Effective automatic recognition of cultured cells in bright field images using fisher's linear discriminant preprocessing

Automatic cell recognition in bright field microscopy is an inherently difficult task due to the immense variability of cell appearance. Yet, it is essential for a highthroughput robotic system. In this paper, we employed a pixel patch decomposition method to detect cultured cells in bright field images. To increase the classification accuracy, we proposed a novel Fisher's Linear Discriminant (FLD) preprocessing approach. This technique was applied to various experimental scenarios utilizing different imaging environments and the results were compared with those for the traditional Principal Component Analysis (PCA) preprocessing. Our FLD preprocessing was shown to be more effective than PCA primarily owing to its ability to maximize the ratio of between-class scatter to within-class scatter. The optimized algorithm has sufficient accuracy and speed for practical use in robotic systems capable of automatic micromanipulation of single cells.