measurement of the optical properties of biological tissue is very important for optical diagnosis and therapeutics. An artificial neural network （ANN）-based inverse reconstruction method is introduced to determine the optical properties of turbid media, which is based on the reflectance （R） and transmittance （T） of a thin sample measured by a double-integrating-spheres system. The accuracy and robustness of the method has been validated, and the results show that the root mean square errors （RMSEs） of the absorption coefficient μa and scattering coefficient μ′ reconstruction are less than 0.01 cm-1 and 0.02 cm-1, respectively. The algorithm is not only very accurate in the case of a lower albedo （～0.33）, but also very robust to the noise of R and T especially for the μ′ reconstruction.
