> I have
> some difficulties in using SPM99. Maybe, my problem is very simple, Since
> the circle in where I can consult is very limited, I acquired the knowledge
> mostly though your website, and little hint in some articles else. I¡¯ve
> read up the information on this website mostly, especially the spm99
> MANUAL, downloaded all data website providing and processed it according to
> the accessary instruct note. Information is scarce for me as my poor basic
> knowledge and intellect. I¡¯ll appreciate your detailed answer.
> Firstly, I depict the steps of my processing using spm99. Hope you¡¯ll
> correct the mistakes.
> In our study, we investigate the respond of the visual cortex stimulated by
> rotating grating with block designing. Acquire data using GE-Signa 1.5T
> MRI. TR:3000ms, a block activate state (or control state): 30s. a session
> have 20 blocks. There are 100 acquisitions in all.
> 1. To batch rename the EPI and structure images, transform the original
> data to *.img using the Acdsee software. (2400 EPI images and 24 T1 images)
> 2. transform all EPI data to 100 volumes (=acquisitions), 24 structure
> images to one volume(*.img & *.hdr) for spm99 using MRIcro software. put
> 100+1 volumes into a directory.
> 3. run matlab, type ¡®spm fmri¡¯ to start spm99. set the directory which
> data locates in as the working directory.
>
> 1st.******************************HDR edit &
> Display**************************** (1). Press ¡°HDR edit¡± (bottom left
> panel), fill in all needed item. A: edit the hdr of structure images
> ¡®set image dimension: 512 512 24
> set voxel dimension: 0.47 0.47 4
> set scalefactor: 1
> set datatype: inte16
> set offsite into file: 0
> set oringal [x,y,z]: 0 0 0
> set description: stucture
> apply to image: select the structure.img and press ¡°done¡±
> quit¡¯
> B: edit the hdr of EPI images
> Press ¡°HDR edit¡±again
> ¡®set image dimension: 64 64 24
> set voxel dimension: 3.75 3.75 4
> set scalefactor: 1
> set datatype: inte16
> set offsite into file: 0
> set oringal [x,y,z]: 0 0 0
> set description: stucture
> apply to image: select the 1-100:I0001.*img and press ¡°done¡±
> quit¡¯
> (2). Press ¡°Display¡±
> A: adjust the original point of structure image.
> Select ¡®1: structure.img¡¯ and press ¡®done¡¯
> Put the crosshair position on anterior-conjunction.
> Press: Reorient image
> B: adjust the original point of EPI image.
> Press ¡°Display¡± again,
> press I0001.*img and expand it, select I0001¡Á001.img, do as former.
>
> 2nd **************************spatial
> preprocessing****************************** The manipulation of these steps
> follows the guidance downloaded from website. (3). Press <REALIGN> in the
> upper left SPM window
> 'Number of subjects' - 1
> 'Num Sessions for subject 1' - 1
> Select all 1-100:I0001.*img and press 'Done'
> Pull down menu - select 'Coregister and Reslice'
> Pull down menu - select 'Sinc Interpolation'
> Pull down menu - select 'Create All Images + Mean image'
> 'Adjust sampling errors' - no
>
> (4). Press <NORMALIZE>
> Pull down menu - 'Determine Parameters and Write Normalised'
> Number of subjects - 1
> 'Image to determine parameters'
> - meanI0001¡Á001.img
> 'Images to write Normalised'
> - meanI0001¡Á001.img AND 2-101:rI0001¡Á*.img
