White Blood Cell Components Separation for Hematological Disease Detection

Abstract

Hematological diseases are blood-related diseases which include leukemia, leucopenia, hemophilia, etc. In which leukemia is severe hematological disease and death rate is very high and hence need proper and faster diagnosis. Leukemia is the formation of giant nuclei or immature increase of leukocytes or WBCs; hence, the need of extracting the shape features and other symptoms must be investigated. This paper suggested and experimented a technique for separating or segmenting white blood cell stain images of leukemia-affected patients. The successful segmentation of an infected cell divides the images into nucleus and outer fluid part cytoplasm, since these two regions contain features that correlate to different forms of leukemia. The algorithm checks and experiments 250 images, and the performance validation using cross-validation techniques shows the nucleus segmentation accuracy as 94% and cytoplasm segmentation accuracy as 78%, respectively.