White Blood Cells Classification Using Convolutional Neural Network Hybrid System

Abstract

Blood and its components play a very important role in human life and it considers the best indicator in determining many biological conditions. Pathologists used optical microscopic blood images to diagnose diseases; some of these diseases are Acquired Immune Deficiency Syndrome (AIDS), Leukemia, after recognizing white blood cells (WBCs) from those images. Recently, a Computer-aided diagnosis system is used for diagnosing blood from its microscopic images. In this research, microscopic WBCs images were classified using a hybrid system where Convolutional Neural Network (CNN) used as features extractor and different machine learning algorithms used as classifiers, then the performances of these classifiers were evaluated to recognize the best of them. These algorithms include Support Vector Machine (SVM), k-Nearest Neighbor (KNN) and Random Forest, for training and test parameters we used five features that were extracted from the images. According to results of performance, the RF performed better than the other methods with a testing accuracy reached 98.7%.