STATISTICAL METHODS TO ANALYZE MASSIVE HIGH-DIMENSIONAL NEUROIMAGING DATA

neuro-image;statistical methods;Biostatistics

Chen, ShaojiePekar, James J. ;

Johns Hopkins University

关键词: neuro-image;    statistical methods;    Biostatistics;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/39524/CHEN-DISSERTATION-2015.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

The statistical analysis of neuroimaging data poses several challenges today, partly dueto their size, high dimensionality and noise. In this work, we address three different methodsfor analyzing massive, high-dimensional and noisy functional magnetic resonance images(fMRI) data. In the first method, parallel computing techniques are combined with anindependent component analysis (ICA) algorithm to decompose resting state fMRI data.The algorithm’s performance is greatly improved compared to existing methods. In thesecond method, a graphical model, referred to as state space model (SSM) is extended byenforcing L-1 and L-2 penalties on parameters. The model scales well to very high dimensionsand can be applied to a vast class of different neuroimaging analysis applications. Inthe third method, a two-stage method is developed to extract information from noisy fMRIdata. We first use functional regression to extract features from fMRI data and then use thefeatures to predicts physical pains that human feels. A support vector machine (SVM) istrained for prediction and it achieves high prediction accuracy.