DefinePK

Abstract

Red blood cells (RBCs) counting in blood smear image is very important to diagnose blood related diseases such as malaria and anemia before a proper treatment can be proposed. The conventional practice for such procedure is executed manually by pathologist under light microscope. However, manual visual inspection is laborious task and depends on subjective assessment which leads to variation in the RBC counting especially when there are many clumped RBC areas. In this paper a computer-aided systems is proposed to automate the process of counting the RBC from the blood smear image by considering an overlapping constraint. Initially RBCs region are extracted from the background by using global threshold method applied on green channel color image. Next, noise and holes in the RBCs are abolished by utilizing morphological filter and connected component labeling. Following that, information from the RBCs’ area is extracted to determine single and overlapping RBC region. The former region can be counted directly while the latter need to be process further to estimate the number of individual cells. In this paper, two estimators which are Distance transform and Hough transform are utilized to count cells in the clumped regions. Eventually, the total RBCs is found by summing up information from the single cell number and from the estimator. The proposed method has been tested on blood cell images and it demonstrates that Hough transform is more reliable to predict number of total RBCs.