FMRLAB_quickstart

1

FMRLAB

Quick-Start Tutorial

Version 2.0 September 10, 2002

? Jeng-Ren Duann & Scott Makeig, 2002 Swartz Center for Computational Neuroscience

Institute for Neural Computation University of California San Diego

Information and downloads: http://sccn.ucsd.edu/fmrlab Questions and feedback: fmrlab@sccn.ucsd.edu.

2

FMRLAB Quick-Start Tutorial

Introduction

1. FMRLAB Installation 1.1 Download FMRLAB

1.2 Unzip and install FMRLAB

1.3 Add the FMRLAB path to the Matlab environment

1.4 Edit the FMRLAB settings file 'icadefs.m' to set ICA defaults 1.5 Download the FMRLAB example data set

2. Functional Image Preprocessing and ICA Decomposition 2.1 Start FMRLAB 2.2 Quitting FMRLAB

2.3 Create an FMRLAB dataset 2.4 Save the FMRLAB dataset 2.5 Remove initial 'dummy' scans 2.6 Perform slice timing adjustment

2.7 Remove off-brain voxels

2.8 Decompose the data with ICA

3. Visualizing Results of ICA Decomposition 3.1 Visualize component regions of activity (ROAs) 3.2 Visualize component maps on structural images 3.3 Find dominant components by maximum Z value 3.4 Find dominant components by PVAF

3.5 Export a selected component 3.6 Spatially normalize the ROA maps

3.7 Produce a maximal intensity projection (MIP) display 3.8 Produce a 2-D slice-overlay display 3.9 Produce a 3-D dead-model rendered display

Appendix – Function List of FMRLAB A.1 Main Files

3

A.2 Functions from ICA Toolbox A.3 Functions from ICA Toolbox

A.4 Function from Supplement of SPM’99 Toolbox

4

Introduction

This document provides step-by-step guidelines for those who want to use FMRLAB, a Matlab toolbox for fMRI data analysis using independent component analysis (ICA) available under the GNU public license from http://sccn.ucsd.edu/frmlab/. This document first describes the procedures for installing the toolbox and then illustrates the procedure for using FMRLAB to analyze fMRI time series using a hands-on example. The example dataset used in this tutorial is also available at http://www.sccn.ucsd.edu/fmrlab/example/.

Theory and Background. For background information about ICA and its application to fMRI data analysis, please refer to the references available at http://www.sccn.ucsd.edu/fmrlab/. FMRLAB has a counterpart, EEGLAB, for analyzing EEG or MEG data using ICA. It also can be freely downloaded from http://www.sccn.ucsd.edu/eeglab/. Some of the visualization functions FMRLAB uses to display ICA results are adapted from functions contained in SPM99, a Matlab-based program for brain imaging visualization and analysis which can be downloaded from http://www.fil.ion.ucl.ac.uk/spm/.

Requirements. FMRLAB runs under core MATLAB (The Math Works, Inc., Natick, MA), version 5.3 or higher. Currently, the (faster) binary version of our infomax ICA routine ‘runica()’ (run from within Matlab using ‘binica()’) has only been compiled under Linux (and FreeBSD). On the visualization side, the SPM99 display functions (e.g., max intensity projection or MIP, 2-D slice overlay, and 3-D rendering) run only under Linux. For other platforms, please refer to SPM99 website and download the proper version of related functions (see list in Appendix). FMRLAB requires at least 256 MB of RAM (more is better) and a Pentium III (or IV) CPU.

Processing Time. The amount of processing required by FMRLAB is relatively modest. For example, applied to a conventional fMRI dataset, FMRLAB requires less than 10 minutes to preprocess and run ICA training using a laptop running Linux with 1.6 GHz Pentium IV CPU and 1GB memory.

Image formats. The image format used in FMRLAB is generic raw (.img) without header and footer. Thus, the experimenter needs to enter the image acquisition parameters (e.g. image height and width, number of slices, FOV, slice thickness, TR, etc) as well as the experimental paradigm (e.g., total number of scans, stimulation onset asynchrony (SOA), etc). FMRLAB provides an editor allowing users to enter this information. To convert fMR images to the FMRLAB format, the FMRLAB distribution includes some MATLAB routines to convert images from different systems (Siemens Symphony, Siemens Magnetom, GE Signa 1.5/2.0 T and Bruker MedSpec S300 3T). Please refer to http://www.sccn.ucsd.edu/fmrlab/ for further details. To display the results of regions of activity

5

(ROAs) using MIP, 2-D slice overlay, or 3-D rendering, the data need to be converted to ANALYZE format (Biomedical Imaging Resource, Mayo Foundation) used by SPM. FMRLAB is equipped with a build-in converter function for this purpose.

