This is interesting research. My sense is that there is both modularity and wave patterns occurring simultaneously during brain activity.
Our understanding of brain activity has traditionally been linked to brain areas – when we speak, the speech area of the brain is active. New research by an international team of psychologists led by David Alexander and Cees van Leeuwen (Laboratory for Perceptual Dynamics) shows that this view may be overly rigid. The entire cortex, not just the area responsible for a certain function, is activated when a given task is initiated. Furthermore, activity occurs in a pattern: waves of activity roll from one side of the brain to the other.
A still-shot of a wave of brain activity measured by electrical signals in the outside (left view) and inside (right view) surface of the brain. The colour scale shows the peak of the wave as hot colours and the trough as dark colours. | © D.A.
The brain can be studied on various scales, researcher David Alexander explains: "You have the neurons, the circuits between the neurons, the Brodmann areas – brain areas that correspond to a certain function – and the entire cortex. Traditionally, scientists looked at local activity when studying brain activity, for example, activity in the Brodmann areas. To do this, you take EEG's (electroencephalograms) to measure the brain’s electrical activity while a subject performs a task and then you try to trace that activity back to one or more brain areas."
In this study, the psychologists explore uncharted territory: "We are examining the activity in the cerebral cortex as a whole. The brain is a non-stop, always-active system. When we perceive something, the information does not end up in a specific part of our brain. Rather, it is added to the brain's existing activity. If we measure the electrochemical activity of the whole cortex, we find wave-like patterns. This shows that brain activity is not local but rather that activity constantly moves from one part of the brain to another. The local activity in the Brodmann areas only appears when you average over many such waves.”
Each activity wave in the cerebral cortex is unique. "When someone repeats the same action, such as drumming their fingers, the motor centre in the brain is stimulated. But with each individual action, you still get a different wave across the cortex as a whole. Perhaps the person was more engaged in the action the first time than he was the second time, or perhaps he had something else on his mind or had a different intention for the action. The direction of the waves is also meaningful. It is already clear, for example, that activity waves related to orienting move differently in children – more prominently from back to front – than in adults. With further research, we hope to unravel what these different wave trajectories mean."
The full text of the study "Traveling waves and trial averaging: the nature of single-trial and averaged brain responses in large-scale cortical signals" is available on the website of Neuroimage: http://www.sciencedirect.com/science/article/pii/S1053811913000633
Video 1 (above): A wave of brain activity measured by electrical signals at the surface of the brain. The electrodes have been implanted into the left hemisphere of a patient with intractable epilepsy, prior to surgical treatment. The two head views show the electrode array viewed from either the outside surface (left view) or the inside surface (right view). The wave takes about 125 milliseconds to traverse the area of cortex shown. The times displayed at the bottom are relative to the subject's voluntary finger movement at zero milliseconds. The travelling wave originates from the back of the cortex and propagates toward the frontal region. The colour scale shows the peak of the wave as hot colours and the trough of the wave as dark colours.
Video 2 (above): A wave of brain activity measured by the magnetic field it generates externally to the head. The left view of the head is shown on the left side of the image and the right view of the head on the right side of the image. This wave takes about 100 milliseconds to traverse the entire surface of the brain. The travelling wave originates on the lower-left of the head and travels to the lower front-right of the head. Most of the magnetic field shown in this video is generated by brain activity close to the surface of the cortex. The times displayed at the bottom are relative to the subject pressing a button at time zero. The colour scale shows the peak of the wave as hot colours and the trough of the wave as dark colours.
Here is the Abstract and Introduction to the original research article summarized above.
Read the whole article.
Traveling waves and trial averaging: The nature of single-trial and averaged brain responses in large-scale cortical signals
Open Access Article
David M. Alexandera, Peter Juricab, Chris Trengovea, Andrey R. Nikolaeva, Sergei Gepshteinc, Mikhail Zvyagintsevd, Klaus Mathiakd, Andreas Schulze-Bonhagee, Johanna Rueschere, Tonio Balle, Cees van Leeuwena.
Analyzing single trial brain activity remains a challenging problem in the neurosciences. We gain purchase on this problem by focusing on globally synchronous fields in within-trial evoked brain activity, rather than on localized peaks in the trial-averaged evoked response (ER). We analyzed data from three measurement modalities, each with different spatial resolutions: magnetoencephalogram (MEG), electroencephalogram (EEG) and electrocorticogram (ECoG). We first characterized the ER in terms of summation of phase and amplitude components over trials. Both contributed to the ER, as expected, but the ER topography was dominated by the phase component. This means the observed topography of cross-trial phase will not necessarily reflect the phase topography within trials. To assess the organization of within-trial phase, traveling wave (TW) components were quantified by computing the phase gradient. TWs were intermittent but ubiquitous in the within-trial evoked brain activity. At most task-relevant times and frequencies, the within-trial phase topography was described better by a TW than by the trial-average of phase. The trial-average of the TW components also reproduced the topography of the ER; we suggest that the ER topography arises, in large part, as an average over TW behaviors. These findings were consistent across the three measurement modalities. We conclude that, while phase is critical to understanding the topography of event-related activity, the preliminary step of collating cortical signals across trials can obscure the TW components in brain activity and lead to an underestimation of the coherent motion of cortical fields.
