Permittivity extraction of glucose solutions through artificial neural networks and non-invasive microwave glucose sensing

•Accurate permittivity model for glucose/water solutions is vital for sensor simulations.•Appropriate RF sensor design is needed for achieving high sensitivity.•The small change in blood glucose levels is challenging to measure.•Challenges and sources of problems in glucose concentration measurement. An accurate low-cost method is presented for measuring the complex permittivity of glucose/water solutions. Moreover, a compact non-invasive RF/microwave sensor is presented for glucose sensing with the reasoning behind design parameters as well as simulation and measurement results. The complex permittivity values of aqueous solutions of glucose were measured with an in-house manufactured open-ended coaxial probe and the values were extracted from the measured complex reflection coefficients (S11) utilizing artificial neural networks. The obtained results were validated against a commercial probe. The values were fitted to the Debye relaxation model for ease of evaluation for a desired glucose concentration at a desired frequency. The proposed permittivity model in this paper is valid for glucose concentrations of up to 16 g/dl in the 0.3–15 GHz range. The model is useful for simulating and validating non-invasive RF glucose sensors.