[Insight-users] Problem with BSpline Registration with Mattes-MI

Serena Fabbri fabbri at u.washington.edu
Thu Jul 9 15:12:59 EDT 2009


Hi Luis, hi All,

I used DeformableRegistration15 and I got a good result.

In order to understand the problem in my code I did some trials:

1)
fixed=MRI, moving=BSpline (grid 6x6x6) registered image 
Bspline grid=13x13x13
I have got the same result already described.

2)
I have changed the optimizer in my code. I have put itkRegularStepGradientDescentOptimizer with the 
same parameters set in DeformableRegistration15.
fixed=MRI, moving=BSpline (grid 6x6x6) registered image
Bspline grid=13x13x13
I have got good result.
(I have checked several slices and i have seen pretty good overlap between fixed and registered image, 
indeed nose structures appear in the correct place).

so I have a question: could  LBFGSB be the reason about the bad result that i got ?
is LBFGSB able to manage only 'few' parameters?

now I have another question not correlated with the previous things.
I'd like to use Normalized Mutual Information and Bspline Transformation, but i get this error 
message:

terminate called after throwing an instance of 'itk::ExceptionObject'
   what():  /Users/physics/InsightToolkit-3.10.0/Code/Algorithms/itkImageToImageMetric.txx:91:
itk::ERROR: NormalizedMutualInformationHistogramImageToImageMetric(0x6690c0): Transform is not 
present
Abort trap

of course in the code there are: 
TransformType::Pointer	  transform		= TransformType::New();
registration->SetTransform( transform );

where is the error? any suggestion?

Tx a lot!

Serena.





On Tue, 7 Jul 2009, Luis Ibanez wrote:

