Detection Of Diseased Brain Based On Medical Images

In today's era,the impact of brain diseases on human health is becoming more and more serious,which makes early detection and diagnosis and treatment become the key topics of current medical research.In order to avoid manual reading,doctors' subjective errors due to fatigue and inexperience,computer-aided diagnosis technology has become the trend of medical development in the current era,which can assist doctors to make better and more accurate judgment and treatment of patients with brain diseases.This article aims to use computer-aided diagnosis technology to automatically classify healthy brain and diseased brain.Using medical brain MRI images,the contrast limit adaptive histogram equalization algorithm is first used to enhance the brain image to make the image texture features more obvious and highlight the focus area.Then the wavelet decomposition is performed on the brain image,on the basis of which the texture features of the image are extracted,and the shape features are combined to perform feature recognition on the brain image.In order to speed up the wavelet transform,it is proposed to use the lifting Harr wavelet transform to decompose the brain image;by comparing and classifying the extracted multiple texture features,the kurtosis is selected as the texture feature for recognition,and then combined with the shape of the image The characteristic Hu invariant moment is used to identify the total characteristic.Finally,RBF kernel support vector machine is selected to classify the brain image,and a scheme based on grid search and harmony search to optimize the parameters of the support vector machine is proposed,resulting in higher classification accuracy.After five-fold cross-validation,a comparison experiment with several commonly used brain detection algorithms is carried out.The experimental results show that this algorithm has a high recognition rate.Convolutional neural networks can automatically extract the characteristics of image features by adjusting the parameters in the convolutional layer and the pooling layer,which is more advantageous than traditional machine learning that requires artificial feature extraction.In order to further improve the accuracy of disease brain detection,this paper uses GoogLeNet deep learning network for training classification.Considering the limited number of brain image samples,a disease brain detection scheme based on GoogLeNet and transfer learning is proposed.The experimental results show that it is more traditional Support vector machine classification,deep learning method is more effective,classification accuracy reached 100%.