Chest Image Recognition Of COVID-19 Based On Deep Learning

The COVID-19 has caused a huge impact on all countries the world.As of October 10,2021,Beijing time,the cumulative number of confirmed COVID-19 cases in the world has exceeded 230 million,and the cumulative death has exceeded 4.8 million.Accurate identification of sick individuals is a prerequisite for effectively isolating the source of infection.How to take scientific measures to reduce missed diagnosis and misdiagnosis has become the top priority in the prevention and control of the COVID19.Compared with the commonly used reverse transcription polymerase chain reaction(RT-PCR)detection,the use of medical imaging pictures such as chest X-ray pictures,chest CT scan pictures has a higher accuracy rate for the identification of COVID-19,which has become an important method for clinical use.The application of artificial intelligence technology to medical imaging has also been favored by many clinicians and artificial intelligence researchers in recent years.Deep learning is an end-to-end artificial intelligence technology.Compared with traditional image classification technology,deep learning method has excellent feature representation ability and does not require manual extraction of features.It has achieved certain achievements in the field of medical image classification.Chest X-ray imaging is fast,low-cost,and widely used in poor areas but is slightly less accurate than CT.In the identification of COVID-19,chest X-rays can be used for primary screening and then CT for identification,which can greatly improve the identification accuracy.Based on this,this paper proposes a neural network based on the classic convolutional neural network VGG16 to identify COVID-19 for the chest X-ray and chest CT image data of new coronary pneumonia.The specific content of this article is as follows:The first chapter is the introduction part.This paper introduces the research background and significance of image and picture recognition for COVID-19,and summarizes the research status of pneumonia identification and COVID-19 from two aspects of machine learning and deep learning.The deficiencies of the current research are pointed out,and the content of this study is introduced.The second chapter is the preparatory knowledge part.First,the relevant knowledge of deep learning is briefly introduced,including neurons,activation functions,loss functions,and optimization functions.Then introduce the basic structure of convolutional neural network and various classic convolutional neural network models.The third chapter is the data preprocessing part.First,the public datasets of COVID-19 chest X-rays and CT scans used in this paper are introduced and some samples are shown,and then the methods of data preprocessing including data normalization,data enhancement,etc.are introduced.Finally,the partition of the dataset is introduced.The fourth chapter proposes a VGG16 network-based dual-branch structure to identify COVID-19 chest X-ray images.First,the related concepts of transfer learning are introduced in detail,and then the experimental environment and experimental evaluation indicators are introduced.Then the model structure proposed in this paper is introduced in detail,and the experimental process and experimental results are also introduced.This chapter conducts experiments on multiple neural network models,using four different optimization algorithms for each model,and selects the best optimization algorithm as the experimental result.The experimental results show that,compared with the classical convolutional neural network,the model proposed in this paper achieves the best results in each evaluation index.Finally,the comparison with the model results proposed by other scholars shows that our proposed model has great advantages.The fifth chapter proposes a neural network fused with multi-scale models based on VGG16 network and long-term and short-term network to identify chest CT scan images of COVID-19.Firstly,the related concepts of recurrent neural network and long-term and short-term network are introduced in detail,then the model structure proposed in this paper is introduced in detail,and then the experimental results are introduced in detail.This chapter conducts experiments on multiple neural network models,using four different optimization algorithms for each model,and selects the best optimization algorithm as the experimental result.The experimental results show that,compared with the classical convolutional neural network,the model proposed in this paper achieves the best results in each evaluation index.Finally,the comparison with the model results proposed by other scholars shows that our proposed model has great advantages.Finally,there is a summary and outlook.the research work of the text is summarized,the problems existing in the research work are analyzed,and finally the research prospects of this direction are prospected.