K-Means Clustering Based on Otsu Thresholding For Nucleus of White Blood Cells Segmentation

Abstract

White blood cells function as the human immune system, and help defend the body against viruses. In clinical practice, identification and counting of white blood cells in blood smears is often used to diagnose many diseases such as infection, inflammation, malignancy, leukemia. In the past, examination of blood smears was very complex, manual tasks were tedious and time-consuming. This research proposes the k-means clustering algorithm to separate white blood cells from other parts. However, k-means clustering has a weakness that is when determining the initial prototype values, so the otsu thresholding method is used to determine the threshold by utilizing global values, then proceed with morphological operations to refine the segmentation image. The results of segmentation are measured by the Positive Predeictive Value (PPV) and Negative Positive Value (NPV) parameters. The results obtained prove that the use of otsu thresholding and morphological operations significantly increase the value of PPV compared to the value of PPV that does not use otsu thresholding. Whereas the NPV value increased but not significantly.