Classification of White Blood Cells Using CNN and Con-SVM

Abstract

White blood cells are one of the most important cells for the human body that make up the human immune system. These cells in the blood can contain important information for the diagnosis of many diseases as well as protecting the body. Nowadays, blood analysis performed by doctors can cause both waste of time and human errors. A support diagnosis system can be created to prevent these errors, to perform rapid diagnosis and to reduce the workload. In this study, 12442 white blood cell images consisting of four types of white blood cells were tried to be classified. These images were first classified with CNN (Convolutional Neural Networks), then classification is performed using Con-SVM (Convolutional features - Support Vector Machines). While an accuracy rate of 83.91% was obtained with the CNN model, an accuracy rate of 85.96% was observed with the Con-SVM model.