Published Paper Details:

Leukemia is a type of malignancy blood cancer occurring in both adults and children due to a rapid development of abnormal white blood cells. In the diagnostic process, abnormal white blood cells (blast cells) play an important role for hematologists. Whereas the treatment is usually allocated based on the leukemia types and subtypes which depending upon the morphological features of the cell under microscope. Thus, the blast cells segmentation is the first step for diagnostic process. Generally, this paper aims to develop a methodology to detect and localize acute leukemia blast cells based on image processing techniques using peripheral blood smear images. To achieve this aim, the blast cells were localized by removing other blood components and then the region of interest was segmented into sub-images. The problem of overlapped cells was also addressed for better classification. This stage was accomplished by using various image processing techniques such as; color transformation, threshold algorithm, mathematical morphological operations, and watershed segmentation. The k-means clustering and mathematical subtraction operation were used to further segment the blast cell into its nucleus and cytoplasm, respectively. Based on the Image Segmentation Quality Scores, the proposed algorithm in this study was able to segment the blast cell region and its nucleus with a high level of accuracy of 97% and 94%, respectively. The proposed segmentation algorithm achieved an overall accuracy of 96% and 93% to discriminate between acute leukemia and normal cell region and their nucleus region, respectively.