Research On The Human Umbilical Vein Endothelial Cells Image Segmentation Algorithms And System Implementation

As the important factor of inflammation, tumor and other diseases, Human umbilical vein endothelial cells have gradually become one of the key subjects in medical research. However, due to the Human umbilical vein endothelial cells belong to a kind of vascular adherent cells, the shape and size of cells are irregular, and the brightness of cells is very uneven. Moreover, there is serious adhesion between the cells. The existing cell image automatic analysis systems are not suitable for the analysis of Human umbilical vein endothelial cells, especially not suitable for the segmentation of the adhesion Human umbilical vein endothelial cells. Segmentation results of the adhesion cells directly affect the extraction and statistical analysis of the subsequent feature parameters of the cells. In order to segment the Human umbilical vein endothelial cells image accurately, this paper proposes two improved watershed segmentation methods according to the characteristics of images of Human umbilical vein endothelial cells in activated and resting states after the endotoxin induced. And the two methods have improved the segmentation accuracy of the cells image effectively:For the segmentation of images of Human umbilical vein endothelial cells in an activated state, an improved watershed segmentation method that based on extracting regional minimum is proposed. Firstly use the top-hat and bottom-hat transformation to enhance the image contrast ratio. Then use H-minima transform and morphological minimum imposed technology to extract the regional minimum. Finally the morphological watershed transform method is used to segment the image that is marked by minimum. The proposed algorithm can segment the images of Human umbilical vein endothelial cells in an activated state. And the accurate segmentation of images of Human umbilical vein endothelial cells in an activated state are realized at last.According to the images of Human umbilical vein endothelial cells in a resting state, a segmentation algorithm based on “seed area” and distance map reconstruction is proposed in this paper. The segmentation of cells images is realized by combing the cells images and corresponding nucleus images. Firstly locate nucleus to the corresponding cell. Then use the improved distance map reconstruction to reconstruct the distance map of cell image, with nuclei as the “seed region” of cell images. Finally the watershed transform is used to segment the distance map. And the accurate segmentation of images of Human umbilical vein endothelial cells in a resting state are realized in the last.At last, the marking, counting as well as extracting of feature parameters of cell images which has segmented are carried out in this paper. And an analysis system of Human umbilical vein endothelial cell images is established.