1. Home
  2. fsl

fsl

 

 

 

What I want to do?
Perform epi-distorition correction over our dwi images using fugue.

Our dataset for this experiment can be downloaded from here (dcm ; converted data; data4fuguePartI; data4fuguePart2; data4fuguePart3 )  which consists of:

     - ax-ap  (SET1_acq1_AP_6)
     - ax-pa (SET1_acq1_PA_7)
     - fieldmap  (fieldmap_16x16x32_14)
     - magnitudes of multi echos (coilsens_multiecho_32)
     - phases of multi echos (coilsens_multiecho_33)

(P.S. the data in data4fugue folder is splited to 3 parts due to the limitation on uploading the files.)


To run fugue command line: 

fugue  -i  b0AP --icorr --unwarpdir=y --dwell=dwellTime  --loadfmap=FieldMap -u result

I set the dwell time (echo spacing) to:    Echo Spacing  (s) = 1 / [(0019,1028) * (0051,100b component #1)]
                                                                                      = 1/[ 5.5019999999999998 * 128 ]= 0.0014

I run fugue considering the following options to set FieldMap (FM):


OPTION1:  the FM in hand (fieldmap_16x16x32_14)

fm.nii

Intensity properties:
Minimum -4096.000000000000000 Maximum 4094.000000000000000 Mean -47.703113902698867 Variance 2603513.713320951443166 Sum -42986230.000000000000000
Geometric properties:
Size: [128, 128, 55]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200004577636719]
Origin: [-92.558097839355469, -76.970733642578125, -57.719863891601562]
 

FM_resample_B0AP_brain.nii.gz in our dataset represents the resampled and brain-masked FM image (the above). Its values are between [-4096 4096]. If I feed this image directly to fugue:

fugue -i ../SET1_acq1_AP_6/b0AP.nii  --icorr  --unwarpdir=y --dwell=0.0014  --loadfmap=./FM_resample_B0AP_brain.nii.gz    -u ../out_fugue_FM_resample_B0AP_brain

The output resutl (out_fugue_FM_resample_B0AP_brain) is not correct.

out_fugue_FM_resample_B0AP_brain.nii

b0AP.nii

That means that we need to map the values of our FM image to proper unit. Would you please guide me what is the proper mapping to do so?

 

 

 

PTION2: compute FM using  fsl_prepare_filedmap

In our dataset, we have the manitude  (mag1.nii mag2.nii mag3.nii mag4.nii) and phase  (pha1.nii pha2.nii pha3.nii pha4.nii) images of 4 echos.

============ mag1.nii ============Intensity properties:
Minimum 0.000000000000000 Maximum 1084.000000000000000 Mean 83.806185404459640 Variance 16717.556370987054834 Sum 65907866.000000000000000
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593799948692322, 1.593799948692322, 3.200000047683716]
 

============ pha1.nii ============Intensity properties:
Minimum -4096.000000000000000 Maximum 4094.000000000000000 Mean -154.429211934407562 Variance 5765004.457419949583709 Sum -121448074.000000000000000
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593799948692322, 1.593799948692322, 3.200000047683716]
 

============ pha2.nii ============Intensity properties:
Minimum -4096.000000000000000 Maximum 4094.000000000000000 Mean -124.518488566080734 Variance 5395802.720908988267183 Sum -97925324.000000000000000
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593799948692322, 1.593799948692322, 3.200000047683716]

============ pha3.nii ============
Intensity properties:
Minimum -4096.000000000000000 Maximum 4094.000000000000000 Mean 26.398208618164062 Variance 5284832.440826209262013 Sum 20760396.000000000000000
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200000762939453]
Origin: [-92.558097839355469, -76.970733642578125, -47.300003051757812]

============ pha4.nii ============
Intensity properties:
Minimum -4096.000000000000000 Maximum 4094.000000000000000 Mean -15.623992919921875 Variance 5366888.233273105695844 Sum -12287208.000000000000000
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200000762939453]
Origin: [-92.558097839355469, -76.970733642578125, -47.300003051757812]

 

The values of our phase images are between [-4096 +4096]. I assume that these values are corresponding to [-\(\pi\) \(+\pi\)]. To make the range of the phase images be between [0 2\(\pi\)], I did the following:

   - step1: map the values to [-\(\pi\) \(+\pi\)] simply by rescaling as follows:

           fslmaths pha1.nii   -div 4096 -mul 3.1416 pha1_range_NegPi_PosPi.nii -odt float

           fslmaths pha2.nii   -div 4096 -mul 3.1416 pha2_range_NegPi_PosPi.nii -odt float

    - step2: add 2\(\pi\) to voxels with negative values.

