Design The System Of Gland Cancer Cell Segmentation Counting Method

With the daily adenocarcinoma cell image dealing work carring out,the pathology department,where the following cells-counting work.At present,the pathologist of the hospital is able to distinguish the adenocarcinoma cells by observing the stained sections under the microscope and using artificial experience,but the identification of adenocarcinoma cells is difficult.Therefore,it has become an urgent need for clinicopathological work to design an automatic segmentation counting method for adenocarcinoma cells to improve the efficiency of clinicopathologists and the accuracy of cell identification.The area of a cell is obtained by the watershed algorithm,and then the features are extracted by the AdaBoost algorithm,and then the region of the adenocarcinoma cell is judged according to the characteristics.Finally,the count of the adenocarcinoma cells is completed and the software is displayed by the software.The segmentation and counting of all overlaps,adhesions and multicore cells are completed as much as possible when the segmentation counting method is designed,so as to realize the automatic segmentation and counting of various adenocarcinoma cells,which has certain clinical significance.In order to realize the automatic segmentation counting method,it is necessary to complete the precise segmentation of cells.For the cell adhesion and overlap of the cells caused by various causes such as operation,the precise segmentation of the cells and the expression of the adhesion and overlap of the cells on the image will directly determine the effect of the cell feature extraction.The segmentation counting method is designed as follows: first,the image is marked to get the result of the watershed which is directly used for the two valued image processing;the edge de-noising,filtering the small noise on the image;the distance transformation,calculating the shortest distance of the background pixel in the global operation,and getting the final approximate distance image;Reconstruction,the boundary information in the image is built in the zero value of the water ridge,the watershed is modeled,the image boundary is detected by the watershed function,and the image boundary is detected.The local area is extracted by different distance transformation parameters in different regions,and the parameters are trained to retune the extracted area.The distance parameter of the whole distance transformation is processed to achieve better segmentation effect.Feature extraction,based on the final image segmentation of adenocarcinoma cells,extracts the shape,gray value and texture features.Cell counts,based on the feature vector of the extracted cell features to train the AdaBoost classifier,and then optimize the final segmentation map.Image segmentation results eliminate the segmentation area of non adenocarcinoma cells and achieve cell counting.The segmentation counting method of the adenocarcinoma cell division can realize the segmentation and counting of many complex adenocarcinoma cell images.The morphological reconstruction can show the structural features of the cells more intuitively through the morphological reconstruction.After the parameter training,the cell segmentation is completed more accurately,and then the AdaBoost algorithm is used to complete the cell characteristics extraction and the software is delivered.The number of cells in the image is visually displayed in the interface.The function and performance of the segmentation counting method are tested,and the test results are basically consistent with the design indexes.