FMRLAB Manual. This document gives a quick-start to FMRLAB only. It gives step-by-step instructions as to how to install the toolbox and get some hands-on experience with its use. Some available FMRLAB functions are not covered in this tutorial. For the details of these and other FRMLAB functions, please refer to the FMRLAB manual, which can also be downloaded from http://sccn.ucsd.edu/fmrlab/.

1. FMRLAB Installation

1.1 Download FMRLAB

The FMRLAB toolbox for fMRI data analysis using ICA can be downloaded at: http://www.sccn.ucsd.edu/fmrlab/ as a file named fmrlab.tgz. Under Microsoft Explorer, click the right mouse button and select “Save link as ….” Under Netscape, press SHIFT + left mouse button to download the toolbox .tgz file and save it to disk.

1.2 Unzip and install FMRLAB

Copy the FMRLAB .tgz file into an FMRLAB directory, for example, “/home/ourlab/matlab/fmrlab”. Use “tar xvfz fmrlab.tgz” to unzip and untar the file. This will save all the necessary files for running FMRLAB in the fmrlab directory.

1.3 Add the FMRLAB path to the Matlab environment

Open the file startup.m using a text editor. Add the line “path( path, FMRLAB_DIR);” to the end of file. Replace FMRLAB_DIR here with the actual pathname of the FMRLAB directory (in our example, ‘/home/ourlab/matlab/fmrlab/’)

1.4 Edit the FMRLAB settings file, ‘icadefs.m,’ to set ICA defaults

Open the file icadefs.m using a text editor. On line 8, replace ICADIR by your FMRLAB directory path (for example ‘/home/lab/matlab/fmrlab’). On line 17, set ICABINARY to ‘FMRLAB_DIR/ica_linux’, using the pathname of your FMRLAB_DIR directory (in our example,

6

‘/home/ourlab/matlab/fmrlab/ica_linux’). 1.5 Download the FMRLAB example data set

The example data set used in this tutorial can be downloaded from: http://www.sccn.ucsd.edu/fmrlab/example/. This data set contains two files, 2dseq_r1 and 2dseq_str. The first is the file of functional images. The second contains the corresponding structural images. The functional images were acquired during a 5-minute experiment in which the subject was shown brief 8-Hz flickering-checkerboard stimulation lasting 0.5 s every 30 seconds (SOA). (See Duann et al., 2002 for details).

The image acquisition parameters for the functional images were: ????????

Image dimensions = 64 x 64 x 5 FOV = 250 mm x 250 mm Slice thickness = 7 mm TR = 0.5 sec

??Total number of scans = 610 (600 time points) ??Dummy scans = 10

The structural scans were T1-weighted images with the same slice positions, number of slices and FOV as the functional scans. However, they were acquired at 256 x 256 resolution to provide more structural details than the functional scans.

For this data set, the structural image acquisition parameters were: ??????

Image dimensions = 256 x 256 x 5 FOV = 250 mm x 250 mm Slice thickness = 7 mm

2. Functional Image Preprocessing and ICA Decomposition

Next we demonstrate, step-by-step, how to use FMRLAB to analyze the example fMRI data set.

2.1 Start FMRLAB

From the MATLAB command line, type fmrlab. If the environmental variables have been set properly, we recommend starting FMRLAB from the directory where the example image data are located.

7

2.2 Quitting FMRLAB

To exit from FMRLAB, select the FMRLAB menu item “Dataset > Quit”. This will clean the workspace, close any figures created by FMRLAB, and clear all the FMRLAB variables, including the FMRLAB global variable structure, FMRI. 2.3 Create an FMRLAB dataset

The procedure for creating an FMRLAB dataset is: (1) Select a functional image file. (2) Enter the image acquisition parameters for the functional scans. (3) Enter the image information for the structural scans. Select Dataset > Create Dataset from the FMRLAB menu. A Select Image File window will pop up (as below) allowing you to select the file containing the functional scans.

