kNN Classification Based Erythrocyte Separation in Microscopic Images of Thin Blood Smear

Abstract

In this proposed work, k-nearest neighbors (kNN) classifier comprising of three features i.e. area, compactness ratio, aspect ratio is used to separate the isolated and compound erythrocytes present in microscopic images of thin blood smear. In the microscopic image of thin blood smear, blood components such as erythrocytes, platelets etc are available which is used for diagnostic approach to blood disorder. In the microscopic image, both the isolated and compound erythrocytes are also present. Compound erythrocyte is formed due to overlapping of two or more erythrocytes. In malaria diagnosis, parasitaemia estimation is done which define the ratio of infected erythrocytes related to total number of erythrocytes in microscopic image. For proper quantification of erythrocyte, erythrocytes need to be separated as isolated cell and compound cell. As isolated cells directly count in the counting system and compound cells are further analysed to determine number of erythrocytes. The proposed method to separate the isolated and compound cell provides an average accuracy of ~0.942. It is observed that the proposed method can effectively separate the cells in comparison to some of the existing methods.