Machine Learning Implementation of a Diabetic Patient Monitoring System Using Interactive E-App

Area of Science:

• Digital Health
• Chronic Disease Management
• Machine Learning in Healthcare

Background:

• Lifestyle factors significantly impact morbidity and mortality, particularly in type 2 diabetes (T2D) and its cardiovascular complications.
• While healthy lifestyle behaviors are crucial for T2D prevention and management, adherence to self-management recommendations remains low.
• Understanding factors influencing T2D self-management is vital for developing effective interventions.

Purpose of the Study:

• To design, develop, and test a mobile application aimed at enhancing diabetes self-management.
• To integrate wearable technology and machine learning for comprehensive health data tracking.
• To improve knowledge regarding lifestyle change factors in T2D prevention and management.

Main Methods:

• Development of a mobile app to monitor dietary intake and health metrics.
• Utilization of Bluetooth-enabled wearable insole devices for tracking carbohydrate intake, blood glucose, medication adherence, and physical activity.
• Construction and evaluation of two machine learning models (SVM and decision tree) for activity recognition (sitting/standing).

Main Results:

• The machine learning models achieved 86% accuracy in recognizing sitting and standing activities.
• A decision tree model was successfully implemented for real-time activity classification within the app.
• The mobile app demonstrated potential for comprehensive tracking of key diabetes self-management indicators.

Conclusions:

• Mobile health applications, incorporating wearable sensors and machine learning, show promise for improving chronic disease self-management.
• Accurate activity classification using machine learning can support personalized lifestyle interventions for type 2 diabetes.
• Further development and testing of such integrated digital health solutions are warranted for effective T2D management.