Brain Mechanism Study Of Migraine Patients Without Aura Based On Multi-scale ICA Methods And Cortical Structural Analysis

Migraine is the most common primary headache,outbreaks most in childhood and reaches its peak in adolescence,and its prevalence in population is around 5% to 10%.Light and acoustic stimulation or daily activities will aggravate headaches,which causes great inconvenience to the patients' life.Functional magnetic resonance imaging(f MRI)was used to study the changes of brain function in patients with migraine,and the structure magnetic resonance imaging(s MRI)was used to measure the morphology and structure of the brain tissues.Previous studies have found that the functional connectivity of the resting state brain network is related to the specificity of the pain sensory transmission process.Therefore,the functional connectivity of the resting state brain network in migraine patients may be abnormal.The brain region activation study also found that the brainstem region was significantly activated during migraine headaches,there may be differences in functional connectivity between brainstem and other regions in migraine patients.In addition,the use of imaging features to construct a classification diagnosis system has been applied in many brain diseases.Thus,this thesis collected functional and structural magnetic resonance images of migraine patients,to investigate the abnormal functional connectivity of resting state brain networks and brainstem and to construct the classification model of migraine patients and normal controls basing on cortical morphological characteristics.This thesis collected 71 subjects'(37 patients,34 controls)resting f MRI,and used the Independent Components Analysis(ICA)and Masked Independent Components Analysis.MICA)to study the abnormality of resting state functional connectivity in patients with migraine without aura: Firstly,basing on the ICA method,we extracted 7 resting state brain networks in patients without aura migraine and studied the abnormality of functional connectivity in the network;Secondly,given the pivotal role of the brainstem region in migraine pathology,we used the brainstem region as a Region of Interest(ROI),and used the MICA method to study group differences of functional connectivity between brainstem sub regions and whole brain voxels.Besides,this thesis collected 120 subjects'(60 patients,60 controls)s MRI,and extracted four types of morphological parameters of the two groups basing on Desikan_killiany 68 brain region template: the average cortical surface area,the average cortical thickness,the average gray matter volume and the mean curvature,then used statistical analysis method to study the structural abnormalities of migraine patients.Thirdly,trained the classification models by using elastic network algorithm,then calculated the correct classification rate(CCR)and area under curve(AUC)of different classification models by using 5 fold cross validation method,and compared their classification abilities.The results are as follows:(1)Among the ICA analysis results,four intrinsic networks with significant group differences include: default mode network,sensorimotor network,executive control network and frontoparietal network(P <0.05,FDR correction).(2)the masked ICA analysis results showed functional connectivity between the medulla oblongata and regions of the occipital lobe of controls was significantly higher than the patients',while the functional connectivity between the anterior superior inferior pons and regions of the temporal lobe and occipital lobe of patients was significantly higher than control's.It indicated that the neuronal activity intensity and brain functional connectivity in patients with migraine have changed,resulting in the damage of process of brain integration of information.(3)The two sample T-test results of the four morphological features all showed significantly regional differences(P <0.05,FDR correction),and the mean curvature showed most regional differences;besides,classification model based on the mean curvature has biggest classification accuracy among all four kinds,reaching at 82.52%,while the classification accuracy of model based on the combination of all morphological features is 84.12%.It shows that the classification ability of mean curvature is better than the cortical surface area,cortical thickness,and gray volume;more importantly,using the complementarity between these features can effectively improve the classification accuracy.In summary,these functional abnormalities in resting brain network and brainstem provided additional evidence of brain dysfunction in the transmission,acceptance and integration of pain stimuli;and the high CCR and AUC of the classification model reflected the reliability of using cortical morphological feature in the clinical diagnosis of migraine.These findings promoted the understanding of the mechanism of migraine diseases,as well as diagnosis information and treatment strategies.