[Insight-users] MRIBiasCorrection

[Luis Ibanez]

Hi Bryn,


1) It is normal for the coefficients to be reported with slightly
    different values each time that the tool is run. This, as you
    pointed out, is due to the internal use of the OnePlusOne
    evolutionary optimizer.


2) Please note that "how big" the changes of the coefficients are,
    is something that must be interpretend in the context of the
    polynomials that these coefficients represent.


For the record please find below the results that I get in my local machine.


When using the following command:

 ./BiasFieldEstimator --input ~/data/BrainWeb/brainweb165a10f17.mha
--class-mean 80  130 --class-sigma 10 10 --use-log yes --input-mask
brainMask.mha --degree 3 --grow 1.01 --shrink 0.95 --max-iteration 20000
--init-step-size 1.0 --schedule 2 2 2

I get the following output:


Loading images...
Images loaded.
Setting multires schedule :1
Estimating the bias field...
--input /home/ibanez/data/BrainWeb/brainweb165a10f17.mha --input-mask
brainMask.mha --class-mean 80 130 --class-sigma 10 10 --use-log yes
--degree 3 --grow 1.01 --shrink 0.95 --volume-max-iteration 20
--inter-slice-max-iteration 4000 --init-step-size 1 --coefficients
0.00324493 0.00285925 0.0112537 -0.00752933 -0.00259406 0.0103765 0.0118277
0.00271833 -0.0157952 0.00323054 -0.0111964 0.0136344 0.0137674 -0.0195429
0.0252928 0.00874817 -0.00472533 0.0132882 0.00348112 -0.00622985

The brainMask.mha was generated by binary thresholding the input image
brainweb165a10f17.mha at a level of 80.


I can then, take the resulting coefficients as input for the ImageCorrector,
with the following command:

./ImageCorrector --input /home/ibanez/data/BrainWeb/brainweb165a10f17.mha
--output correctedImage.mha --degree 3 --use-log yes --coefficients
0.00324493 0.00285925 0.0112537 -0.00752933 -0.00259406 0.0103765 0.0118277
0.00271833 -0.0157952 0.00323054 -0.0111964 0.0136344 0.0137674 -0.0195429
0.0252928 0.00874817 -0.00472533 0.0132882 0.00348112 -0.00622985

This produces a corrected image.


You are correct in that, if I run the estimator multiple time, it will
produce
slightly different results. Again the correctness of each one of the
individual
results must be interpreted in terms of the polynomials that will be used
for
the correction.

See below, subsequence runs of the exact same estimation command:


Loading images...
Images loaded.
Setting multires schedule :1
Estimating the bias field...
--input /home/ibanez/data/BrainWeb/brainweb165a10f17.mha --input-mask
brainMask.mha --class-mean 80 130 --class-sigma 10 10 --use-log yes
--degree 3 --grow 1.01 --shrink 0.95 --volume-max-iteration 20
--inter-slice-max-iteration 4000 --init-step-size 1 --coefficients 0.0134269
0.0202263 0.00169249 0.00282789 0.00010718 0.00300399 -0.00800395
-0.00404422 0.00371428 0.00299892 0.0187534 -0.00332432 0.0125541 0.0121211
0.00732879 0.0076021 0.00305025 -0.00499408 -0.0234599 0.00227258

Loading images...
Images loaded.
Setting multires schedule :1
Estimating the bias field...
--input /home/ibanez/data/BrainWeb/brainweb165a10f17.mha --input-mask
brainMask.mha --class-mean 80 130 --class-sigma 10 10 --use-log yes
--degree 3 --grow 1.01 --shrink 0.95 --volume-max-iteration 20
--inter-slice-max-iteration 4000 --init-step-size 1 --coefficients
0.00798128 0.0262933 0.00123127 -0.00268059 -0.00844952 0.0081207 0.0041786
-0.00708119 0.00602885 0.0119248 -0.00464199 0.00428011 0.0180249 -0.0159507
0.0100633 0.00221582 0.00835673 0.014121 -0.00110316 -0.00652514

Loading images...
Images loaded.
Setting multires schedule :1
Estimating the bias field...
--input /home/ibanez/data/BrainWeb/brainweb165a10f17.mha --input-mask
brainMask.mha --class-mean 80 130 --class-sigma 10 10 --use-log yes
--degree 3 --grow 1.01 --shrink 0.95 --volume-max-iteration 20
--inter-slice-max-iteration 4000 --init-step-size 1 --coefficients
-0.0296475 0.00381279 -0.0104907 0.0032725 0.0110771 0.029399 0.020405
0.00330157 0.0034649 -0.00315351 0.00130834 -0.00658817 0.0145596 0.0191295
0.00672149 0.0054705 -0.00427434 0.00168028 -0.0180597 0.00478804



      Regards,


              Luis



----------------------------------------------
On Tue, Feb 10, 2009 at 12:21 PM, Bryn Lloyd <blloyd at vision.ee.ethz.ch>wrote:

> Hi
>
> I tried to use the MRIBiasCorrection tools in the ITK Application folder.
> However, I am not really sure about their usage.
>
> As explained in the README file, I call BiasFieldEstimator in order to
> estimate the coefficients. The strange this is that I get (very) different
> coefficients each time I run this tool.
> Is this because of the Evolutionary algorithm?
>
> Thanks
> Bryn
>
>
>
>
>
>
>
>
> _____________________________________
> Powered by www.kitware.com
>
> Visit other Kitware open-source projects at
> http://www.kitware.com/opensource/opensource.html
>
> Please keep messages on-topic and check the ITK FAQ at:
> http://www.itk.org/Wiki/ITK_FAQ
>
> Follow this link to subscribe/unsubscribe:
> http://www.itk.org/mailman/listinfo/insight-users
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://www.itk.org/pipermail/insight-users/attachments/20090215/5b4f9dad/attachment.htm>