Automatic white blood cell differentiation.
详细信息
文摘
There are many different classes of white blood cells present in bone marrow smears. A differential count of these various types of cells gives pathologists valuable information regarding the normalcy of the patient and clues to various cancers. It is a tedious task to locate, identify, and count these classes of cells. Besides relieving the already time intensive job from the technologists, an advantage of automation is that many more cells can be inspected, giving rise to better statistical information in the differential counts. We developed a system that can be used to differentiate white cells in bone marrow images. The system is consist of segmentation, feature extraction and classification. Our major concern is focused on segmentation.;We present a new algorithm for object boundary extraction, called the watersnake. They seek solutions by iterative energy minimization. We use a new relaxation method that assigns labels (background, nucleus and cytoplasm) to objects in bone marrow images. Before the labeling procedure, the bone marrow image is divided into patches (homogeneous regions) using the watershed transformation. Patch-based relaxation updates the memberships of each patch based on the relationship with neighboring patches. This approach provides more robust and consistent segmentation, because it uses relatively larger context information than does pixel-based relaxation.