Automated Identification of Red Blood Cells in Optical Microscope Images of Blood Smears Using Various Color Segmentation Methods

Abstract

The identification and counting of red blood cells (RBCs) in microscopic blood images can provide useful information concerning the health of patients. Software-based cell counting has the advantages of objectivity, speed, and convenience over the manual method. Most of the automated RBC counting techniques in literature employed the grayscale or the green component of the red-green-blue (RGB) color images of blood smears. This work focuses on comparing the effect of using different color layers on the performance of software-based RBC counting. Ten color layers were extracted from different color models of blood smear images along with the grayscale conversion. Two comparisons were made: a comparison of contrast and a comparison of RBC counting performance using 52 blood smear images. The RBC contrast in the magenta layer of the cyan-magenta-yellow-key (CMYK) color model was at least 230% higher than that in the other layers. Additionally, our results indicated that using the magenta layer can provide better RBC counting performance when compared to the green, grayscale, and key layers with the p-values of p=0.0283, p<0.0001, and p<0.0001, respectively.