Research On Method Of Segmentation And Recognition Of Overlapping Blood Cells

Blood cell image processing has become an important auxiliary tool for clinical diagnosis and pathological analysis. Currently, in the actual working, cell positioning and recognition still mainly rely on manual operation. It results in heavy workload and dullness. Moreover, cell recognition results depend on the technical level of operator. Therefore, it becomes more and more important to develop an automatic cell classification and analysis system that can accomplish an quantitative analysis of blood cells and computer-aided diagnosis.Blood cells consist of red blood cells, white blood cells and platelets. Although white blood cells are a fewer, they play an important role that they can swallow and clear the bacteria invading the body, remove damaged tissue cells and a variety of foreign debris. Many clinical diseases manifest in the changes of the number and appearance of white blood cells and their nucleuses. So, studying on the number and shape of white blood cells is very important. However, counting and recognition of white blood cells are more difficult due to the overlapping white blood cells. Therefore, if more precise classification can be obtained, overlapping white cells must be separated and recognized.In this paper, main research focuses on as follows:1) Because the gray levels have difference between white blood cell nuclei and other regions of the microscopic images in S component image, threshold segmentation is used to gain nucleus area. Then, region growing algorithm is used to segment cytoplasmic region of white cells. For segmented white cells, a new method based on distance chart and principles of geology is proposed to judge if the white blood cells are overlapped. The experimental results show the proposed method is effective.2) For the overlapping cells, a new segmentation method is proposed based on the shape of adhesive cells. The near-circular characteristic of white cells is used to obtain the separating points of the overlapping cells. Then, separating line can be got and the overlapping cells are separated exactly. The experimental results show that the proposed separation method is effective.3) Various characteristics of white blood cells are analyzed deeply and the morphological features, color features and texture features of cells are extracted for recognition. 4) The application of support vector machine (SVM) in automatic blood cells recognition is studied. The blood cells can be recognized effectively and higher recognition rate is obtained.