Research On AD Diagnosis Model Based On Deep Learning Of Multi-modal Data

Alzheimer’s disease(AD,Alzheimer Disease)is a typical progressive fatal neurodegenerative disease.Mild cognitive impairment(MCI,Mild Cognitive Impairment)is the early stage of AD.Because the pathological information presented by the patient in the early stage is not obvious,it is easy to be mistaken for natural aging,and it will be more difficult to take effective treatment measures as the condition deteriorates.Therefore,early detection and treatment are particularly important.Due to the limitation of medical conditions,early reliance on single-mode data for analysis,the diagnosis effect is not ideal.In recent years,with the development of medical imaging and computer technology,multimodal data and deep learning methods have gained extensive attention in this field.Multimodal brain imaging data usually have higher dimensions and complexity.Looking for an efficient method to extract valuable features and diagnostic effects in complex data sets is the focus of this study.From the perspective of deep learning,the subject effectively processes multimodal data of brain images in order to achieve the purpose of computer-assisted AD diagnosis.main tasks as follows:(1)Preprocessing of brain image data.Considering that the original image data will be artificially shaken during the acquisition process,the shape and size of the brain area of different subjects are different.In order to improve the image quality and signal-to-noise ratio,the necessary pre-processing is performed on the original image data.Use VBM software to perform spatial standardization,image segmentation,modulation and smoothing on structural MRI data;perform head correction,registration,normalization,standardization and normalization on positron emission computed tomography(PET,Positron Emission Tomography)data Smoothing,etc.,lay the foundation for later feature extraction.(2)Multimodal data feature extraction.In order to give full play to the role of multimodal data features in AD diagnosis,the paper completes brain image data feature extraction from three aspects.AAL is used to extract the features of the whole brain from the image;based on the saliency analysis method,the features of the brain regions with large differences are obtained;and the effective features are extracted from the region of interest(ROI)based on the three-dimensional sparse autoencoder.(3)Construction and experiment of diagnostic model.In view of the great success of deep learning in the field of image analysis,this paper has conducted in-depth research on three-dimensional sparse autoencoders and convolutional neural networks,and proposed a AD diagnostic model of convolutional neural networks based on three-dimensional sparseautoencoders.Carrying out relevant comparative experiments,the results show that the proposed model has a better diagnostic effect;on the other hand,it also verifies that the multi-modal data is better than the single-modal data.