Introduction to fMRI Mgr. Petra Zemánková, Ph.D. 26/10/2018 fMRI outline • Advantages • MRI vs. fMRI • Basic neurophysiology • fMRI experiment • Analysis steps • Connectivity • Limitations Why is it so cool? Source: PubMed search • Noninvasive and doesn’t involve radiation • Excelent spatial and good temporal resolution • Tool for imaging entire network of brain regions engaged when performing particular tasks (cognitive/affective/motor functions) • Broad application Zephyr/Science Photo Library 0 500 1000 1500 2000 2500 3000 3500 4000 4500 5000 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 Number of studies with fMRI in the title or abstract Why it is not so cool? • Contraindications • Motion artefacts • Expensive • Surrogate signal - indirect measurement of neuronal activity – hard to interpret MRI vs. fMRI • Anatomic structure • High resolution (1 mm) • One image (volume) • function • Low resolution (3 mm) • Number of images (volumes) in succession (N)MRI principle in a nutshell T1 T2Hydrogen nucleus 1. Hydrogen nuclei (protons) are aligned by a strong magnetic field 2. Protons can be rotated using radio waves 3. Protons oscillate while returning to equilibrium 4. During realignment the nuclei lose energy and a measurable radio frequency (RF) signal 5. Detected signal is used for detailed images of body tissues fMRI principle Blood Oxygen Level Dependent - BOLD signal www.ijbem.org neural activity blood flow deoxyhemoglobin MR signal Hemodynamic Response Function (HRF) (Malonek and Grinvald, 1996) http://mriquestions.com fMRI raw data Slices (2D) Volume (3D) Acquisition time 1,5-3 s One measuring session Number of images depending on measuring time t One voxel t 4D matrix fMRI setup MR compatible screen Stimulus computer Spectrometer computer Mirror Response box Head coil http://www.howstuffworks.com/ B0 fMRI experiment fMRI experiment issues We don’t know the baseline BOLD signal level measurement of resting state or other control condition Very small signal change repetition of conditions many times, acquisition of lot of data to detect the difference Low-frequency fluctuations and BOLD signal drift alternations of experimental states during stimulation fMRI experiments - design Blocked Event-related BOLDsignal BOLDsignalstimulation stimulation Experimental design – advantages/disadvantages Blocked • Series of stimuli during time epoch(16-60s) • Simple and powerful • Easy to analyze • Very good detection power • Longer blocks may evoke heterogenous neuronal activity • Estimating relative changes between tasks Event related • Reactions to single stimuli • Lower statistic power • Estimation of HRF possible • Longer measurement necessary • More complicated data analysis Sequence parametres • T2* weighted, echo-planar BOLD images • Typical parametres on 1,5 T MR: • resolution: 64x64 (128x128) • FOV = 220 mm  3,4375 x 3,4375 mm2 • Slice thickness: 3 - 7 mm • Bigger thickness – better SNR • Number of slices: 16 - 32 • Volume acquisition time • Repetition time (TR) • 1,5 – 4 s • Stronger fields possible • 3T, 7T, (9T, 11T, ...) • Stronger field  better SNR X worse susceptibility to artefacts Preprocessing of the data Motion correction Slice-timing correction normalization smoothing Motion correction • Motion artefacts are the most problematic in fMRI • Statistical parametric mapping methods work with single voxels independently • The same voxel must correspond with the same spot in the brain Motion correction • Realignment to the first (referential) image • Searching for optimal parameters (translation, rotation) Slice timing correction • Ideal case – acquisition of the whole brain volume in an infinitely short time • Scanning of individual slices between 100 and 500 ms • First and last layer from 2 to 4s apart • Different slices different HRF • Can be corrected by weighted distance from referential slice and referential time time0 1. scan 2. scan Actual acquisition time Ref. slice (time)Linear interpolation of two points on ref. Normalization • Neuroscience research X clinical application • Transformation into standardized template (MNI, Talairach) • Linear transformations • Translation • Rotation • Size change • Nonlinear deformations Spatial smoothing • Filtration of rapid changes in signal intensity • Various filter shapes – most frequent Gaussian kernel • reduces noise in the image (increase signal to noise ration - SNR) • improves spatial distribution of the data • better fit in group comparison • Risk of loosing activated regions • data transformation – not working with the raw data Anatomical images • Normalization and co-registration with functional scans • Segmentation and rendering (3D view) • Used for display of activation maps Statistical analysis • Methods without model • ICA/PCA • Methods with model (hypothesis about the shape of the BOLD signal) • Correlation • T-test • ANOVA • AnCova • Linear regression • Multiple regression • F-test • etc.. Cases of GLM (general linear model) GLM Measured signal Regressors/modeled signals parameters residuals Y = X*b + e Hypotheses testing • T-test about β parametres • T-test is run on every voxel across the brain • Supratreshold voxels mark regions where tested effect was significant P < 0.001 T=3.1 Movement of left hand Multiple comparisons problem Neural correlates of interspecies perspective taking in the post-mortem Atlantic Salmon: An argument for multiple comparisons correction (Bennett et al., 2009) Uncorrected results Correction for multiple testing Without correction With correction • Sig. level for one voxel • When n results displayed, probability od false positive results increases n-times! • p < 0.001 • Sig. level for whole sample • FEW – family wise error • FDR – false discovery rate • p < 0.05 Functional organization in the brain Functional specialization • Specialized and spatially separated modules • Regions of interest Functional integration • Specific function is characterized by interconnection of relevant regions • Large-scale neuroimaging • networks Connectivity • Structural • Functional • Effective Structural connectivity • Mapping of anatomical connections between cortical region • white matter organization • Diffusion tensor imaging (DTI) • Static description of the system Functional connectivity • Correlation between large distance neurophysiological events • Indirect assumption about relationship between the regions • Seed analysis, ICA r = 0.70 r = 0.02 Effective connectivity • Causal directed influences between neurons or neuronal populations • Indirect assumption of connection between regions and how these are modulated by external stimulation • PPI, DCM External modulation Change of correlation in response to external stimulation More examples of fMRI applications • Resting state fMRI – large scale networks • Real time fmri - neurofeedback • multimodal imaging EEG- fMRI • Hyperscanning • Physiological mesurment ECG, breathing, galvanic response… (Carie and de Falco, 2015) fMRI limitations • Artifacts - misleading BOLD signal changes (head motion, ECG, respiration) • BOLD response to a stimulus is delayed, which restricts maximum temporal resolution achievable • Vasculature – implications for spatial resolution • Conceptual limitations (Menon and Kim, 1997) Beware – the main magnetic field is always ON! Thank you for attention…