Automated Detection and Quantification of Erythrocytes and Leukocytes from Giesma Stains of Blood Smear

Automated differential blood quantification is a fundamental prerequisite for the development of a computer-assisted diagnostic system for microscopic hematology. Digitized images of Giesma stained blood smears enable ease of analysis due to cell coloration. Nevertheless cell clustering adversely affects detection making quantification a challenging task. In this paper, we present a simple yet effective technique for detection and quantification of erythrocytes and leukocytes from Giesma stained images of blood smears. Contrary to colour based segmentation, we split the RGB images into their constituent channels. Green channel is then enhanced using histogram equalization. A key step of our proposed technique is the use of different binarization schemes for erythrocyte and leukocyte detection. Cell clustering is then removed using morphological operations. Later contour based detection is used for cell localization and size based segmentation is employed for quantification. With enhanced preprocessing, our proposed scheme yielded 70% and 99% accuracies for erythrocyte and leukocyte quantification respectively.