Segmentation of white blood cells using integrated process of improved spectral angle mapper, gram-schmidt orthoganalization with K-means clustering

Abstract

White blood cells (leucocytes)are main constituents of blood, plays a major role in immune system. They rescuing our body from foreign materials and infectious diseases. All five types of WBCs are origin from multipotent cell bone marrow also called hematopoietic stem cells. These five types are grouped into granulocytes and agranulocytes. Leukemia a blood cancer occur due to abnormal development of leukocytes(WBCs). In united states 62130 people receive leukemia diagnosis in 2017 and 24500 deaths occurs by this disease its effects are even more in undeveloped countries. Counting and identifying of white cells in images of microscope is so tedious, time taking and the results depends on experience of hematologist. The proposed algorithm is a combination of spectral-spatial method for segmenting cytoplasm and nucleus of WBCs from the microscopic, spectral and hyperspectral images. Here we use an Integrated method of improved spectral angle mapper(ISAM) with Gram-Schmidt orthogonalization process and k-means segmentation process. The resulting spatial spectral information of WBCs gives an accuracy of above 95% for both nuclei and cytoplasm.