Comparison of Red Blood Cells Counting using two Algorithms: Connected Component Labeling and Backprojection of Artificial Neural Network

摘要

Accurate determination of the number of red blood cells (RBCs) is important, which leads to correct diagnosis to the patient. In this paper a comparison of counting procedure using two different algorithms is reported. They are the connected component labeling, based on morphological processing, and the back-projection of artificial neural network. The comparison is based on 23 images of blood samples taken from hematological microscopic system which is equipped with a Panasonic CCTV camera. Results are then benchmarked using a hematology analyzer Sysmex KX-21. Counting accuracies of 87,74 % and 86.97 % are obtained for the respective compared algorithms.