>
>
> Hi Serena,
>
> Please try running the example:
>
>
>   Insight/Examples/Registration/
>             DeformableRegistration15.cxx
>
>
> with your input images,
>
>
> and let us know if the results are similar or
> better than what you get with your current
> code.
>
>
>   Thanks
>
>
>       Luis
>
>
> -----------------------
> Serena Fabbri wrote:
>> Hi All,
>> 
>> I am registering brain MRI and CT image with BSpline (Non Rigid 
>> Registration), Mattes MI and LBFGSB optimization.
>> 
>> fixed=MRI   (176,256,160); (1,1,1)mm; (0,0,0)mm
>> moving=CT (176,256,160); (1,1,1)mm; (0,0,0)mm
>> 
>> CT image is been previously Rigid registered to MRI
>> 
>> I have started with 6x6x6 grid (3x3x3 nodes on the fixed image), MI sample 
>> =20% statistics and the registered image is pretty good.
>> In order to improve the result I have used a fine grid: 13x13x13 (10x10x10 
>> nodes on the fixed image), MI sample =20% statistics, and the result is:
>> in general the registration is good because some details are better 
>> overlapped, but  the slices containing the nose structures are distorted. In 
>> particular nasal septum is deviated in the left side and the other bone 
>> appears too much smoothed.
>> I have read some previous e-mails 
>> (http://www.itk.org/pipermail/insight-users/2008-
>> December/028417.html) in archives and an Ibanez advice was to check if there 
>> are enough samples for population every cell of the BSpline grid. I think 
>> that it is enough samples because i get the same result with 50% samples.
>> 
>> I have tried to set different number of nodes in x and y direction. I have 
>> used 6x13x13 grid and 20% sample too, and i get exactly the same registered 
>> image previous described (Below I put the Iteration-
>> MattesMI value ).
>> 
>> 
>> can anyone give me some advice?
>> maybe it is necessary to set different number of nodes in x and y direction. 
>> Could it be a reasonable explanation?
>> 
>> any suggestion will be appreciate.
>> 
>> Thanks.
>> 
>> Serena.
>> 
>> 
>> Iter  MattesMI
>> 0   -0.350671   0   -0.351995   0   -0.357607   0   -0.377623   0 
>> -0.402784   0   -0.402784   1   -0.435035   1   -0.435035   2   -0.430811 
>> 2   -0.466589   2   -0.466589   3   -0.491583   3   -0.491583   4 
>> -0.499865   4   -0.499865   5   -0.53353   5   -0.53353   6   -0.533558   6 
>> -0.543796   6   -0.543796   7   -0.559133   7   -0.559133   8   -0.567699 
>> 8   -0.567699   9   -0.577058   9   -0.577058   10   -0.586574   10 
>> -0.586574   11   -0.59515   11   -0.59515   12   -0.600187   12   -0.600187 
>> 13   -0.608752   13   -0.608752   14   -0.614033   14   -0.614033   15 
>> -0.619695   15   -0.619695   16   -0.623183   16   -0.623183   17 
>> -0.627047   17   -0.627047   18   -0.630555   18   -0.630555   19 
>> -0.634089   19   -0.634089   20   -0.638822   20   -0.638822   21 
>> -0.642822   21   -0.642822   22   -0.648013   22   -0.648013   23 
>> -0.651278   23   -0.651278   24   -0.65333   24   -0.65333   25   -0.656569 
>> 25   -0.656569   26   -0.659193   26   -0.659193   27   -0.661128   27 
>> -0.661128   28   -0.663331   28   -0.663331   29   -0.66513   29   -0.66513 
>> 30   -0.666392   30   -0.666392   31   -0.667716   31   -0.667716   32 
>> -0.668881   32   -0.668881   33   -0.669772   33   -0.669772   34 
>> -0.671548   34   -0.671548   35   -0.672711   35   -0.672711   36 
>> -0.674103   36   -0.674103   37   -0.674998   37   -0.674998   38 
>> -0.676283   38   -0.676283   39   -0.677051   39   -0.677051   40 
>> -0.678061   40   -0.678061   41   -0.678757   41   -0.678757   42   -0.67975 
>> 42   -0.67975   43   -0.680577   43   -0.680577   44   -0.681572   44 
>> -0.681572   45   -0.682606   45   -0.682606   46   -0.683061   46 
>> -0.683061   47   -0.683633   47   -0.683633   48   -0.684106   48 
>> -0.684106   49   -0.684759   49   -0.684759   50   -0.685085   50 
>> -0.685085   51   -0.685878   51   -0.685878   52   -0.686366   52 
>> -0.686366   53   -0.686903   53   -0.686903   54   -0.687805   54 
>> -0.687805   55   -0.688129   55   -0.688129   56   -0.688648   56 
>> -0.688648   57   -0.689047   57   -0.689047   58   -0.689372   58 
>> -0.689372   59   -0.690139   59   -0.690139   60   -0.690221   60 
>> -0.690221   61   -0.690577   61   -0.690577   62   -0.69085   62   -0.69085 
>> 63   -0.691391   63   -0.691391   64   -0.691557   64   -0.691557   65 
>> -0.691489   65   -0.691692   65   -0.691692   66   -0.691924   66 
>> -0.691924   67   -0.692416   67   -0.692416   68   -0.692831   68 
>> -0.692831   69   -0.692933   69   -0.692933   70   -0.693209   70 
>> -0.693209   71   -0.693661   71   -0.693661   72   -0.69374   72   -0.69374 
>> 73   -0.694059   73   -0.694059   74   -0.694317   74   -0.694317   75 
>> -0.694812   75   -0.694812   76   -0.694946   76   -0.694946   77 
>> -0.695411   77   -0.695411   78   -0.695788   78   -0.695788   79 
>> -0.695936   79   -0.695936   80   -0.69627   80   -0.69627   81   -0.696409 
>> 81   -0.696409   82   -0.696002   82   -0.696418   82   -0.696418   83 
>> -0.696533   83   -0.696533   84   -0.697105   84   -0.697105   85 
>> -0.697287   85   -0.697287   86   -0.697609   86   -0.697609   87 
>> -0.697957   87   -0.697957   88   -0.697482   88   -0.697993   88 
>> -0.697993   89   -0.698045   89   -0.698045   90   -0.698389   90 
>> -0.698389   91   -0.699091   91   -0.699091   92   -0.699072   92 
>> -0.699127   92   -0.699127   93   -0.699169   93   -0.699169   94 
>> -0.699422   94   -0.699422   95   -0.699545   95   -0.699545   96 
>> -0.700181   96   -0.700181   97   -0.70052   97   -0.70052   98   -0.700755 
>> 98   -0.700755   99   -0.700869   99   -0.700869   100   -0.699019   100 
>> -0.700874   100   -0.700874   101   -0.697901   101   -0.700846   101 
>> -0.700874   101   -0.700875   101   -0.700875   101   -0.700874   101 
>> -0.700875   101   -0.700875   101   -0.700875   101   -0.700875   101 
>> -0.700875   102   -0.682951   102   -0.700818   102   -0.700878   102 
>> -0.700878   103   -0.700522   103   -0.700962   103   -0.700962   104 
>> -0.700956   104   -0.701143   104   -0.701143   105   -0.700847   105 
>> -0.701264   105   -0.701264   106   -0.699357   106   -0.701293   106 
>> -0.701293   107   -0.700802   107   -0.701318   107   -0.701318   108 
>> -0.701101   108   -0.701454   108   -0.701454   109   -0.701117   109 
>> -0.701525   109   -0.701525   110   -0.700914   110   -0.701622   110 
>> -0.701622   111   -0.701429   111   -0.701679   111   -0.701679   112 
>> -0.702049   112   -0.702049   113   -0.702067   113   -0.702067   114 
>> -0.702424   114   -0.702424   115   -0.702911   115   -0.702911   116 
>> -0.702993   116   -0.702993   117   -0.703354   117   -0.703354   118 
>> -0.70348   118   -0.70348   119   -0.703623   119   -0.703623   120 
>> -0.703758   120   -0.703758   121   -0.703861   121   -0.703861   122 
>> -0.703992   122   -0.703992   123   -0.704031   123   -0.704031   124 
>> -0.704308   124   -0.704308   125   -0.704431   125   -0.704431   126 
>> -0.704565   126   -0.704565   127   -0.704793   127   -0.704793   128 
>> -0.704626   128   -0.704884   128   -0.704884   129   -0.7044   129 
>> -0.704907   129   -0.704907   130   -0.705365   130   -0.705365   131 
>> -0.705579   131   -0.705579   132   -0.705919   132   -0.705919   133 
>> -0.706196   133   -0.706196   134   -0.706419   134   -0.706419   135 
>> -0.706565   135   -0.706565   136   -0.706604   136   -0.706604   137 
>> -0.706731   137   -0.706731   138   -0.706708   138   -0.706768   138 
>> -0.706768   139   -0.706699   139   -0.70692   139   -0.70692   140 
>> -0.706833   140   -0.706975   140   -0.706975   141   -0.706896   141 
>> -0.707044   141   -0.707044   142   -0.706378   142   -0.70706   142 
>> -0.70706   143   -0.705602   143   -0.707069   143   -0.706376   143 
>> -0.707055   143   -0.707068   143   -0.707069   143   -0.707069   143 
>> -0.707069   144   -0.0399776   144   -0.677537   144   -0.706919   144 
>> -0.707062   144   -0.707068   144   -0.707069   144   -0.707069   144 
>> -0.707069   144   -0.707069   144   -0.707069   144   -0.707069   144 
>> -0.707069   144   -0.707069   145   -0.614007
>> 
>> 
>> 
>> 
>> 
>> 
>> 
>> _____________________________________
>> 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
>> 
>





More information about the Insight-users mailing list