AimsCorrectionEpi

[clenglet]

Hi,

I'm running BrainVisa 2.3.0 on a Linux system (RH9).
When I try to perform EPI distortions correction on DW Images with the command:

AimsCorrectionEpi
-t2 $DATA/subjectx_t2_diffusion.ima
-dw $DATA/subjectx_raw_dw_diffusion.ima
-o $DATA/subjectx_EPIcorrdw_diffusion.ima -verbose 1

I get the output below and the process does not stop (I let it running several hours!).
Am I missing anything ? Is there some important parameters I should try to adjust ?

Thank you very much for your help.
-Christophe

reading T2 ($DATA/subjectx_t2_diffusion.ima) and DW ($DATA/subjectx_raw_dw_diffusion.ima) data : check consistency of T2 and DW images :
check that images are squared :
check that images have same x and y resolutions :
processing echoplanar distortion correction : ---------------------------------------------------------
S0=1.00787 T0=0.0124595 T1=1.73681e-05 Cost=-0.0918971 [32771]
0 0 0.01
0 1 0
4.43845e-07 7.20238e-07 -3.12323e-11

---------------------------------------------------------
S0=1.00786 T0=0.00224415 T1=-8.60203e-08 Cost=-0.0890827 [16386]
0 0 0.01
0 1 0
-1.53631e-07 -4.38457e-08 1.68067e-12

---------------------------------------------------------
S0=1.00799 T0=0.0244311 T1=-6.85667e-05 Cost=-0.0911285 [32771]
0 0 0.01
0 1 0
-0.00336907 1.02243e-10 2.92854e-05

---------------------------------------------------------
S0=1.01164 T0=-0.0123958 T1=4.53628e-08 Cost=-0.0853254 [16386]
0.05 0 0
0 1 0
0 0 0.01

---------------------------------------------------------
S0=1.00781 T0=-5.94438e-05 T1=-5.84405e-08 Cost=-0.0786784 [16386]
0.05 0 0
0 1 0
0 0 0.01

---------------------------------------------------------
S0=1.00787 T0=0.0722861 T1=0.000257366 Cost=-0.0959818 [32771]
0.05 0 0
0 1 0
0 0 0.01

......

[Yann Cointepas]

There was a bug in AimsCorrectionEpi in thread management about one year ago. The command used to hang from times to times. Since 2.3.0 package was released on december 18th 2003, I suggest that you download a more recent package. To date the latest stable version is 2.3.2.

[clenglet]

Thanks for the reply.
I moved to version 2.3.2 but the problem stays unsolved.

I don't know exactly how to interpret the output of AimsCorrectionEpi (see below) but it looks to me that the algorithm oscillates around some solution since the Cost (the objective function to be optimized I suppose ?) alternatively takes the values ~ -0.08 and ~ -0.17.

Is that possible ? I reproduced this behavior on 2 different sets of DWI.

-Christophe

---------------------------------------------------------
S0=1.05219 T0=0.0421709 T1=6.06864e-05 Cost=-0.0847086 [131075]
0 0 0.01
0 1 0
-8.87155e-07 -7.16845e-07 -1.03158e-09

---------------------------------------------------------
S0=1.02124 T0=4.80336e-05 T1=-4.6592e-05 Cost=-0.163373 [131075]
0 0 0.01
0 1 0
3.81222e-12 8.61981e-15 -8.36116e-15

---------------------------------------------------------
S0=1.06122 T0=-2.69358e-05 T1=4.48536e-05 Cost=-0.079542 [131075]
0 0 0.01
0 1 0
-1.75042e-06 -5.63352e-15 -1.2825e-09

---------------------------------------------------------
S0=1.02445 T0=0.0309543 T1=0.00108053 Cost=-0.175908 [131075]
0 0 0.01
0 1 0
-0.000137237 -0.000172012 -6.31932e-06

---------------------------------------------------------
S0=1.02034 T0=0.000662239 T1=-2.22631e-05 Cost=-0.0801666 [131075]
0 0 0.01
0 1 0
-0.028995 -8.8921e-11 -2.25676e-06

---------------------------------------------------------
S0=1.02102 T0=-0.00660788 T1=-1.97438e-05 Cost=-0.171831 [131075]
0 0 0.01
0 1 0
4.1979e-06 -1.45966e-05 -2.32822e-09

---------------------------------------------------------
S0=1.05711 T0=0.00427238 T1=0.00432758 Cost=-0.0996268 [131075]
0 0 0.01
0 1 0
-0.000255175 -1.72333e-05 -1.9328e-05

---------------------------------------------------------
S0=1.02292 T0=0.0140241 T1=0.00390879 Cost=-0.175223 [131075]
0 0 0.01
0 1 0
0.000171771 0.000105111 2.96787e-05

[Jean-Francois Mangin]

Yes your output is very strange...
But nothing like that never happened before, so we have no idea yet.
Is it the case for the first slice of the volume in both cases?
Cyril Poupon made the guess it could happen with black slices...
Otherwise, did you have a look to the slice leading to the problem?
Could you provide us with one of this slice?