Image Segmentation Approach for Acute and Chronic Leukaemia Based on Blood Sample Images

Abstract

The uncontrolled development of abnormal white blood cells (blast) in bone marrow is the starting point where leukaemia cancer begins. It does not die like it should. Instead, it goes on dividing and crowding out the healthy blood cells, making it difficult for these healthy bloods to function normally. The diagnosis process from haematologist consumes a lot of time. Therefore, a good segmentation procedure is necessary in order to successfully identify and diagnose leukaemia cells automatically from blood sample images. This paper proposes segmentation procedures which consist of contrast enhancement, extraction of hue component information, as well as segmentation based on Otsu’s thresholding and watershed technique. The experimental results shows that the proposed segmentation procedure has successfully segmented 200 images consisting of acute and chronic leukaemia with average accuracy, sensitivity and specificity obtained of 98.90%, 82.14% and 99.49%, respectively. The result of segmentation performance achieved shows the significant of this approach. Based on results obtained, this segmentation technique is applicable to segment both acute and chronic leukaemia images with good segmentation performance.