Move the mouse cursor to 2dseq_r1 (the functional scan file for the sample data set) and click the Open button at the bottom of the figure. Next, a Functional Image Information window will pop up (as below) allowing you to enter the image acquisition parameters for the functional scans. Fill all the fields with the values given above (see Section 1.5).

After filling in the correct values in all the fields, press the OK button at the bottom of the window. Next, the Structural Image Information window will pop up (as below) to allow you to enter the necessary structural image parameters.

8

Press the (top center) continuation button […], a File Selection dialog will be brought up for user to select file from the list, then it will fill in the filename as well as the pathname automatically. Refer to the structural image acquisition parameters (Section 1.5 above) and enter the correct values in the other fields of this window. Keep the Flip Image checkbox checked. This will flip the images to the neurology standard (“right is right”). If the sequence of structural images begins with the slice closest to the top of the head, check the Re-sort Image checkbox. This will convert the images to begin with bottom slice (as per the requirements of SPM99) for spatial normalization and visualization. The completed window should look like that shown below.

9

2.4 Save the FMRLAB dataset

To save the dataset you have just created, select Dataset > Save Dataset from the FMRLAB menu. A file selection window will pop up allowing you to input the output filename. Be sure to append the extension .fmr to the output dataset file. Then, click the “Save” button to save the dataset to disk.

2.5 Remove initial ‘dummy’ scans

The number of dummy scans was specified during the create-dataset procedure (see 2.3 above). To remove those scans from the fMRI time series, select Process > Remove Dummies from the menu.

2.6 Perform slice timing adjustment

To adjust the slice timing (using interpolation to make the acquisition times for the slices as synchronous as possible), select Process > Slice Timing. A Slice Timing Window will pop up (as below) allowing you to specify the sequence in which the functional slices were acquired. There are four possible selections: Interleaved, Ascending, Descending and User Defined. The default is Interleaved.

To use User Defined mode, enter the actual slice sequence, for example, “2 4 6 8 10 1 3 5 7 9,” into the text entry field of the Slice Timing Window. Then press OK to start the slice timing adjustment

10

process. During this process, a progress-bar window will pop up to indicate its progress. When the timing adjustment process is done, FMRLAB will close both the Slice Timing Window and the progress bar.

2.7 Remove Off-Brain Voxels

To remove off-brain voxels from the ICA training data, select Process > Extract Brain from the menu. The Extract Brain Voxels window will pop up. Before you start the voxel removal process, press the Preview button (bottom right) to load the functional images.

The images will be displayed in the top row of the window. The bottom row shows the results of using the (not-yet specified) voxel intensity threshold. The figure below shows the Extract Brain Voxels window after the functional images are loaded. The whole functional images are displayed in the first row. Because the threshold (appearing in the text entry field at the bottom of this window) is 0, the thresholded images shown in the bottom row are identical to the whole (top) images.

16

color-coded epoch time courses in acquisition order to form a color-coded 2-D BOLD image. Clicking on the Time text selection box and selecting the Component + BOLD item, a text entry window asking for the experiment epoch length and SOA will appear (as shown in the figure above). When these values are specified (followed by pressing OK), the BOLD image of the component time course will be displayed as the right lower panel below.

3.2 Visualize component maps on structural images

FMRLAB provides a function to overlay component ROA maps on the structural images. If you input structural images, you can also overlay the component ROAs on top of them by selecting Visualize > Component ROAs > On Structural Images from the FMRLAB menu. An example is shown in the figure below. The display features are the same as for mapping ROAs on the functional images (3.1 above). Although the user can also use this function to browse the component ROAs, the structural images must be available. Also, the structural maps appear much more slowly than maps on the functional images since the ROAs must be interpolated to fit the image dimensions of the structural scans. To save time, we recommend first mapping component ROAs on functional images, then

17

creating and saving ROA maps on structural images only for selected components of interest.

3.3 Find dominant components by maximum z value

Another FRMLAB component search method selects, for every voxel, the component having the maximum (z-value) weight at that voxel. The assigned component participates most strongly in generating the BOLD signal at the specific voxel. This component may also have the highest correlation coefficient between the back-projected component time course and the whole voxel time course, though this need not always be the case. Color-coding the dominant component for each voxel allows the user to graphically select voxel regions dominated by a single component. To bring up this search image, select Visualize > Dominant Component > by Max Z.

