Signal Processing for Diffuse Correlation Spectroscopy with Recurrent Neural Network of Deep Learning

Diffuse correlation spectroscopy (DCS) is a newly developed near-infrared technology for tissue blood flow measurement. We have proposed an Nth-order linear (NL) algorithm to extract the blood flow index (BFI) in heterogeneous tissue with arbitrary geometry. A critical procedure included in NL algorithm is the linear regression, through which the DCS data are denoised. In this study, we proposed a recurrent neural network (RNN) regression model of deep learning to perform linear regression. For model evaluation, the DCS data at different source-detector separations (i.e., 1.5, 2.0, 2.5 and 3.0 cm) were generated by computer simulations and processed by RNN and the support vector regression (SVR) that was previously proposed by us. The results show that the accuracy of RNN regression model is at the same level as SVR approach at 3.0 cm source-detector separation. However, the RNN model exhibits better robustness than SVR method. The outcomes derived from this study will improve the signal processing for DCS blood flow measurements.

[1]  D. Boas,et al.  Noninvasive optical measures of CBV, StO2, CBF index, and rCMRO2 in human premature neonates' brains in the first six weeks of life , 2010, Human brain mapping.

[2]  A. Yodh,et al.  In vivo cerebrovascular measurement combining diffuse near-infrared absorption and correlation spectroscopies. , 2001, Physics in medicine and biology.

[3]  Qingming Luo,et al.  MCVM: MONTE CARLO MODELING OF PHOTON MIGRATION IN VOXELIZED MEDIA , 2010 .

[4]  Dennis E. Koppel,et al.  Statistical accuracy in fluorescence correlation spectroscopy , 1974 .

[5]  T. Floyd,et al.  Validation of diffuse correlation spectroscopy for muscle blood flow with concurrent arterial spin labeled perfusion MRI. , 2007, Optics express.

[6]  Scott D. Stevens,et al.  Influences of tissue absorption and scattering on diffuse correlation spectroscopy blood flow measurements , 2011, Biomedical optics express.

[7]  R. Behmanesh,et al.  Control chart forecasting: A hybrid model using recurrent neural network, design of experiments and regression , 2012, 2012 IEEE Business, Engineering & Industrial Applications Colloquium (BEIAC).

[8]  Guoqiang Yu,et al.  Diffuse optical monitoring of repeated cerebral ischemia in mice , 2011, Optics express.

[9]  Iman Rahimi,et al.  Using Combination of Optimized Recurrent Neural Network with Design of Experiments and Regression for Control Chart Forecasting , 2012 .

[10]  Lian He,et al.  Simultaneously Extracting Multiple Parameters via Fitting One Single Autocorrelation Function Curve in Diffuse Correlation Spectroscopy , 2013, IEEE Transactions on Biomedical Engineering.

[11]  Lakhmi C. Jain,et al.  Recurrent Neural Networks: Design and Applications , 1999 .

[12]  R. Ichord,et al.  Diffuse optical monitoring of hemodynamic changes in piglet brain with closed head injury. , 2009, Journal of biomedical optics.

[13]  G. Guo,et al.  Approaches to denoise the diffuse optical signals for tissue blood flow measurement. , 2018, Biomedical optics express.

[14]  Yu Shang,et al.  A Nth-order linear algorithm for extracting diffuse correlation spectroscopy blood flow indices in heterogeneous tissues. , 2014, Applied physics letters.

[15]  Geoffrey E. Hinton,et al.  Speech recognition with deep recurrent neural networks , 2013, 2013 IEEE International Conference on Acoustics, Speech and Signal Processing.

[16]  Chong Huang,et al.  Simultaneous measurement of deep tissue blood flow and oxygenation using noncontact diffuse correlation spectroscopy flow-oximeter , 2013, Scientific Reports.

[17]  Yu Shang,et al.  Extraction of diffuse correlation spectroscopy flow index by integration of Nth-order linear model with Monte Carlo simulation. , 2014, Applied physics letters.

[18]  Jürgen Schmidhuber,et al.  Offline Handwriting Recognition with Multidimensional Recurrent Neural Networks , 2008, NIPS.