Can Laser Speckle Contrast Imaging be a quantitative measurement using machine learning?

Abstract

Laser speckle contrast blood flow imaging (LSCI) analyzes the spatial or temporal statistical characteristics of laser speckle patterns to obtain a signal proportional to blood flow. As it can achieve real time blood flow imaging at high resolution with simple instrumental setup, LSCI has been widely applied in both clinic and research. The lack of quantitative blood flow measurement is the primary limitation. Efforts have been made to optimize LSCI from the aspects of measurement accuracy and linearity such as multi-exposure laser speckle contrast imaging (MELSCI) and AI-based (artificial intelligence) LSCI. This paper reviews LSCI in terms of basic principles, system development and quantitative measurement of LSCI. The application of machine learning in LSCI is discussed in detail. By comparing the estimated perfusion results of LSCI, MELSCI and LDI (laser Doppler blood flow imaging), we propose that using machine learning to correct LSCI to MELSCI has great potential for improving measurement linearity while retaining system simplicity.