Detection Of A? Deposition In MR Images Based On Pixel Feature Learning Algorithm

Alzheimer's disease,also known as senile dementia,is a progressive degenerative disease of nervous system occurring in or pre the geratic period,which affects the health and happiness of more than 50 million people around the world.Amyloid ?-protein deposition plays a vital role in the pathogenesis of AD and also is an important prevention and treatment target for AD,therefore early detection of A? deposition in the brain is the key to early diagnosis of AD.Magnetic resonance imaging is a perfect imaging technology.It not only can precisely and quantitatively reflect the changes in structure and function occurring in different brain tissues,but also can represent metabolic concentration.MRI is inexpensive,noninvasive,has no radiation and no tracer,and has high resolution,so it has been widely applied in clinical applications for diagnosis of AD.However,the public research on detecting A? deposition's information is very rare yet.Aiming at this problem,this thesis proposed pixel feature learning method,thereby constructing detection algorithm for A? deposition's information in brain MR image,which establishes the relevant theoretical basis and methodoloty.The main contents are as follows:(1)Completed images acquisition and preprocessing: Firstly through the MRI technology respectively scanning the APP double transgenic mouse and control group,then the brain T2 weighted structure images were obtained;Secondly,deal with the mice head,get brain histological slices and form histological images,and then matched with brain MR images of mice to make up image pair.(2)The research on detecting of A? deposition information based on pixels feature selection method from brain MR images was studied and proposed: Firstly,the brain region was segmented from brain MR images of APP mouse model;Secondly,the pixels in the segmented brain region were extracted according to row as a feature vector;Thirdly,multiple feature learning classification algorithms were conducted on the extracted features,and the optimal feature subset was obtained.By repeating the same processing above,several optimal feature subsets were obtained,then the final optimal feature subset was obtained by voting mechanism;Finally,based on the final selected feature subset and elastic mapping method,the corresponding pixels on MR images were found and marked for showing the information about A? plaque deposition.(3)The research on detecting of A? deposition information based on pixels feature grouping method from brain MR images was studied and proposed:Firstly,the brain region was segmented from brain MR images of APP mouse model;Secondly,the pixels in the segmented brain region were extracted according to row as a feature vector;Thirdly,feature grouping algorithms were conducted on the extracted features,then multiple classification models were trained based on the grouped features;According to the output classification results,the optimal feature group was obtained.;Finally,based on the optimal feature subset and elastic mapping method,the corresponding pixels on MR images were found and marked for showing the information about A? plaque deposition.In this paper,the problem of detection of A? plaque deposition information can be converted to the problem of classification of AD and CTL based on selected optimal pixels features.The classification accuracy based on the pixels selected by the proposed detection algorithms can achieve about 87%,which realized the noninvasive detection of A? plaque deposition based on brain MR images.This paper provided new research ideas and methods to reference for early diagnosis of AD,and also provided theoretical basis and methodology to promote the development of related clinical and practical research.