Reconstructing optical parameters from double-integrating-sphere measurements using a genetic algorithm

Abstract

For the reconstruction of physiological changes in specific tissue layers detected by optical techniques, the exact knowledge of the optical parameters μa, μs and g of different tissue types is of paramount importance. One approach to accurately determine these parameters for biological tissue or phantom material is to use a double-integrating-sphere measurement system. It offers a flexible way to measure various kinds of tissues, liquids and artificial phantom materials. Accurate measurements can be achieved by technical adjustments and calibration of the spheres using commercially available reflection and transmission standards. The determination For the reconstruction of physiological changes in specific tissue layers detected by optical techniques, the exact knowledge of the optical parameters μa, μs and g of different tissue types is of paramount importance. One approach to accurately determine these parameters for biological tissue or phantom material is to use a double-integrating-sphere measurement system. It offers a flexible way to measure various kinds of tissues, liquids and artificial phantom materials. Accurate measurements can be achieved by technical adjustments and calibration of the spheres using commercially available reflection and transmission standards. The determination of the optical parameters of a material is based on two separate steps. Firstly, the reflectance ρs, the total transmittance TsT and the unscattered transmittance TsC of the sample s are measured with the double-integrating-sphere setup. Secondly, the optical parameters μa, μs and g are reconstructed with an inverse search algorithm combined with an appropriate solver for the forward problem (calculating ρs, TsT and TsC from μa, μs and g) has to be applied. In this study a Genetic Algorithm is applied as search heuristic, since it offers the most flexible and general approach without requiring any foreknowledge of the fitness-landscape. Given the challenging preparation of real tissue samples it comes as no surprise that these are subject to various uncertainties. In order to perform a robust parameter reconstruction samples of different thickness are used. This adds a further, strong restriction to the potential results from the heuristic reconstruction algorithm.