> 'Template image' - EPI.img
> Pull down menu - 'Bilinear interpolation'
>
> (5). Press <DEFAULTS>
> Pull down menu - 'Spatial Normalization'
> Pull down menu - 'Defaults for Writing Normalized'
> Pull down menu - '-78:78 -112:76 -50:85 (Default)'
> Pull down menu - '1 1 1'
> Press <NORMALIZE>
> Pull down menu - 'Write Normalized Only'
> '# Subjects' - '1'
> 'subj 1 - Normalisation parameter set'
> - meanI0001¡Á001_sn3d.mat
> 'subj 1 - Images to write normalised'
> -01: structure.img
> Pull down menu - 'Bilinear Interpolation'
> Press <DEFAULTS>
> Pull down menu - 'Reset All'
> (6). Press <SEGMENT>
> 'number of subjects' - '1'
> 'Select MRI(s) for subject 1'
> - nstructure.img
> 'Are they spatially normalized?'
> - 'yes'
> Pull down menu
> - 'Lots of inhomogeneity correction'
> Pull down menu
> - 'Save inhomogeneity corrected images'
> (7). Press <Xtract Brain>
> 'Select gray and white matter images'
> - nstructure_seg1.img
> - nstructure_seg2.img
> Pull down menu
> - 'Save Extracted Brain and Rendering'
> (8). Press ¡°Smooth¡±
> smoothing {FWHM in mm}: 6
> select 1-100:nrI0001¡Á*.img
> Press done
>
> 3rd ****************Model specification and parameter
> estimation************* (9). Press ¡°fMRI models¡±
> Pull down menu: Would you like to... - "specify a model"
> Interscan interval - 3
> Scans per session - 100
> Number of conditions or trials - 1
> Condition or trial name - 'trail 1'
> Stochastic design - "no"
> SOA (Stimulus Onset Asynchrony) - "Fixed"
> SOA (scans) for active - 20
> First trial (scans) - 10
> Parametric modulation - "none"
> Are these trials - "epochs"
> Select type of response... (PullDown) - fixed response (Box-Car)
> Convolve with HRF - "yes"
> add temporal derivatives - "no"
> epoch length {scans} for active - 6
> interactions among trials (Volterra) - "no"
> user specified regressors - '0'
> Press "fMRI models"again
> Pull down menu: Would you like to... - "estimate a specified model"
> Select the fMRIDesMtx.mat from (1) and press "Done"
> Select the 100 snrI0001¡Á*.img
> Remove Global effects - "scale"
> High-pass filter? - "specify"
> Cutoff period - '120'
> Low-pass filter? - "hrf"
> Model intrinsic correlations - "none"
> Estimate - "now"
>
>
> 4th *************************Results
> assessement********************************* (10) Press "Results" in the
> SPM MenuWindow
> Select the SPM.mat that was produced in previous step and press 'Done'
> Type 'aaa' in the name area (at the top)
> Select "t-contrast"
> Type a '1' in the contrast are (in the middle)
> (The contrast is depicted above the design matrix)
> Press "OK"
> Select the newly defined 'activation' contrast & press "done"
> Mask with other contrasts - "no"
> Title for comparison - 'activation'
> Corrected height threshold - "yes"
> Threshold - '0.05'
> Extent threshold - '0'
> The MIP appears in the graphics window.
> Press "volume" in the "p-values" panel
> This displays a summary p-values for all local maxima
> Drag the red cursor on the MIP to the interesting cortex
> Press 'plot'
> PullDown menu: Plot... - "fitted and adjusted responses"
> PullDown menu: Which contrast? - "activation"
> PullDown menu: Plot against - "Scan or Time"
>
>
> Thank you very much for point out mistakes and correct it.

