Joint lab meeting

Characterizing neural oscillations and functional connectivity between tonic pain and control conditions within frequency bands and lengths of epochs

Wenxin (Anny) Su

Analysis overview:

Preprocess: average reference; 50 Hz notch filter; band pass filter [0.1,100]; spatial filter (Laplacian transform); chunking data into 2s / 10s epochs

Brain activity (power spectral density) & functional connectivity (index matrix transformed into connectivity strength)

Statistical analysis: cluster-based permutation test(neighbours around 4 in average, 2000 times permutation, p = 0.05, two-tailed, results were corrected by surrogate data )

Research aim and hypothesis:

Capture pain-related biomarker by comparing

pain(H) vs control(BL1O) => Pain-related brain response (perceptual + emotional)

pain vs warm(W) conditions => Pain-related brain response only in emotion

36 subjects; 62 channels; epochs longer than 2 seconds reach stationarity rate over 97% in average across all conditions.

3 conditions (H; W; BL1O) * 3 frequency bands (theta; alpha; beta) * 2 epoch lengths (2s & 10s)

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

theta (4-8 Hz)

alpha (8-13 Hz)

Local brain activity analysis:

local relative PSD

We calculated relative PSD within theta, alpha, and beta frequency bands for each subject and channel in each condition and length of epoch.

Average relative PSD value for each channel was calculated for each subject in each condition and length of epoch within each frequency band.

Cluster-based permutation tests were performed for all 36 subjects in H vs BL1O and H vs W comparison for each frequency band and length of epoch.

Corrected t maps were shown left with color bar from [-4,4]. Topographies without significant cluster were presented with reduced opacity.

Oscillatory brain activity

Local brain activity analysis:

local absolute PSD

We calculated absolute PSD within theta, alpha, and beta frequency bands for each subject and channel in each condition and length of epoch.

Average absolute PSD value for each channel was calculated for each subject in each condition and length of epoch within each frequency band.

Cluster-based permutation tests were performed for all 36 subjects in H vs BL1O and H vs W comparison for each frequency band and length of epoch.

Corrected t maps were shown left with color bar from [-4,4]. Topographies without significant cluster were presented with reduced opacity with uncorrected t values.

Global brain activity analysis:

absolute PSD & relative PSD

In each condition and lengths of epochs, oscillatory brain activity was assessed between 1 to 100 Hz using a fast Fourier transformation and multitapers with 1 Hz smoothing for each epoch and electrode.

The power spectra were averaged acorss epochs, electrodes and subjects, resulting in a global absolute PSD for each condition and lengths of epoch.

To eliminate inter-subject variability, we computed global relative PSD by dividing the absolute PSD by the summed power across the whole frequency range for each subject.

beta (14-30 Hz)

2-seconds epochs

10-seconds epochs

2-seconds epochs

10-seconds epochs

Local relative PSD t maps

Local absolute PSD t maps

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

H vs BL1O

H vs W

theta (4-8 Hz)

Functional connectivity

alpha (8-13 Hz)

beta (14-30 Hz)

10-seconds epochs

2-seconds epochs

10-seconds epochs

2-seconds epochs

10-seconds epochs

2-seconds epochs

10-seconds epochs

2-seconds epochs

PLI t maps

WPLI t maps

Coherence t maps

PLV t maps

Correlation t maps

In temporal domain, Pearson's correlation was calculated for each pairs of channels for each frequency band.

Connectivity strength for each channel was defined by the average correlation value of such channel to all other channels.

In each length of epoch and frequency band, comparison with connectivity strength between H and other conditions was analysed by cluster-based permutation test.

Tle last index in the phase domain was Weighted PLI.

The WPLI was invented with weighting the contribution of phase difference by the magnitude of its imaginary part in which real synchrony at zerophase lag was also discarded.

The value of WPLI was the nonzero absolute value of the mean across all transformed phase differences by signum function, ranges also from 0 to 1.

In the frequency domain, magnitude-squared coherence was calculated for each pairs of channels for each frequency band, and the maximum coherence value was selected. Coherence ranges from 0 to 1 with 1 means two signals are completely similar in the frequency domain.

Connectivity strength for each channel was defined by the average coherence value of such channel to all other channels.

In each length of epoch and frequency band, comparison with connectivity strength between H and other conditions was analysed by cluster-based permutation test.

Another index in the phase domain was phase lag index.

The PLI measures the asymmetry of the distribution of phase differences between two signals using the imaginary part of the phase difference.

The value of PLI was the absolute value of the mean across all transformed phase differences by signum function, which categorises phase differences from 0 to pie into positive 1 and from -pie to 0 into negative 1.

The value of PLI ranges from 0 to 1, in 1 if all phase differences are valued positive (or all negative) which indicates a consistent, nonzero phase difference between two signals.

In the phase domain, PLV was calculated for each pairs of channels for each frequency band. The value of PLV was an average phase difference calculated from the extracted phase by Hilbert transform between two signals in the frequency range of interests.

PLV ranges from 0 to 1. If the phase difference varies little, therefore the PLV value is close to 1.

Connectivity strength for each channel was defined by the average PLV value of such channel to all other channels.

In each length of epoch and frequency band, comparison with connectivity strength between H and other conditions was analysed by cluster-based permutation test. The same as PLI and WPLI.

Conclusion

1. Clusters of channels were found significantly different between H vs BL1O conditions and H vs W conditions. The previous one has larger and easily detectable results than the latter one.

2. Notably, distinctive patterns among each frequency band could be observed in all assessments. Stable Functional connectivity differences could be found in the alpha band.

3. The relative longer term of the chunking method led to smaller PSD values and phase connectivity measurements.

Thanks for you listening :)