Abstract

Info

Author(s):: JI Chunyang; XU Xiulin; WANG Yan; School of Medical Instrument and Food Engineering, University of Shanghai for Science and Technology, Shanghai 200093, China

Keywords:: ; deep neural network; convolutional neural network; artificial intelligence; tumor cell; review

Abstract:: Abstract: Deep neural network (DNN), as the main branch of artificial intelligence, is a computer program based on the imitation of the way of how human brain thinks, aiming to simulate the way the human brain processes information for classifying or predicting things. The universality of DNN includes self-learning, self-adaptation and associative memory. DNN can perform various tasks even without a priori background. In recent years, DNN has received extensive attention from domestic and international medical communities. Moreover, some major breakthroughs have been made in accurately classifying the automatic recognition of digital images of tumor cells. DNN gains experience through intensive learning, which enables doctors to provide patients with an accurate treatment strategy. Herein the latest research progress of DNN in tumor cell recognition is reviewed, and the principles of convolutional neural network, deep belief network, generative adversarial network and deep residual network as well as their applications are elaborated. The neural networks based on different models are compared, and the accuracy and performance of various models in application are analyzed. Finally, the problems and future development trends of DNN in tumor cell recognition are pointed out.