Conference proceeding
Comparison between spatial and temporal independent component analysis for blind source separation in fMRI data
2011 4th International Conference on Biomedical Engineering and Informatics (BMEI), Vol.2, pp.690-692
10/2011
DOI: 10.1109/BMEI.2011.6098493
Abstract
Independent component analysis (ICA) is an exploratory method for analyzing spatial and temporal properties of fMRI data and requires no explicit temporal model, necessary for conventional fMRI analysis. Two varieties of ICA are employed to achieve maximal independence component in space or time yields for functional MRI (fMRI) analysis: spatial ICA (sICA) and temporal ICA (tICA). sICA is widely studied and used in signal separation of fMRI data. In this study, we compared the performance of sica and tICA to extract and separate signals with spatial and temporal independence based on simulated data. Our results reveal that sICA is able to extract and separate relatively highly independent signals. tICA can fulfill the separation of mutually independent component signal in time course and classify the temporally corresponding signal as one group in spite of having a spatially independent component. The results suggest that tICA can be applied to detect a special signal overlapping with the physiological signals by evoking other activations using the special signal.
Details
- Title: Subtitle
- Comparison between spatial and temporal independent component analysis for blind source separation in fMRI data
- Creators
- Xin Gao - Suzhou Institute of Biomedical Engineering and TechnologyTao Zhang - Suzhou Institute of Biomedical Engineering and TechnologyJinhu Xiong - University of Iowa
- Resource Type
- Conference proceeding
- Publication Details
- 2011 4th International Conference on Biomedical Engineering and Informatics (BMEI), Vol.2, pp.690-692
- Publisher
- IEEE
- DOI
- 10.1109/BMEI.2011.6098493
- ISSN
- 1948-2914
- eISSN
- 1948-2922
- Language
- English
- Date published
- 10/2011
- Academic Unit
- Roy J. Carver Department of Biomedical Engineering
- Record Identifier
- 9984383293702771
Metrics
7 Record Views