pain caused by the procedure that is often expected multiple times a day.
The availability and advancements of smart devices, such as smartphones, have made the monitoring of diabetes-related features more accessible. Many studies have examined this much welcomed technology [
Artificial intelligence (AI) is a broader term that encompasses machine learning (ML). Technically, ML is a subset of AI, often loosely used interchangeable buzzwords. As a high-level definition, AI is anything related to making machines smarter (eg, computational search algorithms). ML, on the other hand, is an AI system that can self-learn via an algorithm, and as a result, such a system becomes smarter without human intervention over time (eg, classifying an outcome) [
Many studies have been conducted on AI-based WDs for diabetes. Exploring the features of AI-based WDs reported in these studies is important for developers, patients, health care providers, and researchers to identify the recent advances and challenges in this area. Although several reviews were conducted in this area, (1) they were focused on smartphones and AI for diabetes [
This scoping review was carried out to satisfy this study’s goals of exploring features of AI-driven wearable technologies for diabetes. In order to construct a complete scoping review, the
The article search for this review began by identifying all relevant studies using 7 electronic databases: MEDLINE, PsycINFO, EMBASE, IEEE Xplore, ACM Digital Library, Web of Science, and Google Scholar. We scanned the first 100 hits retrieved by searching Google Scholar. The reason being Google Scholar typically returns several items that are sorted by relevance to the search topic. Bibliographic collection was conducted from October 25 to October 30, 2021. The reference lists of the included articles were then searched for additional sources. We also checked relevant articles that cited the included studies using Google Scholar’s “cited by” tool (forward reference list checking).
A number of different sets of keywords were designed to search databases depending on each database’s search term limit; as IEEE and Google Scholar have term limits, search queries were truncated based on the required limit. We considered the research topics included in the database to complete our search queries. We combined
Studies were chosen based on the criteria in
Publications that are in the English language.
Peer-reviewed articles including proposals.
Population with or suspected to have diabetes. No restrictions regarding their age, gender, and ethnicity.
Commercial, medical, or prototypes but with condition wearable device and uses artificial intelligence (AI).
Wearable usable by individual person not with help of clinical staff or plugged in to hospital setting.
Wearables using methods for diabetes analysis are to be noninvasive.
Any study that does not contain AI as an intervention.
People with other diseases, health care providers, and caregivers as population.
Not a wearable device (example artificial implant or body infused).
Studies opting statistical measures only, for analysis of collected data.
Sensors or tracking devices infused inside a person’s body.
Wearable devices that need professional sittings or hospital sittings.
This review’s studies were selected in 2 steps. In the first stage, 2 reviewers (AA and SA) independently reviewed the titles and abstracts of all retrieved papers. In the second phase, the same reviewers individually read the whole texts of the papers included in the first step. Rayyan (Qatar Computing Research Institute, Hamad Bin Khalifa University) [
AA and SA constructed the data extraction form, as shown in
SA synthesized the extracted data using the narrative approach, aggregating the data using tables and text and nonstatistical techniques. For being more precise, we presented the search results followed by general features of the studies, finally describing characteristics of the WDs and AI technologies. We described the general features of WDs (eg, device placement, type, and operating system [OS]) and their technical features (ie, features of sensors, such as sensors used, sensing approach, and primary measurements). The AI features were addressed based on the models used, the evaluation metrics, and their applications.
Having searched 7 bibliographic databases, this study returned 3872 citations. As shown in
PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) flow chart of the study selection process. EC: exclusion criteria; IC: inclusion criteria.
General features of included studies (n=37).
