基于神经网络脑功能活动探测-生物医学工程专业论文.docx

ABSTRACTfMRIdata,anewspatialdistanceandtemporalcorrelationwasintroduced.Itismore reasonablethattakingadvantageoftheintrinsicfeatureoftemporalandspatial information.HiddenMarkovmodel(HMM),aspecialneuralnetworkmethodwasappliedin thefMRIdataprocessing.Inthispart,atwo-stateHMMbasedonBayesian probabilisticmethodologywasintroducedtolocatebrainfunctionalactivationregions andtorevealstatetransitionofthetimecourseinanalyzingevent-relatedfMRIdata. Loglikelihoodmapsforallvoxelswerefirstlycalculatedbyfirst-ordercontinuous densityHMM,wheretheprobabilitydensitiesfunctionwasassociatedwithsimple on-offstatestorevealactivatingregions.Astatetransitionmapforeachgivenvoxel wasthencharacterizedbythefMRItimecourseintensitychangethatdescribedthe neuralresponse to thespecial stimuli.Keywords：fMRIdataprocessing,neuralnetwork,independentcomponent analysis, self-organizing mapping, hidden Markov model目录目录第一章 绪论.........................................................11.1 功能磁共振成像技术概述......................................11.1.1 磁共振的成像原理......................................11.1.2 脑功能磁共振成像......................................21.1.3 功能磁共振成像特点和其应用............................21.1.4 功能磁共振数据的处理方法..............................31.2 神经网络方法探测脑功能活动的概况............................61.2.1 神经网络..............................................61.2.2 神经网络方法和脑成像的交叉融合........................61.3 本论文的研究内容和结构安排..................................7第二章 空间独立成分分析多任务fMRI数据82.1 独立成分分析(ICA) 8ICA简介8ICA的基本原理8ICA假设条件9ICA算法10ICA的分类102.2 多任务的功能磁共振.........................................112.3 本研究的目的和意义.........................................112.4 数据获取和数据分析..........................错误!未定义书签。2.4.1 实验任务.............................................112.4.2 数据获取.............................................122.4.3 数据分析.............................................122.5 数据结果和讨论.............................................132.5.1 结果.................................................132.5.2 讨论.................................................15