The final images following the above steps are shown here:

out prelude:

============ pha1_0To2Pi.nii.gz ============
Intensity properties:
Minimum 0.000000000000000 Maximum 6.281665802001953 Mean 3.088931463426006 Variance 3.305485068272457 Sum 2429234.548645040951669
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200000762939453]
Origin: [-92.558097839355469, -76.970733642578125, -47.300003051757812]

 

============ pha2_0To2Pi.nii.gz ============
Intensity properties:
Minimum 0.000000000000000 Maximum 6.281665802001953 Mean 3.114725006161946 Variance 3.515247404286431 Sum 2449519.416045951191336
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200000762939453]
Origin: [-92.558097839355469, -76.970733642578125, -47.300003051757812]

 

============ pha3_0To2Pi.nii.gz ============
Intensity properties:
Minimum 0.000000000000000 Maximum 6.281665802001953 Mean 2.999124929375997 Variance 3.569161460006509 Sum 2358607.816459024325013
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200000762939453]
Origin: [-92.558097839355469, -76.970733642578125, -47.300003051757812]

 

============ pha4_0To2Pi.nii.gz ============
Intensity properties:
Minimum 0.000000000000000 Maximum 6.281665802001953 Mean 3.040485530163449 Variance 3.527428407162822 Sum 2391135.116457501426339
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200000762939453]
Origin: [-92.558097839355469, -76.970733642578125, -47.300003051757812]

 

Then, phase difference of the first two echos is computed as:

fslmaths pha2_0To2Pi.nii.gz  -sub     pha1_0To2Pi.nii.gz   sub_pha2_pha1_0To2Pi.nii -odt float

 

============ sub_pha2_pha1_0To2Pi.nii.gz ============
Intensity properties:
Minimum -6.270927906036377 Maximum 6.281665802001953 Mean 0.025793543054518 Variance 6.347289289345887 Sum 20284.867651450913399
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200000762939453]
Origin: [-92.558097839355469, -76.970733642578125, -47.300003051757812]

 

since fsl_prepare_fieldmap is expecting at least 90% of 0 to 4096; I follow the same steps mentioned in the above (adding 2\(\pi\) to negative values and then rescaling).

And the following is the output of fsl_pep:

 

fsl_prepare_fieldmap SIEMENS  sub_pha2_pha1_0To2Pi_cor4NegValues.nii  mag1_brain.nii.gz  out_fslprep.nii.gz 2.63

 

============ ../out_fslprep.nii.gz ============
Intensity properties:
Minimum -3484.412597656250000 Maximum 2522.610839843750000 Mean -28.863866216069862 Variance 82732.142100835044403 Sum -22699468.036036252975464
Geometric properties:
Size: [128, 128, 48]
StartIndex: [0, 0, 0]
Spacing: [1.593750000000000, 1.593750000000000, 3.200000762939453]
Origin: [-92.558097839355469, -76.970733642578125, -47.300003051757812]

and feed the computed FM (res_fsl_prep.nii.gz) to

fugue -i ./data4fugue/b0Aap.nii.gz --dwell=0.0014 --loadfmap=./data4fugue/res_fsl_prep.nii.gz -u  ./data4fugue/resFugue_option2_ypos --unwarpdir=y

 

============ ../out_fugue_fslPrep.nii.gz ============
Intensity properties:
Minimum 0.000000000000000 Maximum 4095.000000000000000 Mean 182.933863771004752 Variance 142380.295101588475518 Sum 1198875369.609656810760498
Geometric properties:
Size: [256, 256, 100]
StartIndex: [0, 0, 0]
Spacing: [0.796875000000000, 0.796875000000000, 1.599998474121094]
Origin: [-92.558097839355469, -76.970733642578125, -50.519859313964844]

 

 

OPTION3: compute FM using two gradient echos; The phase images need to be in RAD; so,              

Given the computed phases in rad (pha1_0To2Pi.nii.gz  and pha2_0To2Pi.nii.gz), I run prelude to unwarp them:

prelude -a mag1_brain.nii  -p pha1_0To2Pi.nii.gz    -o  pha1_rad_unwrapped_rad.nii

Outputs of the prelude cmd are shown in the following:

 

 

and then, get the FM in rad/s:

fslmaths ./data4fugue/pha2_resampledOnb0Aap__unwrapped_rad.nii.gz -sub ./data4fugue/pha1_resampledOnb0Aap__unwrapped_rad.nii.gz -mul 1000 -div 2.67 ./data4fugue/res_fieldmap_rads_option3 -odt float

============ ./data4fugue/res_fieldmap_rads_option3.nii.gz ============
Intensity properties:
Minimum -6612.202636718750000 Maximum 4763.383789062500000 Mean -25.735924166325969 Variance 486072.281850565457717 Sum -168662952.616433858871460
Geometric properties:
Size: [256, 256, 100]
StartIndex: [0, 0, 0]
Spacing: [0.796875000000000, 0.796875000000000, 1.599998474121094]

and then, run fugue:

fugue -i ./data4fugue/b0Aap.nii.gz --dwell=0.0014 --loadfmap=./data4fugue/res_fieldmap_rads_option3.nii.gz -u  ./data4fugue/resFugue_option3_ypos --unwarpdir=y

/resFugue_option3_ypo (which is not correct)