Group-wise sparse representation of brain states reveal network abnormalities in mild traumatic brain injury

Mild traumatic brain injury (mTBI) is a leading public health care burden. Recent research has shown that the functional impairment in mTBI patients could be captured by resting state fMRI (rsfMRI) at network level. Moreover exploring brain response to mTBI over time at large scale network level can help physicians better diagnose brain injury and order appropriate rehabilitation plan. Therefore, there is a need for methodological innovation that could assess brain impairment in rsfMRI data and further define biomarkers for network changes. In this paper, we propose a novel group-wise sparse representation of brain states (GSRBS) approach, based on rsfMRI data, to explore the effect of mTBI on functional networks across different groups and longitudinal stages. Specifically, a dictionary of brain networks is learned from the volumes of rsfMRI data, and at each time point these networks are linearly and sparsely combined to realize a brain state. Our results showed that group-wise statistical difference on the network composition of brain states could be found between healthy controls and mTBI patients at two different temporal stages.