A noble technique for detecting anemia through classification of red blood cells in blood smear

Anemia is responsible for various health hazards. Anemia decreases and also alters the shape of red blood cells (RBCs) present in our blood. Different type of RBC shapes account for different type of anemia. Automated blood cell analyzers can detect anemia and provide RBC, WBC and platelet count but anemia type identification, which requires classification of RBCs is carried out manually. The classification of RBCs provides invaluable information to pathologists for diagnosis and treatment of various types of anemia. The manual visual inspection is tedious, time consuming, repetitive and prone to human error. In this paper we have performed the automated classification of RBCs as falling into one of the anemia type. The segmentation and classification of the RBCs are the most important stages. The proposed system identifies RBCs using intensity ratio transformation followed by centroid contour distance for segmentation of RBC. Due to the large RBC shape variations, a shape independent framework for identification and segmentation is required. The proposed method can successfully separate the agglomerates of RBCs in spite of grouping of non-uniform RBC shapes. Two geometric features are used to distinguish between normal and anemic RBCs: Aspect Ratio and Fourier Descriptors. The Euclidean distance measure is used as a criterion to determine the similarity degree between the templates and testing samples. Also the presence of high number of nucleated RBCs (NRBCs) in severe anemic patients gives erroneous WBC count in automated cell-analyzers and requires correction which is carried out manually. This paper also presents the automated NRBC count and provides automatic solution of WBC count correction obtained from automated hematology analyzers.