An Impedance and Multi-Wavelength Near-Infrared Spectroscopy IC for Non-Invasive Blood Glucose Estimation

摘要: A multi-modal spectroscopy IC combining impedance (IMPS) and multi-wavelength near-infrared (mNIRS) is proposed for high precision non-invasive glucose level estimation. combination of IMPS mNIRS can compensate the estimation error to improve its accuracy. The circuit measures dielectric characteristics tissue using RLC resonant frequency estimate level. To accurately find frequency, a 2-step sweep sinusoidal oscillator (FSSO) proposed: 1) 8-level coarse switching (f STEP = 9.4 kHz) in 10–76 kHz, 2) fine analog range 18.9 kHz. During sweep, adaptive gain control loop stabilizes output voltage swing (400 mV p-p ). accuracy mNIRS, three wavelengths, 850 nm, 950 1,300 are used. For highly accurate estimation, measurement data combined by an artificial neural network (ANN) external DSP. ANN method reduces mean absolute relative difference 8.3% from 15% IMPS, 15–20% 80–180 mg/dL blood occupies 12.5 mm 2 0.18 µm 1P6M CMOS technology dissipates peak power 38 mW with maximum radiant emitting 12.1 mW.