Automatic Image Segmentation Scheme for Counting the Blood Cell Nuclei with Megaloblastic Anemia

The red blood cell count in blood smears provides vital information that helps in the diagnosis many of many diseases, such as anemia, leukemia etc. However, locating, identifying and counting blood cells manually are a time-consuming task. This task can be carried out by an automatic image analysis system, of which segmentation is a critical step. This paper presents an automatic segmentation method for calculating the blood cell nuclei in thin blood smears with megaloblastic anemia. After removing the background and noises in the blood images, the proposed method analyzes the changes in the RGB color spaces of the nucleus and applies histogram normalization to segment the nucleus out. For calculating the nucleus of the blood cells, the minimum distance between each object pixel and the boundary pixels can be measured by using distance transform to effectively find the center points of the object. Moreover, the center point information is then adapted to an expanding circle to separate each nucleus from the object. The proposed method eventually analyzes the number of nuclei in the blood smears, utilizing the calculation of the quantity of nuclei and recognition of megaloblastic anemia.