White Blood Cells Nuclei Localization Using Modified K-means Clustering Algorithm and Seed Filling Technique
The presence of White Blood Cells (WBCs) in the body of human has a great role in the protection of the body against many pathogens. The recognition of the WBC is the first important step to diagnose some particular diseases. The pathologists usually use an optical microscope to recognize WBCs, but, this process is a quite tedious, time-consuming, error prone, very slow, and expensive. In addition, it needs experts with long practice in this field. For these reasons, a computer assisted diagnostic system that helps pathologists in the process of diagnosis can be effective, easy and safe. This research is devoted to develop a system based on digital image processing methods to localize WBCs nuclei. The proposed system involved a collection of pre-processing and segmentation algorithms that are capable of allocating the nuclei in different shapes of WBCs from a microscope images. To accomplish this task, a combination of local enhancement using histogram statistics, modified k-means clustering, normalization, convert to binary image using a suitable global threshold, islands removing and holes filling based on seed filling technique, and nucleus localization algorithms were performed. The features of WBCs images in the tested dataset make the WBC nuclei extraction process representing a great challenge. The test results indicate promising ability to completely isolate the nucleus from other parts of the cell. The analysis presents a high similarity between the ground truth samples and the results obtained by the proposed method. The precision percentage of the proposed method applied on the tested dataset images is 97.21% and F-score percentage is 96.23%.