First, a ROAMAP Display window will pop up (as below) allowing you to input the lower-bound z value to use in the display. Entering a higher threshold will make the resulting figure simpler (with less “chickenpox” noise).

18

After the lower-bound threshold is set, a second ROAMAP Display window pops up to ask you if you want to show all components (option 1), or just the components selected as of interest in the component browser (option 2). You can also choose (option 3) to show all components except those on the “reject” list.

After the proper parameters are assigned, FMRLAB will calculate the z value of each voxel weights according to the activation (ui,j = Wi,k * xk,j, where i is the component number, j is the voxel index, and k is the time point of the fMRI time series) of each component, and will assign to every voxel a maximal z value (shown in the left panel of the figure below). To every voxel, the function will also assign a component having the maximum z value (as in the right panel below). We may call this component the “defining component” of the voxel.

Clicking on one of the thumbnail slice images of this map (on right, above) will pop up an enlarged

19

slice image in another window (as below), allowing the user to select a voxel of interest. Normally, we click on a voxel in a color-connected region, for example, as shown by the white arrow below) since most relevant hemodynamic processes for cognitive research may be those that affect the BOLD signals of geometrically connected voxel regions rather than of isolated voxels. After the voxel component of interest is selected, the pop-up window will be closed. On the command line FMRLAB will indicate the number of the component selected.

FMRLAB then computes the Region of Activity (ROA) of the component by searching across all voxels and finding those whose z-normalized component weights are higher than the default threshold. The function then determines the mean whole-BOLD-signal time course for the voxels in the ROA. We call this the ROA raw-mean time course. In the left panel of the figure below, the black trace shows the ROA raw-mean time course, and the red trace the back-projected time course of the defining component. The blue and green traces show the back-projected component time courses of the 2nd and 3rd components accounting most strongly for the ROA raw-mean time course variance.

The right panel shows the user the ROA of the specified defining component and the four components that account maximally for the variance of the ROA raw-mean time course. This panel also gives the pvaf for these four components. For example, below the pvaf of the defining component (IC16) is 66.7%. The second-largest pvaf (IC 47) is 22.0%. the third (IC8), 16.7% and the fourth (IC85), 10%. Note that the ROA raw-mean time course pvaf by the sum of these four independent components is 88.2% (shown under the ROA map). Note: Since spatially independent components need not have orthogonal time courses, the pvaf of the backprojected sum of two or more components is not usually the same as the sum of the individual pvaf’s of the individual component backprojections.

20

3.4 Find dominant components by PVAF

Another way to find dominant components is by ranking components by pvaf by assigning to every pixel the component maximally accounting for the variance of its raw time course. Select Visualize > Dominant Component > By PVAF from the menu. The system will first bring up a window allowing you to select the ICA components you want to consider in constructing the map. There are three options: Entering 1 will use all components in the analysis. If 2, only the browser-selected components will be considered. Entering 3 will cause the function to consider all components except those on the component-browser “Reject” list.

The two figures below show a maximal pvaf map (left) and a maximal z-value map (right). In the z-value map color shows the indices of the most highly weighted components. Here, dark red corresponds to the first (and largest) component, dark blue to the 100th (and smallest) component. These maps show some patches of connected voxels that are dominated by the same ICA component. These are often areas (and components) of functional interest.

26

Positive z values will be assigned to warm (red | orange | yellow) colors and the negative z values cold colors (green | blue | indigo). The third parameter (above) identifies the range of slice coordinate (in mm, in Talairach space) to display. It can be given by a 1-D vector (such as [2 3.5 5 …]), or in Matlab [start:gap:end] format. Example: Entering [–26:2:20] with axial orientation, tells FMRLAB to display slices from z-axis position –26 mm to +20 mm with a 2-mm gap. A typical 2-D slice-overlay display is shown below.

27

3.9 Produce a 3-D head-model rendered display

3-D rendering is probably the most popular and well-accepted format for displaying fMRI results. FMRLAB uses the 3-D template provided with SPM99 to overlay the region of activity (ROA) map onto the SPM’99 3-D template brain. The ROA maps must first be spatially normalized to standard Talairach coordinates using the spatial normalization function (see Section 3.6 above).

