FMRI: Biological Basis and Experiment Design Lecture 23: GLM in multiple voxels Errata Final project review Keeping track of space Multi-voxel GLM Before.

Slides:



Advertisements
Similar presentations
2nd level analysis – design matrix, contrasts and inference
Advertisements

1 st Level Analysis: design matrix, contrasts, GLM Clare Palmer & Misun Kim Methods for Dummies
Kriging.
The General Linear Model Or, What the Hell’s Going on During Estimation?
Designing a behavioral experiment
Opportunity to Participate EEG studies of vision/hearing/decision making – takes about 2 hours Sign up at – Keep checking.
Rapid-Presentation Event-Related Design for fMRI
1Neural Networks B 2009 Neural Networks B Lecture 1 Wolfgang Maass
Efficiency in Experimental Design Catherine Jones MfD2004.
Opportunity to Participate
The General Linear Model (GLM)
FMRI: Biological Basis and Experiment Design Lecture 17: Baselines, SNR & CNR Visualizing data Trial triggered average Baselines SNR and CNR –Block vs.
Vectors and Vector Equations (9/14/05) A vector (for us, for now) is a list of real numbers, usually written vertically as a column. Geometrically, it’s.
Spike-triggering stimulus features stimulus X(t) multidimensional decision function spike output Y(t) x1x1 x2x2 x3x3 f1f1 f2f2 f3f3 Functional models of.
FMRI: Biological Basis and Experiment Design Lecture 16: Final projects Experiment design Data simulation Data analysis 1 light year = 5,913,000,000,000.
FMRI: Biological Basis and Experiment Design Lecture 24: GLM in all its glory.
FMRI: Biological Basis and Experiment Design Lecture 22: GLM 101 Which linear equations? Design matrix Solution assuming HIRF, single voxel  +==+
Laurent Itti: CS599 – Computational Architectures in Biological Vision, USC Lecture 7: Coding and Representation 1 Computational Architectures in.
FMRI: Biological Basis and Experiment Design Lecture 26: Significance Review of GLM results Baseline trends Block designs; Fourier analysis (correlation)
Self-organizing Learning Array based Value System — SOLAR-V Yinyin Liu EE690 Ohio University Spring 2005.
Signal and Noise in fMRI fMRI Graduate Course October 15, 2003.
Efficiency – practical Get better fMRI results Dummy-in-chief Joel Winston Design matrix and.
The General Linear Model
How To Do Multivariate Pattern Analysis
Functional Spectroscopy of Brain Activation Following a Single light Pulse: Examinations of the Mechanism of the Fast Initial Response. J.Henning,C.Janz,O.Speck,and.
Basics of Functional Magnetic Resonance Imaging. How MRI Works Put a person inside a big magnetic field Transmit radio waves into the person –These "energize"
Position Reconstruction in Miniature Detector Using a Multilayer Perceptron By Adam Levine.
The future of MINC Robert D. Vincent
MAE126B Report Writing Lecture Refer also to MAE126A Report Writing Lecture.
Signal and noise. Tiny signals in lots of noise RestPressing hands Absolute difference % signal difference.
MVPA Tutorial Last Update: January 18, 2012 Last Course: Psychology 9223, W2010, University of Western Ontario Last Update:
FMRI Methods Lecture7 – Review: analyses & statistics.
Biological Modeling of Neural Networks: Week 12 – Decision models: Competitive dynamics Wulfram Gerstner EPFL, Lausanne, Switzerland 12.1 Review: Population.
Statistical learning and optimal control: A framework for biological learning and motor control Lecture 4: Stochastic optimal control Reza Shadmehr Johns.
A comparison of methods for characterizing the event-related BOLD timeseries in rapid fMRI John T. Serences.
Functional Brain Signal Processing: EEG & fMRI Lesson 14
FEAT (fMRI Expert Analysis Tool)
The General Linear Model (for dummies…) Carmen Tur and Ashwani Jha 2009.
1 Lecture 5 Post-Graduate Students Advanced Programming (Introduction to MATLAB) Code: ENG 505 Dr. Basheer M. Nasef Computers & Systems Dept.
1 Lecture 3 Post-Graduate Students Advanced Programming (Introduction to MATLAB) Code: ENG 505 Dr. Basheer M. Nasef Computers & Systems Dept.
FMRI Modelling & Statistical Inference Guillaume Flandin Wellcome Trust Centre for Neuroimaging University College London SPM Course Chicago, Oct.
The General Linear Model
The General Linear Model Guillaume Flandin Wellcome Trust Centre for Neuroimaging University College London SPM fMRI Course London, May 2012.
Analysis of FMRI Data: Principles and Practice Robert W Cox, PhD Scientific and Statistical Computing Core National Institute of Mental Health Bethesda,
The general linear model and Statistical Parametric Mapping II: GLM for fMRI Alexa Morcom and Stefan Kiebel, Rik Henson, Andrew Holmes & J-B Poline.
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.
Group Analyses Guillaume Flandin SPM Course London, October 2016
State Space Representation
FMRI experimental design and data processing
The General Linear Model
While Loops BIS1523 – Lecture 12.
The General Linear Model (GLM): the marriage between linear systems and stats FFA.
CSCI1600: Embedded and Real Time Software
Attention Narrows Position Tuning of Population Responses in V1
Advances in Event-Related fMRI Design
Optimization of Designs for fMRI
Michael S Beauchamp, Kathryn E Lee, Brenna D Argall, Alex Martin 
The General Linear Model
State Space Analysis UNIT-V.
Bayesian Methods in Brain Imaging
Adaptive multi-voxel representation of stimuli, rules and responses
Michael S Beauchamp, Kathryn E Lee, Brenna D Argall, Alex Martin 
Linear Transformations
Chapter 3 General Linear Model
The General Linear Model
The General Linear Model (GLM)
The General Linear Model
Basics of fMRI and fMRI experiment design
CSCI1600: Embedded and Real Time Software
Presentation transcript:

