Simultaneous investigation of local and distributed functional brain connectivity from fMRI data

In this paper we propose a new method for simultaneous assessment of local and distributed functional brain connectivity from functional magnetic resonance imaging (fMRI) data. Our method maps fMRI time series from brain voxels into a high dimensional feature space where in clustering and vector quantization are performed in order to aggregate voxels with similar temporal evolution and segregate those with low temporal correlation. An iterative algorithm is proposed based on the Eigen structure of the geodesic distance matrix of the high dimensional manifold for selecting the number of clusters. The choice is then verified by unfolding the manifold using ISOMAP. A combined connectivity index (CCI) is then defined for every brain region based on the percentage of its voxels connected to each other and to voxels from other regions. The CCI maps local and distributed connectivity on a continuum from 0 to 1. This method was applied to resting state fMRI data obtained from humans in one of the two states: normal awake state and a sedated state induced by propofol anesthesia. Our results demonstrate that propofol anesthesia makes the connectivity in the brain become more local and less distributed as compared to awake state. In addition, our results show that the distributed connectivity between the thalamus and the cortex is greatly impacted by anesthesia, lending support to the thalamo-cortical disconnection hypothesis. However, our results indicate the thalamus was still connected to the cortex via the amygdala, providing a pathway for sensory information to reach the cortex during sedation. In the cortex, the connectivity also tended to become more local than distributed, supporting the cortico-cortical disconnection hypothesis. Therefore, our conclusions reconcile both the hypotheses of anesthetic action and altered states of consciousness, and demonstrate the utility of our approach for addressing important neuroscientific questions.