IEEE Signal Processing - May 2018 - 89

Versus

(a)

(c)

Onset

500

Time (ms)

(b)

(d)

FIGURE 4. The application of EEG source connectivity to tracking the dynamic reshaping of functional brain networks during visual object recogniton.
(a) BNS 1, (b) BNS 2, (c) BNS 3, and (d) BNS 4. Brain regions show significant differences (p < 0.05) in terms of clustering coefficient over time periods
(obtained using a k-means clustering algorithm as described in [39]) between the two categories of objects (meaningful images such as those at top left
versus meaningless ones). Orange red indicates uncorrected data for multiple comparisons, and gold indicates corrected data for multiple comparisons
using the false discovery rate approach. These results show the power of the EEG source connectivity method to track very short cogitive tasks (< 1 s) and
to reveal brain regions that associate the meaning to visual objects recognized in the human brain.

(with closed or open eyes). In this context, EEG allows for the
tracking of the temporal dynamics of resting state networks at a
subsecond timescale, a result that is not attainable using fMRI.
In these studies, the results showed the key role of some specific
brain regions, such as the posterior cingulate cortex and the prefrontal cortex (forming the so-called default mode networks), in
maintaining efficient temporal communication in the whole brain.
Other studies have focused on assessing the temporal transitions
between the main resting-state networks, such as the visual, audio,
and dorsal attentional networks [4]. Recently, the EEG source connectivity method was also used to track the task-related networks,
i.e., following the trajectory of the information processing in the
human brain from the beginning to the end of a short-duration
(subsecond) cognitive task. The brain network reconfiguration
was tracked during visual, motor, and memory tasks [3], [34], [35].
Using clustering algorithms (such as a k-means algorithm), these

studies showed that any cognitive function can be decomposed
into a set of brain network states (BNSs) that reflect the underlying cognitive processes (e.g., visual or semantic processing and
access to memory).
In Figure 4, we report some novel results showing the performance of the EEG source connectivity method within the
context of a visual cognitive task. We presented two categories
of visual stimuli on a screen: meaningful images (e.g., animals
and tools) and meaningless ones (scrambled). We asked the
participants (N = 20) to name the presented visual stimuli.
By using a combination of the wMNE and PLV (see the sections "Reconstruction of EEG Sources" and "Functional and
Effective Connectivity") computed over the trials (n = 120),
we obtained functional networks in the EEG gamma band
(30-45 Hz). We then applied a k-means clustering algorithm
to segment the EEG responses, which led to identifying four

IEEE Signal Processing Magazine

|

May 2018

|

89
Table of Contents for the Digital Edition of IEEE Signal Processing - May 2018