Features | Studies, n (%) | Study ID | |
|
|||
|
2019 | 10 (27) | S4, S8, S10, S16, S18, S20, S21, S24, S29, S30 |
|
2020 | 9 (24) | S3, S7, S11, S13, S15, S17, S19, S22, S35 |
|
2021 | 8 (22) | S5, S9, S12, S14, S25, S27, S28, S33 |
|
2018 | 6 (16) | S1, S6, S23, S34, S36, S37 |
|
2017 | 3 (8) | S2, S26, S32 |
|
2016 | 1 (3) | S31 |
|
|||
|
IEEE | 21 (57) | S1, S3, S5, S9-S11, S13-S18, S20, S24, S26, S28, S29, S31, S32, S36, S37 |
|
Elsevier | 3 (8) | S2, S12, S22 |
|
MDPI | 3 (8) | S6-S8 |
|
ACM | 2 (5) | S21, S35 |
|
Other (JMIR, IET, ICST, Confluence, BMJ Publishing Group, SPIE, Telemedicine and e-Health, SAGE) | 8a (22) | S4, S19, S23, S25, S27, S30, S33, S34 |
|
|||
|
United States | 7 (19) | S13, S14, S21, S27, S30, S31, S34 |
|
China | 5 (14) | S5, S15, S18, S19, S37 |
|
India | 5 (14) | S12, S17, S23, S25, S32 |
|
Pakistan | 2 (5) | S3, S20 |
|
Switzerland | 2 (5) | S6, S35 |
|
Bangladesh | 2 (5) | S10, S24 |
|
Other (Korea, Colombia, Canada, Morocco, Mexico, Italy, Macedonia, Sri Lanka, United Kingdom, Russia, Taiwan, Philippines, Saudi Arabia, Germany) | 14b (38) | S1, S2, S4, S7, S8, S9, S11, S16, S22, S26, S28, S29, S33, S36 |
|
|||
|
Journal articles | 26 (70) | S1-S20, S22, S23, S27, S28, S33, S34 |
|
Conference proceedings | 11 (30) | S21, S24-S26, S20-S32, S35-S37 |
|
|||
|
Both T1Dc and T2Dd | 9 (24) | S1, S5, S6, S8, S10, S11, S14, S24, S36 |
|
T2D | 7 (19) | S2-S4, S15, S16, S21, S29 |
|
T1D | 5 (14) | S13, S22, S30, S34, S35 |
|
T1D, T2D, and prediabetes | 2 (5) | S12, S25 |
|
T1D and prediabetes | 1 (3) | S17 |
|
Prediabetes | 1 (3) | S27 |
|
Not specified | 12 (32) | S7, S9, S18, S19, S20, S23, S26, S28, S31, S32, S33, S37 |
a1 study for each publication.
b1 study for each country.
cT1D: type 1 diabetes.
dT2D: type 2 diabetes.
Study design features (n=37).
Features | Studies, n (%) | Study ID | ||||
|
||||||
|
Blood glucose estimation (predictions) | 10 (27) | S3, S18, S19, S21, S23, S25, S27, S28, S32, S33, S37 | |||
|
Glucose level monitoring | 10 (27) | S7-S11, S15, S20, S24, S26, S30 | |||
|
Diagnostic solution | 5 (14) | S4, S29, S33, S34, S35 | |||
|
Diabetes classification | 4 (11) | S12-S14, S17 | |||
|
Self-administration and monitoring | 4 (11) | S1, S5, S6, S31 | |||
|
Prevention | 2 (5) | S2, S16 | |||
|
Other disease predictions, detection, and monitoring (hypoglycemia and foot temperature) | 2 (5) | S22, S36 | |||
|
||||||
|
Not mentioned | 31 (84) | S1-S22, S24-S26, S29, S30, S34-S37 | |||
|
Mentioned | 6 (16) | S23, S27, S28, S31, S32, S33 | |||
|
||||||
|
Private | 25 (68) | S2, S3, S5, S7, S8-S19, S21, S22, S24, S26, S29, S34-S37 | |||
|
Public | 4 (11) | S1, S4, S6, S25 | |||
|
Not mentioned | 2 (5) | S20, S30 | |||
|
||||||
|
|
|||||
|
|
Children and young adults (≤18) | 1 (3) | S8 | ||
|
|
Adult (19-65) | 18 (49) | S2-S5, S8, S10, S13, S15, S16, S17, S19, S21, S22, S27, S29, S31, S33, S34 | ||
|
|
Older adult (>65) | 6 (16) | S2, S4, S15, S21, S22, S33 | ||
|
|
Not mentioned | 19 (51) | S1, S6, S7, S9, S11, S12, S14, S18, S20, S23-S26, S28, S30, S32, S35-S37 | ||
|
|
|||||
|
|
Male | 10 (27) | S2, S3, S5, S13, S15, S17, S18, S27, S29, S34 | ||
|
|
Female | 10 (27) | S2, S3, S5, S13, S15, S17, S18, S27, S29, S34 | ||
|
|
Not mentioned | 27 (73) | S1, S4, S6-S12, S14, S16, S19-S26, S28, S30-S33, S35-S37 | ||
|
|
|||||
|
|
Yes | 14 (38) | S1, S4, S5-S7, S10, S12, S14, S15, S18, S19, S21, S27, S34, S36 | ||
|
|
No | 15 (41) | S1, S5, S6, S8-S10, S12, S14, S18, S19, S27, S29, S31, S33, S36 | ||
|
|
Not mentioned | 17 (46) | S2, S3, S11, S13, S16, S17, S20, S22-S26, S29, S30, S32, S35, S37 |
aNumbers do not add up as participants in some studies belong to more than one age group.
