Computer Aided Diagnosis System Based On Medical Imaging Of Brain Pathology

Brain cancer(BC)and Alzheimer’s disease(AD)are recognized by the World Health Organization as serious diseases that damage human life and health.Rapid and accurate diagnosis is a key factor in saving countless lives.The diagnosis methods can be divided into symptom diagnosis and computer-aided diagnosis.Symptom diagnosis refers to judging the probability of illness by analyzing the patient’s condition,which is inefficient.Computer aided diagnosis(CAD)is a way to judge the location of lesions by analyzing brain medical images,which is most commonly used in clinic because of its fast speed and high efficiency.The basis of computer-aided diagnosis mainly depends on the computing power of the computer,so as to find the location of the disease and improve the efficiency and accuracy of the final judgment.However,with the development of the times,a large number of different types of medical imaging data are increasing.Analyzing and judging the increasing data and images not only requires doctors to have excellent experience,but also time consumption is a big problem due to the increase of the amount of data.In order to meet the needs of medical diagnosis,radiomics based on deep learning has become the product of the development of the times.It not only contains the advantages of high efficiency,strong plasticity and universality of deep learning,but also can extract more features from images than computer-aided diagnosis.On this basis,based on the depth network,this paper proposes the radiomics method of brain tumor magnetic resonance imaging(MRI)image,so as to assist doctors to quickly and accurately obtain the exact location of the patient’s brain tumor.The main research work is carried out around the following aspects:For the prediction task of brain MRI,the traditional prediction model is inefficient and needs high-quality professional doctors.In model training,due to the small amount of data or the lack of graphic information,the model will be over fitted.Therefore,for the brain MRI image segmentation task,a method of brain image segmentation using deep learning is proposed,this method includes the concepts of stationary wavelet transform(SWT)and growth convolutional neural networks(GCNN).The important goal of this work is to improve the accuracy of traditional systems.In this paper,support vector machines(SVM)and convolutional neural networks(CNN)are compared and analyzed.The experimental results can divide the overall structure of the tumor,and this method is better than the results of the two classical networks in terms of accuracy,peak signal-to-noise ratio,mean square error and other performance indexes.This paper improves on the traditional classical network U-Net.Similar to the growth convolution network in Chapter3,it also adds hole convolution,residual network module and attention mechanism module.The final result is better than Chapter3.It can not only locate the location of brain tumors and assist doctors to reduce the probability of misjudgment,but also divide three regions of brain tumors to know the future trend of the disease.Finally,through medical statistical analysis,this paper tracks the general trend of the image markers of tau protein in the brain of patients with mild cognitive impairment(MCI)through the brain MRI images of patients with mild cognitive impairment(MCI),so as to determine the specific causes of mild cognitive impairment when they turn into Alzheimer’s disease.