Abstract:Prediabetes is a common health condition that often goes undetected until it progresses to type 2 diabetes. Early identification of prediabetes is essential for timely intervention and prevention of complications. This research explores the feasibility of using wearable continuous glucose monitoring along with smartwatches with embedded inertial sensors to collect glucose measurements and acceleration signals respectively, for the early detection of prediabetes. We propose a methodology based on signal processing and machine learning techniques. Two feature sets are extracted from the collected signals, based both on a dynamic modeling of the human glucose-homeostasis system and on the Glucose curve, inspired by three major glucose related blood tests. Features are aggregated per individual using bootstrap. Support Vector Machines are used to classify normoglycemic vs. prediabetic individuals. We collected data from 22 participants for evaluation. The results are highly encouraging, demonstrating high sensitivity and precision. This work is a proof of concept, highlighting the potential of wearable devices in prediabetes assessment. Future directions involve expanding the study to a larger, more diverse population and exploring the integration of CGM and smartwatch functionalities into a unified device. Automated eating detecting algorithms can also be used.

What problem does this paper attempt to address?

The main problem that this paper attempts to solve is the early detection of prediabetes in order to achieve timely intervention and prevent its further development into type 2 diabetes (T2D) and its complications. Specifically, the author proposes a method based on wearable devices (such as continuous glucose monitors and smart watches), using signal processing and machine - learning techniques to analyze acceleration signals and glucose measurements, so as to accurately classify whether an individual is prediabetic or not. ### Main Problems and Background 1. **Challenges of Prediabetes**: - Prediabetes usually has no obvious symptoms, resulting in about 80% of cases remaining undiagnosed. - Traditional diagnostic methods (such as fasting blood - glucose tests, oral glucose tolerance tests, and glycated hemoglobin tests) have limitations, for example, they are unable to capture instantaneous blood - glucose fluctuations, are affected by factors such as anemia or ethnicity, and rely on single or a few measurement results. 2. **Research Objectives**: - Propose a non - invasive method using wearable devices (such as continuous glucose monitors and smart watches) to detect prediabetes. - Through advanced signal processing and machine - learning techniques, analyze acceleration signals and glucose measurements to accurately distinguish between normoglycemic and prediabetic individuals. ### Method Overview 1. **Data Collection**: - Collected continuous glucose monitoring data and acceleration signals from smart watches of 22 participants. - The participants wore these devices for 14 days in their daily lives and recorded diet information. 2. **Feature Extraction**: - Proposed two feature sets: one is based on the dynamic modeling of the human glucose homeostasis system (Θh), and the other is based on the glucose curve parameters of the CGM device (Θc). - The Θh feature set includes parameters extracted from coupled differential and integral equations, which describe the dynamic changes of blood glucose. - The Θc feature set includes parameters extracted from the glucose curve, such as the average blood - glucose level, standard deviation, and post - prandial blood - glucose gradient, etc. 3. **Classification Model**: - Used support vector machines (SVM) to classify the extracted features and evaluate the performance of the model in distinguishing between normoglycemic and prediabetic individuals. ### Results and Discussion - The experimental results show that the SVM model combining the Θh and Θc feature sets achieved high accuracy and sensitivity (0.86 and 0.92 respectively), indicating that this method has significant potential in prediabetes detection. - Compared with traditional laboratory blood tests, although this method is slightly inferior in specificity, it performs excellently in sensitivity and does not rely on invasive tests, and is suitable for large - scale screening. ### Conclusion This study shows that using wearable devices for prediabetes detection is a promising method, which can achieve high - precision and high - sensitivity detection while avoiding the burden of traditional laboratory tests. Future research will further expand the data set and improve device integration to improve the reliability and applicability of the method.

Prediabetes detection in unconstrained conditions using wearable sensors

Noninvasive Hypoglycemia Detection in People With Diabetes Using Smartwatch Data

Blood Glucose Monitoring Using Non-Invasive Features of Wearable Devices and Machine Learning

Real-time Non-invasive Blood Glucose Monitoring using Advanced Machine Learning Techniques

Assessing Elevated Blood Glucose Levels Through Blood Glucose Evaluation and Monitoring Using Machine Learning and Wearable Photoplethysmography Sensors: Algorithm Development and Validation

Detection of Physical Activity Using Machine Learning Methods Based on Continuous Blood Glucose Monitoring and Heart Rate Signals

Digital Biomarkers for Personalized Nutrition: Predicting Meal Moments and Interstitial Glucose with Non-Invasive, Wearable Technologies

Exploring EEG Signals for Noninvasive Blood Glucose Monitoring in Prediabetes Diagnosis

Using Wearable Activity Trackers to Predict Type 2 Diabetes: Machine Learning–Based Cross-sectional Study of the UK Biobank Accelerometer Cohort

Predicting Type 2 Diabetes Metabolic Phenotypes Using Continuous Glucose Monitoring and a Machine Learning Framework

Machine Learning-Based Diabetes Detection Using Photoplethysmography Signal Features

Lifestyle Profiling Using Wearables and Prediction of Glucose Metabolism in Individuals with Normoglycemia or Prediabetes

Pioneering diabetes screening tool: machine learning driven optical vascular signal analysis

Detection and Classification of Unannounced Physical Activities and Acute Psychological Stress Events for Interventions in Diabetes Treatment

Sensing interstitial glucose to nudge active lifestyles (SIGNAL): feasibility of combining novel self-monitoring technologies for persuasive behaviour change

Gait acceleration-based diabetes detection using hybrid deep learning

Smartphone based wearable sweat glucose sensing device correlated with machine learning for real-time diabetes screening

Multi-modal Predictive Models of Diabetes Progression

Precision Medicine and Artificial Intelligence: A Pilot Study on Deep Learning for Hypoglycemic Events Detection based on ECG

Explainable Early Prediction of Gestational Diabetes Biomarkers by Combining Medical Background and Wearable Devices: A Pilot Study With a Cohort Group in South Africa

Discriminative Information Added by Wearable Sensors for Early Screening - a Case Study on Diabetic Peripheral Neuropathy.