Study Of Neuro-fingerprint Based On Precise Analysis Of Brain MRI

Chronic brain diseases progressively disrupt the central nervous system during the maturing phase and aging phase of neurons,respectively.The long course nature and heterogeneity of these diseases,as well as unclear pathogenesis and lack of effective treatment,make early diagnosis and intervention a critical issue.As an invasive visualization technique in clinics,MR images with multiple modalities can help radiologists and neurologist to better evaluate the structural and functional changes of the brain due to pathology.However,the theoretical system of neuroimage quantitative analysis for chronic brain diseases is sti ll lacking.Besides,there is a lack of association between image features across multiple image modalities,leading to inefficiency and misdiagnosis in clinics.Therefore,this study has proposed neuro-fingerprint technique based on precise analysis of brain MRI,which provides an integrated and extensible analytical system for chronic brain diseases like Alzheimer's disease and first-epsisode psychosis in a data-driven manner.Under this framework,we explored the association among multi-modal image features and extend quantitaive analysis for conplex human brain network for different types of brain connectomes.Our research findings are promising to solve clinical problems in multiple domains including pathological mechanism investigation,early screening,disease subtyping and disease course tracking,thus providing evidence and reference for precise medicine in chronic brain disease.The content of this paper is consisting of the following parts.(1)Atlas preselection strategy of the atlas-based brain image segmentation was studied.Considering the current atlas preselection strategies have no common standard,this paper proposed two novel preselection approaches to rank the existing atlases according to the anatomical similarity between the target image and each atlas,and further established a test framework to evaluate performance of each approaches with different age groups,different granularities of the atlas and different atlas numbers.The results show that these proposed strategies are superior to the conventional ones in terms of both computational efficiency and segmentation accuracy,improving the overall spatial preciseness of the neuro-fingerprinting technique.(2)Multi-modal MR neuro-fingerprints were studied.Since the single-modal MR image lacks sensitive markers for first-episode psychosis,this study leveraged multi-modal images to investigate the features of this disease.The atlas-based analysis was employed to reduce image dimensions,leading to improvement of the effect size of features and discriminative power for the disease.Plus,the external database was utilized to validate the reproducibility of the derived multi-modal image features.The results show that the projection fiber,association fiber and the functional connectivity between the thalamus and the sensorimotor cortex are abnormal in patients.(3)The brain anatomical covariance neuro-fingerprints were studied.Given that preclinical Alzheimer's disease have no obsvious MR image markers,this study extended anatomical covariance analysis in two directions: multi-structural and multi-modality.The group-based clustering and the statistical analysis on the covariance matrix were conducted to investigate the brain connectome.The results show that the intrinsic physiological covariance network of patients was disrupted by-amyloid deposition,indicating that neuron degeneration may occur before the onset of symptoms.(4)The brain functional network neuro-fingerprints were studied.Due to the heterogeneity of this disease,this study employed factor analysis to investigate the various cognitive factors,and further leveraged canonical correlation analysis and hierarchical clustering to subgroup patients into different b iotypes.The proposed analysis approach can explain the association between the neurophysiology abnormality and cognitive phenotypes.The results show that different biotypes of first-episode psychosis can have various expression levels in terms of brain functional network and cognitive phenotypes.(5)The brain structural network neuro-fingerprints were studied.The pathology of Alzheimer's disease would disrupt the brain structural network,however,how to precisely detect the abnormal one from the numerous structural connectivity remains a challenging issue.This paper utilize the multivariate distance matrix regression model to screen the potential abnormal connectivity in a high-throughput manner,in order to investigate the aberrant network pattern associated with the pathology.The results show that a wide range of brain connectome reorganization occurs in cortical-subcortical areas in patients,indicating of the damage on the local neuron circuit.In brief,this study quatitatively analyzed MR images of two typical brain choronic diseases using the proposed neuro fingerprinting technique.The research findings would facilitate the understanding of etiology and pathogenesis of brain choronic diseases,and provide crucial reference for precise prevension,diagnosis and therapy in the future.