Improved DenseNet And Its Application In Medical Image Analysis

In this paper,the Dense Net neural network algorithm is improved,and the U-Net network is introduced to detect and study the new coronavirus pneumonia and ovarian cancer,and a large amount of medical image data is automatically cut,leaving the target lesions.It is hoped that the computer can learn from the data.The identified pathological features can provide better data conditions for the learning of neural networks,and can also achieve efficient identification of computer intelligent detection,and obtain better classification results to achieve the goal of machine intelligent diagnosis.This paper mainly completes the following work:1.The relevant medical theoretical knowledge of the new coronavirus was introduced,including the early,middle and late symptoms of infection,and chest X-ray manifestations.As well as related medical theoretical knowledge and clinical manifestations of ovarian cancer,we can better deal with medical images of new coronavirus pneumonia and ovarian cancer.In terms of data processing,image enhancement is firstly performed to improve the brightness in the data to help the characteristics and locations of the lesions to be better displayed;then image cutting,especially the ovarian cancer data set,separates the lesions in the ovarian cancer data.,discard the data that affects the model training and improve the accuracy.2.It systematically introduces several theoretical knowledge of deep learning,as well as various modules in the neural network structure,especially the basic theory of Dense Net neural network,which provides a theoretical basis for the construction and transformation of structural models.3.The network structure of the intelligent diagnosis algorithm based on ovarian cancer is constructed,and the U-Net network is used to realize the automatic cutting of data on the basis of the original neural network,and a new network structure U-Dense Net is obtained.Fusion,Dense Net169 is used in the layer selection of Dense Net network,and compared with other deep learning network models such as Dense Net,Res Net,VGG16,etc.The experimental results show that the results of the U-Dense Net network structure are the best.The data set based on ovarian cancer will also be used for experiments with existing algorithms,and the experimental results will be compared with the results of U-Dense Net network.The results show that U-Dense Net can achieve high-accuracy detection for ovarian cancer data.For the purpose of extracting features and accurately identifying ovarian cancer lesions,this model can be extended to other medical image analysis,which has practical significance for the diagnosis of the medical industry.It is of great clinical significance to help patients find other tumor diseases early,treat them early,and relieve the suffering of patients.4.The network structure of the intelligent diagnosis algorithm based on the new coronavirus pneumonia was constructed.In order to further verify the detection results of the U-Dense Net model,the model was applied to the data set of the new coronavirus pneumonia,and compared with Res Net,VGG16,etc.Multiple experiments show that Dense Net169 is suitable for the detection and analysis of new coronavirus pneumonia,and has the best accuracy,so U-Dense Net also has high applicability in other diseases.