Application Of Deep Learning In Medical Images

Clinical medical images are of great importance to the diagnosis and treatment of diseases.The common types of medical images are CT images,X-ray images,ultrasound images,MRI and so on.In recent years,deep learning is becoming more and more preferred in medical imaging analysis,because it has the advantages of strong learning ability and good effect.At present,deep learning has made a lot of achievements in medical imaging analysis,which has promoted the great progress of medical related technology.Under this background,this paper mainly studies two applications of deep learning in medical imaging:(1)Because the X-ray used in CT scanning is harmful to human body,most CT images are obtained at low dose,which leads to the decline of CT image quality and brings some inconvenience to doctors in practical application.In order to realize CT image reconstruction under low dose,this paper constructs a network for low-dose CT image reconstruction based on GAN network.The U-Net structure is used in the generator,the parameters are optimized and modified,and the L1 loss function is added.The idea of Patch GAN is used in the discriminator.The constructed network can effectively remove artifacts,retain complete details and edges,and the reconstruction time is short,the reconstructed image quality has been greatly improved.(2)The lung images of the human body will change significantly after infection with COVID-19.In order to realize the automation of COVID-19 detection and reduce the pressure on medical staff,this paper constructs a deep learning network model based on Efficient Net.The model is completely automated without manual feature extraction,only need to input a lung image,the model can output the category and probability,it can achieve the effect of rapid automatic diagnosis of a lung image.Compared with the methods proposed in some literatures,the accuracy of this method has been further improved.In the current situation of COVID-19’s grim situation,it can provide some reference and help for the current COVID-19 control and detection work.This paper constructs two deep learning network models,through the improvement of the network structure,and then after training and adjustment to obtain the best parameters.The final model shows excellent performance on the dataset of this paper,which can greatly improve the efficiency of medical institutions and save the time of doctors and patients.At the same time,it also makes a certain contribution for integrated development of deep learning and medical images.