Optimized Laser Speckle Rheology Measurement Based on Speckle Pattern’s Gamma Correction and Neural Network

Abstract

Laser speckle rheology (LSR) is a powerful technique for probing the dynamic properties of complex fluids and biological tissues. However, multiple scattering in turbid samples remains a significant challenge, limiting its accuracy and requiring extensive calibration. In this study, a neural-network–assisted Gamma correction is introduced, in which the recorded speckle intensities are re-weighted to restore the true g₂(t) decay, eliminating the need for lengthy calibration. The neural network predicts the optimal Gamma value for speckle intensity correction across different sample concentrations. Once corrected, the speckle patterns are analyzed to compute the autocorrelation function and extract the complex modulus |G*(ω)|. Experimental results show that the neural network achieves a maximum absolute error of 0.006 in Gamma prediction, requires only 5 min to train, and computes each Gamma value in just 0.000273 s. These results not only ensure rapid processing but also provide highly accurate Gamma-corrected speckle patterns, leading to the precise calculation of |G*(ω)|. By removing the necessity for laborious calibration procedures, this approach ensures rapid and accurate rheological characterization of complex fluids.