1. 报告人：蔡世民 合作者：禚钊，乔赫元，傅忠谦，周佩玲 电子科学与技术系
Cluster structure and localization of brain functional networks based on the ERP signals of auditory task
报告人：蔡世民
合作者：禚钊，乔赫元，傅忠谦，周佩玲
电子科学与技术系

2. Outline
Introduction
Data Acquisition
Phase Synchronization
Results

3. Introduction What is brain functional networks?
A brain functional network can be derived from the
physiological signals such as EEG,MEG, ECoG, and
fMRI.
Nodes: ROIs (fMRI) or channels (EEG,MEG,ECoG).
Edges : correlation (interaction) between ROIs or
channels.

4. Introduction (cont.) Construction of large-scale brain functional
networks
--Pearson correlation coefficient
--Correlation coefficient based on Wavelet transform
--Mutual information
--Nonlinear interdependence
--Phase synchronization based on Hilbert transform

5. Introduction (cont.)

6. Introduction (cont.)
Brain functional networks posses some common structures of complex networks
--small-world property (D. S. Bassett, Neuroscientist 12, 512, 2006)
--scale-free property (V. M. Eguiluz, et al. PRL 94, , 2005)
--Hierarchical organization (C. S. Zhou, et al. PRL 97, , 2006)

7. Data Acquisition
Five persons were asked to distinguish between synonymous and non-synonymous word pairs (the second word presented 1 second after the first) they heard.
Data epochs were extracted from 2 sec before the second word onset to 2 sec after the second word onset.
Sampling rate (Hz) 200.
T = 0s. Start recording
T = 1s. First word
T = 2s. Second Word.
T = 4s. Stop recording.

8. Data Acquisition (cont.)
61-channel ERP signal.
Letters refer to the main areas of the cortex:
F: the frontal (额叶),
T: left and right temporal (颞叶),
P: the parietal (顶叶),
O: the occipital (枕叶),
C : central,
FP: frontopolar(额极),
AF: anterior frontal(前额叶).

9. Data Acquisition (cont.)
The testee was cued to move a particular
figure by displaying the corresponding word, such as “thumb”;
Each cue lasted two seconds following
an another two seconds resting period;
Band pass filtered between 0.15 and 200 Hz, and sampled at 1000 Hz;
The experiment lasted 400 seconds for echa testee.

10. Data Acquisition (cont.)
Sketch of ECoG recording

11. Phase Synchronization
Phase of real value time series
bivariate phase synchronization index
If two time series are complete phase synchronized, this value will be the maximum.

12. Generating Networks
Divide data into four parts:1st ,2nd ,3rd and 4th second.
resting state: 1st and 4th seconds;
auditory task state: 2nd and 3rd seconds
Fixed mean degree for four parts
mean degree: 4-30,increased
by 2.
The thresholds as a function
of mean degree ⟨k⟩.

13. Generating Networks (cont.)
Divide data into two parts:
-- task state: 1st 2 seconds
-- resting: 2nd 2 seconds
Fixed mean degree for
two parts
--mean degree: 4-30,increased
by 2.
ECoG

14. Results
Networks show different property during rest and task state for EEG

15. Networks show different property during
Results (cont.)
Networks show different property during
rest and task state for ECoG

16. Results (cont.)

17. Results (cont.)
Networks show small-world property
ECoG
EEG

18. 6.Conclusion and outlook
The diversity of topology between the resting and task states suggests the variance of correlations among the functional modules.
The larger cluster coefficients during task mean that the correlations of cortex regions are more localized in the large-scale brain functional networks
The connectivity of networks under task state presents a better performance than that under resting state via the estimation of giant components’ sizes.
Moreover, the mean path lengths of brain functional networks confirm the small world property.
Future work will focus on the location of community during the cognitive process and the relationship between the large-scale functional networks and micro-scale neural dynamics via diffusion
tensor imaging