Fig. 5: In-vivo evaluation of the calibration method and wearable device.

a Schematic of the OGTT procedure, illustrating the relationship between the measurements and each data point related to blood glucose. Inset: Optical images of a female healthy participant (left), a male diabetic patient (middle) wearing the device, and the device applied to the arm during the OGTT (right). Scar bars, 1 cm. OGTT, oral glucose tolerance test. Results of OGTT for one healthy participant (b) and one diabetic patient (c). (Left): Variation of extracted glucose concentration, Na⁺ concentration, and skin surface pH during OGTT; (Right): Results of blood glucose predictions from the wearable device under four conditions: without calibration, with Na⁺ calibration, and with cooperative Na⁺ and pH calibration. d–f Clarke’s error grid analysis of predicted glucose concentrations versus reference glucose concentrations without calibration (d), with Na⁺ calibration (e), and with cooperative Na⁺ and pH calibration (f) for 6 healthy participants (n = 60 data points). h–j Clarke’s error grid analysis of predicted glucose concentrations versus reference glucose concentrations without calibration (h), with Na⁺ calibration (i), and with cooperative Na⁺ and pH calibration (j) for 15 diabetic patients (n = 70 data points). g, k Clarke’s error grid analysis of the OGTT, including percentage distribution across different zones and MARD for 6 healthy participants (g) and 15 diabetic patients (k). MARD, mean absolute relative difference.