ATLAS-BASED ANALYSIS OF SINGLE-SUBJECT FUNCTIONAL MAGNETIC RESONANCE IMAGING STUDIES ON TINNITUS AND MUSIC IMPROVISATION

Functional Magnetic Resonance Imaging;Atlas-based analysis;Large Deformation Diffeomorphic Metric Mapping;Biomedical Engineering

Lee, DavidRatnanather, J. Tilak ;

Johns Hopkins University

关键词: Functional Magnetic Resonance Imaging;    Atlas-based analysis;    Large Deformation Diffeomorphic Metric Mapping;    Biomedical Engineering;   

Others  :  https://jscholarship.library.jhu.edu/bitstream/handle/1774.2/38113/LEE-THESIS-2015.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: JOHNS HOPKINS DSpace Repository

【 摘 要 】

Atlas-based analysis (ABA) was introduced as a potential supplement to functional mag-netic resonance imaging (fMRI) techniques. In contrast to the conventional voxel-basedanalysis (VBA), the time series was defined for each parcel, a set of voxels grouped accord-ing to an anatomical atlas of the brain. The parcels were objectively and accurately definedwithin the subject’s native space by deforming the atlas of brain to the subject’s brain usingLarge Deformation Diffeomorphic Metric Mapping (LDDMM). VBA and ABA were appliedto single-subject studies of tinnitus and music improvisation, and their performances werecompared qualitatively, and quantitatively in terms of test-retest reliability and predictionaccuracy. Furthermore, the relevance of the neural substrates identified from ABA wasinvestigated based on the current literature on the two cognitive phenomena. ABA facili-tated interpretation of whether an isolated activation or suppression represents the generalresponse of the parcel as a whole. Quantitative analysis showed higher test-retest reliabilityand prediction accuracy, suggesting that ABA reduces variability between sessions, and en-hances the distinct features of a response pattern. A drawback of ABA was observed whencomplex voxelwise response pattern was present within a parcel, whose detail was averagedand overlooked. Therefore, ABA is expected to improve with development of parcellationatlas representing deeper level of cytoarchitecture and neuroanatomy. While at the intro-ductory step, ABA, with high reliability and prediction accuracy granted by anatomicallyrelevant dimension reduction, is anticipated to contribute to development of multimodaland personalized analysis of human cognition.