1 / 33

Functional MRI: Design & analysis

Functional MRI: Design & analysis. Dr. N. Jade Thai. fMRI = functional magnetic resonance imaging. Introduction. Recap – 4D data the basics Basic designs The Haemodynamic time series Functional & effective connectivity Defining ROI Structural equation modelling (SEM)

diata
Télécharger la présentation

Functional MRI: Design & analysis

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. Functional MRI: Design & analysis Dr. N. Jade Thai

  2. fMRI = functional magnetic resonance imaging

  3. Introduction • Recap – 4D data the basics • Basic designs • The Haemodynamic time series • Functional & effective connectivity • Defining ROI • Structural equation modelling (SEM) • Independent components analysis

  4. Typical fMRI Setup

  5. Basic Designs • Block design • Event-related design • Mixed-design A A A A + B B B B A + B + B + B + A+A A + B + A + B + A+B Ab A Ab + BaBB

  6. ROI Time Course fMRI Signal (% change) ~2s Condition Time Condition 1 Statistical Map superimposed on anatomical MRI image Condition 2 ... Region of interest (ROI) ~ 5 min Activation Statistics Functional images Time

  7. Comparing patients & control groups Understanding neural mechanisms of cognitive difficulties in CFS/ME Thai & Crawley , University of Bristol fMRI pilot study using a Rapid Serial Visual Processing task (RSVP) on patients with CFS/ME and 4 controls. The main objective of the pilot was to ascertain The figure on the left shows brain areas significantly activated during performance of the RSVP task. Group fixed effects analysis statistical parametric map for 4 CFS/ME patients in yellow and 4 controls in blue, overlaid on the MNI brain template, threshold at p<0.01 correcting for multiple comparisons. (FWE). The performance on the tasks between patients and controls were similar, with a mean accuracy 0.78 for both groups and mean reactions times of 412ms (patients) 564ms (controls). fMRI results showed differential neural activations between patients and controls despite similar performance. Successful performance in the RSVP requires sustained attention and working memory, CFS/ME patients activated brains regions associated with components of WM whereas the controls performance is supported by brain regions associated with visual processing and executive function. The additional activations seen in patients suggest the recruitment of additional compensatory brain regions.

  8. Behavioural Data • Reaction times • Psychophysics • Accuracy measures • Behavioual measures of effects of priming

  9. FUNCTIONAL MRI STUDY OF NEGATIVE PRIMING IMPLICATIONS FOR FRONTAL DISINHIBITION N. J. Thai 1, T. L. Hodgson 2, P. Andres 3, & C. Guerrini 4 1Neurosciences, Aston University, Birmingham, United Kingdom, 2Psychology, Exeter University, Exeter, United Kingdom, 3Psychology, University of Plymouth, United Kingdom, 4Psychology, Hull University, Hull, United Kingdom

  10. Negative priming versus control Frontal-parietal network BILATERAL PRECENTAL GYRUS AREA 6 LEFT FRONTAL EYE FEILDS AREA 8 LEFT INFERIOR FRONTALAREA 47 L R LEFT INFERIOR PARIETAL AREA 40 BILATERAL SUPERIOR PARIETAL AREA 7 L N.J. Thai et al (2007) ISMRM

  11. Negative priming effect 2.2 2.2 2.0 2.0 1.8 1.8 Error Free Subjects LEFT INFERIOR PARIETAL AREA 40 R2 = 0.81 p<0.001 N.J. Thai et al (2007) ISMRM

  12. The role of the lateral prefrontal cortex and anterior cingulate in stimulus-response association reversals Parris et al (2007) JCN

  13. Left Middle frontal gyrus (BA46) Right Inferior frontal gyrus (BA47) Anterior cingulate gyrus (BA 32) Dorso-medial thalamus 2.2 2.2 2.2 2.2 2.0 2.0 2.0 2.0 Error Makers 1.8 1.8 1.8 1.8 Error Free Subjects 1.6 1.6 1.6 1.6 p<0.025 1.4 1.4 1.4 1.4 1.2 1.2 1.2 1.2 R2 = 0.002 ns 1.0 1.0 R2 = 0.91 p<0.0001 1.0 1.0 R2 = 0.79 P<0.05 R2 = 0.59 P<0.05 .8 .8 .8 .8 .6 .6 .6 .6 0 10 20 30 40 0 10 20 30 40 Subject Group 0 10 20 30 40 0 10 20 30 40 % Response Errors

  14. Do we think alike?Do our brains work in the same way? • Individual differences • Over generalisation • Reliability and reproducibility • Intersubject synchronisation • Hasson et al (2004) • 5 subjects freely viewing half an hour of a movie • Striking level of voxel-by-voxel synchronisation between subjects • Primary, secondary visual cortices and association cortices

  15. Inter-subject synchronisation

  16. Post central sulcus

  17. Structural equation modelling • SEM developed in economics, psychology & social sciences • ANOVA, multipe regression model individual observations = minimize the sum of square differences between observed and dependant variables. • SEM = covariance structure NOT variables individually • Minimizing the difference between the observed covariances and the implied model

  18. GLM statistical, linear model: X1 × β1 ε Y = + + X2 x β2 fMRI Signal = Design Matrix (predictors) xBetas + Residuals Our Data = “what we CAN x “how much of it + “what we explain” we CAN explain” CANNOT explain” • Statistical significance is basically a ratio of explained to unexplained variance Y=X.β +ε

