Diffuse Correlation Spectroscopy Analysis Implemented on a Field Programmable Gate Array.

Wei Lin,Thomas F Floyd,Chia Chieh Goh,James Barsi,David R Busch

doi:10.1109/access.2019.2938085

Abstract

Diffusive correlation spectroscopy (DCS) is an emerging optical technique that measures blood perfusion in deep tissue. In a DCS measurement, temporal changes in the interference pattern of light, which has passed through tissue, are quantified by an autocorrelation function. This autocorrelation function is further parameterized through a non-linear curve fit to a solution to the diffusion equation for coherence transport. The computational load for this non-linear curve fitting is a barrier for deployment of DCS for clinical use, where real-time results, as well as instrument size and simplicity, are important considerations. We have mitigated this computational bottleneck through development of a hardware analyzer for DCS. This analyzer implements the DCS curving fitting algorithm on digital logic circuit using Field Programmable Gate Array (FPGA) technology. The FPGA analyzer is more efficient than a typical software analysis solution. The analyzer module can be easily duplicated for processing multiple channels of DCS data in real-time. We have demonstrated the utility of this analyzer in pre-clinical large animal studies of spinal cord ischemia. In combination with previously described FPGA implementations of auto-correlators, this hardware analyzer can provide a complete device-on-a-chip solution for DCS signal processing. Such a component will enable new DCS applications demanding mobility and real-time processing.

Highlights

Diffusive correlation spectroscopy (DCS) is a relatively new optical technology to probe microvascular blood flow [1]
We adopted the Nelder-Mead method because it is used in our offline DCS analysis and it involves only simple math computations implementable in Field Programmable Gate Array (FPGA)
The phantom study demonstrated that the DCS system with the hardware analyzer produced the consistent data compared to the independent DCS system with offline DCS data analysis

Summary

INTRODUCTION

DCS is a relatively new optical technology to probe microvascular blood flow [1]. In the typical implementation, the tissue of interest is illuminated with a long-coherence length near infra-red (NIR) laser and the temporal fluctuation of interference patterns formed on the tissue surface are detected. The DCS data analysis typically requires two steps: (1) calculating a temporal autocorrelation function of the scattered light intensity and (2) fitting the measured autocorrelation to a theoretical model, i.e. a solution of the correlation diffusion equation, to extract a blood flow index. Multi-tau correlation was first invented by Schatzel et al [22] and has been implemented in software [23], [24] and in hardware using FPGA technology [25] Both traditional and fast DCS measurements require a non-linear fit of the autocorrelation curve at each time point of measurement to a solution of the correlation diffusion equation. This forms a computational bottleneck, as these fitting algorithms require significant computation resources. Coupled with previous work utilizing FPGAs to calculate the correlation function, this has opened the path to integrate the entire data processing necessary for DCS on a single silicon chip

METHODS

FPGA COMPUTATION

DISCUSSION

INFINITE GEOMETRY

Findings

SEMI-INFINITE GEOMETRY