Pattern Recognition And Its Applications Based On Brain MRI Data

Functional magnetic resonance imaging(fMRI)has been one of the most important tools for clinical diagnosis and medical research owing to it is non-invasive and non-ionising,they provide brain imaging tools for scientists to reveal brain structure and function changes.This paper focuses on extracting features from functional magnetic resonance data to provide biomarkers for disease or therapeutic effects.Combined with the traditional feature extraction methods(such as local area consistency and low frequency oscillation amplitude)and the current popular dynamic network feature extraction method to extract features.The details are as follows:1.Using Regional Homogeneity(ReHo)and Amplitude of low-frequency fluctuation(ALFF)to find the difference between brain regions in patients with perimenopausal syndrome(MPS)and normal subjects,Thereby providing a reliable biomarker for the disease.Medical research now shows that perimenopausal syndrome is likely due to abnormal cerebellum-controlled hormone release.In order to justify the theory,we validate the previous hypothesis by analyzing the resting magnetic resonance data of the acquired perimenopausal patients.We believe that ReHo and AlFF are a potential tools to study MPS.So we hope that our results will improve the pathophysiology of MPS.2.In order to understand the relationship between the different states of the brain,we use handedness fMRI data to research the relationship between the two states.The brain activation region was obtained by using the generalized liner model(GLM)and single T test,and select 12 regions of interest(ROI)as seeds.In order to get a brain network,we Calculate the functional connectivity(FC)between the corresponding ROI regions in the resting state.Since both the task data and the resting data are time series data,each subject can get a dynamic brain network.The clustering methods is used to cluster the dynamic brain network of each sample,resulting in the division of the brain network in each subject.All subjects dynamic brain network data are clustered together to get seven class centers.Each individual with the group on the group on the 7 class center to calculate the correlation value,so that left hand and right hand have to get around 7 classes.Fianllly,use the two-sample T test to check the corresponding category,and then thtough the FDR get the difference between the brain.We hope tofind a clustering algorithm to solve the current dynamic network clustering dilemma,in order to further explore the real brain network to provide new avenues.