Classification of White Blood Cells from Microscopic Images

Abstract

Image processing has become an important pillar in modern age of technology. It plays a significant role in medical field. It makes the diseases diagnosis and analysis more accurate and speedup this process. Human health is usually measured from the immune system of body which is natural defense system against infections and invaders. Human immune system contain white blood cells (WBC‟s) which are good indicator of many diseases like bacterial infections, AIDS, cancer, spleen, etc. White blood cells have been further classified into four major classes such that monocytes, lymphocytes, eosinophils, neutrophils on the base of their nucleus, shape and cytoplasm. Traditionally during tests in laboratories, pathologist and hematologist analysis of these cells in blood is done through microscope and then classified manually as normal or abnormal. Counting is also done manually. This hard work takes more time and increase the chance of human error. In last decade, some research has been done on this field of medical science in order to automate the process. In this research, a method is introduced to automatically classify the sub types of white blood cells and compare the results to previous techniques. The proposed method includes: preprocessing, color pallet based segmentation, hybrid features (pattern and shape based features) extraction and neural network. The classification accuracy of 96.5% is achieved through the proposed method.