Wall-Adherent Cells Segmentation Based on BP Neural Network

Anti-virus experiment in vitro is a common way to screen and identify antiviral drugs. Most of the cells in the experiments are wall-adherent. Segmenting, recognizing and counting this wall-adherent cells in micrograph successfully can cut down the time and cost of the experiment. It is much difficult to segment this kind of image because of the wall-adherent cells characteristics. In this paper, some works have been done to use BP neural network to segment wall-adherent cells. Firstly, three layers BP network has been designed and features are extracted for segmentation. The weights and thresholds of the network are determined by training it with hundreds of samples. Other three typical wall-adherent cell images are respectively input into the designed BP network to test its segmentation performance. In the segmentation test result images, the cells are well segmented not only in clear region, but also in polluted region and small fracted region. The test results show that the designed BP network is much effective for wall- adherent cell image segmentation.