---




Displaying the Results

Dealing with dynamic range problem in Siemens 3D anatomic scans

3D T1-weighted anatomic scans acquired on the Siemens scanner tend to suffer from a problem with their dynamic range. This occurs because artifacts inherent to the MRI process appear as a cluster of extremely bright voxels at the edge of the field of view. These generally lie outside the head, and do not obscure the anatomy. However, they are often five times as bright as brain tissue voxels, and cause some visualization programs, such as register, to allocate too few gray levels to the range of intensities arising within the brain. You can get around this by using `mincmath' to clamp the intensity levels in the volume. After looking at the image, you should be able to spot the artifact and determine it's intensity.  The constant value that you can use to clamp the intensity, should be chosen in such a way that the bright spot is cut out, but the values inside the brain are not clamped. For example if you chose your constant value to be 300
( this is just an example, each image has to be checked separately) you will say:

mincmath -clamp -const2 0 300 smith_john_19971024_1_2_mri.mnc.gz \
        smith_john_19971024_1_2_mri_clamped.mnc
 

Initial inspection of the activation image

Assuming that your experiment consisted in 120 time frames and the standard gradient-echo EPI protocol was used. To take a look at the activation maps that you just obtained, it is suggested that you follow the next  steps:
Load the anatomic and activation images in `register,' e.g.

register smith_john_19971024_1_105235_mri_MC_ tstat.mnc smith_john_19971024_1_2_mri.mnc.gz &

or, if you would like to look at the images in Talairach space, you can load in register the tmaps and the anatomical file, after you resampled both in Talairach space, using
 the corresponding transformation.

set register to `synced'

make sure that activation map is shown in `spectral' colormap

set upper window level for activation map to +10

set lower window level for activation map to -10


This will result in regions of zero activation appearing green, with positive, significant, responses showing up red to white. Motion artifacts or spurious foci will generally be easy to spot on maps displayed in this way. It is worth making a habit of looking at all images in this standard way. Note that the window levels given here are specific to 120 frame acquisitions.

You can then proceed to apply the threshold computed in Matlab as described above.


---

Summary of analysis steps (example only for one subject, one dynamic run)

mkdir /data/fmri/analysis/yourname

mkdir /data/fmri/analysis/yourname/john_smith

mkdir /data/fmri/analysis/yourname/john_smith/dynamic

cd /data/fmri/analysis/yourname/john_smith

cp ~fmrixfer/images/smith_john_19971024_1_2_mri.mnc.gz .

cp ~fmrixfer/images/smith_john_19971024_1_??????_mri.mnc.gz ./dynamic

cd /data/fmri/analysis/yourname/john_smith/dynamic

fmr_preprocess  -fwhm 6 -target 3 smith_john_19971024_1_105235_mri.mnc.gz

start Matlab

df = fmristat(INPUT_FILE , FRAME_TIMES , SLICE_TIMES , EVENT_TIMES , S,[CONTRAST , EXCLUDE , OUTPUT_FILE_BASE , NUM_STATS , MEAN_LAG , 
SD_LAG , FWHM_RHO, N_POLY )

input_file = 'smith_john_19971024_1_105235_mri_MC.mnc';

frametimes =  [0     3     6     9    12    15    18    21    24    27    30    33 ... 351   354   357];

slicetimes = [ 0.140  0.240  0.340  0.440 .... 2.540 ];

eventimes = [

     9      27
    45     63
    81     99
  117    135
  153    171
  189    207
  225    243
  261    279
  297    315
  333    351];  for event related design

S = [ 

           0     0 
           0     0 
           0     0 
           1     0 
           1     0 
           1     0 
           0     0 
           0     0 
           0     0 
           0     1 
           0     1 
           0     1 
           0     0 
           0     0 
           0     0 
           1     0 
           1     0 
           1     0 
            ......... 
           ];    for block design

 contrast = [1 -1];

 exclude = [1 2 3];

 output_file_base = 'smith_john_19971024_1_105235_mri_MC';

 num_stats = 4;

 MEAN_LAG = 6;

 SD_LAG = 3;

 FWHM_RHO = 15;

 n_poly = 3;

thresh = tstat_threshold( SEARCH_VOLUME , FWHM , VOXEL_VOLUME , DF , SIGNIFICANCE )

search_volume = 1200000;

fwhm = 6;

voxel_volume = 38.4521;

df = 113;

significance = 0.05;

thresh = 4.9

quit Matlab

cd /data/fmri/analysis/yourname/

register smith_john_19971024_1_105235_mri_MC_ tstat.mnc smith_john_19971024_1_2_mri.mnc.gz &

E-mail:

<vali@bic.mni.mcgill.ca>

Office phone:

(514) 398-8547

Office fax:

(514) 398-2975

---

Valentina PETRE <vali@bic.mni.mcgill.ca>

created: 20 July 1999
last modified: 17 August 1999

Previous		Back to table of contexts