A simple and accurate method for white blood cells segmentation using K-means algorithm

Abstract

White Blood Cells (WBCs) counting provides invaluable information for diagnosis of different disease. Automatic counting is helpful for improving the hematological procedure. First step in automation; segmentation; is crucial for subsequent steps; feature extraction and classification. In this paper, WBCs segmentation using K-means Clustering (KMC) is proposed. First, RGB image is converted to a*L*b*. Next, data in a* and b* are fed to KMC with proper Initial Seed Points (ISP) to extract nuclei. Then, nuclei are subtracted from prime image and data in L* is fed to KMC with suitable ISP to estimate the background. Next, both nuclei and background are subtracted from prime image and residual image is enhanced and converted to L*a*b*. Next, data in b* are fed to KMC with appropriate ISP to segment cytoplasm and finally entire cell. We achieved an average of 6.46% Segmentation Error and 93.71% Jaccard Similarity Index which are desired in segmentation.