An Adaptive Coherent Optical Processor for Cell Recognition and Counting

An adaptive coherent optical processor, using discrete sampling of the Fourier spectrum, is described. The discrete sampling method was used to realize a system for recognizing and counting reticulated red blood cells using conventional blood film slides as the input field. This process is based on the assumption that a family of cells of distinct morphology will have a unique Fourier spectrum and that the intensity of the spectrum from the individual cells is additive when the population is large, randomly located, and nonoverlapping. Intensity measurements made at discrete spatial frequencies provide information on the number of cells of each type present. Inference of the counts is made through a linear estimation model obtained by a least squares regression of the intensity measurements against a set of known counts. If applied to the class of spectra for which the system model is designed, the estimation error is equal to or better than equivalent hand counting methods and is of the same order as that inherent in any least squares regression process.