fMRI: Biological Basis and Experiment Design Lecture 23: GLM in multiple voxels Errata Final project review Keeping track of space Multi-voxel GLM Before After x

Erratum: ICE9 % Repeat the "experiment" a few times %%!!! Wrong way... shouldn't run the same design again and again - then %%the experimental conditions aren't nearly as randomized as they could %%be! nScans = 10; clear data for n = 1:nScans data(n,:) = caoSimulateData(neural,TR,BOLDamp,SNR,1); end % "Analyze" data... % %Check out code that does TTA... TTAlength = 20; [TTA TTAsem count] = caoComputeTTA(design,data,TTAlength); TTA_minus_baseline = TTA - repmat(TTA(:,1),[1 size(TTA,2)]); subplot(2,1,1) errorbar(repmat((0:(TTAlength-1))',[1 size(TTA,2)]),TTA,TTAsem); title(['nScans = ' num2str(nScans)]); axis tight subplot(2,1,2) errorbar(repmat((0:(TTAlength-1))',[1 size(TTA,2)]),TTA_minus_baseline,TTAsem); title(['nScans = ' num2str(nScans)]); axis tight legend(num2str(count))

Erratum: ICE9 % Repeat the "experiment" a few times % Right way - new trial order each repetition nScans = 10; fixedLength = 500; design = zeros(nScans,fixedLength); neural = zeros(nScans,fixedLength); data = zeros(nScans,fixedLength); for n = 1:nScans % For this, I'd built in a flag that would make the design vector the % same length each time (adding 0's at end) to deal with the fact that % randomized ISI's yield unpredictable scan lengths design(n,:) = caoMakeER(waitUpFront,nStimTypes,ISIrange,nStim,500); for nt = 1:length(design) if design(n,nt) % if there's a stimulus neural(n,nt) = relResp(design(n,nt)); % pick the appropriate response end data(n,:) = caoSimulateData(neural(n,:),TR,BOLDamp,SNR,1); end

Basic assignment, final project (all projects need these components) Abstract (250 words) Background - experiment design and neuroscience question motivating the experiment (1 page, references) Methods (2 – 4 pages) –MR Methods section (resolution, TR, TE, protocol) –Prediction of neural activity –Prediction of BOLD response (noiseless model) –Simulated data (BOLD + noise - physiological and/or thermal) –ROI selection –GLM Results (1 – 3 pages) –Paragraph describing simulation output + 1 figure –Paragraph describing GLM results + 1 figure Discussion (1 - 2 pages) –3 things that might not be as expected if you ran the experiment

