Identification Of Leukemia Diseases Based On Microscopic Human Blood Cells Using Image Processing

Abstract

White blood cell cancer or what is often referred to as leukemia is a very dangerous disease. Until now the process of recognizing white blood cell cancer or leukemia is still done conventionally, which method can cause a diagnosis difference if done by different medical personnel. To answer these problems, a desktop-based application will be built. This application is made to help the process of identification and classification of types of leukemia using the technique of processing microscopic images of human blood cells. By doing several approaches, such as pre-processing, using the median filtering method, conversion color (RGB to HSV) to clarify the blood image and tresholding to get the image pattern of blood cells that have. Segmentation will be carried out to separate between objects that will be taken with unnecessary objects. This system will use extracted form features from each cell in microscopic blood images. This feature retrieval will be used as a classifier input which is divided into two classification classes, namely Acute Lympotic Leukemia (ALL) and Acute Myolegenous Leukemia (AML). With this method an accuracy of 80% can be generated for the detection of ALL cells separately (one cell), 100% for AML detection separately (one cell) and 90% for cell detection throughout (many cells).