My research interests are mainly about understanding brain functional integrations and the organization of brain networks. My efforts can be summarized as three themes on how three factors modulate brain functional connectivity: 1) different task conditions, 2) other brain regions, 3) and neurophysiological factors. I use functional MRI (fMRI) as my main research tool, but also pursue combining different neuroimaging modalities including positron emission tomography (PET) and electroencephalography (EEG). My research philosophy is to utilize divergent research methodologies to approach my scientific questions. I have been using multiple methods to study brain connectivity, including psycho-(physio-)physiological interaction (PPI), dynamic causal modeling (DCM), independent component analysis (ICA), network analysis, sliding window method, and meta-analysis. 

1. Functional connectivity modulated by tasks

It has been generally accepted that brain functions could be better understood in terms of function integration than simple functional localizations. I started to appreciate this notion in my very first research project in graduate school. We used fMRI to study the neural substrates of the Fist-Edge-Palm (FEP) motor sequencing task, which has been clinically linked to prefrontal lobe functions. By using conventional analysis, however, we did not find any increased activations in the prefrontal cortex in the FEP task than simple motor tasks. We then performed psychophysiological interaction (PPI) analysis, and showed that the connectivity was increased between the right middle/inferior frontal gyrus and sensiromotor cortex in the FEP task (Rao et al., 2008). The right prefrontal regions modulate the primary sensorimotor cetex during the complex motor task, even though their activations remain unchanged.

The PPI approach is limited by the task that has been designed in an experiment. To overcome this limitation, I dug into neuroimaging literatures to estimate task related connectivity using a meta-analytic approach, i.e. examining how likely two regions are co-activated in different fMRI experiments. I obtained coactivation patterns of regions across the whole brain using a large literature database. The likelihood of coactivation between two regions was highly correlated with their functional connectivity in resting-state. However, the network of coactivation was organized in a more efficient way, with higher global efficiency, lower local efficiency, and lower modularity. The results suggest different network configurations when the brain is at task compared with at rest (Di, Gohel, Kim, & Biswal, 2013). 

Di X, Fu Z, Chan SC, Hung YS, Biswal BB, Zhang Z (2015). Task-related Functional Connectivity Dynamics in a Block-designed Visual Experiment. Front. Hum. Neurosci 9:543.
Di X, Gohel S, Kim EH and Biswal BB (2013). Task vs. Rest - Different Network Configurations between the Coactivation and the Resting-State Brain Networks. Front. Hum. Neurosci. 7:493.
Rao H, Di X, Chan RC, Ding Y, Ye B, Gao D (2008). A Regulation Role of Prefrontal Cortex in The Fist-Edge-Palm Task: Evidence from Functional Connectivity Analysis. Neuroimage 41(4), 1345-1351.

2. Functional connectivity modulated by other regions

A question arises from studies in theme 1: where does “task modulation” come from? Theoretically, the “task” must be encoded in a region, so that the activity of the region modulates the connectivity between two other regions. This motivates me to perform a series of studies on the modulation of connectivity by a third region using physiophysiological interaction (also as PPI). In one study, I applied PPI analysis on resting-state fMRI data by using the default mode network and frontoparietal network as seeds, and found that the salience network was a critical network that modulated the functional interaction between the default mode network and frontoparietal network (Di & Biswal, 2014). Extending the PPI idea, I directly identified time points when a network had higher or lower activity, and compared correlation differences between the higher activity and lower activity conditions among regions across the whole brain (Di & Biswal, 2015). I found that the salience network, among other networks, had the largest effect on large scale brain connectivity. Connectivity was higher when the salience network had higher activity, mainly between one region from the default mode network and one region from the frontoparietal network. These convergent findings may be helpful to explain why the negative connectivity between the default mode network and frontoparietal network is highly variable. And the nonlinear interactions among the “triple networks” may provide new insight on the neural mechanisms of mental disorders, such as schizophrenia and autism spectrum disorder. 

I further systematically studied modulatory interactions across the whole brain to characterize general properties of this phenomenon (Di & Biswal, in press). I found that across the whole brain, there were more negative modulatory interactions than positive ones. This means that in more cases, increased activity in one region is accompanied by decreased connectivity between two other regions. Although it is a little counterintuitive, it is reasonable because more negative modulatory interactions will make the whole brain as a system stable. In addition, I found that modulatory interactions were more representative in certain systems, for example among the cerebellum, basal ganglia and motor system, and between the default mode network and visual systems. These results have implications on dynamic interactions in brain, and provide a roadmap for our future studies on modulatory interactions. 

Di X, Biswal BB (2015). Characterizations of resting-state modulatory interactions in human brain. J Neurophysiol 114(5):2785-96.
Di X, Biswal BB (2015). Dynamic Brain Functional Connectivity Modulated by Resting-State Networks. Brain Structure and Function 220(1):37-46.
Di X, Biswal BB (2014). Modulatory Interactions between the Default Mode Network and Task Positive Networks in Resting-State. PeerJ 2:e367.
Di X, Biswal BB (2013). Modulatory interactions of resting-state brain functional connectivity. PLoS ONE 8(8): e71163.

3. Functional connectivity modulated by neurophysiological factors

Functional connectivity measured from fMRI does not only reflect bona fide neuronal interactions, but is also affected by underlying neurophysiological factors, such as neurovascular coupling. I have shown that neurovascular coupling as reflected by fMRI activations of breath holding was highly correlated with task activations (Di, Kannurpatti, Rypma, & Biswal, 2013). To extend this idea, my further work demonstrated that brain functional connectivity, as measured using independent component analysis (ICA) and correlation coefficients, were both correlated with the amplitude of low frequency fluctuations, a measure reflected neurovascular coupling (Di, Kim, et al., 2013). These results suggest that differences in functional connectivity measured using fMRI may be due to differences in underlying neurovascular coupling, but not real neuronal interactions. Therefore, taking into account of neurophysiological factors is critical when performing cross-sectional comparisons of functional connectivity. 

Di X, Kim E, Huang C, Tsai S, Lin C and Biswal BB (2013). The influence of the amplitude of low-frequency fluctuations on resting-state functional connectivity. Front. Hum. Neurosci. 7:118.
Di X, Kannurpatti SS, Rypma B, Biswal BB, (2013). Calibrating BOLD fMRI activations with neuro-vascular and anatomical constraints. Cereb Cortex 23(2):255-263.