Building blocks, final project (weekly assignments, 2 nd half of semester) Abstract (250 words) Background - experiment design and neuroscience question motivating the experiment (1 page, references) Methods (2 – 4 pages) –MR Methods section (resolution, TR, TE, protocol) – WA 8 –Prediction of neural activity – WA7 –Prediction of BOLD response (noiseless model) – WA7 –Simulated data (BOLD + noise - physiological and/or thermal) – WA8 –ROI selection –GLM – WA10 & WA11 Results (1 – 3 pages) –Paragraph describing simulation output + 1 figure – WA8 –Paragraph describing GLM results + 1 figure – WA10 & WA11 Discussion (1 - 2 pages) –3 things that might not be as expected if you ran the experiment

Presentation schedule Groups –Denkinger –Han, Honwanishkul, Roher –Haut, Johnson –Kruse, White –Kwon –Liu, Schmidt –Seo, Vizueta –Theodorou –Thompson Times –May 2 9:45 – 10:05 Kwon 10:05 – 10:25 Theodorou 10:25 – 10:45 Denkinger 10:45 – 11:15 Big group HHR –May 4 9:45 – 10:05 Thompson 10:05 – 10:25 Big group 10:25 – 10:45 Seo, Vizueta 10:45 – 11:15 Haut, Johnson

A 1,1 A 2,1 A 3,1 A 4,1 A 1,2 A 2,2 A 3,2 A 4,2 A 1,3 A 2,3 A 3,3 A 4,3 A 1,m A 2,m A 3,m A 4,m Linear model for BOLD in a single voxel A = x = x1x2x3x4x1x2x3x4 y = y1y2y3ymy1y2y3ym Ax = y Design matrix, [m x n] - m time-points - n stimulus types Data [m x 1] - response through time Responses [n x 1] - for each stimulus, a scalar (single number) representing how well that voxel responds to that stimulus...

A 1,1 A 2,1 A 3,1 A 4,1 A 1,2 A 2,2 A 3,2 A 4,2 A 1,3 A 2,3 A 3,3 A 4,3 A 1,m A 2,m A 3,m A 4,m Linear model for BOLD in multiple voxels A = x = x 1,1 x 2,1 x 3,1 x 4,1 y = y 1,1 y 2,1 y 3,1 y m,1 Ax = y Design matrix, [m x n] - m time-points - n stimulus types Data [m x p] - response through time Responses [n x p] - for each stimulus, a scalar (single number) representing how well that voxel responds to that stimulus x 1,2 x 2,2 x 3,2 x 4,2 x 1,p x 2,p x 3,p x 4,p y 1,2 y 2,2 y 3,2 y m,2 y 1,p y 2,p y 3,p y m,p...

Keeping track of voxels 2D (or 3D) array of data 1D vectorX & Y (& Z) coords =+

Fake experiment – distributed neural activity Brain (3 types of responses) 3 types of stimuli; 3 types of "neuron" (voxel?)

Fake experiment – simulated data Brain (3 types of responses) voxels (Baseline trends, noise dominate visualization)

Design Matrix - Stimulus 1 - Stimulus 2 - Stimulus 3 - Linear drift - Half-cycle cos drift cycle cos drift

Design Matrix + (sim.) data X= ? time voxels

(An aside on rows and columns) Why do [nTpts nVoxels] instead of [nVoxels nTpts]? –Typical experiment 6 scans, 200 tpts per scan --> nTpts = brain, 64 x 64 x 30 voxels --> nVoxels = 122,880 –Size of A [nTpts nVoxels]: [122, ] [nVoxels nTpts]: [ ,880] –Size of A T A [nTpts nVoxels]: [ ] [nVoxels nTpts]: [122, ,880] !

Estimate of voxel responses X= time voxels x est = x +  = (A T A) -1 A T (y +  )

Estimate of voxel responses Response to Stim1Response to Stim3Response to Stim2 Estimated Input

WA10 – Build design matrix & GLM analysis tool