[Insight-users] Re: Mail on Registration.
Luis Ibanez
luis.ibanez at kitware.com
Fri, 09 Apr 2004 08:28:15 -0400
Hi Srivalli
If the top levels of the pyramid are taking
all the alloted iterations that means that
they are not converging, and therefore there
is no point in continuing with the lower levels
(Those with higher resolution).
You *must* tune parameters by solving one level
at a time.
Here are some suggestions:
1) Replace the transform with the
VersorRigid3DTransform
http://www.itk.org/Insight/Doxygen/html/classitk_1_1VersorRigid3DTransform.html
2) Replace the optimizer with the
VersorRigid3DTransformOptimizer
http://www.itk.org/Insight/Doxygen/html/classitk_1_1VersorRigid3DTransformOptimizer.html
3) Are you using the CenteredTransformInitializer ?
(you should use it for initializing the transform)
If you put your MRI and SPECT images in our ftp incoming
directory we can give them a try for tunning the parameters.
(that will be at ftp public.kitware.com ). Please use a common
fileformat, such as Analyze or MetaImage.
Regards,
Luis
----------------------
valli gummadi wrote:
>
> Hi All,
> I am posting the same message again and again because I am really struck to Proceed further in my Registration application. And I am in need of this registration process.And,I could achieve the registration process when both datasets are in same view.only problem is when both are in different view.Any help is favourable to me.Thank you in advance.
>
> I have been working in Registration of MRI and SPECT Images of same
> patient Brain.
> MR dimensions are 256 X 256 X 120 and 0.859382 \ 0.859375 \ 1.60000.
> My moving image is SPECT image.Actual dimensions are 128 X 128 X 47 with voxel dimensions 1.802125\1.802125\3.60425.
> SPECT image undergoes a ForceCubic operation and Its dimensions changes to 128X 128 X 95 and voxel dimesnions to 1.802125 \ 1.802125 \ 1.802125.
> And it is in Transverse View.
> As MR image is in Coronal view i changed the Spect to coronal before
> giving it to the Registration. So,After changing its view to coronal its dimensions become
> 128 X 95 X 128 and 1.802125 \ 1.802125 \ 1.802125.SPECT with these dimensions is given to the registration process.
> I used the following itk classes:
> MattesMutualInformationImageToImageMetric,RecursiveMultiResolutionPyramidImageFilter,
> QuaternionRigidTransform,RegularStepGradientDescentOptimizer,RecursiveMultiResolutionPyramidImageFilter.
>
> And the input parameters are :
> Translate scale value is 1/10000, iterations are 150, scales =3D 1, max step length = 4.0, min step length = 0.005, bins = 35, samples =
> 1000, levels = 3.
>
> I added observer to the process and i am getting the following values :
> First Level:
> 0 -0.320374 [0.0100438, -0.0350144, 0.0105636, 0.986362, =
> -0.553508, -0.00411211, 0.831858]
> 1 -0.321063 [0.030953, -0.0382486, -0.00937501, 0.999558, =
> -0.00125347, -0.440459, 1.5415]
> 2 -0.371695 [0.0659527, -0.0315038, -0.0302047, 0.974523, =
> 0.0133357, -1.43741, 1.60113]
> 3 -0.389297 [0.0748733, -0.0149963, -0.0518501, 0.992978, =
> 0.962366, -1.64413, 1.8366]
> 4 -0.473205 [0.0765801, -0.0308507, -0.0461271, 1.0002, 0.831316, =
> -1.52403, 2.30357]
> 5 -0.489634 [0.0867224, -0.0318137, -0.0522786, 1.01073, 1.15721, =
> -1.6065, 2.67335]
> -------------------------------
> 20 -0.526699 [0.0869593, -0.0339316, -0.0559707, 1.01556, 1.18742, =
> -1.66461, 2.72034]
> 21 -0.526875 [0.086819, -0.0340573, -0.056085, 1.01558, 1.19139, =
> -1.65994, 2.72518]
> 22 -0.527021 [0.0867698, -0.0341474, -0.0562826, 1.0156, 1.19699, =
> -1.65843, 2.73041]
> 23 -0.527126 [0.0866777, -0.0342297, -0.0564333, 1.01563, =
> 1.20222, -1.65538, 2.73534]
> -----------------------------------------------
> 47 -0.528611 [0.0852474, -0.0356566, -0.058518, 1.015, 1.25176, =
> -1.57546, 2.82544]
> 48 -0.528646 [0.0852217, -0.0356863, -0.0585624, 1.015, 1.25251, =
> -1.57258, 2.82797]
> 49 -0.528681 [0.0852033, -0.0357116, -0.0586277, 1.01498, =
> 1.25373, -1.57004, 2.83068]
> ---------------------------------------------
> 144 -0.525427 [0.0845718, -0.0363607, -0.0603638, 1.01422, =
