Research On Image Analysis Method Of Alzheimer’s Disease Based On Functional Brain Networks And Machine Learning

Alzheimer’s disease(AD)is one type of neurodegenerative disease and mild cognitive impairment(MCI)has been implicated as a preclinical phase of AD.Patients with AD in clinical have been demonstrated a series of symptoms of brain function damage,such as progressive memory impairment and cognitive ability decline.The functional image of brain contains time series,which can reflect the patient’s brain function information,and is of great significance to the scientific research and clinical diagnosis of Alzheimer’s disease.With the continuous improvement of computer hardware architecture and the computing power,machine learning has made great progress in the field of image classification.However,medical functional images are high-dimensional images which has a three-dimensional or four-dimensional structure,and the sample size is limited This kind of data can easily cause the Curse of Dimensionality Problem when processed by machine learning models.Therefore,how to use functional images for auxiliary diagnosis is a problem worthy of study.Researches have shown that the brain networks of patients with Alzheimer’s disease has abnormal changes,but the distinguishing ability brought about by these changes are unclear.If the abnormal changes in the brain networks can be utilized to distinguish AD and MCI from Normal Control(NC),it will provide a new approach for the diagnosis of Alzheimer’s disease.In addition,brain function images of different modalities can reflect human brain function information from diverse perspectives and obtain more features than single modality,which also contributes to the improvement of diagnostic accuracy.In response to the above-mentioned problems and actual clinical needs,this thesis mainly uses brain networks and machine learning analysis methods to study functional Magnetic Resonance Imaging(fMRI)and Diffusion Tensor Imaging(DTI)data of patients with Alzheimer’s disease.The main content of this article are mainly in the following aspects(1)Organize fMRI and DTI data,convert data with DICOM format to NIFTI format in batches and desensitize the data.Then use existing Alzheimer’s researches conclusions to select seeds region.(2)A method of Alzheimer’s disease analysis based on fMRI brain networks is proposed.Construct a voxel-wised brain networks by processing fMRI data,and then perform statistical analysis on the brain networks to find out the brain regions with significant differences in the brain networks of AD,MCI and NC.Take the brain networks values of these brain regions as the features of the subjects,the dimensionality reduction and classification are performed subsequently.High classification accuracy is achieved in the experiment,which proves that the local brain networks features with hippocampus as the seed can be used as a biomarker of Alzheimer’s disease,and also verify the effectiveness of this analysis method.(3)A functional brain networks fusion classification framework based on sparse discriminant analysis is proposed.The framework first preprocesses fMRI and DTI data respectively,and constructs brain networks based on the region of interest.Brain networks features are extracted for feature selection and feature fusion on the two modality.Finally the Sparse Discriminant Analysis algorithm is used for dimensionality reduction and classification.The diagnosis effect acquired from this framework.is better than that of the single modality.The experimental results prove the effectiveness of the functional brain networks for modality fusion and the superiority of the Sparse Discriminant Analysis method.