► We analyzed data from the MEG, EEG and ECoG, quantifying several signal components.
► We compared topographies of activation across large-scale cortex.► The topography of evoked responses was primarily a function of within-trial phase.► Within-trial phase topography was modeled as traveling waves.► Traveling waves explained more signal than the trial-averaged phase topography.
A wealth of evidence links cross-trial averaged, evoked response (ER) measures to various stages of perception, cognition and action. However, the mechanisms responsible for the ER are poorly understood. Here we relate ERs to cortical traveling waves (TWs). The functional significance of TWs, at the columnar scale (Eckhorn et al., 2004 and Nauhaus et al., 2009), Brodmann area (Freeman and Barrie, 2000, Prechtl et al., 1997, Rubino et al., 2006 and Takahashi et al., 2011) and in large scale cortex (Alexander et al., 2006b,Ito et al., 2007, Klimesch et al., 2007 and Massimini et al., 2004) is the subject of a growing literature. The goal of this study is to analyze the composition of the ER, considered as a large-scale pattern of TW activity.
The ER is the trial-average of the measured time-series. ER measurements across the entire scalp are typically used in source localization techniques, for both event-related potentials and event-related fields (Liu et al., 1998 and Pascual-Marqui et al., 1994). An underlying assumption is that ERs reflect the magnitude and location of brain activity and its time-course. Areas or intervals with low ER magnitude therefore tend to be ignored. The ER, along with other cross-trial measures such as coherence (Hipp et al., 2011 and Lachaux et al., 1999), targets brain activity that consistently summates across trials. Brain activity that does not consistently summate across trials is not considered functionally important; sites of low phase consistency are generally considered to reflect activation uncorrelated with the task. This activity is assumed to behave randomly across trials (Britten et al., 1992 and Ray and Maunsell, 2011).
Though technically challenging, it is possible to fit the equivalent current dipole to individual trial data and then compute the average source location, or range of locations, indicated by these single trial fits (Liu and Ioannides, 1996). The resulting evidence suggests that the sequence of activations in the trial-averaged signal does not accurately reflect the sequence of activations within individual trials. In the present research we consider the sequence of activations in the cortex by analyzing TWs within trials.
The issue of trial-averaging also arises in discussions of the relative importance of amplitude and phase to the ER. Two alternative ideas of ER generation have emerged in the literature. In the evoked model, changes in amplitude contribute to the ER, both directly and by selectively enhancing the amplitude of some phase components (Mäkinen et al., 2005 and Shah et al., 2004). In the phase resetting model, only the phase of the signal will affect the ER, via changes in cross-trial phase locking initiated by experimental events (Gruber et al., 2005 and Makeig et al., 2002). In this discussion of ER generation, the relative effects of phase and amplitude on evoked responses are often considered only for a limited number of recording sites (Barry et al., 2004, Gruber et al., 2005 and Fell, 2007 cf. Makeig et al., 2002). Consistent with this approach, oscillatory components that have been isolated using independent components analysis also summate in a manner that suggests a topography of activity arising from localized cortical sources (Makeig et al., 2002). However, if there are large-scale topographical relationships in the way phase and amplitude interact, they may be missed when only sites with maximum ER magnitude are considered. In particular, in the trial averaged signals, TWs are often not clearly in evidence. However, this does not preclude TWs from being prominent in the unaveraged signal. This we investigate here. The topography of phase might be explained better by TWs at the single trial level than by any component of the trial-averaged signal.
An alternative to cross-trial measurement is to characterize the amplitude and phase prior to collation of signals across trials. This may be achieved by considering the topography of these components at the single-trial level. Useful information can be extracted from single trials even without analyzing the Fourier components of the signal, for example, Arieli et al. (1996) considered the topography of local field potentials in cat visual cortex. They found that the spatial pattern of within trial activity dominated the signal; the ER was merely a relatively small component. The preceding within-trial activity was an excellent predictor for within-trial activity at the latency of the ER; a much better predictor than was the trial-averaged ER topography. Thus the within trials topography of activity is highly structured, not simply noise on top of an ER (Arieli et al., 1996).
