High Dimensional Tests for Functional Networks of Brain Anatomic Regions
There has been increasing interests in learning resting-state brain functional connectivity of autism disorders using functional magnetic resonance imaging (fMRI) data. The data in a standard brain template consist of over 200,000 voxel specific time series for each single subject. Such an ultra-high dimensionality of data makes the voxel-level functional connectivity analysis (involving four billion voxel pairs) lack of power and extremely inefficient. In this work, we introduce a new framework to identify functional brain network at brain anatomic region-level for each individual. We propose two pairwise tests to detect region dependence, and one multiple testing procedure to identify global structures of the network. The limiting null distributions of the test statistics are derived. It is also shown that the tests are rate optimal when the alternative networks are sparse. The numerical studies show the proposed tests are valid and powerful. We apply our method to a resting-state fMRI study on autism and identify patient-unique and control-unique hub regions. These findings are consistent with autism clinical symptoms.