When the normalized ROA maps have been created, select Visualize > 3-D Display to start the 3-D rendering. First, the Pick a File Window will pop up (below, left), allowing you to specify the ROA map to display (normally a file named, again, something like “nroa_005.img”). When ready, click the Open button to close the file selection window and bring up the 3-D Rendering window (right below). The top parameter input is the lower-bound z-value threshold, which is used to ignore insignificant voxels in the ROA map. The second entry specifies the translucency of the color display. Translucency allows the viewer to “see through” the brain to activations within the outer brain surface. Typical values for translucency are 0.25, 0.5, 0.75, 1 or NaN. The lower the value, the more opaque the brain template. To display without translucency, use [NaN] (Matlab for “not a number”). The third option allows the user to specify which 3-D brain template to use. Possible values are 1 (SPM96 template), 2 (subject-average template) or 3 (single-subject template).

After these inputs are complete, press OK to begin the 3-D rendering, which will produce a figure like that below.

28

For other functions in the FMRLAB toolbox, see the FMRLAB manual. We hope that you will enjoy exploring the complexity of BOLD data sets using FMRLAB, and that in so doing, you may make exciting discoveries about what hemodynamics may tell us about how human brain dynamics support experience and behavior.

Jeng-Ren Duann

Scott Makeig

La Jolla 9/2002

29

Appendix – Function List of FMRLAB

A. 1 Main Files: fmrlab.m fmrlab.mat license.txt boldimage.m clear_fmri_global.m clear_workspace.m dilation.m erosion.m execute_ica.m export_result.m

main function of FMRLAB toolbox

MAT file to keep necessary parameters for FMRLAB toolbox GNU license

image the intertrial dynamics of BOLD signal

clear FMRI data structure from the working environment

clean up the workspace by closing all the opened windows by FMRLAB perform dilation on input image (used in extract_brain_ui()) perform erosion on input image (used in extract_brain_ui()) execute ICA with GUI for users to specify parameters

export region of activity (ROA) maps to ANALYZE format for further visualization

extract_brain_by_edit.m set threshold value for removing off-brain voxels by key in value in edit box extract_brain_ui.m GUI for user to remove the off-brain voxels extract_brain_ui.mat fmri_bpfilter.m get_status.m ica_linux jr_color.m jr_normalization.m jr_render.m load_dataset.m make_blobs.m map_on_fmri.m map_on_struc.m modify_param.m modify_struc_info.m progressbar.m pvafmap_ui.m

MAT file to keep the necessary fields for extract_brain_ui() perform ideal high/band/low-pass filter on fMRI time courses get current status of FMRI data structure main program of binary ICA

specify the colormap used to display the functional ROAs 3D normalize ROA map to standard brain template

3D rendering of ROA map on 3D standard brain template provided by SPM99

load FMRI data structure up to the working space

read spatially normalized ICA ROA map and convert it to the data structure used to in 3D rendering processes

component browser by overlaying ROA onto 2D slices of functional images with interactive graphic user interface

component browser by overlaying ROA onto 2D slices of structural images with interactive graphic user interface

modify necessary parameters for data analysis and visualization modify parameters of structural images

progress bar showing the progress of the running program

display percentage variance accounted for (pvaf) map with graphic user 30

interface

read_analyze_hdr.m read_structure.m remove_dummy.m reselect_fmri.m rm_slice.m roamap_ui.m roaproj_ui.m

read header file of images saved in ANALYZE format read structural images according to the specified parameter remove dummy scans from the fMRI time series data select new fMRI data set with the same parameters remove noisy slices from fMRI data display ROA maps with graphic user interface

