Advanced applications of the GLM: - PowerPoint Presentation

Slide1

Advanced applications of the GLM:

Cross-frequency coupling

Charité

- University Medicine Berlin Department of Neurology

University College London Wellcome Trust Centre for Neuroimaging

Bernadette van Wijk

SPM course for MEG & EEG 2017Slide2

Why care about cross-frequency coupling?

Brain

needs to integrate

information at different

frequencies and time scalesUnexplored mechanisms of information processing1Slide3

Axmacher

et al. (2009)

PNAS

Jensen & Colgin (2007) TiCS

Spatial navigation

Working memory

Canolty

et al. (2006)

Science

Florin

&

Baillet

(2015)

Neuroimage

Resting state

Modulation by behavioural task

Modulation by pathology

d

e

Hemptinne

et al.

(

2013)

PNAS

2Slide4

Jirsa & Müller, 2013; Frontiers in Comp Neurosci

Frequency

combinations Within a recorded

signal (brain region)Between

two different signals (

brain

regions

)

Many

combinations!

Various forms of cross-frequency coupling

delta-alpha

theta-beta

theta-gamma

alpha-gamma

beta-gamma

etc.

3Slide5

Amplitude

Amplitude

Envelope correlation

Dynamic causal modelling

Phase

Phase

Phase

A

mplitude

Phase

Frequency

n

:

m

phase locking index

Bispectrum

Bicoherence

Modulation

Index

Entropy

Vector

length

GLM

Phase locking envelope/

phase

Correlation envelope/

signal

Event-related PAC

…

Amplitude

Frequency

Frequency

Frequency

Cross-Frequency Coupling

Correlation

Low-frequency

phase modulates

High-frequency

amplitude

Optimization cross-frequency estimates

Instantaneous frequency tracking

Modelling non-sinusoidal waves

Selection of high-amplitude time bins

Spatial filters

4Slide6

PAC: How to detect it

How to detect PAC

Berman et al., 2012; Brain Connectivity

Bandpass filtering

Phase:

narrow bandAmplitude:

bandwidth

should

include

carrier

frequency

±

modulating

frequency

Extraction

of time series

for

phase

and

amplitude

5Slide7

How to detect PAC

Assessment of

distribution

amplitude as function of phaseStatistical evaluation (often bootstrapping)

6Slide8

Parametric

approach

No surrogate time series

Based on mean vector lengthPAC: How to detect it

General Linear Model

Penny et al. (2008)

J

Neurosci

Methods

Van Wijk et al. (2015)

J

Neurosci

Methods

7Slide9

Van Wijk et al. (2015)

J

Neurosci

Methods

Parametric

approach

No surrogate time series

Based

on

mean

vector

length

General Linear Model

8Slide10

Easy

to include other predictors:

amplitude correlations

non-linearitiesconfounding factorsPAC: How to detect it

General Linear Model

Van Wijk et al. (2015)

J

Neurosci

Methods

9Slide11

Noise level

ρ

GLM

permutations

Slightly lower statistical power for GLM

GLM vs permutation tests: simulations

van Wijk et al. (2015)

J

Neurosci

Methods

10Slide12

GLM 200

permutations

<7min 159minGLM ~24x faster to computevan Wijk et al. (2015) J Neurosci MethodsGLM vs permutation tests: real data11Slide13

PAC: How to detect it

GLM for PAC within SPM

12Slide14

Compute amplitude time series and phase time series for frequencies of interest

Create two separate files because of filter settings

First step: Time-frequency analysis

Amplitude

Phase

13Slide15

Data set with amplitude time series

Data set with phase time series

Add other time series as

regressors

Select phase or

amplitude as

regressors

Second step: Cross-frequency coupling

14

Select epoch size here (for stats) Slide16

Results

Figure is plotted + .

nii

images saved

Phase Frequency

Amplitude Frequency

Phase Frequency

15Slide17

Get in touch for

an example batch and/or script:

vanwijk.bernadette@gmail.com

Or ask for help during the practical session on Wednesday16