| Title |
AI Based Multivariate Analysis of Electric Field Application Conditions for Optical Non-Invasive Glucose Concentration Measurement |
| Authors |
Jooran Yang ; Jinwoo Song ; Jin-Gyu Kim ; Hyungpyo Kim |
| DOI |
https://doi.org/10.5207/JIEIE.2026.40.5.380 |
| Keywords |
Electric field assisted optical sensing; Multiwavelength; Non-invasive glucose sensing; Near-infrared; Random forest |
| Abstract |
This in vitro study evaluated the effects of externally applied electric-field conditions on multiwavelength optical signals and the feasibility of AI-based glucose concentration prediction. Aqueous glucose samples were measured using 650, 850, and 950nm light sources and a thermopile detector while varying the electrode configuration, excitation frequency, voltage, and waveform. Electric-field excitation increased the mean output from 0.905 to 1.155 V, corresponding to a 27.6% increase, and the four-electrode configuration produced an 18.8% higher mean output than the two-electrode configuration. The highest concentration linearity was obtained at 950 nm, 6 V and 240kHz (R² = 0.973). In group-wise cross-validation, the Random Forest model achieved the best predictive performance (R² = 0.808, MAE = 36.1mg/dL and RMSE = 62.9mg/dL). These results indicate that an external electric field may serve as an auxiliary stimulus that modulates optical signals by altering the dielectric and interfacial properties of the sample. |