bNumbers do not add up as participants in some studies were diabetic and nondiabetic.
General features of wearable devices (n=37).
Features | Studies, n (%) | Study ID | |
|
|||
|
Prototype | 22 (59) | S1, S3-S5, S8-S11, S16, S17, S20, S23, S24, S26, S28-S33, S36, S37 |
|
Commercial | 15 (41) | S2, S6, S7, S12-S15, S18, S19, S21, S22, S25, S27, S34, S35 |
|
|||
|
Smart clothes | 1 (3) | S1 |
|
Smart socks | 1 (3) | S31 |
|
Smart watch | 8 (22) | S2, S7, S14, S15, S18, S19, S28, S35 |
|
Smart watch and wearable sensor | 2 (5) | S21, S24 |
|
Smart wristband | 9 (24) | S4, S6, S12, S13, S25, S27, S30, S33, S34 |
|
Smart wristband, smart footwear, and smart neckband | 2 (5) | S23, S32 |
|
Wearable sensor | 14 (38) | S3, S5, S8-S11, S16, S17, S20, S22, S26, S36, S37 |
|
|||
|
Body | 1 (3) | S1 |
|
Chest | 1 (3) | S11 |
|
Finger | 5 (14) | S3, S8, S17, S20, S26 |
|
Foot | 6 (16) | S5, S9, S16, S29, S31, S36 |
|
Hand | 1 (3) | S10 |
|
Wrist | 18 (49) | S4, S6, S7, S12-S15, S18, S19, S24, S25, S27, S28, S30, S33-S35, S37 |
|
Wrist and arm | 1 (3) | S21 |
|
Wrist or thigh | 1 (3) | S2 |
|
Wrist, foot, and neck | 2 (5) | S23, S32 |
|
Arm and body | 1 (3) | S22 |
|
|||
|
Actigraph | 1 (3) | S21 |
|
Arduino Nano | 1 (3) | S24 |
|
Basis Peak | 1 (3) | S34 |
|
FreeStyle Libre Flash | 2 (5) | S22, S35 |
|
Medtronic Zephyr | 1 (3) | S22 |
|
Dexcom G4 Platinum (Professional) | 1 (3) | S21 |
|
Empatica E4 | 6 (16) | S12, S13, S14, S25, S27, S35 |
|
Glutrac | 3 (8) | S15, S18, S19 |
|
Mi band 2 | 1 (3) | S6 |
|
Raspberry Pi Zero | 2 (5) | S8, S16 |
|
Pebble | 1 (3) | S2 |
|
Custom | 2 (5) | S26, S28 |
|
Not mentioned | 18 (49) | S1, S3-S5, S7, S9-S11, S17, S20, S23, S29-S33, S36, S37 |
|
|||
|
Android | 3 (8) | S2, S8, S16 |
|
iOSc | 2 (5) | S9, S11 |
|
Microsoft | 1 (3) | S31 |
|
Raspberry Pi OSd | 1 (3) | S24 |
|
iOS and Android | 16 (43) | S6, S7, S12-S15, S17-S20, S22, S23, S26, S28, S30, S32 |
|
Any desktop OS | 3 (8) | S25, S27, S29 |
|
Any smartphone OS | 1 (3) | S29 |
|
Not mentioned | 11 (30) | S1, S3-S5, S10, S21, S33-S37 |
|
|||
|
Smartphone | 16 (43) | S1, S6, S7, S11-S15, S17-S20, S23, S28, S30, S32 |
|
Database servers (Hbase and Hadoop or Spark) | 1 (3) | S33 |
|
Adapter | 1 (3) | S4 |
|
Smartphone or PC | 2 (5) | S25, S27 |
|
None | 17 (46) | S2, S3, S5, S8-S10, S16, S21, S22, S24, S26, S29, S31, S34-S37 |
|
|||
|
Cloud (MongoDb, Database server, Google) | 18 (49) | S1, S6, S7, S11-S15, S17-S19, S23, S25, S27, S28, S30, S32, S33 |
|
PC (laptop, desktop, or Microsoft Surface) | 4 (11) | S4, S20, S29, S31 |
|
Raspberry Pi | 1 (3) | S24 |
|
Smart devices (smartphone, tablet, or PC) | 6 (16) | S5, S8, S9, S16, S22, S26 |
|
None | 8 (22) | S2, S3, S10, S21, S34-S37 |
|
|||
|
Bluetooth | 19 (51) | S2, S5, S6, S9, S11-S15, S18-S20, S22, S25-S28, S30, S31 |
|
Internet (Wi-Fi or cellular or mobile network) | 6 (16) | S1, S7, S8, S16, S17, S33 |
