A low-power approach to optical glucose sensing via polarisation switching

Abstract

High-precision polarimetry is crucial for sensing and imaging applications, particularly for glucose monitoring within the physiological range of 50 to 400 mg/dl. Traditional approaches often rely on polarisation modulation using magneto-optic or liquid crystal modulators, which require high voltages or currents, limiting their practicality for wearable or implantable devices. In this work, we propose a polarisation-switching technique that alternates between two discrete polarisation states, offering a low-power alternative with miniaturisation potential. Using this method, we achieved a Mean Absolute Relative Difference of 7.7% and a Standard Error of Prediction of 9.6 mg/dl across the physiological glucose range, comparable to commercial continuous glucose monitors. Our approach demonstrates a limit of detection of approximately 40 mg/dl, with measurements performed in phosphate-buffered saline spiked with glucose. This work establishes polarisation switching as a viable alternative for glucose sensing, providing a foundation for future development of wearable and implantable glucose monitoring systems. By eliminating power-intensive components, our approach addresses key limitations of traditional polarimetric methods, paving the way for more accessible and energy-efficient diabetes management technologies.

Keywords: CGM; Glucose sensing; Optical biosensor; Polarimetry; Polarisation switching.

Fig. 1

Schematics of the experimental setup. A green diode laser (532 nm) served as the light source, which is linearly polarised using a Glan-Thompson polariser (LP). The beam is split into two arms by a 50:50 non-polarising beamsplitter (BS). Polarisation modulation is achieved by linear polarisers set to (). The modulated beams are recombined using a second beamsplitter and pass through the sample cuvette. The transmitted light is analysed using another linear polariser (LP) and detected by a photodiode (PD). Additional components include mirrors (M), a neutral density filter (ND), and a lens (L) for alignment, intensity control, and focusing.

Fig. 2

(a) Sample waveform recorded for 49.0 and 401.7 mg/dl glucose solutions. (b) Corresponding Fast Fourier Transform (FFT) plot. For clarity, only a selected section of the waveform and its corresponding FFT are displayed.

Fig. 3

Glucose dependence extracted from the sum of odd harmonic intensities up to . The linear fit achieves an of 0.95 with an SEP of 26.37 mg/dl (x-error). The error bars in the y-axis represent the standard deviation of 15 independent measurements.

Fig. 4

Ratio of the DC component of the measured signal to the sum of odd harmonics as a function of glucose concentration. Each data point represents the average of 15 independent measurements, with y-errors indicating the standard deviation. The data (black dots) is fitted using the non-linear equation, achieving a fit quality of 0.997.