Research On Medical Image Classification Based On Deep Learning

With the emergence of various kinds of medical instruments,more and more medical images have appeared in the clinical diagnosis.Due to the wide variety of medical images and the complex structure of human body,it is difficult to extract medical images artificially,and the adaptive ability is poor.The classification effect needs to be improved.In recent years,the deep learning method has become a hot research field in machine learning,and it has also achieved preliminary results in the application of medical image.The deep learning method does not require artificial over intervention,and can extract the image features abstractly,which is simpler than the traditional method,and the learning ability is stronger.Therefore,this paper applies the deep learning model to the classification of benign and malignant pulmonary nodules in medical images,and further improves the accuracy of classification of benign and malignant pulmonary nodules by using different methods and models in deep learning.This paper mainly includes the following aspects:Firstly,starting from the convolutional neural network,we analyzed the shortcomings of traditional classification methods for pulmonary nodules,and proposed a convolution neural network and support vector machine for the classification of benign and malignant solitary pulmonary nodules.Aiming at the classification problem of pulmonary nodules,we use the advantage of convolutional neural network feature extraction to extract image features of pulmonary nodules.After principal component analysis and dimensionality reduction,we use particle swarm optimization support vector machine to classify and identify.In this paper,we use LIDC-IDRI database to test lung nodule images.The convolutional neural network and support vector machine presented in this paper have achieved 92.75% accuracy in the classification of benign and malignant pulmonary nodules.Then from the deep belief network,the advantages and disadvantages of the BP algorithm in the deep belief network are analyzed.Aiming at the problem that BP algorithm may fall into local extreme value during fine-tuning process,this paper proposes a deep belief network improved by artificial bee colony algorithm,which improves the problem of BP fine-tuning,and then applies it to the classification of benign and malignant pulmonary nodules.By changing the number of nodes in the network hidden layer for comparison experiments,the improved deep belief network proposed by the artificial bee colony algorithm in this paper has achieved good classification results in the classification of small samples or simple network models.