Within trials topographical information has also been critical for detecting TWs, either by amplitude deflections or by phase estimates (Alexander et al., 2006b, Eckhorn et al., 2004, Massimini et al., 2004 and Rubino et al., 2006). The present research tested the prevalence of TWs as an instance of highly structured activity patterns within trials, also in order to assess the strength of such patterns against trial-averaged components such as ERs and phase coherence. Whereas the former collates signals across space but within trials, the latter, more commonly used measures, collates signals for individual sites but across trials.
Theory and simulation of cortical mechanisms have predicted the existence of large-scale TWs (Nunez and Srinivasan, 2006 and Wright et al., 2001). Essentially, the global resonances that account for the 1/f spectra of cortical activity are also associated with TW dynamics. Using the electroencephalogram (EEG) and magnetoencephalogram (MEG), global cortical waves have been shown to arise at a variety of frequencies, from the sub-delta through to gamma bands (Alexander et al., 2006b, Alexander et al., 2009, Ito et al., 2007,Massimini et al., 2004, Ribary et al., 1991 and Sauseng et al., 2002). These waves are typically of long wavelength, with a spatial period of the order of 10 to 20 cm. Over one temporal cycle of the wave, wave peaks typically traverse the entire EEG/MEG recording array.
The functional significance of TWs has been established by noting their close correspondence with the latency topography of known visual and auditory ERP components, such as the P1–N1 complex, P2–N2 complex, as well as the P3b (Alexander et al., 2006b, Alexander et al., 2009, Anderer et al., 1996, Fellinger et al., 2012 and Klimesch et al., 2007). For these event-related potential (ERP) components, the latency, temporal frequency and task-dependency of evoked TW components are consistent with latency, temporal frequency and task-dependency of the corresponding ERPs.
ERPs aside, more evidence needs to be brought to bear on the functional relevance of large-scale TWs at the single-trial level. Some progress has been made toward this goal in analyzing wave activity differences across defined brain states, such as rest (Ito et al., 2005 and Ito et al., 2007), deep sleep (Massimini et al., 2004) and working memory (Fellinger et al., 2012 and Sauseng et al., 2002). Single-trial TWs have also been used to uncover genetic differences in brain activity (Alexander et al., 2007), differences across age groups and clinical groups and in correlations with clinical symptoms (Alexander et al., 2006a, Alexander et al., 2006b, Alexander et al., 2008 and Alexander et al., 2009).
Large-scale patterns of activity in the scalp EEG are partly a function of blurring due to volume conduction of the dura, skull and other tissues; to a lesser extent blurring effects also apply to MEG and electrocorticogram (ECoG) measurements. However, a number of analyses clearly indicate that TWs measured in the EEG are not an artifact of volume conduction. TWs can still be detected in the EEG using sparse electrode arrays with a minimum electrode separation of 10 cm (Alexander et al., 2009); the spatial resolution of the measurement is matched to the spatial resolution of the signal to discount blurring artifact. TWs in EEG can also be detected by measuring latency delays rather than spatial patterns, per se (Alexander et al., 2006b, Fellinger et al., 2012, Manjarrez et al., 2007, Massimini et al., 2004, Nauhaus et al., 2009 and Patten et al., 2012). Since volume conduction effects are essentially instantaneous, to explain away the motion of the apparent waves requires a more complicated hypothesis than volume conduction alone provides (see Ray and Maunsell, 2011, for one such hypothesis; c.f. Nauhaus et al., 2012). To further address the issue of blurring by volume conduction, in the present research we analyzed data from a range of imaging modalities: MEG, EEG and ECoG. We chose these modalities because they have different effective spatial resolutions: approximately 4 cm, 10 cm and 1 cm for MEG, EEG and ECoG, respectively (Bullock et al., 1995 and Srinivasan et al., 2007). Establishing the ubiquity of TWs using each of these measurement modalities would add further evidence against the argument that TWs arise as an artifact of volume conduction.
The results of this study show that phase plays the major role in the ER topography, more-so than amplitude, consistently across imaging modalities. We also observed that the topography of trial-averaged phase, while correlated with ER topography, co-occurs with ubiquitous episodes of TW activity at the single-trial level. In the three data sets analyzed here, the topography of within-trial phase was better described by TWs than by trial-averaged phase, suggesting a loss of information in the latter case. Linking TWs back to ER topography, we show that at some event-related times and frequencies, the ER topography can be approximated as a trial-average of TW components estimated at the single trial level. The existence of smooth spatial gradients of phase within trials, i.e. TWs, is entirely consistent with spatial and temporal variations in cross-trial phase locking. We propose that ER magnitudes are partly the product of TWs that summate and cancel differentially across measurement sites.