Sensing of White Blood Cells Using Residual Network and a Comparison Study using Various Pre-Trained Network
White Blood Cells(WBCs) play a significant role in defending our body from diseases. Leukemia also called blood cancer is normally identified by testing the blood smear using a microscope by the pathologist. Both traditional methods and automated techniques are used for blood counting. In traditional methods better accuracy is achieved by the skills of technicians who are involved in diagnostic procedures. Automated techniques which are very expensive are not affordable by most of the hospitals or laboratories. Therefore, to overcome these difficulties, and for speedy diagnosis deep learning systems are employed. In this paper identification of leukemic cells whether normal or cancerous is detected through ResNet50 where an accuracy of 99.61% is achieved. A performance analysis through GoogLeNet, ResNet152, ResNet101, VGG16and VGG19 is also performed. The classification is done through the softmax layer.
Keywords - Leukemia, Deep Learning, Convolutional Neural Network(CNN), Residual Networkfirst, second, and thirdlevel headings