Retrieval of particle characteristics with high-order neural networks: application to scanning flow cytometry

The problem of retrieval of homogeneous spherical particles characteristics by the data on the intensity of scattered light is considered. To solve this problem the high-order neural networks method is used. The algorithms to determine radius and refractive index of particle using the multidot high-order neural networks are proposed. The nets to retrieve particle's radius and refractive index by the data on the intensity of scattered light in a limiting range of available for measurement angles are constructed. The neural networks are trained in the range of radius from 0.5 up to 15.5 microns and refractive index from 1.02 to 1.2, respectively. Dependence of the retrieval errors on particle characteristics and the neural network structure is estimated.