|
Internet (Wi-Fi or cellular or mobile network) and Bluetooth | 2 (5) | S23, S32 |
|
Wired | 2 (5) | S24, S29 |
|
Removable media | 1 (3) | S4 |
|
N/Ae | 7 (19) | S3, S10, S21, S34-S37 |
aNumbers do not add up as some studies used more than one wearable device.
bNumbers do not add up as the WD in one study worked on 2 operating systems.
ciOS: iPhone operating system.
dOS: operating system.
eN/A: not applicable.
Technical features of wearables (n=37).
Feature | Studies, n (%) | Study ID | |
|
|||
|
Blood glucose | 15 (41) | S3, S8, S10, S15, S17-S22, S24, S26, S28, S30, S37 |
|
Physiological | 2 (5) | S1, S10 |
|
Heart rate, heart rate variability, or interbeat interval of the heart | 9 (24) | S6, S11, S14, S22, S23, S32-S35 |
|
Galvanic skin response | 9 (24) | S12-S14, S23, S25, S27, S32, S34, S35 |
|
Blood volume pulse | 6 (16) | S12-S14, S25, S27, S35 |
|
Acceleration | 6 (16) | S12-S14, S25, S27, S35 |
|
Plantar pressure | 5 (14) | S5, S9, S23, S29, S32, S33 |
|
Temperature (skin, foot, shoe, air, or ambient) | 10 (27) | S12, S13, S16, S23, S25, S27, S32, S34-S36 |
|
Step count | 2 (5) | S7, S16 |
|
Other (sedentary behaviors, pulse wave information, inertial data, weight, humidity, activity patterns, frequency of food intake and water, and ankle edema quantification) | 8 (22) | S2, S4, S9, S16, S21, S23, S31, S32 |
|
|||
|
Blood glucose | 27 (73) | S1, S3, S6-S8, S10-S28, S30, S33, S37 |
|
Plantar pressure | 3 (8) | S5, S9, S29 |
|
Heart rate or heart rate variability | 4 (11) | S28, S33, S34, S35 |
|
Other (sedentary behavior, pulse wave, edema, general diabetes symptoms, temperature, sleep quality, step counts, and GSR) | 7 (19) | S2, S4, S31, S32, S34-S36 |
|
|||
|
Opportunistic | 28 (76) | S1, S2, S5, S7, S11-S14, S16-S29, S31-S33, S35-S37 |
|
Participatory | 9 (24) | S3, S4, S6, S8-S10, S15, S30, S34 |
|
|||
|
Accelerometer | 5 (14) | S2, S13, S14, S21, S27 |
|
Photoplethysmography | 12 (32) | S3, S10, S12-S15, S19, S20, S24, S25, S27, S28 |
|
Galvanic skin response | 8 (22) | S10, S13, S14, S23, S24, S27, S32, S34 |
|
Near infrared | 5 (14) | S3, S17, S18, S28, S37 |
|
Electrocardiography | 3 (8) | S11, S18, S22 |
|
Continuous glucose monitoring | 2 (5) | S21, S22 |
|
Bluetooth | 1 (3) | S6 |
|
Pressure sensors | 7 (19) | S5, S9, S23, S29, S32, S33, S36 |
|
Infrared thermopile | 3 (8) | S13, S14, S27 |
|
Temperature sensor | 6 (16) | S7, S16, S23, S24, S32, S36 |
|
Optical heart rate sensor | 2 (5) | S23, S32 |
|
Vibration sensor and flex sensor | 2 (5) | S23, S32 |
|
Motion sensor | 2 (5) | S7, S31 |
|
Others (physiological sensors, pulse sensor, blood glucose level sensor, Raspberry Pi camera, humidity sensor, step count sensor, weight sensor, stretch sensor, and optical sensor) | 6 (16) | S1, S4, S7, S8, S16, S31 |
aNumbers do not add up as WDs in many studies were used to measures many biomarkers.
