A fast-FENICA method on resting state fMRI data

Nizhuan Wang,Weiming Zeng,Lei Chen

DOI: https://doi.org/10.1016/j.jneumeth.2012.05.007

2012-07-30

Abstract:For resting-state fMRI data, independent component analysis (ICA) is an excellent method which enables the decomposition of high-dimensional data into discrete spatial and temporal components. Fully exploratory network ICA (FENICA), a fully automated and purely data-driven ICA-based analysis for group assessment of resting-state networks, was proposed by Schöpf et al. (2010). FENICA is a novel and effective group assessment method, but it is not without limitations, such as those related to memory and time costs in running. Here we present Fast-FENICA, which is based on an energy sifting algorithm for interested networks, a linear candidate networks formation strategy and a correlation coefficients ranking algorithm of network matrix. It is demonstrated that the energy sifting algorithm for interested networks and linear candidate networks formation strategy can transform the stubborn computing time and memory cost limitations of FENICA from a quadratic level to a linear level and thus speed up the group evaluation. Furthermore, the correlation coefficients ranking algorithm can further increase the calculation speed and float up the consistent networks effectively. In comparison to FENICA, the hybrid data and true data experimental results demonstrate that Fast-FENICA not only contributes to the practicability and efficiency without decreasing the detecting ability of functional networks, but also ranks the common functional networks based on the whole spatial consistency at a group level. This proposed effective group analysis method is expected to have wide applicability.