Architecture-Level Evaluation of Standalone Wearable Systems for Real- Time Non-Invasive Blood Glucose Inference

Abstract Non-invasive glucose monitoring is shifting from analyte-centric optical sensing toward system-level physiological inference based on autonomic and metabolic state variables. Building on a systems physiology framework in which glycemic dynamics are inferred from multiparametric signals such as heart rate variability, tissue oxygenation, physical activity, and circadian rhythm, this study compares two hardware deployment architectures: smartphone-dependent wearable systems and standalone display-integrated wearable devices. The comparison is conducted across latency, power consumption, data-processing hierarchy, privacy, human–machine interaction, and long-term clinical adherence. The analysis shows that standalone devices combining local edge computation on low-power microcontrollers with integrated OLED feedback can reduce dependence on continuous Bluetooth Low Energy transmission, shorten the real-time inference pathway, and provide low-friction visual feedback directly on the wearable node. Although integrated displays introduce additional power demand, dark-mode OLED rendering, event-driven screen activation, adaptive sampling, and microcontroller-level power management can mitigate this burden. Compared with smartphone-dependent systems, the standalone architecture provides a more robust deployment pathway for real-time non-invasive glucose inference by improving immediacy, autonomy, privacy, and user engagement. These findings suggest that edge-enabled standalone wearables may provide a suitable architecture for future continuous physiological monitoring and proactive metabolic self-management.