| |
 |
UNIVERSITY OF BUCHAREST
FACULTY OF PHYSICS Guest
2024-11-22 2:10 |
 |
|
|
|
Conference: Bucharest University Faculty of Physics 2003 Meeting
Section: Electricity and Biophysics
Title:
Artifacts and Noise Detected in Functional Magnetic Resonance Imaging of Biomedical Time Series by Spatiotemporal Independent Component Analysis
Authors:
Radu Mutihac, Razvan Raicu
Affiliation:
mutihac@fpce4.fizica.unibuc.ro, razvan@fpce1.fizica.unibuc.ro
University of Bucharest, Faculty of Physics, PO Box MG-11, 76900 Bucharest, Romania
E-mail
Keywords:
Abstract:
The independent component analysis (ICA) fully characterizes the electrophysiological data by separating them into sparse maps, or spatial modes, and associated time courses. Employing an ICA algorithm capable of looking for non-sparse as well as sparse maps can find maps that all are sparse. Many of these maps can be identified with known artifacts, such as blood vessel pulsations, head movements, slow drifts, and imperfections of the recording equipment. These highly spatially structured signals are not easily modeled by a priori estimates as required by hypothesis-driven methods. A basic assumption in ICA, that the maps are spatially independent, does not preclude the possibility of spatial overlap because maximal independence can be achieved with overlap in high-dimensional spaces. Spatiotemporal analysis of fMRI data (sequences of brain images) has been carried out to discriminating between signals from distinctly active brain regions, and artifacts and noise were separated from actual brain activity based on their statistical properties and physiological relevance.
Key Words: Functional magnetic resonance imaging (fMRI), blood oxygenation level dependent (BOLD), independent component analysis (ICA), statistical parametric mapping (SPM), linear transforms, higher-order statistics.
|
|
|
|