There is one area where things could be improved. For estimating spatial
 normalisation parameters by matching an EPI image to an EPI template, it is
 og=ften better to go into the defaults, and change the spatial normalisation
 settings such that the estimation step does not use any brain-masking.
 Before step (4), you would:

hit <Defaults>
select "Spatial Normalisation"
select "Defaults for Parameter Estimation"
Then turn off "Mask brain when registering?"

Also, you don't seem to have coregistered the anatomical and functional data.
  Therefore spatial normalisation params estimated from a functional image
 can not be applied to an anatomical image.

Another thing you may want to try is to write the original images spatially
 normalised. This involves specifying the I0001¡Á*.img files, rather than the
 rI0001¡Á*.img files. It is also better if you change the defaults first, so
 that sinc interpolation is used to resample the original images. Doing it
 this way means that there is only one resampling step, as the rigid-body
 transformations from the realign step are combined with the nonlinear warps
 from the spatial normalisation step.

I can not say too much about the statistical design, as this isn't my area of
 expertise.

% 'Mask brain when registering?'
% Applies a weighting mask to the template(s) during the parameter
% estimation. By default, weights in and around the brain have
% values of one whereas those clearly outside the brain have values
% of zero. This is an attempt to base the normalization purely upon
% the shape of the brain, rather than the shape of the head (since
% low frequency basis functions can not really cope with variations
% in skull thickness).
% The option is now available for a user specified weighting image.
% This should have the same dimensions and mat file as the template
% images, with values in the range of zero to one.

> I¡¯ll consult you for another problems:
> 1. How to reorient the original point of EPI images?
> The EPI image isn¡¯t like structure image, its resolution rate is too low
> to identify the A-C point. It impossible to set crosshair position on A-C
> point more accurately.

The AC does not need to be identified exactly. Providing it is within about
3cm, then registration, spatial normalisation etc, should work OK.

> 2. how to use ¡°coregister¡± ?
> It has not provided in the examples of your website. Only one time have I
> learn this usage in an article written by your institute. [Speed-Dependent
> Motion-Sensitive Responses in V5: An fMRI Study. D. Chawla,1 J. Phillips,
> C. Buechel, R. Edwards, and K. J. Friston. NEUROIMAGE 7, 86¨C96 (1998)].
> ¡°first volume (Friston et al., 1995a). A mean image was created using the
> realigned volumes. A structural MRI, acquired using a standard 3-D
> T1-weighted sequence (1 3 1 3 3 mm voxel size), was coregistered to this
> mean (T2*) image. This ensured that the functional and structural images
> were in the same space. Finally, the structural image was spatially
> normalized (Friston et al., 1995a) to a standard template (Evans et al.,
> 1993; Talairach and Tournoux, 1988), using a nonlinear transformation. This
> nonlinear deformation employs spatial basis functions as described in
> Friston et al. (1995). The transformation, mapping the structural T1 MRI
> scan onto the template, was applied to the fMRI data.¡±
> Can you demonstrate it in my test more detailly? Thanks!

Maybe you should do this before step (5):

hit <Coregister>
specify 'number of subjects' 1
specify 'Which option?': 'Coregister only'
specify 'Modality of first target image?' 'target - EPI'
specify 'Modality of first object image?' 'object - T1 MRI'
specify 'select target image for subject 1' meanI0001¡Á001.img
specify 'select object image for subject 1' structure.img
specify 'select other images for subject 1' none

> 3. how to process multi-subjects of block fMRI dataset analysis?
> Your website provides single-subject data of epoch (block) fMRI dataset,
> multi-subject of event-related dataset and advanced event-related fMRI
> example dataset. In my study, I want to coregister 9 subjects to one. Can
> you guide me how to carry it out? Thanks!

There are currently two possible ways. There is the fixed effects analysis,
 whereby you simply enter all the data into a single big design matrix
 (specifying the number of subjects, so that subject effects are modelled).

The alternative is the mixed or random effects way of doing the analysis.
 The principle here is that the con_*.img files from a number of single
 subject analyses are entered into a second statistical analysis.

Either method has its own advantages and disadvantages. I am not a
 statistician, so I can not say too much. If you join the mailing list (see
 http://www.fil.ion.ucl.ac.uk/spm/help/ ), then the statistical experts on
 the list would be more than happy to answer your queries.

Best regards,
-John

--
Dr John Ashburner.
Functional Imaging Lab., 12 Queen Square, London WC1N 3BG, UK.
tel: +44 (0)20 78337491 or +44 (0)20 78373611 x4381
fax: +44 (0)20 78131420 http://www.fil.ion.ucl.ac.uk/~john