ROA back-projection to fine the back_projected ICA time courses and mean time course of the ROA voxels and calculate the PVAF for a specified component

find the mean time course of the ROA voxels save FMRI data structure as .fmr file in disk

construct FMRI data structure as global variable in current workspace for further analysis

set_fmri_global() with interactive graphic user interface

select structural images into FMRI data structure and set the necessary parameters

call show_actslice() and display normalized ICA ROA maps onto normalized 2D structural image of individual subjects or 2D brain template in a slice-by-slice manner

display normalized ICA ROA maps onto the rendered 3D brain templates provided by SPM99

show_actslice.m

overlay the activation map onto the structural image. Both structural images and activation map should be normalized to the standard brain space (Talairach space) with SPM

display normalized ICA ROA maps on mip template provided by SPM99 show parameters of image acquisition and analysis in main window adjust image inhomogeneity due to different acquisition timing for each slice graphic user interface of slice_timing()

necessary information needed for slice_timing_ui()

spatially smooth image slices to remove the spiky noise due to slignal lose in image acquisition

temporally smooth fMRI time courses with 3 time-point averaging compact version of subplot()

roatc_ui.m save_dataset.m set_fmri_global.m set_fmri_global_ui.m set_struc_info.m show_2d.m

show_3d.m

show_mip.m

show_parameters.m slice_timing.m slice_timing_ui.m slice_timing_ui.mat spatial_smooth.m temporal_smooth.m tightsubplot.m

31

A.2 Functions from ICA Toolbox binica.m Matlab function to interface stannd alone binary version ICA (excutable by C) binica.sc cbar.m eegfilt.m floatread.m floatwrite.m icadefs.m sbplot.m scale.m textsc.m

A.3 Functions from SPM’99 Toolbox mip.mat template file for maximal intensity projection (MIP) display render_single_subj.mat

3D rendered brain template from single subject to render the resulting ROA maps

render_smooth_average.mat 3D rendered brain template from smoothed averaged brain to render the resulting ROA maps

render_spm96.map 3D rendered brain template from SPM96 to render the resulting ROA maps

spm_affsub3.m spm_atranspa.m spm_atrnaspa.mexlx spm_chi2_plot.m spm_conv_vol.m spm_conv_vol.mexlx spm_create_image.m spm_dctmtx.m spm_figure.m spm_get.m spm_get_space.m spm_global.m spm_global.mexlx

highest level subroutine involved in affine transformations Multiplies the transpose of a matrix by itself - a compiled routine mex file for spm_atranspa() in Linux.

display a plot showing convergence of an optimization routine convolves a mapped volume with a three dimensional separable function mex file for spm_conv_vol() in Linux create an image file

creates basis functions for Discrete Cosine Transform setup and callback functions for Graphics window user interface for filename selection

get or set the best guess for the space of the image

returns the global mean for a memory mapped volume image mex file for spm_global() in Linux

script file to keep initial values for ICA training showing color bar

(high|band|low)-pass filter fMRI time courses using two-way least-square FIR filtering (Signal Processing Toolbox needed) read floating-point binary data from a file write floating-point binary data into a file define ICA defaults

create axes in arbitrary subplot grid positions and sizes

scales an image such that its lowest value attains newMin and its highest value attains newMax

print text at the specified location in Matlab figure

32

spm_hread.m spm_hwrite.m spm_list_files.m spm_list_files.mexlx spm.m

spm_matrix.m spm_platform.m spm_project.m spm_project.mexlx spm_sample_vol.m spm_sample_vol.mexlx spm_slice_vol.m spm_slice_vol.mexlx spm_smooth.m spm_str_manip.m spm_type.m spm_unlink.m spm_unlink.mexlx spm_vol_ecat7.m spm_vol.m

spm_vol_minc.m spm_write_plane.m spm_write_sn.m T1.hdr T1.img

reads a header of ANALYZE formatted image write a header of ANALYZE formatted image lists files and directories

mex file for spm_list_files() in Linux

Statistical Parametric Mapping (startup function) returns an affine transformation matrix

platform specific configuration parameters for SPM forms maximium intensity projections mex file for spm_project() in Linux

returns voxel values from a memory mapped image mex file for spm_sample_vol() in Linux

returns a slice through a memory mapped image mex file for spm_slice_vol() in Linux 3 dimensional convolution of an image miscellaneous string manipulation options

translates data type specifiers between SPM & Matlab representations routine for silently deleting files on disk mex file for spm_unlink() in Linux

get header information etc. for ECAT 7 images get header information etc for images

get header information etc. for MINC images write a transverse plane of image data write out normalized images

header file of SPM99 T1 template images image file of SPM99 T1 template images

A.4 Function from Supplement of SPM’99 Toolbox slice_overlay.m overlay functional map onto structural images in standard Talairach coordinates

本文来源：https://www.bwwdw.com/article/9mq6.html