Related Experiment Video
Updated: Apr 18, 2026
Identification of Disease-related Spatial Covariance Patterns using Neuroimaging Data
Published on: June 26, 2013
Characterizing nonlinear relationships in functional imaging data using eigenspace maximal information canonical
Li Dong1, Yangsong Zhang1, Rui Zhang1
1The Key Laboratory for NeuroInformation of Ministry of Education, Center for Information in BioMedicine, High-Field Magnetic Resonance Brain Imaging Key Laboratory of Sichuan Province, School of Life Science and Technology, University of Electronic Science and Technology of China, Chengdu 610054, China.
A new method, eigenspace maximal information canonical correlation analysis (emiCCA), effectively identifies both linear and nonlinear relationships in neuroimaging data. This technique offers superior performance over existing methods for analyzing brain function and connectivity.
Area of Science:
- Neuroimaging analysis
- Brain connectivity
- Data-driven methods
Background:
- Linear methods like canonical correlation analysis (CCA) are common for neuroimage analysis.
- Exploring nonlinear processes in brain function requires more flexible techniques.
Purpose of the Study:
- Introduce eigenspace maximal information canonical correlation analysis (emiCCA) for capturing linear and nonlinear relationships.
- Evaluate emiCCA's performance against existing CCA methods.
- Apply emiCCA to functional magnetic resonance imaging (fMRI) data.
Main Methods:
- Developed a novel unsupervised, data-driven method: emiCCA.
- Validated emiCCA through simulations comparing it to linear CCA and kernel CCA.
- Implemented an emiCCA framework for fMRI data processing.
Main Results:
- Simulations showed emiCCA outperformed linear and kernel CCA.
- Analysis of motor execution fMRI data revealed one linear network (primary motor cortex) and several nonlinear networks (supplementary motor area, insula, cerebellum).
Conclusions:
- emiCCA is a powerful technique for uncovering both linear and nonlinear relationships in complex datasets.
- The identified networks suggest contributions to hand movement execution.
- emiCCA shows promise for advanced neuroimaging data exploration.
