SENSOR LEVEL ANALYSIS AND SOURCE LOCALISATION in M/EEG METHODS FOR DUMMIES 2013-2014 Mrudul Bhatt & Wenjun Bai.

Slides:



Advertisements
Similar presentations
Wellcome Dept. of Imaging Neuroscience University College London
Advertisements

Dynamic causal Modelling for evoked responses Stefan Kiebel Wellcome Trust Centre for Neuroimaging UCL.
EEG/MEG Source Localisation
VBM Voxel-based morphometry
STATISTICAL ANALYSIS AND SOURCE LOCALISATION
Gordon Wright & Marie de Guzman 15 December 2010 Co-registration & Spatial Normalisation.
M/EEG forward problem & solutions Brussels 2011 SPM-M/EEG course January 2011 C. Phillips, Cyclotron Research Centre, ULg, Belgium.
MEG/EEG Inverse problem and solutions In a Bayesian Framework EEG/MEG SPM course, Bruxelles, 2011 Jérémie Mattout Lyon Neuroscience Research Centre ? ?
Introduction to Functional and Anatomical Brain MRI Research Dr. Henk Cremers Dr. Sarah Keedy 1.
OverviewOverview Motion correction Smoothing kernel Spatial normalisation Standard template fMRI time-series Statistical Parametric Map General Linear.
Topological Inference Guillaume Flandin Wellcome Trust Centre for Neuroimaging University College London SPM Course London, May 2014 Many thanks to Justin.
Coregistration and Normalisation By Lieke de Boer & Julie Guerin.
Preprocessing: Coregistration and Spatial Normalisation Cassy Fiford and Demis Kia Methods for Dummies 2014 With thanks to Gabriel Ziegler.
Overview Contrast in fMRI v contrast in MEG 2D interpolation 1 st level 2 nd level Which buttons? Other clever things with SPM for MEG Things to bear in.
ECE 472/572 - Digital Image Processing Lecture 8 - Image Restoration – Linear, Position-Invariant Degradations 10/10/11.
Multiple testing Justin Chumbley Laboratory for Social and Neural Systems Research Institute for Empirical Research in Economics University of Zurich With.
07/01/15 MfD 2014 Xin You Tai & Misun Kim
Multiple testing Justin Chumbley Laboratory for Social and Neural Systems Research Institute for Empirical Research in Economics University of Zurich With.
Multiple comparison correction Methods & models for fMRI data analysis 18 March 2009 Klaas Enno Stephan Laboratory for Social and Neural Systems Research.
Multiple comparison correction Methods & models for fMRI data analysis 29 October 2008 Klaas Enno Stephan Branco Weiss Laboratory (BWL) Institute for Empirical.
Independent Component Analysis (ICA) and Factor Analysis (FA)
Preprocessing II: Between Subjects John Ashburner Wellcome Trust Centre for Neuroimaging, 12 Queen Square, London, UK.
Introduction to SPM SPM fMRI Course London, May 2012 Guillaume Flandin Wellcome Trust Centre for Neuroimaging University College London.
General Linear Model & Classical Inference
Co-registration and Spatial Normalisation
2nd Level Analysis Jennifer Marchant & Tessa Dekker
Multiple Comparison Correction in SPMs Will Penny SPM short course, Zurich, Feb 2008 Will Penny SPM short course, Zurich, Feb 2008.
Random Field Theory Will Penny SPM short course, London, May 2005 Will Penny SPM short course, London, May 2005 David Carmichael MfD 2006 David Carmichael.
Preprocessing of FMRI Data fMRI Graduate Course October 23, 2002.
Source localization for EEG and MEG Methods for Dummies 2006 FIL Bahador Bahrami.
Source localization MfD 2010, 17th Feb 2010
Coregistration and Spatial Normalisation
Multiple comparisons in M/EEG analysis Gareth Barnes Wellcome Trust Centre for Neuroimaging University College London SPM M/EEG Course London, May 2013.
EEG/MEG Source Localisation SPM Course – Wellcome Trust Centre for Neuroimaging – Oct ? ? Jérémie Mattout, Christophe Phillips Jean Daunizeau Guillaume.
EEG/MEG source reconstruction
Contrasts & Inference - EEG & MEG Himn Sabir 1. Topics 1 st level analysis 2 nd level analysis Space-Time SPMs Time-frequency analysis Conclusion 2.
Classical Inference on SPMs Justin Chumbley SPM Course Oct 23, 2008.
**please note** Many slides in part 1 are corrupt and have lost images and/or text. Part 2 is fine. Unfortunately, the original is not available, so please.
1 Experimental Design, Contrasts & Inference - EEG & MEG Joseph Brooks (ICN) Maria Joao (FIL) Methods for Dummies 2007 Wellcome Department For Neuroimaging.
EEG/MEG source reconstruction
Spatial Smoothing and Multiple Comparisons Correction for Dummies Alexa Morcom, Matthew Brett Acknowledgements.
The General Linear Model (for dummies…) Carmen Tur and Ashwani Jha 2009.
SPM Pre-Processing Oli Gearing + Jack Kelly Methods for Dummies
Methods for Dummies Second level Analysis (for fMRI) Chris Hardy, Alex Fellows Expert: Guillaume Flandin.
Multiple comparisons problem and solutions James M. Kilner
Multimodal Brain Imaging Wellcome Trust Centre for Neuroimaging, University College, London Guillaume Flandin, CEA, Paris Nelson Trujillo-Barreto, CNC,
Topological Inference Guillaume Flandin Wellcome Trust Centre for Neuroimaging University College London SPM Course London, May 2015 With thanks to Justin.
The general linear model and Statistical Parametric Mapping II: GLM for fMRI Alexa Morcom and Stefan Kiebel, Rik Henson, Andrew Holmes & J-B Poline.
MfD Co-registration and Normalisation in SPM
M/EEG: Statistical analysis and source localisation Expert: Vladimir Litvak Mathilde De Kerangal & Anne Löffler Methods for Dummies, March 2, 2016.
General Linear Model & Classical Inference London, SPM-M/EEG course May 2016 Sven Bestmann, Sobell Department, Institute of Neurology, UCL
1 Jean Daunizeau Wellcome Trust Centre for Neuroimaging 23 / 10 / 2009 EEG-MEG source reconstruction.
Methods for Dummies M/EEG Analysis: Contrasts, Inferences and Source Localisation Diana Omigie Stjepana Kovac.
Yun, Hyuk Jin. Theory A.Nonuniformity Model where at location x, v is the measured signal, u is the true signal emitted by the tissue, is an unknown.
Topological Inference
Statistical Analysis of M/EEG Sensor- and Source-Level Data
The general linear model and Statistical Parametric Mapping
2nd Level Analysis Methods for Dummies 2010/11 - 2nd Feb 2011
Methods for Dummies Random Field Theory
Multiple comparisons in M/EEG analysis
M/EEG Statistical Analysis & Source Localization
Keith Worsley Keith Worsley
Dynamic Causal Model for evoked responses in M/EEG Rosalyn Moran.
Topological Inference
The general linear model and Statistical Parametric Mapping
Dynamic Causal Modelling for M/EEG
M/EEG Statistical Analysis & Source Localization
Multiple testing Justin Chumbley Laboratory for Social and Neural Systems Research Institute for Empirical Research in Economics University of Zurich.
Multiple testing Justin Chumbley Laboratory for Social and Neural Systems Research Institute for Empirical Research in Economics University of Zurich.
Mixture Models with Adaptive Spatial Priors
Presentation transcript:

SENSOR LEVEL ANALYSIS AND SOURCE LOCALISATION in M/EEG METHODS FOR DUMMIES Mrudul Bhatt & Wenjun Bai

M/EEG SO FAR  Source of Signal  Dipoles  Preprocessing and Experimental design

3 Statistical Analysis Source Reconstruction

Statistical Analysis 1. Sensor level analysis in SPM 2. Scalp vs. Time Images 3. Time-frequency analysis

 Data is time varying modulation of EEG/MEG signal amplitude (or frequency specific power) at each electrode or sensor.  Interested in statistical significance of condition specific effects (observed at some peri-stimulus time or at a particular sensor) at sensors  Need to control FWER - the probability of making a false positive over the whole search space - AKA multiple comparisons problem  FWER scales with number of observations  Bonferroni too conservative due to assumption of independence between neighbouring samples  Can circumvent issue if space/time of interest is specified a priori  Average data over pre-specified sensors or time bins of interest - produces one summary statistic per subject per condition  If this is not possible can use topological inference

Topological inference Based on RFT RFT provides a way of adjusting p-values for the fact that neighbouring sensors are not independent due to continuity in the original data Provided data is smooth, RFT correction is more sensitive than a bonferroni correction This is the method used in SPM

Steps in SPM  Data transformed to image files (NifTI)  Procedurally identical to 1st level analysis in PET or 2nd level in fMRI after this  Analysis assumes one summary statistic image per subject per condition

