[Insight-users] Re: Reg Registration

Thu, 01 May 2003 13:39:42 -0400

Hi Prasad,

1) Your first registration process, for same
    modality is consistently going up in the
    metric value.

    Questions:

    a) What image metric are you using ?
    b) Did you set the optimizer to MinimizeOn() ?

    Note that some metrics are to be maximized
    while others are to be minimized.
    This is described in the software guide

    http://www.itk.org/ItkSoftwareGuide.pdf

    Section 7.9 pdf-page 223

2) The example on Multimodality registration
    will typically use all the iterations.
    This is due to the high degree of noise
    in the mutual information metrics.

    You have to be cautious with the output results
    since they do not necessarily mean that the
    registration have succeed. They just mean that
    it ran out of iterations.

    I would suggest you to use the MattesMutual
    Information and the RegularStep gradient
    descent optimizer. Again starting with low
    values of the step size (around 0.1).

    Connect an observer to your registration
    process, so you can track the evolution
    of the transform.

    This combination is illustrated in the
    Software guide on section 7.4.2, pdf-page 185.
    Note the profile of the metric plots for this
    metric.

    You may want to plot the values you obtain
    for your registration process.

Regards

    Luis

--------------------------
Prasad Sarma wrote:
> I Luis,
>   
>   Thanks for your prior suggestions.   
> I have two questions reg my registraion,
> 1) for registration of same modalities
> I have  64X64X19 images of type unsigned char and as 
> you suggested before I have set the step max length 
> to 0.2 instead of 4.00 and min length to .01 in case 
> of optimizer as same as Registration3 ex. 
> But the optimizer values seem to be increasing 
> instead  of decreasing but after certain iterations 
> the program terminates .
> 
> =====================================
> register.exe my3Dimage_1.mha my3Dimage_3.mha
> diff_3.mha
> 
> 0 = 304.683 : [0.118135, 0.102378, 0.124751]
> 1 = 329.421 : [0.235509, 0.206601, 0.248691]
> 2 = 360.055 : [0.352232, 0.313168, 0.371243]
> 3 = 409.358 : [0.468207, 0.422398, 0.49215]
> 4 = 471.761 : [0.583296, 0.534679, 0.611094]
> 5 = 545.159 : [0.700481, 0.645193, 0.729644]
> 6 = 631.209 : [0.816897, 0.758638, 0.846168]
> 7 = 733.127 : [0.932424, 0.875486, 0.960186]
> 8 = 858.061 : [1.04688, 0.996389, 1.07101]
> 9 = 943.857 : [1.15614, 1.11613, 1.18816]
> 10 = 945.48 : [1.26266, 1.24148, 1.30192]
> 11 = 958.901 : [1.36535, 1.37405, 1.41091]
> 12 = 991.283 : [1.46232, 1.51625, 1.51278]
> 13 = 1037.01 : [1.54743, 1.63598, 1.6485]
> 14 = 1101.85 : [1.67249, 1.7403, 1.76459]
> 15 = 1182.01 : [1.79718, 1.84811, 1.87785]
> 16 = 1282.44 : [1.92141, 1.96026, 1.98736]
> 17 = 1410.81 : [2.04497, 2.07805, 2.09156]
> 18 = 1447.68 : [2.14697, 2.20453, 2.20818]
> 19 = 1422.29 : [2.24345, 2.34025, 2.31896]
> 20 = 1426.5 : [2.33081, 2.48914, 2.41995]
> 21 = 1450.95 : [2.40033, 2.65769, 2.50215]
> 22 = 1488.4 : [2.46535, 2.76159, 2.66019]
> 23 = 1550.99 : [2.52068, 2.87284, 2.81691]
> 24 = 1635.81 : [2.65507, 2.96376, 2.93384]
> 25 = 1735.76 : [2.79106, 3.05978, 3.04469]
> 26 = 1783.74 : [2.89485, 3.17768, 3.1685]
> 27 = 1798.86 : [2.9919, 3.30506, 3.28832]
> 28 = 1831.56 : [3.07633, 3.44863, 3.39903]
> 29 = 1875.3 : [3.1242, 3.62507, 3.48015]
> 30 = 1903.94 : [3.16103, 3.82066, 3.49977]
> 31 = 1937.61 : [3.18572, 4.01874, 3.51226]
> 32 = 1958.31 : [3.23925, 4.11143, 3.6812]
> 33 = 2001.88 : [3.27727, 4.2098, 3.85114]
> 34 = 2070.58 : [3.28074, 4.31793, 4.01935]
> 35 = 2106.42 : [3.32121, 4.41555, 4.18916]
> 36 = 2091.87 : [3.31204, 4.5188, 4.3602]
> 37 = 2101.07 : [3.3203, 4.71457, 4.40029]
> 38 = 2129.94 : [3.30707, 4.91304, 4.42109
> 39 = 2167.35 : [3.26877, 5.10934, 4.4221]
> 40 = 2151.61 : [3.28284, 5.30553, 4.45827]
> 41 = 2162.55 : [3.19958, 5.4842, 4.42442]
> 42 = 2170.67 : [3.26362, 5.43428, 4.48278]
> 43 = 2174.71 : [3.24258, 5.40198, 4.45093]
> 44 = 2166.72 : [3.26115, 5.39213, 4.46445]
> 45 = 2168.35 : [3.25323, 5.3886, 4.45545]
> 46 = 2166.57 : [3.2478, 5.39958, 4.45296]
> =====================================
> I don't know whether this is the problem with the 
> 
> image or with my parameter setting still.
>  
> 2)When I use the  multimodality registraion for an 
> image of 200X200X240(moving)T1 image dimension with a 
> 64X64X19 (fixed) EPI image using the Registraion2 ex:
> (both have the pixel spacing  1 1 1)
> I use the same parameters for the Standard Deviation
> (.4) for the fixed and Moving image parametes 
> optimizer learing rate as 20.00 and munber of 
> 
> iter=200.
> 
> For higher values of learing rate of optimizer I am 
> getting the following final o/p
> Result =
>  Translation X = 2.96702
>  Translation Y = 5.23111
>  Translation Z = 0.628723
>  Iterations    = 200
>  Metric value  = 0.0309063
> and 
> 
> I get  the  run time error when run this program (sth 
> like report to Microsoft :-) ) after I finish all 200 
> iterations
> 
> I am using VTK and FLTK and am integrating this 
> program with that. 
> 
> Instead If I use even learning rate at 1:00 I still
> get the run time error. I tried changing the Standrad
> deviation  values for the Gaussian Kernel and no of 
> iterations still iot was useless. 
> 
> The result of multimodality registration was as below
> for steparate 1.0
> 
> Result =
>  Translation X = 0.0979025
>  Translation Y = 0.211213
>  Translation Z = 0.0136685
>  Iterations    = 200
>  Metric value  = 0.071418
> 
> I am not able to find the reason.
> Could you pl tell me what is the reason for such 
> errors? 
> 
> 
> Thanks alot 
> Prasad
> 
> 
> 
> __________________________________
> Do you Yahoo!?
> The New Yahoo! Search - Faster. Easier. Bingo.
> http://search.yahoo.com
> 