Morphometric Analysis Of Hippocampal Subfields In Alzheimer’s Disease

Alzheimer’s disease(AD)is a chronic disease.Its clinical manifestations are generally cognitive impairments,and it is the most common and high-grade in elderly diseases.Mild Cognitive Impairment(MCI)is also a cognitive disorder and is a condition between the normal elderly and dementia.At this stage,because of the slow process of AD,the clinical symptoms in the early stage of onset are not very obvious.Once the patient has apparent cognitive impairment,many conditions have entered the late stage and can not be effectively treated.Faced with this situation,the early diagnosis and prediction of AD,can delay the progress of the disease,it is the focus of research and attention in recent years.A fundamental bottleneck in current research is the need for human interaction to draw subfield boundaries in images.This severely limits the amount of data that can be analyzed,as manually dividing the subfields in an ultra-high resolution image can be an extremely time consuming process.In addition,the artificial rendering is affected by the variability between the inside of the observer and the observer,which confounds the subsequent statistical analysis of the result.Therefore,this paper proposes an automatic segmentation algorithm based on the Bayesian model,and the construction of an in vitro map for the algorithm.Compared with the traditional in vivo map,the in vitro map can significantly improve the accuracy of the segmentation results.Specific work as follows:(1)T1 weighted MRI brain images of 310 patients in the ADNI database and healthy controls were selected.55 cases of Alzheimer’s disease,180 cases of mild cognitive impairment and 75 healthy control group.As well as demographic information of the sample’s two clinical scales(MMSE and ADAS-Cog)score results,age,gender,and education level.(2)The selected T1 images were skull removed based on the Brain Surface Extractor(BSE)algorithm.For skull images after culling,each gray matter voxel of the image was anatomically labeled according to the volume of each brain region in the established map to fill the underlying cortical structure to create sub-cortical quality.Then the image is normalized in intensity,that is,the gray value of the input image is normalized.Rigidly register each input T1 image with the constructed map so that each image is at a reference point.Finally,the left and right hemispheres were surface extracted,the segmentation results were obtained by extracting the rough surface contour and smoothing,and the brain regions of the segmentation results were labeled.(3)A hippocampus subdivision based on a fully automated method of hippocampal anatomical atlas and MRI data generation models.First of all,to realize the construction of the atlas,the magnetic resonance imaging data for establishing the atlas of the statistical graph is composed of the in vitro and in vivo MRI training data,and the mathematical framework for establishing the atlas includes the basic model based on the tetrahedral grid representation The Bayesian algorithm is introduced to optimize the parameters of tetrahedral mesh.Second,the segmentation algorithm of the in vivo MRI scan was analyzed using the constructed map.This method focuses on simulating the spatial distribution of the hippocampal region and surrounding brain structures(base segmentation)and learning from labeled training data.Automated segmentation of hippocampal MRI in vivo.(4)The statistical analysis of the experimental results was conducted to study the relationship between the volume of hippocampal subdomains and AD.The contents of this paper are briefly summarized,the results show that,AD patients,MCI patients compared with healthy people,the volume of the hippocampus significantly shrink,the specific performance of the hippocampus of each sub-volume there are varying degrees of atrophy,For the early diagnosis of AD has important significance.And the future research is prospected.