Fractal analysis of fMRI data P. Ciuciu 1,2 1: CEA/NeuroSpin/LNAO 2: IFR49.

Slides:



Advertisements
Similar presentations
Joint Detection-Estimation of Brain Activity in fMRI using Graph Cuts Thesis for the Master degree in Biomedical Engineering Lisbon, 30 th October 2008.
Advertisements

A Theory For Multiresolution Signal Decomposition: The Wavelet Representation Stephane Mallat, IEEE Transactions on Pattern Analysis and Machine Intelligence,
Uncertainty and Information Integration in Biomedical Applications Claudia Plant Research Group for Bioimaging TU München.
INTRODUCTION Assessing the size of objects rapidly and accurately clearly has survival value. Thus, a central multi-sensory module for magnitude assessment.
HST 583 fMRI DATA ANALYSIS AND ACQUISITION Neural Signal Processing for Functional Neuroimaging Emery N. Brown Neuroscience Statistics Research Laboratory.
STAT 497 APPLIED TIME SERIES ANALYSIS
fMRI data analysis at CCBI
Automatic Identification of ROIs (Regions of interest) in fMRI data.
Dissociating the neural processes associated with attentional demands and working memory capacity Gál Viktor Kóbor István Vidnyánszky Zoltán SE-MRKK PPKE-ITK.
Overview of Neuroscience Tony Bell Helen Wills Neuroscience Institute University of California at Berkeley.
1 Haskins fMRI Workshop Part III: Across Subjects Analysis - Univariate, Multivariate, Connectivity.
Template for KyaTera presentations Nelson L. S. da Fonseca Optical Internet Laboratory - OIL Unicamp – IC, Campinas, SP, Brazil
Signal and Noise in fMRI fMRI Graduate Course October 15, 2003.
Rosalyn Moran Wellcome Trust Centre for Neuroimaging Institute of Neurology University College London With thanks to the FIL Methods Group for slides and.
The UNIVERSITY of NORTH CAROLINA at CHAPEL HILL Adaptive Weighted Deconvolution Model to Estimate the Cerebral Blood Flow Function in Dynamic Susceptibility.
Lecture 24: Cross-correlation and spectral analysis MP574.
Informational Network Traffic Model Based On Fractional Calculus and Constructive Analysis Vladimir Zaborovsky, Technical University, Robotics Institute,
From Localization to Connectivity and... Lei Sheu 1/11/2011.
Measuring Functional Integration: Connectivity Analyses
Multivariate Approaches to Analyze fMRI Data Yuanxin Hu.
Large scale models of the brain Institut des Sciences du Mouvement Viktor Jirsa Theoretical Neuroscience Group Anandamohan Ghosh Rolf Kötter Randy McIntosh.
Network modelling using resting-state fMRI: effects of age and APOE Lars T. Westlye University of Oslo CAS kickoff meeting 23/
Contrasts & Inference - EEG & MEG Himn Sabir 1. Topics 1 st level analysis 2 nd level analysis Space-Time SPMs Time-frequency analysis Conclusion 2.
Types of Scaling Session scaling; global mean scaling; block effect; mean intensity scaling Purpose – remove intensity differences between runs (i.e.,
Dynamic Causal Modelling Will Penny Wellcome Department of Imaging Neuroscience, University College London, UK FMRIB, Oxford, May
Low-Dimensional Chaotic Signal Characterization Using Approximate Entropy Soundararajan Ezekiel Matthew Lang Computer Science Department Indiana University.
Fractal Dimension and Maximum Sunspot Number in Solar Cycle R.-S. Kim 1,2, Y. Yi 1, S.-W. Kim 2, K.-S. Cho 2, Y.-J. Moon 2 1 Chungnam national University.
Modelling, Analysis and Visualization of Brain Connectivity
Introduction: Brain Dynamics Jaeseung Jeong, Ph.D Department of Bio and Brain Engineering, KAIST.
The brain at rest. Spontaneous rhythms in a dish Connected neural populations tend to synchronize and oscillate together.
FMRI and MR Spectroscopy. BOLD BOLD=Blood Oxygenation Level Dependant contrast Neurons at work use oxygen (carried by hemoglobin) 1-5 s. delay, peaks.
The General Linear Model Guillaume Flandin Wellcome Trust Centre for Neuroimaging University College London SPM fMRI Course London, May 2012.
Dynamic Connectivity: Pitfalls and Promises
UMCG/RuG BCN - NIC Journal club 25 Apr. ’08 A method for functional network connectivity among spatially independent resting-state components in schizophrenia.
False Discovery Rate for Functional Neuroimaging Thomas Nichols Department of Biostatistics University of Michigan Christopher Genovese & Nicole Lazar.
[Chaos in the Brain] Nonlinear dynamical analysis for neural signals Jaeseung Jeong, Ph.D Department of Bio and Brain Engineering, KAIST.
The General Linear Model Christophe Phillips SPM Short Course London, May 2013.
The General Linear Model Guillaume Flandin Wellcome Trust Centre for Neuroimaging University College London SPM fMRI Course London, October 2012.
BOLD functional MRI Magnetic properties of oxyhemoglobin and deoxyhemoglobin L. Pauling and C. Coryell, PNAS USA 22: (1936) BOLD effects in vivo.
Strategy for EEG/fMRI fusion Thomas Vincent 1,2 Neurospin 1: CEA/NeuroSpin/LNAO 2: IFR49 December 17, 2009.
SCHUBERT Kick-Off meeting SCaling in HUmain Brain Evoked and Rest acTivity Philippe Ciuciu
HST 583 fMRI DATA ANALYSIS AND ACQUISITION
Non-linear Realignment Using Minimum Deformation Averaging for Single-subject fMRI at Ultra-high Field Saskia Bollmann1, Steffen Bollmann1, Alexander.
Effective Connectivity: Basics
The general linear model and Statistical Parametric Mapping
Neural mechanisms underlying repetition suppression in occipitotemporal cortex Michael Ewbank MRC Cognition and Brain Sciences Unit, Cambridge, UK.
The General Linear Model
Ch9 Random Function Models (II)
Effective Connectivity
Application of Independent Component Analysis (ICA) to Beam Diagnosis
Signal fluctuations in 2D and 3D fMRI at 7 Tesla
The General Linear Model (GLM)
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
The General Linear Model
Signal and Noise in fMRI
Dynamic Causal Modelling
OCNC Statistical Approach to Neural Learning and Population Coding ---- Introduction to Mathematical.
Introduction to Connectivity Analyses
Dynamic Causal Modelling for M/EEG
Volume 79, Issue 4, Pages (August 2013)
Scalable Fast Rank-1 Dictionary Learning for fMRI Big Data Analysis
The General Linear Model
Hierarchical Models and
The General Linear Model (GLM)
Effective Connectivity
The General Linear Model
The General Linear Model
The General Linear Model
Will Penny Wellcome Trust Centre for Neuroimaging,
Selective and coherent activity increases due to stimulation indicate functional distinctions between episodic memory networks by Sungshin Kim, Aneesha.
Presentation transcript:

