A Study On Rapid Automatic Recognition Of Circulating Tumor Cells

With the developing of modern science and technology, cell analysis technology has been rapid development, it has been researched for many years in the field of cell image analysis which is used for disease diagnosis and treatment. In recent years, lots of technology breakthroughs make this field become a frontier of medical research. The circulating tumor cell recognition and counting also applies to the treatment of neoplastic diseases clinical for a few years. Thanks to the computer image processing and recognition technologies,the cell recognition technology is becoming more automatic and efficient.Manual recognizing and counting the circulating tumor cell, although sometimes still employed,suffers from being time consuming and sometimes unreliable.With application of circulating tumor cells recognition in various types of cancer treatments deepening,a rapid and automated method to recognize and count circulating tumor cells is in great demand.It has been a challenge for cancer research that how to find a rapid,accurate and automated way to recognize and count circulating tumor cells from blood microscopic image.With the basic knowledge of image processing we have analyzed the microscopic image and cell features in low magnification microscopy image.Combining with the characteristics of our microscopic imaging system we proposed a rapid and automated method to recognize circulating tumor cells.The main work of this thesis is divided into several parts:1) According to the large resolution of microscopic cell image in this thesis a block processing method is used to solve the problem of large consumption of memory.Considering the cell adhesion problems we studied the watershed segmentation algorithm,furthermore,a marked watershed segmentation algorithm which is a much more effective solution to split adherent cells is proposed.2) By virtue of the specificity characteristics of circulating tumor cell in nucleus and membrane we compared with the segmentation error of some different segmentation algorithms.We found the improved active contour segmentation algorithm is more efficient in nucleus segmentation and membrane segmentation.3) The number of the suspicious circulating tumor cell is sometimes more with our recognition method.By computing the feature space of the suspicious cell and comparing with the artificial marked circulating tumor cell we got the similarities about the suspicious circulating tumor cells.With the similarities we sort the suspicious cells by ascending count in order to help circulating tumor cell recognition.4) Making use of Matlab language we designed the software.The software is sample to operate and is efficiency to circulating tumor cell recognition.In this thesis, we propose a cell processing recognition method. The experiments shown that the method not only preliminary implements the tumor circulating cell recognition in microscopic images,but also reduces the labor cost and improves the detection efficiency and accuracy.The method we proposed is of important application value.