One of the reasons Independent component analysis (ICA) becoming so popular is that ICA is a promising tool for signal process application, such as Functional magnetic resonance imaging (fMRI) data processing. However, there are still some problems to be solved. Most ICA algorithms are not stable in fMRI data processing. This paper presents a novel composite ICA algorithm integrating fixed-point algorithm and natural gradient algorithm for brain activity localization in Functional magnetic resonance imaging (fMRI) data. The new composite ICA algorithm has overcome the drawbacks of the both algorithms, providing more accurate and fast detection of weak fMRI functional signals. Simulations show great performance improvement compared with correlation analysis and Automated functional neuro imaging (AfNI) software.
展开▼
