ABSTRACT
The identification of acute leukemia blast cells in colored microscopic images is a challenging task. Usually, visual assessment for microscopic images of blood samples is performed. However, considering the quick advances in utilizing different image processing techniques, rapid and more accurate assessment can be achieved. Therefore, this paper proposes an enhanced automatic method to detect Acute Lymphoblastic Leukemia (ALL) utilizing microscopic blood sample images.
Our proposed methodology includes Color-based segmentation using the K-means clustering technique with morphological feature extraction algorithms and cell classification. The proposed method was tested on blood microscopic images from the ALL-IDB1 database, University of Milano, Italy. ALL_IDB1 is comprised of 108 blood cell images (healthy and leukemia) in which the lymphocytes are labeled by expert oncologists. The accuracy achieved was 97.2%. This relatively high accuracy suggests that the colour-based method for segmentation was more efficient and acceptable compared to the traditional thresholding methods and the methods that do not consider the overlapped cells.
