Particle swarm optimization based blood image segmentation for automated leukemia detection

Leukemia is a disease that affects blood forming cells in the body. The cell growth in acute leukemia disease occur rapidly and uncontrollable. Therefore, early detection of the disease is necessary for proper treatment management. Recently, computer-aided detection and diagnosis (CAD) approaches have been developed to assist medical staff interpreting medical images. Image segmentation is one of the important role in CAD for diagnosing and verifying the disease like such as leukemia. The conventional method of image segmentation was quite lack in performance. This is due to the inconsistency of background intensity of the image used where the data provided cannot accurately collect for the process to be done. To overcome such a problem, this paper presents the method for lymphocyte image segmentation by using a Particle swarm optimization (PSO). A microscopic blood images were used to identify the white blood cell (WBC) that has a leukemia or not. The blood sample slide image was segmented to remove the Red Blood Cell (RBC) and the unwanted background and will leave only the White Blood Cell (WBC) image. For the segmentation technique, the proposed method is based on the S-component (Saturation) of HSI (Hue, Saturation, Intensity) colour model. The S-component is obtained and fed into the PSO to perform the segmentation process. The study also proposed a new method that utilized centroids obtained from K-means as initial centroid for PSO, called hybrid K-means-PSO. The results of the proposed methods are benchmarked against the most commonly used method, K-means clustering to evaluate its effectiveness in segmenting WBC images. Simulation results indicated that both PSO and hybrid K-means-PSO methods have a better accuracy compared to K-means with the highest accuracy obtained is up to 98.86%.