Article Preview
Top1. Introduction
At present, computer aids have been rigorously emerging in digital pathology field, including segmentation, identification and classification of cell/ nuclei. The microscopic blood image analysis finds it usefulness in differential and complete count of White blood cells (WBCs) and components, hemiparasite and hematological diseases diagnosis, content-based image retrieval systems, medical decision support systems, computer aided diagnosis, treatment follow up, guided surgery, and many more.
When a doctor suspects any disease, firstly suggests a complete blood test or differential blood test based on the severity of symptoms. Complete blood test results in the total count of platelets, erythrocytes (red blood cells) and leukocytes (white blood cells). A Differential blood test classifies and counts the five major types of leukocytes. Two broad classifications of WBCs are Agranulocytes and Granulocytes. Agranulocytes are Lymphocytes and Monocytes. Granulocytes are Neutrophils, Basophils and Eosinophils. The WBCs is a vital component of the human blood system that develops and maintains an immune system in the body. The WBCs count determines the human health status, which is usually in the range of 4,500-10,000 for a normal adult. The count of WBCs is very significant for diagnosis of diseases and its detection should be done at the earliest. Leukopenia is a result of low WBCs count and may detect the presence of Human Immune Virus (HIV), autoimmune disorders, liver and spleen diseases, and radio therapy to name a few. Similarly, Leukocytosis is a result of high WBCs count that may predict the presence of anemia, allergies, pregnancy, tissue damages, asthma, blood cancer, and many more. On the other hand, Leukemia (unusual and abnormal production of WBCs beyond expectation) is a deadly disease which can develop in any organ of body, directly affecting blood cells, leading to cancers. Continuous monitoring of the WBCs count must be done through blood tests to analyze the effect of treatment on patients, which stresses the need for development of techniques for fast and real time automatic detection and recognition techniques.
Numerous Image Processing (IP) and Computer Vision (CV) based software solutions are available for automatic WBCs image analysis. Segmentation often follows representation of isolated objects for faster analysis in later processing steps. The classification is the process of labeling the segmented objects according to some well-defined features, after successful segmentation of WBCs from background and other articrafts. The WBCs is classified into its 5 constituent types: Neutrophils, Basophils, Eosinophils, Lymphocytes and Monocytes. To perform classification, unique distinguishable set of features that represent individual WBCs types should be extracted from images. Two broad categories of WBCs are granulocytes and a-granulocytes. Cells are characterized by the presence of number of nuclei lobes, granules and shape of nucleus. Neutrophils have 3-5 blobbed nucleuses with small granules. Eosinophils have red granules with blobbed nucleus. The granules are larger in eosinophils than neutrophils. Basophils are featured by a lobed nucleus filled with large black and blue granules that cover entire nucleus. Lymphocytes are round shaped and Monocytes with horse-shoe shaped nucleus respectively. The nucleus to cytoplasm ratio is high in monocytes compared to lymphocytes (De et al., 2007). Hence WBCs are featured by variance in color, intensities, saturation levels, shape and texture. Features like average area, bounding box, and centroid are used to identify and segment the cells.