Determination of the optical properties of two-layered turbid media from spatially resolved diffuse reflectance by SVM

Noninvasive determination of tissue optical properties is essential for clinical applications in medical diagnostics and therapeutics. In recent years, several methods were successfully introduced to deduce the optical properties of semi-infinite tissue model from spatially resolved (SR) diffuse reflectance. However, biological tissue is in fact not homogeneous and usually exhibits a complicated layered structure. The previous methods are not always efficient for the layered-structure tissue model. In this paper, we introduced a new method to determine the optical properties of the two-layer medium from the steady-state spatially resolved diffuse reflectance, which is based on the theory of support vector machine (SVM). The method was validated using the Monte Carlo algorithm generated reflectance from a two-layer model that consists of a 5mm thick top layer and a semi-infinite bottom layer. The training and predicting time of SVM are 20s and 5s respectively. The predictive errors of the proposed method were less than 2% for the top-layer optical properties and less than 4% for the bottom-layer optical properties, showing that the SVM method has a higher accuracy and a shorter training time comparing with other methods. The principle to deal with regression estimation problems with SVis briefly introduced firstly. Then, the phantom experiment set and the results are described. In the end, some limitations and strategies are also discussed.