Answers are grounded in this paper and cite their sources.
Non-invasive wearable sensing of multiple physiological analytes is limited by microwave antenna detuning under wrist body-loading. Existing radio-frequency (RF) glucose-sensing antennas operate at a single frequency and rarely report quantified body-loading compensation. This work contributes two elements, both evaluated entirely in simulation. First, a triple-band microstrip patch antenna whose operating frequencies at 2.45, 5.8, and 8.5 GHz are selected from the electromagnetic penetration depth into distinct wrist tissue compartments rather than from communication band allocation. The selection is derived from the Cole-Cole dispersion of the tissue composite and cross-validated between independent finite-element and finite-difference solvers. Active PIN (positive-intrinsic-negative) diode matching restores the reflection coefficient lost to body-loading at the shortest penetration depth, where glucose permittivity sensitivity is highest. Second, a four-stage signal pipeline coupling sigmoid-weighted recursive least-squares (RLS) motion rejection, cross-modal feature extraction, a normalised seven-state Kalman observer, and INT8-quantised gradient-boosted inference on an ARM Cortex-M33 processor, with each stage isolated by ablation. Evaluation across 1,000 Dalla Man Monte Carlo virtual subjects yielded a simulation-derived glucose reconstruction mean absolute relative difference (MARD) of 9.3% within the synthetic Dalla Man/forward-simulation framework on 200 held-out subjects. Simulation-based Clarke Error Grid analysis placed 91.3% of estimates in Zone A. All reported figures quantify inversion fidelity within the generating pipeline, not ISO 15197:2013-certified clinical accuracy. No hardware was fabricated and no human data were used.
No takes yet. Share an insight, caveat, or question.
Praween Kumar Nishad (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: