Diffuse Reflectance Spectroscopy: Using Multivariate analysis method for determination of tissue optical properties

Abstract: Diffuse reflectance Spectroscopy is a non-invasive and real-time technique used both in research and clinical studies for purposes such as identifying tumors and monitoring their response to therapy. Here, a compact, cost-effective and portable experimental setup is used in order to acquire the diffuse reflectance spectra from tissue-like liquid phantoms. Two fiber optic probes with different source-detector separations are used for collecting the diffuse light. A phantom preparation protocol is proposed in order to construct a data set of diffuse reflectance spectra from phantoms with different tissue chromophores compositions. Nonlinear least-squares support vector machines (LS-SVM) regression technique within the Multivariate analysis (MVA) framework is employed in order to extract the optical properties of the tissue-like phantoms. Validation measurements of the liquid phantoms demonstrate a higher prediction accuracy for larger number of training samples. Percentage error of <2% is observed when testing 1 sample in both reduced scattering coefficient and blood volume fraction models. The reduced scattering coefficient can be estimated with higher accuracy in the models constructed with the data collected using both probes. Ways to improve the regression models’ performance are proposed along with suggestions for future work.