Dynamic functional connectivity

Dynamic functional connectivity (DFC) refers to the observed phenomenon that functional connectivity changes over a short time. Dynamic functional connectivity is a recent expansion on traditional functional connectivity analysis which typically assumes that functional networks are static in time. DFC is related to a variety of different neurological disorders, and has been suggested to be a more accurate representation of functional brain networks. The primary tool for analyzing DFC is fMRI, but DFC has also been observed with several other mediums. DFC is a recent development within the field of functional neuroimaging whose discovery was motivated by the observation of temporal variability in the rising field of steady state connectivity research. Dynamic functional connectivity (DFC) refers to the observed phenomenon that functional connectivity changes over a short time. Dynamic functional connectivity is a recent expansion on traditional functional connectivity analysis which typically assumes that functional networks are static in time. DFC is related to a variety of different neurological disorders, and has been suggested to be a more accurate representation of functional brain networks. The primary tool for analyzing DFC is fMRI, but DFC has also been observed with several other mediums. DFC is a recent development within the field of functional neuroimaging whose discovery was motivated by the observation of temporal variability in the rising field of steady state connectivity research. Functional connectivity refers to the functionally integrated relationship between spatially separated brain regions. Unlike structural connectivity which looks for physical connections in the brain, functional connectivity is related to similar patterns of activation in different brain regions regardless of the apparent physical connectedness of the regions. This type of connectivity was discovered in the mid-1990s and has been seen primarily using fMRI and Positron emission tomography. Functional connectivity is usually measured during resting state fMRI and is typically analyzed in terms of correlation, coherence, and spatial grouping based on temporal similarities. These methods have been used to show that functional connectivity is related to behavior in a variety of different tasks, and that it has a neural basis. These methods assume the functional connections in the brain remain constant in a short time over a task or period of data collection. Studies that showed brain state dependent changes in functional connectivity were the first indicators that temporal variation in functional connectivity may be significant. Several studies in the mid-2000s examined the changes in FC that were related to a variety of different causes such as mental tasks, sleep, and learning. These changes often occur within the same individual and are clearly relevant to behavior. DFC has now been investigated in a variety of different contexts with many analysis tools. It has been shown to be related to both behavior and neural activity. Some researchers believe that it may be heavily related to high level thought or consciousness. Because DFC is such a new field, much of the research related to it is conducted to validate the relevance of these dynamic changes rather than explore their implications; however, many critical findings have been made that help the scientific community better understand the brain.Analysis of dynamic functional connectivity has shown that far from being completely static, the functional networks of the brain fluctuate on the scale of seconds to minutes. These changes are generally seen as movements from one short term state to another, rather than continuous shifts. Many studies have shown reproducible patterns of network activity that move throughout the brain. These patterns have been seen in both animals and humans, and are present at only certain points during a scanner session.In addition to showing transient brain states, DFC analysis has shown a distinct hierarchical organization of the networks of the brain. Connectivity between bilaterally symmetric regions is the most stable form of connectivity in the brain, followed by other regions with direct anatomical connections. Steady state functional connectivity networks exist and have physiological relevance, but have less temporal stability than the anatomical networks. Finally, some functional networks are fleeting enough to only be seen with DFC analysis. These networks also possess physiological relevance but are much less temporally stable than the other networks in the brain. Sliding window analysis is the most common method used in the analysis of functional connectivity, first introduced by Sakoglu and Calhoun in 2009, and applied to schizophrenia. Sliding window analysis is performed by conducting analysis on a set number of scans in an fMRI session. The number of scans is the length of the sliding window. The defined window is then moved a certain number of scans forward in time and additional analysis is performed. The movement of the window is usually referenced in terms of the degree of overlap between adjacent windows. One of the principle benefits of sliding window analysis is that almost any steady state analysis can also be performed using sliding window if the window length is sufficiently large. Sliding window analysis also has a benefit of being easy to understand and in some ways easier to interpret.As the most common method of analysis, sliding window analysis has been used in many different ways to investigate a variety of different characteristics and implications of DFC. In order to be accurately interpreted, data from sliding window analysis generally must be compared between two different groups. Researchers have used this type of analysis to show different DFC characteristics in diseased and healthy patients, high and low performers on cognitive tasks, and between large scale brain states. One of the first methods ever used to analyze DFC was pattern analysis of fMRI images to show that there are patterns of activation in spatially separated brain regions that tend to have synchronous activity. It has become clear that there is a spatial and temporal periodicity in the brain that probably reflects some of the constant processes of the brain. Repeating patterns of network information have been suggested to account for 25–50% of the variance in fMRI BOLD data. These patterns of activity have primarily been seen in rats as a propagating wave of synchronized activity along the cortex. These waves have also been shown to be related to underlying neural activity, and has been shown to be present in humans as well as rats. Departing from the traditional approaches, recently an efficient method was introduced to analyze rapidly changing functional activations patterns which transforms the fMRI BOLD data into a point process. This is achieved by selecting for each voxel the points of inflection of the BOLD signal (i.e., the peaks). These few points contain a great portion of the information pertaining functional connectivity, because it has been demonstrated, that despite the tremendous reduction on the data size (> 95%), it compares very well with inferences of functional connectivity obtained with standard methods which uses the full signal. The large information content of these few points is consistent with the results of Petridou et al. who demonstrated he contribution of these 'spontaneous events' to the correlation strength and power spectra of the slow spontaneous fluctuations by deconvolving the task hemodynamic response function from the rest data. Subsequently, similar principles were successfully applied under the name of co-activation patterns (CAP). Time-frequency analysis has been proposed as an analysis method that is capable of overcoming many of the challenges associated with sliding windows. Unlike sliding window analysis, time frequency analysis allows the researcher to investigate both frequency and amplitude information simultaneously. The wavelet transform has been used to conduct DFC analysis that has validated the existence of DFC by showing its significant changes in time. This same method has recently been used to investigate some of the dynamic characteristics of accepted networks. For example, time frequency analysis has shown that the anticorrelation between the default mode network and the task-positive network is not constant in time but rather is a temporary state.Independent component analysis has become one of the most common methods of network generation in steady state functional connectivity. ICA divides fMRI signal into several spatial components that have similar temporal patterns. More recently, ICA has been used to divide fMRI data into different temporal components. This has been termed temporal ICA and it has been used to plot network behavior that accounts for 25% of variability in the correlation of anatomical nodes in fMRI.