Laser-induced photoacoustic glucose spectrum denoising using an improved wavelet threshold translation-invariant algorithm

Abstract

In this paper, the photoacoustic spectroscopy (PAS), a novel, hybrid, noninvasive imaging modality, is used to monitor the change of glucose concentration. A laser light with wavelength of 532nm is irradiated into the sample, The PA signal is detected by the ultrasonic transducer based on the thermal-expansion effect, which can reflect the biochemical properties of the biological tissues e.g., blood. Although PAS has better resolution and deeper penetration than others, the PA signal is unavoidably polluted by some noises. In order to remove these noises to improve the precision and the signal-to-noise ratio(SNR) of the glucose concentration, an improved wavelet threshold function is proposed. Meanwhile, to eliminate the Pseudo-Gibbs phenomenon, the improved threshold function combines translation-invariant(TI) algorithm to further improve the denoising effect. This threshold function can be looked as the compromise between Donoho's soft-threshold function and hard-threshold function. Experimental results show that the denoising result of improved function is better than that of Donoho's threshold functions and the semisoft-threshold function proposed by Gao, etc. The improved function with combined TI algorithm denoise not only gets higher SNR and lower root-mean-square error(RMSE), but makes the PA glucose spectral smoother. The SNR increases from 27.3658 to 44.7060, the RMSE decreases from 0.272 to 0.017.