Abstract
The authors report a novel application of a neural network to perform curve fitting to simulated experimental data, obviating the need for lengthy non-linear least-squares fitting procedures. This application arose in the non-invasive determination of tissue optical properties, namely the transport scattering coefficient, mu 's and the absorption coefficient, mu a. This problem is relevant in clinical applications such as photodynamic therapy of cancer where the optical properties are used for the calculation of light fluence distributions in tissue, and for measuring the concentration of endogenous or exogenous chromophores. One approach to determine this optical coefficients noninvasively is to use a pencil beam of light to irradiate the tissue and to measure the multiply scattered light remitted from the tissue at different radial distances, rho , from the pencil beam on the tissue surface.