Fractal analysis of fMRI data P. Ciuciu 1,2 1: CEA/NeuroSpin/LNAO 2: IFR49

2 /23 12/17/2009 Outline I. Introduction II. Analysis of scale invariance in fMRI time series III. Fractal connectivity IV. Conclusions

3 /23 12/17/2009 The BOLD signal at rest Low frequency content of the resting BOLD signal: ➢ Physiological (cardio-respiratory cycles) artifacts ➢ Direct consequence of neocortical neuronal ongoing activity ➢ Vascular processes? ➢ Power spectrum at rest exhibits a 1/f law ➢ Does a long range-coherence in this activity reflect functional connectivity? [Biswal et al, MRM, 1995; Lowe et al, NIM, 1998] [Thurner et al, Phys. A 2003; Shumizu et al, NIM 2004; Fadili et al, NIM 2002; Ciuciu et al, JSTSP 2008] [Lowe et al, NIM 1998; Xiong et al, HBM 1999; Cordes et al, AJNR 2000] [De Luca, NIM 2006; Goldman et al, NR 2002; Leopold CC, 2003; Laufs et al, NIM 2003] [Kiviniemi et al, MRM 2000; Wise, NIM 2004 ]

4 /23 12/17/2009 Functional connectivity Study of resting state networks (RSNs) ➢ Multivariate exploratory analyses: PCA, spatial ICA, or other transformation ➢ Analysis based on measurements (EEG, fMRI signals) ➢ Lack of a coherent statistical framework ➢ Does brain dynamics reflect SOC, stochastic or chaotic systems? [Beckmann&Smith, IEEE TMI, 2004; De Luca et al, NIM 2006; Shimizu et al, NIM 2004; Perlbarg et al, ISBI'08; Varoquaux, MICCAI-WS09, sub. To NIM] [Werner, J Phys Paris 2008; Bedard and Destexhe, 2008, Destexhe and Contreras, Science 2006; Bedard et al, Phys. Rev Let, 2006 ;Piękniewski&Schreiber, NN in press] Need to probe scale invariance and to work on deconvolved neuronal signatures in fMRI or on reconstructed sources in MEG/EEG

5 /23 12/17/2009 Beyond classical analyses ➢ Statistical issues ➢ GLM-based framework no longer valid: loss of independence between “signal” and “noise” Long-memory correlation structure affects estimator performance ➢ Modulation of the scaling properties with stimulation ➢ Relative deactivations of RSNs during tasks [ Marre et al, 2008; Shimizu et al, NIM 2004; Ciuciu et al, JSTSP, 2008] [ Kincses, 2008] [ Abry et al, IEEE IT 2002] Need to define a proper statistical framework

6 /23 12/17/2009 Outline I. Introduction II. Analysis of scale invariance in fMRI time series III. Fractal connectivity: a novel functional connectivity test IV. Conclusions and perspectives

