Investigation of Efficient Implementation of Local Constrained Canonical Correlation Analysis for fMRI

Mingwu Jin¹, Rajesh Nandy², Dietmar Cordes¹

In previous work, local constrained canonical correlation analysis (cCCA) methods were proposed in order to avoid model overfitting and loss of specificity. In this work, we further investigate the performance, efficiency and possible improvement of region-growing based cCCA (cCCA-RG) methods. Using simulated data, we compare the estimation power of different cCCA-RG methods as well as the exhaustive search method (cCCA-ES). The detection power is also investigated upon real fMRI data. Our results demonstrate that cCCA-RG can significantly improve the detection power within an acceptable period of computation time.