|
Abstract:
|
Correlation analysis between two groups of time series is common in many fields, for example in analysis of functional magnetic resonance imaging (fMRI) data. The most widely used approach in fMRI is probably to compute Pearson's correlation between the group-mean temporal vectors, averaged across the (spatial) variables in each group. This approach does not account for the continuity of the time series and the inhomogeneity in the variables. In this paper, we propose a spatial and temporal correlation analysis (STC) that addresses these two issues simultaneously. It integrates the functional correlation and canonical correlation analysis (CCA) in a unified optimization-based framework. This allows, for example, varying contributions of spatial variables and increased signal strength. Simulation results show the proposed method outperforms other competing methods. Applying to a fMRI dataset, we identify the connection strength between brain regions and the inhomogeneous functions within regions.
|
