Analysis of leukemia development based on marrow cell images

To analyze the development of leukemia in bone marrow, a C-Mean fuzzy clustering recognition method is proposed to identify cells from sliced marrow images so that the movement of promyelocytes to the blood can be tracked. In our image processing, red cells, leukocytes, megakaryocyte, and cytoplasm carry the same color because of the failure of staining. We use the gray value distribution, the texture patterns to construct a cluster based evaluation groups to establish the minimum object function value which when it fall in the desired category, indicates the matching of the objects and thus we can obtain the number of the specific cells in the picture. Multiple pictures are tested using this method. The 80% recognition rate can be easily reached. There are cases when the pictures are poorly illuminated or blurred which can hardly be recognized, these kinds of pictures are manually removed because of the limitation of digital picture processing.