  19. Functional interactions MTL activations Experimental stage co-ordinates Z-values x y z Sample stimuli encoding Right parahippocampal gyrus (BA 38) 15 -39 0 2.85 15 sec delay interval Right parahippocampal gyrus (BA 38) 21-36 -6 2.93 45 sec delay interval Right Hippocampus 33 -15 -15 2.83 Right Entorhinal cortex(BA28) 24-15-30 2.86 DMS retrieval stage Right Perirhinal cortex(BA 36) 27-42-9 2.97 DNMS retrieval stage Right middle temporal gyrus(BA21) 48-51 9 3.11 Results from the random effects analysis of regions of theoretical interest found to be significantly activated for sample stimulus encoding, delay period retention and retrieval for (DMS) or (DNMS). Regions of activation are displayed on representative axial slices of the MNI template brain from SPM99.The colour bar represents z-value of local maxima of voxels in a given brain region of t (9) > 2.58 corresponding to p<0.005 uncorrected for near whole brain volume. N.J. Thai et al (2001)

  20. e1 x1 x3 e3 x2 e2 y = b0 + bx + e Y=

  21. Functional connectivity Neuroanatomical model The neuroanatomical model, Talairach co-ordinates of each ROI location of extracted time course for SEM. All ROI are in the right hemisphere. Brain Region DLPFC =Dorsolateral PFC (BA46/9) 21 36 21 OFG =Orbitofrontal cortex (BA11) 15 24 -21 P = Parietal cortex (BA7) 6 -44 50 IT = Inferior temporal gyrus (BA20) 63 -12 -15 PH = Parahippocampal gyrus (BA38) 21 -36 -6 PRC = Perirhinal cortex (BA36) 27 -42 -9 ERC= Entorhinal cortex (BA28) 24 -15 -30 H = Hippocampus proper 33 -15 -15

  22. SEM Modelling • Voxel time course of local maxima was extracted for each ROI and grouped according to six conditions • A correlation matrix including all 8 ROIs was then computed for each of the 12 subjects according to conditions • A Meta-analysis was then performed to combine the correlation matrices of all 12 subjects according to the six conditions, thus accounting for inter-subject variability. • Finally the model was specified in LISREL 8.51to test for functional interactions. SHORT MATCH (15sec delay interval) SHORT NON-MATCH (15sec delay interval) LONG MATCH (45sec delay interval) LONG NON-MATCH (45sec delay interval) SHORT PERCEPTUAL MOTOR CONTROL LONG PERCEPTUAL MOTOR CONTROL

  23. Results: 15 sec delay DMS & DNMS Match Non-Match The model fit for DMS and DNMS with 15 second delay intervals was poor. DMS with 15 second delay interval ² = 72.90, df= 15, IFI = 0.69 RMSEA = 0.08. DNMS with 15 second delay intervals ² = 50.02, df= 15, IFI = 0.72 RMSEA =0.069 However a number of the path coefficients were significant at p<0.05 and are displayed above. N.J. Thai et al (2002)

  24. 45 second delay DMS and DNMS Match Non-Match The neuroanatomical model was a good fit for both DMS and DNMS trials with 45 second delay intervals. DMS with 45secs delay interval ² = 12.37, df= 15, IFI = 0.98 RMSEA =0.016. For the DNMS trials with 45secs. delay interval, ² = 26.21, df = 15, IFI = 0.98, RMSEA = 0.042. Significant path coefficients between brain regions are displayed above N.J. Thai et al (2002)

  25. Summary of results • longer delay interval allow the formation of a memory (deep encoding) of the sample stimulus and interactions between MTL & PFC • MTL hierarchically organised neural system supporting the processing of perceptual information. • The orbitofrontal cortex in inhibitory control (Rolls et al 1994; Hornak et al1996; Dias et al 1997; Kawashima et al 1996; Casey et al 1997; Konishi et al1999) • The role of the Dorsolateral prefrontal cortex monitoring the selection of visual information • Interaction across multiple memory systems (Kim &Baxter 2001;Poldrack &Packard, 2003)

  26. Psychophysiological interactions Psychophysiological interactions (PPI) is a method of connectivity analysis of fMRI which was developed to allow for the detection of interactions between brain regions specifically in response to cognitive/sensory processes (Friston et al., 1997).

  27. Effective connectivity • Neural correlates of motivation/ brain reward systems • N-back working memory

  28. Limitations • Functional connectivity only shows if there is an interaction between the 2 regions it can not tell you the direction of influence. • Both feed forward & feedback connections are modelled and weighted equally.

  29. Dynamic causal modelling • Effective connectivity • fMRI indirect measure of neuronal activity • DCM distinguishes between neuronal level and haemodynamic level • “forward model” solution that estimates the effective connectivity from the neurophysiological time series. • Using neuronal parameters as time constraints. • Limitation • Neurodynamics must be predefined to account for complex interactions within and between brain regions

  30. Summary • The analysis and the design are highly dependant on each other • Think about your hypothesis, what behavioural/ cogntive/ sensory functions/processes or neural networks are you investigating should influence your design and therefore your analysis strategy. • Go beyond the blob!

More Related