Support Vector Machine For Recognition Of Cancer Cells In Microscopic Cell Image

The recognition of cancer cell is one of the popular research topics in the field of digital image processing and pattern processing. The extraction of the characteristics of cancer cell as well as the classification of the cell based on its characteristics is one of the key steps in the recognition of cancer cell. The support vector machine based on statistical learning theory is a strong pattern recognition tool under the condition of small samples learning. In this thesis, the cell color image is segmented to extract the characteristics of single cell and cell group by using the excellent classification performance of the support vector machine. Based on this, the support vector machine is employed to classify cell.In order to calculate and analyze the characteristics of cell color image, firstly, the image should be segmented. Because the cell images to be processed are colorful, processing the color images directly can simplify the program and improve the efficiency of characteristics extraction. This thesis adopts the M-SVM to segment the single cell color image into three values and employs the SVM to segment the cell group image into two values. The best support vector machine model has been obtained by analyzing and evaluating the experimental results.According to the pathological change of the characteristics of the cell, the single normal mesothelial cell differs from the adenocarcinoma in the color and the shape. The discriminating characteristics of the single cells are given in the color and the shape. This thesis also propose specific algorithm to extract corresponding characteristics of single cells. Single cells are classified by making use of SVM, and the problem of the recognition of single cells is solved efficiently. The suspicious general cells have remarkable shape when they are in the shape of group. For example, they present in the shape of private soldier, the spinal column, the plum blossom and so on. Research on the shape of cell group can be replaced by research on the shape of nucleus, which can be described by image moment. As a result, the shape of the cells group can be classified by using M-SVM, and the suspicious cells group can be recognized.