Computer Aided Classification Based On Lung Image Recognition

In the field of medical image recognition,with the continuous expansion of medical image databases,the wide application of classical convolutional neural networks has jointly promoted the rapid development of the computer image field.Computer-aided medical image recognition and diagnosis is one of the core application directions of computer images.This technology aims to quickly and accurately find lesions to improve the accuracy of diagnosis.Among them,computer-aided medical image recognition and diagnosis is also used in lung diseases.However,the current research on computer-assisted lung disease classification is mostly focused on tumor diseases,and common lung diseases are less involved.The model designed in this paper is designed for five common lung diseases: lung infiltration,emphysema,atelectasis,pneumonia and tuberculosis are classified to make up for the vacancy in this direction.After comparing traditional machine learning and convolutional neural network-assisted identification and diagnosis of lung images,this article found many problems with existing methods.Traditional machine learning is limited to the accuracy threshold and needs to manually extract image features,using convolutional neural networks alone.The network is limited to the small amount of medical image data and cannot be adapted to complex neural networks.Therefore,this paper chooses to combine convolutional neural networks and traditional machine learning methods to identify and diagnose lung images.The convolutional neural network chooses the classic VGG16 The model is improved,and the traditional machine learning model uses the gradient boosting tree model.The specific process mainly includes the following steps: acquiring lung medical image resources,performing data and expansion,and performing image data enhancement on the image data set through oblique cutting and rotation.The image data set is divided into training set and test set at a ratio of 8:2.The convolutional neural network uses the migration model of the classic neural network VGG16 to train and fine-tune the last three output layers,and improve the loss function among them to increase the clustering effect of the model.The image features extracted by the convolutional neural network are input as input data into the gradient boosting tree model for classification.Finally,the overall model is designed to combine multiple branches to form a strong classifier to improve the accuracy of classification and the stability of the model.The accuracy and robustness of the model proposed in this paper are verified and analyzed by completing the comparison simulation experiment between different models.Finally,the model is applied to the medical image recognition and diagnosis software,and the model is further analyzed through three sets of examples.verification.The relevant experimental results prove that the computer-aided lung image diagnosis method proposed in this paper has better classification accuracy and model stability than the current advanced model CheXNet.