bNumbers do not add up as some studies used more than one measure.
cNumbers do not add up as WDs in most studies used more than one sensor.
Diabetes type with regards to wearable device type. PreD: prediabetes; T1D: type 1 diabetes; T2D: type 2 diabetes.
For the purpose of this study, we categorized the ML algorithms into 4 categories (classification models, regression models, neural network–based models, and optimization algorithms) and those that were not clearly specified by the study authors were categorized as black boxes (ie, studies that mention they make use of ML or AI but do not specify any further details of algorithms used). Many ML technologies were reported that come under these headings (refer to
Artificial intelligence (AI)– and machine learning (ML)–related features (n=37).
Features | Studies, n (%) | Study ID | ||||
|
||||||
|
|
|||||
|
|
Support vector machine | 13 (35) | S1, S2, S4, S5, S9, S12, S13, S25, S29, S30, S33, S34, S36 | ||
|
|
Random forest | 12 (32) | S2, S4, S5, S7, S11, S14, S15, S18, S27, S29, S36, S37 | ||
|
|
K-nearest neighbors | 7 (19) | S5, S9, S12, S13, S25, S29, S31 | ||
|
|
Naive Bayes | 5 (14) | S2, S7, S13, S31, S36 | ||
|
|
Decision tree | 4 (11) | S1, S13, S31, S35 | ||
|
|
Ensemble learning or ensemble—boosted trees | 2 (5) | S1, S13 | ||
|
|
Logistic regression | 2 (5) | S2, S11 | ||
|
|
J48 | 2 (5) | S2, S7 | ||
|
|
Linear discriminant analysis or linear discriminant | 2 (5) | S4, S13 | ||
|
|
Gradient boosting decision trees | 2 (5) | S5, S35 | ||
|
|
AdaBoost classifier | 1 (3) | S5 | ||
|
|
ZeroR | 1 (3) | S7 | ||
|
|
OneR | 1 (3) | S7 | ||
|
|
Simple logistic regression | 1 (3) | S7 | ||
|
|
Gaussian Process classifier | 1 (3) | S29 | ||
|
|
C4.5 | 1 (3) | S33 | ||
|
|
Linear ridge Classifier | 1 (3) | S14 | ||
|
|
Extreme gradient boost | 1 (3) | S12 | ||
|
|
|||||
|
|
Linear regression | 2 (5) | S3, S16 | ||
|
|
Support vector regression or Fine Gaussian support vector regression | 1 (3) | S3 | ||
|
|
Random Forest regression | 1 (3) | S15 | ||
|
|
AdaBoost regression | 1 (3) | S15 | ||
|
|
Multilayer Polynomial regression | 1 (3) | S17 | ||
|
|
Ensemble—boosted trees | 1 (3) | S3 | ||
|
|
Exponential Gaussian process regression | 1 (3) | S20 | ||
|
|
|||||
|
|
Artificial Neural Network | 5 (14) | S1, S2, S8, S26, S36 | ||
|
|
Long short-term memory | 4 (11) | S6, S13, S21, S34 | ||
|
|
Convolutional Neural Network | 3 (8) | S10, S22, S24 | ||
|
|
Deep neural networks | 3 (8) | S11, S13, S22 | ||
|
|
Recurrent Neural Network | 2 (5) | S21, S34 | ||
|
|
Multilayer Perceptron | 2 (5) | S6, S29 | ||
|
|
|||||
|
|
Sequential minimal optimization | 1 (3) | S7 | ||
|
|
L1 norm optimization | 1 (3) | S19 | ||
|
|
Particle swarm optimization | 1 (3) | S23 | ||
|
MLa black box | 3 (8) | S19, S23, S32 | |||
|
||||||
|
Blood glucose level forecasting | 12 | S6, S8, S16, S18, S20, S22, S24, S25-28, S34 | |||
|
Blood glucose monitoring | 4 | 11, S30, S32, S37 | |||
|
Classify patients with diabetes (normal, diabetic, and prediabetic) | 12 | S3, S4, S5, S6, S7, S12, S14, S21, S23, S29, S32, S36 | |||
|
Classify other diseases (patients with hypertension or hypoglycemia) | 2 | S33, S35 | |||
|
Evaluation of a developed system | 3 | S2, S10, S13 | |||
|
Feature selection | 2 | S3, S5 | |||
|
Performance validation | 3 | S1, S9, S15 | |||
|
Optimize sensors results | 3 | S16, S17, S19 | |||
|
Predictions for step count, shoe removal time, or serum glucose | 2 | S16, S17 | |||
|
Edema monitoring | 1 | S31 |
aNumbers do not add up as most studies developed more than one AI algorithms.
