Online Meal Detection Based on CGM Data Dynamics

2025-06-29

AI Summary

This study addresses the low accuracy and poor interpretability of online meal detection in continuous glucose monitoring (CGM) data. We propose a dynamic mode decomposition (DMD)-based feature extraction framework that models temporal glucose dynamics to capture physiologically meaningful intrinsic modes—such as rise rate, peak morphology, and recovery pattern—thereby constructing a time-varying, clinically grounded feature set. Integrated with a lightweight temporal classifier, the framework enables end-to-end real-time meal detection. Evaluated on a multicenter dataset spanning diverse CGM devices and individuals, our approach achieves a 12.6% average improvement in F1-score over conventional threshold- or static-statistics-based methods. Crucially, the extracted features exhibit strong alignment with established postprandial glucose kinetics, ensuring explicit clinical interpretability. This work provides a robust, deployable algorithmic foundation for closed-loop dietary management in digital therapeutics for diabetes.

Technology Category

• Data Mining & Knowledge Management: Knowledge Acquisition from the Web
• Machine Learning: Time-Series/Data Streams
• Computer Vision: Multi-modal Vision

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation

Abstract

We utilize dynamical modes as features derived from Continuous Glucose Monitoring (CGM) data to detect meal events. By leveraging the inherent properties of underlying dynamics, these modes capture key aspects of glucose variability, enabling the identification of patterns and anomalies associated with meal consumption. This approach not only improves the accuracy of meal detection but also enhances the interpretability of the underlying glucose dynamics. By focusing on dynamical features, our method provides a robust framework for feature extraction, facilitating generalization across diverse datasets and ensuring reliable performance in real-world applications. The proposed technique offers significant advantages over traditional approaches, improving detection accuracy,

Problem

• Detect meal events using CGM data dynamics
• Improve accuracy and interpretability of meal detection
• Provide robust feature extraction for diverse datasets

Innovation

• Dynamical modes from CGM data for meal detection
• Improved accuracy and interpretability of glucose dynamics
• Robust feature extraction for diverse datasets