Morphology Based Detection of Abnormal Red Blood Cells in Peripheral Blood Smear Images

Abstract

Red blood cells are the most abundant type of blood cells in the human body, delivering oxygen to body tissues. The count of these vital cells is often the first step done in analyzing a patient’s pathological condition. Normal RBC’s are biconcave in shape with a central pale area, and any deviation in size, shape, volume, structure or color represents an abnormal cell. Such abnormalities are detected by viewing the blood-smear images through a microscope, a time consuming and error-prone method. This process can be automated by analyzing the individual cells in a peripheral blood smear image and segmenting the cells using appropriate segmentation techniques. The proposed study aims at Morphologybased detection of abnormal red blood cells in peripheral blood smear images, based on their size and shapes. Abnormalities such as Anisocytosis, Macrocytosis and Microcytosis are detected based on the size of the RBCs. Variations in the shape of RBCs couldindicate various abnormalities. Convex hull based detection of speculated RBCs, is carried out in Acanthrocytosis and Echinocytosis. The condition Eliptocytosis, where some of the RBCs turn elliptical is detected using Houghman Transform. In the abnormality called Rouleaux the RBCs appear as stack of coins, which are detected by applying a watershed algorithm to individual stacks and counting the number of cells in the stack. Sickle cell anemia is another common condition in people, where few RBCs are sickle or crescent shaped and this shape is determined using the roundness factor. Codocytes resemble a bull’s eye, and can be identified by examining if the segmented RBCs have rounded areas within the cell. Dacrocytes are tear drop RBCs, which can be detected by analyzing the extreme points of the cell. The experiment was conducted for fifty images and the success rate achieved was 80%.