bNumbers do not add up as AI algorithms in some studies were used for more than one application.
Statistical evaluation of artificial intelligence and machine learning algorithm (n=37).
Characteristic | Value | Study ID | |
|
|||
|
≤80 | S6, S33 | |
|
81-90 | S15, S21, S28, S35, S36 | |
|
91-95 | S1, S9, S13, S15, S22, S25, S29 | |
|
>95 | S4, S5, S7, S12, S14, S30, S31 | |
|
|||
|
≤80 | S35 | |
|
81-90 | S4, S6, S25, S33 | |
|
91-95 | S9, S22 | |
|
>95 | S5, S7 | |
|
|||
|
≤85 | S35 | |
|
86-90 | S9, S22 | |
|
91-95 | S5, S25 | |
|
>95 | S4, S7 | |
|
|||
|
≤91 | S22 | |
|
>91 | S35 | |
|
|||
|
≤74 | S37 | |
|
75-80 | S19, S10 | |
|
81-90 | S18, S28 | |
|
>90 | S3, S8 | |
|
Not mentioned | S24 | |
|
|||
|
≤80 | S6, S33 | |
|
81-90 | S9 | |
|
91-95 | S25 | |
|
>95 | S2, S7 | |
|
|||
|
<5 | S19, S21 | |
|
5-15 | S17 | |
|
>15 | S27 | |
|
|||
|
8 (22) | S3, S8, S16, S17, S19, S21, S27, S37 | |
|
|||
|
Artificial Neural Network | 2 (5) | S8, S26 |
|
Convolutional Neural Network | 3 (8) | S10, S22, S24 |
|
Deep Neural Networks | 4 (11) | S14, S17, S21, S28 |
|
Support Vector Machine | 6 (16) | S4, S9, S25, S29, S30, S33 |
|
Random Forest | 7 (19) | S2, S5, S15, S18, S27, S36, S37 |
|
Long Short-Term Memory | 1 (3) | S13 |
|
Decision Trees or Gradient Boosting Decision Trees | 2 (5) | S31, S35 |
|
K-Nearest Neighbors | 1 (3) | S31 |
|
Multilayer Perceptron | 1 (3) | S6 |
|
OneR | 1 (3) | S7 |
|
Ensemble | 1 (3) | S1 |
|
Support Vector Regression | 1 (3) | S3 |
|
Not mentioned | 6 (16) | S11, S19, S20, S23, S32, S34 |
Artificial intelligence (AI) or machine learning (ML) models used with regard to wearable device placement and measurement studied. CM: classification model; NN: neural network; RM: regression model.
This was the first study of its kind to the best of our knowledge, considering the amount of features we were able to extract from each publication. The features extracted should give researchers insight not only into the technologies that are readily available commercially but also into what is possible in the future with studies we identified that developed prototypes. Our findings shed light on this emerging field, which is still in its infancy. This is further highlighted by the fact that 59% (22/37) of the studies that met our inclusion criteria were prototypes; we were only able to identify 41% as commercially available (as demonstrated in
This review was conducted according to the PRISMA-ScR; therefore, it can be considered a high standard. Two reviewers independently conducted the study selection and data extraction. We believe this to be the first of its kind study focusing on WDs targeting diabetes using AI approaches and were unable to identify previous scoping reviews in the literature that has as an exhaustive list of features extracted in this field. A combination of expert research computer scientists and research medical practitioners allowed us to explore the current technologies in depth and highlight gaps in the research community. The most popular databases in the health care and information technology fields were searched; furthermore, Google Scholar with forward and backward reference list checking allowed an exhaustive search of the literature, reducing the risk of publication bias.