> 1.22735, -1.23997, 2.96168]
> 145 -0.525436 [0.0845504, -0.0363369, -0.06038, 1.01423, 1.22755, =
> -1.2361, 2.96218]
> 146 -0.52494 [0.0845361, -0.0363315, -0.0603705, 1.01426, =
> 1.22752, -1.23235, 2.96328]
> 147 -0.52495 [0.0847016, -0.036275, -0.0605773, 1.01413, 1.22964, =
> -1.23218, 2.96654]
> 148 -0.525374 [0.0846794, -0.0363122, -0.0605381, 1.01414, =
> 1.22845, -1.23069, 2.96692]
> 149 -0.525412 [0.0846573, -0.036348, -0.0605014, 1.01415, =
> 1.22734, -1.22913, 2.96732]
> Second Level:
> 0 -0.418512 [0.0887451, -0.0189369, -0.077675, 1.00696, 2.20275, =
> -1.37414, 3.13128]
> 1 -0.418945 [0.0836208, -0.0236872, -0.0630348, 1.01192, 2.22691, =
> -0.917966, 3.33384]
> 2 -0.413912 [0.0652313, -0.0161832, -0.0486405, 1.00361, 2.10126, =
> -0.505873, 3.08141]
> 3 -0.371655 [0.0629191, -0.00723327, -0.0532294, 1.01451, =
> 2.54857, -0.291537, 3.14261]
> 4 -0.361762 [0.0715741, -0.0131997, -0.059176, 1.02901, 2.72654, =
> -0.369328, 3.60296]
> 5 -0.367962 [0.0714853, -0.0257059, -0.0513781, 1.02656, 2.41281, =
> -0.219652, 3.96205]
> 6 -0.388251 [0.0817307, -0.0310034, -0.0601911, 1.01481, 2.18368, =
> -0.662284, 3.99708]
> 7 -0.41388 [0.0881204, -0.03356, -0.0638268, 0.991765, 1.87106, =
> -1.04525, 3.92623]
> ----------------------------
> 53 -0.444609 [0.0904033, -0.0317655, -0.0649183, 0.986647, =
> 2.28652, -1.12252, 4.21855]
> 54 -0.444668 [0.0903786, -0.0317895, -0.0648504, 0.986712, =
> 2.29139, -1.12121, 4.22452]
> 55 -0.444666 [0.0904111, -0.0317457, -0.064888, 0.986666, =
> 2.29744, -1.12326, 4.22902]
> 56 -0.44467 [0.0903941, -0.0317951, -0.0648093, 0.986712, 2.3017, =
> -1.12212, 4.23546]
> 57 -0.44468 [0.0904166, -0.0317319, -0.064856, 0.986649, 2.30811, =
> -1.12411, 4.23945]
> 58 -0.444699 [0.0903994, -0.0317871, -0.0647737, 0.986687, =
> 2.31225, -1.12307, 4.24599]
> ----------------------------------
> 146 -0.445491 [0.0900287, -0.0319401, -0.0639129, 0.985568, =
> 2.73939, -1.19044, 4.6637]
> 147 -0.445511 [0.0900287, -0.031949, -0.0639108, 0.985541, =
> 2.74461, -1.19175, 4.66937]
> 148 -0.445525 [0.0900161, -0.0319461, -0.0639007, 0.985548, =
> 2.75007, -1.19225, 4.67494]
> 149 -0.445532 [0.090016, -0.0319676, -0.0638845, 0.985523, =
> 2.75507, -1.19321, 4.68086]
> Third Level:
> 0 -0.302895 [0.112791, -0.0365895, -0.0844165, 1.00896, 3.33379, =
> -1.60939, 5.38112]
> 1 -0.322343 [0.0992007, -0.0304807, -0.0782828, 1.00517, 3.25472, =
> -1.32882, 4.97522]
> 2 -0.313556 [0.101683, -0.0331957, -0.0747196, 1.01122, 3.1463, =
> -1.1917, 5.15378]
> 3 -0.328914 [0.0990012, -0.0300732, -0.0743876, 1.0092, 3.13413, =
> -1.19278, 5.02946]
> 4 -0.330295 [0.0997371, -0.0316464, -0.0740325, 1.00807, 3.0786, =
> -1.21719, 5.0444]
> 5 -0.332905 [0.100176, -0.0317067, -0.0743968, 1.00936, 3.0995, =
> -1.2133, 5.06726]
> ---------------------------
> 35 -0.338694 [0.100619, -0.031991, -0.0749333, 1.01022, 3.03826, =
> -1.2607, 5.02906]
> 36 -0.33863 [0.100617, -0.0320048, -0.0749167, 1.01019, 3.03661, =
> -1.26134, 5.02824]
> 37 -0.338678 [0.100626, -0.0320129, -0.0749159, 1.01021, 3.03511, =
> -1.26227, 5.02739]
> 38 -0.338677 [0.10062, -0.0319927, -0.0749215, 1.01021, 3.03419, =
> -1.26311, 5.02589]
> -------------------------
> 59 -0.338709 [0.100639, -0.0320017, -0.0748891, 1.01031, 3.00697, =
> -1.27972, 5.00324]
> 60 -0.338664 [0.100631, -0.0319735, -0.0749054, 1.01029, 3.00644, =
> -1.28063, 5.0016]
> 61 -0.338709 [0.100639, -0.0320041, -0.0748886, 1.01031, 3.00458, =
> -1.28112, 5.00129]
> 62 -0.338667 [0.10063, -0.0319716, -0.074908, 1.0103, 3.00421, =
> -1.28205, 4.9996]
> 63 -0.33871 [0.100645, -0.0320181, -0.074882, 1.01033, 3.00229, =
> -1.28225, 4.99989]
>
> First and Second Levels took all the 150 iterations and at third levels process came out at 65th iteration. I tried to change the tranlate scale value to less than 1000. Becuase for mr volumes 100 is sufficient.But, I am gettign the itk errorr message "All the sampled points are moved outof the image regios.
> Please anybody explain me the better way to achieve good registration
> output.
>
> Regards,
> Srivalli.
>