Creating Summary Statistics: Conversion to images Data converted to an image by generating a scalp map for each time frame and stacking over peristimulus time Scalp maps are generated from using the 2D sensor layout (specified in data set) and linear interpolation between sensors (64 pixels each spatial direction suggested) 3D image files (space x space x time) If time-window of interest is known in advance we can average over this are and create a 2D spatial image

9

Time-Frequency Data In principle can apply topological inference for n dimensions In SPM 8 its limited to 3 dimensions If data has time-frequency components it must be reduced from 4D (space x space x time x frequency) to 3/2D Reduce data by averaging over frequency (3D) or spatial channels (2D time-frequency image) When averaging over frequency, bandwidth must be specified and a new data set is produced and is exported in the same way images in the time domain are

11

Smoothing  Smoothing: prior to 2nd level/group analysis -multi dimensional convolution with Gaussian kernel.  Important to accommodate spatial/temporal variability over subjects and ensure images conform to assumptions. Multi-dimensional convolution with Gaussian kernel

13 EEG analysis steps Epoching D/A conversion Digital filtering Baseline correction Artifact reduction Single trial averaging Re-referencing Grand averaging Plotting, spline and CSD maps Quantification Statistical evaluation

14 EEG/MEG source localization The purpose of source localization The hurdle prevent us to accurately localize the source : Inverse Problem A little recap: The advantage of EEG compare to fMRI: Superior Temporal resolution, with the cost of inferior spatial resolution

15 Why it is so challenging? Smearing and distortion

16 Inverse Problem Data Y Current density J Forward problem (well-posed) Inverse problem (ill-posed)

Analogy to understand the inverse problem

18 How We Deal with Inverse Problem 1.Setting up Assumptions(Constraints) 2.Two Basic Approaches A. Discrete Source Analysis B. Distributed Source Analysis Anatomical constraints Functional constraints Final Product: Reconstructed Source EEG/MEG Data ill-posed inverse problem

19 Constraints Assumptions about the nature of the sources Three Types of Constraints: 1. Mathematic Constraints( e.g., minimum norm, maximum smoothness, optimal resolution, temporal independence) 2. Anatomical Constraints (e.g., Normally use the subject’s MRI scan, if not, it is possible to use standardized MRI brain atlas (e.g., MNI) can be be warped to optimally fit the subject's anatomy based on the subject's digitized head shape.) 3. Functional Constraints (e.g.,

20 Discrete vs. Distributed Source Model Discrete source analysisDistributed source analysis Current dipole represents an extended brain area Each current dipole represents one small brain segment Number of sources < number of sensorsNumber of sources > number of sensors The leadfieldmatrix has more rows (number of sensors) than colums (number of sources) The leadfieldmatrix has more colums than rows Result: Source model and source waveforms Result: 3D Volume image for each time point

21 Algorithms associated with each analysis Discrete source analysisDistributed source analysis 1.Parametric Dipole source fitting: (a)Uncorrelated noise model (b)Correlated noise model (c)Global minimization 1.Spatial scanning and beamforming: independently scan for dipoles within a grid containing candidate locations (i.e., source points) All (a)(b)(c) algorithms converges to a local minima in the multidimensional space of parameters, the optimal parameters (each corresponding to a dimension) are found. The algorithms estimates five nonlinear parameters per dipole: the x, y, and z dipole position values, and the two angles necessary to define dipole orientations in 3D space. 2. Distributed MAP- based estimation assume dipoles at all possible candidate locations of interest within a grid and/or mesh called the sourcespace (e.g., source-points in grey matter) and then solve the underdetermined linear system of equations Take Home Message 1: No cure-it-all Approach

22 SPM Pipeline for source localization One kind reminder: Source Localization(source reconstruction) is a computationally intense procedure. If you get “out of Memory” error message, try more powerful computer

Step 1: Mesh

MRI template MRI – individual head meshes (boundaries of different head compartments) based on the subject’s structural scan Template – SPM’s template head model based on the MNI brain

Step 2: Coregister Co-register

Step 3: Forward Model

Step 4: Invert (The most crucial)

WHAT DO WE GET

Comparison between fMRI and MEG on Temporal Resolution Take Home Message 2: Source Localization is not perfect, being cautious in drawing any inferences related to location and strengthen of the source

31 REFERENCES  Tolga Esat Ozkurt-High Temporal Resolution brain Imaging with EEG/MEG Lecture 10: Statistics for M/EEG data  James Kilner and Karl Friston Topological Inference for EEG and MEG. Annals of Applied Statistics Vol 4:3 pp  Vladimir Litvak et al EEG and MEG data analysis in SPM 8. Computational Intelligence and Neuroscience Vol 2011  MFD 2012/13