Only studies published between 2015 and 2021 in the English language were included. Furthermore, we did not use Medical Subject Headings terms in our search; therefore, we may have overlooked some relevant studies. We excluded devices that could be classified as WDs, such as electroencephalogram and ECG machines, which limited their use in hospital settings. As our focus was AI, we excluded any study of WDs and diabetes that had a statistical measurement not considered an AI approach. Although we included a large number of features and some effectiveness measures, we fall short of critically assessing the quality of each of the included studies—this goes beyond the scope of our review—and we hope to cover this in a full systematic review in the near future on the same topic.
WDs hold great potential for the self-monitoring of diabetes-related parameters, and their ability to be paired with a range of smart devices, including smartphones and general connectivity to clouds, allows the continuous collection of data from many biosensors that measure vitals and biosignals without user interference. The fact that they can be worn in a stylish and fashionable manner has potential for wider acceptance than other technologies, such as CGMs. Although many studies have used WDs for diabetes, we found that ML is still lacking in a sizable number of these studies. With the limited number of studies that reported the use of ML, we see great promise, largely owing to the accuracy levels of the ML algorithms reported in
Another area for exploration is the use of the internet of things (IoT); in our search, we found a handful of studies making use of IoT. Most IoT papers describe the IoT architecture for diabetes management without specifying the sensors or WDs actually used or implemented, and do not go into much (if any) detail about any ML deployed. There are many opportunities in this domain; none of the studies were found to make good use of developed commercial technologies such as Alexa, Google Home, and Apple watches, which are readily available. The possibilities here are endless, using a combination of data gathered from sensors at the WDs with other patients and personal data in real-time with IoT. This brings along with its own caveats and the need to incorporate questions of privacy and data sovereignty arising from the mass data storage in cloud-based systems and the many interconnected devices and hospital datacenters; there are issues that need to be considered with the use of data and individual consent. There are also problems regarding the scope of an individual’s consent to use their data, as well as potential accountability if the data are mishandled. There are dangers associated with AI algorithms and their misdiagnoses, dangerous advice, or recommendations that do not correspond to the required standard of care. Data security breaches or the reidentification of previously deidentified data may have unintended repercussions. Furthermore, other ethical issues need to be considered, such as accessibility, although commercial WDs that are easily and cheaply available may not be affordable for the masses in low-income countries. A multidisciplinary effort is required, including but not limited to engineers, medical practitioners, and legal experts.
We investigated and reported the current state of WDs and their features for the purpose of diabetes that use ML approaches. Considering the availability of consumer-grade biosensors, we see great advancement potential in this domain, replacing hospital setting, invasive devices, especially when it comes to monitoring glucose levels. Further clinically significant studies are needed to instill confidence and validate WD use as well as the application of ML algorithms on WD data. Researchers and those wanting to develop AI-based WDs can use our review to understand where the gaps are in this emerging field. We encourage readers to use more data and delve deeper into the studies we have identified in order to establish, validate, and repeat studies that showed high accuracy. There is still much work needed, and we feel our review has provided the most extensive work so far summarizing WDs that use ML for people with diabetes to date. Finally, researchers will also benefit from our study as they can embark on longer and better populated systematic studies scrutinizing the benefits of WDs as data gathering, monitoring, prediction, classification, and recommendation devices in the context of diabetes. We envisage several follow-up studies, starting with a full systematic review from our own group.
PRISMA (Preferred Reporting Items for Systematic Reviews and Meta-Analyses) checklist.
Full search terms and strings table.
Data extraction form.
Study reference table.
Data extraction sheet.
Wearable technologies status with regard to wearable device technologies.
artificial intelligence
continuous glucose monitoring
electrocardiogram
internet of things
machine learning
operating system
Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews
support vector machine
type 1 diabetes
type 2 diabetes
wearable device
The authors would like to thank Weill Cornell Medicine-Qatar for making this study possible.
None declared.