Adaptive Neuro Fuzzy Inference System for White Blood Cell Classification

Abstract

Dean, Department of Computer Science and Engineering, SNS College of Technology, CoimbatoreABSTRACTThis paper describes the practical application of Adaptive Neuro-Fuzzy Inference System (A NFIS) framework forclassification of white blood cells. Differential count of the several types of white blood cells (WBC) in bone marrowsmears is applied to assist find infection, anemia and leukemia or to observe the process of treatment. In this paperwe represent a fuzzy Inference system approach to discover edge inside color bone marrow microscopic pictures, toget a solution for iteration level for image processing. In this paper, we investigate whether selective informationabout the nucleus only is sufficient to classify white blood cells. This is essential because partitioning of nucleusis very much simpler than the partitioning of the total cell, particularly in the bone marrow where the whiteblood cell density is really high.Keyword: Anfis, white Blood Cells, Fuzzy Logic, Neural Network, Segmentation, ClassificationINTRODUCTIONThe traditional method for an expert is to use a microscope to select an area of interest in a bone marrow slide,detect a white blood cell, classify the cell based on his knowledge and experience, increase the count of thecorresponding cell class, and repeat the cycle until all cells in the area of interest are counted. Research and studyabout leukocyte or white blood cell are realized a is an important issue in Hematology for a long time[2].It also recognized as a one of many challenging and complicate investigation in medical image processing andrelated field both cell component segmentation and cell classification for blood cell counting. White blood cell isthe most important cell in an immune system and can be found generally in the whole body [1]. Its function is toeliminate a foreign body in blood circulation system. In general, an increasing or decreasing in a number of totalwhite blood cells from normal level can be used statistically as an efficient primary indicator for an infection,inflammatory or disease in a human body. This figure can also be used for medical cure or follow up treatment anddrug. From the reason described above , the counting of white blood cell in order to specify a cell type and gainan amount of white blood cell in blood sample is necessary and very important. A normal stained blood sampleslide will typically compose of red blood cell, white blood cell and plasma as a background. Each WBC cell hasnaturally different in both size and shape of nucleus. Some complications usually found during manual blood cellcounting are mostly due to an unclear boundary between membrane both cytoplasm and plasma or cytoplasmand nucleus[4]. Although the blood cell counting is realized as an important tank in blood componentdiagnosis, however this is a very time consuming process and quite inherent biasing from an expertise skilland experience, person by person. These cause an implicit result intentionally [13].An automatic blood cell counter by using computer vision technique can help to perform this medical test rapidlyand accurately. Most available commercial automatic white blood cell analysis composed of three main stepsincluding cell component segmentation, feature extraction and cell type classification [9]. Among those describedprocessing steps, blood cell segmentation plays an important role as the first essential step of blood cell countingprocess to separate a composition of white blood cell into nucleus region and cytoplasm region. A correction andaccuracy of the following cell classification is affected from this segmentation .Several segmentation techniqueswere introduced and applied into white blood cell images and organ cell [3].Hea