7 /23 12/17/2009 Scale invariance Evidence: Covariance under dilation operation Covariance under a change of scale The subpart and the whole are statistically indistinguishable No characteristic scale of time Implications: Non stationarity Long range dependence [Abry, et al, Lois d’échelles, fractales et ondelettes. Traité IC2, Lavoisier 2002]

8 /23 12/17/2009 Models of scale invariance Self-similarity: Multifractality:

9 /23 12/17/2009 Self-similarity and wavelets

10 /23 12/17/2009 High temporal resolution fMRI BOLD impulse response ~ 20 s Current EPI acquisitions in fMRI: Time of repetition ~ 1 or 2 s. EPI 2D Acquisition, TR = 2s (s) First high temporal resolution fMRI results obtained with a new rapid imaging method: localized EVI parallel sequence (s) EVI 3D Acquisition, TR = 200 ms robust to motion artifacts [Rabrait, Ciuciu et al, JMRI 2008]

11 /23 12/17/2009 Voxelwise Multifractal analysis Log scale diagrams based on Wavelet-leaders Time scale (s) Clear scale invariance from 1.5 to 15 s. [Ciuciu et al, ISBI’07]

12 /23 12/17/2009 Voxelwise Multifractal spectrum Quantify the reproductibility of each scaling exponent Evidence for a multifractal behavior [Ciuciu et al, IEEE JSTSP 2008]

13 /23 12/17/2009 Voxelwise Multifractal spectrum Influence of the activation level Modulation of MF properties between task-related TS, residuals and resting state signal [Ciuciu et al, IEEE JSTSP 2008] Strongly activated voxel Weakly activated voxel

14 /23 12/17/2009 Ongoing vs. evoked activity Estimation of first order cumulant of [Ciuciu et al, IEEE JSTSP 2008] Rest Activ Region 1Region 2 Region 3

15 /23 12/17/2009 Ongoing vs. evoked activity Estimation of second order cumulant of RestActivRestActiv RestActiv Region 1Region 2 Region 3 [Ciuciu et al, IEEE JSTSP 2008]

16 /23 12/17/2009 Summary Multifractal analysis exhibits long memory in EVI fMRI data Multifractal analysis makes feasible the comparison of ongoing (resting state) and evoked activity Evidence of change in WL-based cumulants with activation ➢ Activation induces an increase in self-similarity ➢ Activation induces a decrease in multifractality (lower irregularity)

17 /23 12/17/2009 Outline I. Introduction II. Analysis of scale invariance in fMRI time series III. Fractal connectivity IV. Conclusions and perspectives

18 /23 12/17/2009 Fractal connectivity Fractal connectivity: Particular model where the LFs of the interspectrum of each pair of process components are determined by the autospectra of the components Adequate model for exhibiting a complex network organization (eg RSN) from the LF signatures of its components

19 /23 12/17/2009 Bivariate fractal connectivity Bivariate long memory model: [Achard et al, Phys Rev E 2008] Long memory process: Coherence function:

20 /23 12/17/2009 Bivariate fractal connectivity [Wendt et al, sub to ICASSP'09] Fractal connectivity: Intuition: ➢ A single mechanism in the system uniquely controls the independent and joint properties of the multivariate dataset Test in the wavelet domain:

21 /23 12/17/2009 Testing fractal connectivity Fisher Z statistic of : : Test of fractal connectivity Statistic for the UMPI test of Eq. of means of GRVs [Wendt et al, sub to ICASSP'09]

22 /23 12/17/2009 Outline I. Introduction II. Analysis of scale invariance in fMRI time series III. Fractal connectivity IV. Conclusions and perspectives

23 /23 12/17/2009 Conclusions & Perspectives Exploratory analysis for exhibiting RSNs ➢ Extend univariate analysis to a multivariate perspective Comparison of evoked and ongoing activity ➢ Modulation of scale-invariance properties ➢ Identify inhibition/excitation mechanisms Fractal connectivity: ➢ Analysis of functional connectivity in the original space ➢ Analysis from the reconstructed neuronal time series Functional connectivity and plasticity mechanisms

24 /23 12/17/2009 Multifractality and wavelets Scaling exponents: Estimation of MF quantities with confidence intervals [Wendt et al, IEEE SP 2007]

25 /23 12/17/2009 Introduction Low frequency content of the resting BOLD signal: ➢ Physiological (cardio-respiratory cycles) artifacts ➢ Direct consequence of neocortical neuronal ongoing activity ➢ Long range-coherence in this activity reflects functional connectivity Study of resting state networks (RSNs) ➢ Exploratory analyses: PCA, ICA, or space transformation ➢ Link with a statistical framework Consequences ➢ Modulation of the LF content with evoked activity? ➢ Statistical issues: GLM-based approaches no longer valid [Thurner et al, Phys. A 2003; Shumizu et al, NIM 2004; De Luca, NIM 2006] [Biswal et al, MRM, 1995; Lowe et al, NIM, 1998]