©2010 International Journal of Computer Applications (0975 - 8887)
Volume 1 – No. 20
INTENSITY BASED IMAGE REGISTRATION
BY MAXIMIZATION OF MUTUAL
INFORMATION
R.Suganya,
K.Priyadharsini,
Dr.S.Rajaram,
Thiagarajar college of
Engg., Madurai,
TamilNadu,India
Thiagarajar college of
Engg.,Madurai,
TamilNadu,India .
Thiagarajar college of
Engg.,Madurai,
TamilNadu,India .
Abstract:
Biomedical image registration, or geometric alignment of two-dimensional and /or three-dimensional
(3-D) image data, is becoming increasingly important in diagnosis, treatment planning, functional
studies, and computer-guided therapies and in biomedical research [1]. Registration is an important
problem and a fundamental task in image processing technique. In the medical image processing fields,
some techniques are proposed to find a geometrical transformation that relates the points of an image to
their corresponding points of another image. In recent years, multimodality image registration
techniques are proposed in the medical imaging field. Especially, CT and MR imaging of the head for
diagnosis and surgical planning indicates that physicians and surgeons gain important information from
these modalities. In radiotherapy planning manual registration techniques performed on MR image and
CT images of the brain. Now-a-days, physicians segment the volume of interest (VOIs) from each set of
slices manually. However, manual segmentation of the object area may require several hours for
analysis. Furthermore, MDCT images and MR images contain more than 100 slices. Therefore, manual
segmentation and registration method cannot apply for clinical application in the head CT and MR
images. Many automatic and semiautomatic image registration methods have been proposed [2]. The
main techniques of image registration are performed by the manual operation, using Landmark and using
voxel information. In this paper, an automatic intensity based registration of head images by computer
has been employed by applying maximization of mutual information. The primary objective of this
paper is to increase accuracy of the registration and reduce the processing time. Experiments show our
algorithm is a robust and efficient method which can yield accurate registration results.
Keywords: Image registration, Mutual information, Medical imaging, Multimodality image.
1
©2010 International Journal of Computer Applications (0975 - 8887)
Volume 1 – No. 20
1. Introduction:
Medical
images
are
becoming
increasingly widely used healthcare and in biomedical research and a very wide range of
imaging modalities is now available, such as CT,
MRI, PET, SPECT, and so on. In some clinical
scenarios, the information from several different
imaging modalities should be integrated to
deduce useful clinical conclusions. Image
registration aligns the images and so establishes
correspondence between different features
contained on different imaging modalities,
allows monitoring of subtle changes in size or
intensity over time or across a population and
establishes correspondence between images and
physical space in image guided interventions.
Image registration is the process of
determining the spatial transform that maps
points from one image (defined as the moving
image) to homologous points on an object in the
reference image (called as the fixed image). The
similarity of the two images will be calculated
and investigated after each transform until they
are matched.
Of the multitude of image registration
similarity measures that have been proposed over
the years, mutual information is currently one of
the most intensively researched measures. This
attention is a logical consequence of both the
favourable characteristics of the measure and the
good registration results reported. Mutual
information is an automatic, intensity-based
metric, which does not require the definition of
landmarks or features such as surfaces and can
be applied in retrospect [2]. Furthermore, it is
one of the few intensity-based measures well
suited to registration of multimodal images.
Unlike measures based on correlation of gray
values or differences of gray values, mutual
information does not assume a linear relationship
among the gray values in the images. However,
the mutual information registration function can
be well-defined, containing local maxima.
Another drawback of MI is that it is calculated
on a pixel-by-pixel basis, applying that it tasks
into account only the relationship between
corresponding individual pixels and not those of
each pixel’s respective neighbourhood. As a
result, much of the global spatial information
inherent in images is not utilized. In addition, it
is a time-consuming work, especially for high
resolution images, because mutual information of
the two images must be calculated in each
iteration.
In this paper, a novel
automatic registration of head
computer, which obtained CT and
employing maximization of mutual
and reduce the processing time.
method for
images by
MR images
information,
Medical image datasets
Target image
CT scan
Reference image
MRI scan
Preprocessing:
Image size,pixel spacing,noise
Registration using mutual
information,optimization,interpolation
MRI
registered on
CT scan
Image
Reconstruction
2
©2010 International Journal of Computer Applications (0975 - 8887)
Volume 1 – No. 20
Figure 1: Block diagram for multimodality
image registration
Where R is a specified region.The centroid
denoted by (xc, yc) is calculated by
2. MUTUAL INFORMATION AND IMAGE
REGISTRATION
(xc,yc)= (m10/m00,m01/m00)
In this paper, we develop a multimodal image registration technique, which is
based on CT and MR imaging of head. When an
image superposes onto another one, which
obtained multi-modality, several preprocessing
techniques are needed for image registration. For
multi-modal image registration, the relation
between the sizes of image is usually not same.
In the preprocessing step, we remove some noise
and normalize the size of image. In the next step,
initial point is detected by using center of gravity
of each image as an initial registration. Finally,
the registration is performed employing
maximization of mutual information. The details
of each method are shown as follows.
2.1 Pre-processing and initial registration
The images that obtained from
different modalities may have different image
size, pixel spacing and a number of slices. In the
preprocessing step, image size, pixel spacing is
normalized and image noise is reduced for
registration. To align the multi-modal images,
center of gravity is used for initial registration. In
this paper, we assume that the MR image is
target image and CT image is reference image. In
this step, the center of gravity on multi-modal
images has been calculated as follows.
Let a digital image be denoted by f(x,
y) where x and y are bounded positive integers.
The (i+j) th order moment mij of f(x,y) is defined
by
Mij=∑ xjyj f(x,y)
(x ,y)εR
The target image is transformed onto the x-y
coordinates of the derived center of gravity. It
must be the same coordinates as an initial
position for registration.
The most popular technique of interpolation is
linear interpolation, which defines the intensity
of a point as the weighted combination of the
intensities of its neighbours.
2.2 Summation of intensity projection with
weight
In order to reduce the computational
time, three dimensional (3-D) image data is
projected as a two dimensional (2-D) image data.
It is similar to Maximum Intensity Projection
(MIP) method which is applied in medical
imaging field as 3-D image represent technique.
We make the six 2-D image from 3-D image set
for image registration by calculating the
Summation of Intensity Projection (SIP) with
weight on each direction. The weight depends on
the distance from the screen. Xk (k=1,2,…n) is
assumed as the change in signal density in 3-D
space. The expression of SIP with weight
describe as follows.
N
S= ∑ xkwk
K=1
Where S is the value of density with projected to
screen and wk shows weight value.
2.3 Final Registration
3
©2010 International Journal of Computer Applications (0975 - 8887)
Volume 1 – No. 20
Final registration is performed by using the
mutual information method. For two images A
and B, mutual information I(A,B) can be defined
as follows [12].
I(A,B)=∑pA,B (a,b)log pA,B(a,b)/pA(a).pB(b)
The interpretation of this equation measures the
distance between the joint distribution of image
pixel value pA,B(a,b) and the joint distribution in
case of independence of the image, pA(a), pB(b).
In this paper, the target image is transformed for
maximization of mutual information. When
mutual information is calculated in the images,
the intensities of corresponding points of two
images a and b are used for making 2-D
histogram h(a,b). It can be thought that the
probability distribution function can be obtained
by dividing the histogram by N, where N is a
number of pixels. It can be expressed as follows:
pA,B (a,b)=h(a,b)/N
pA(a) = ∑ pA,B (a,b)
b
pB(b)=∑ pA,B (a,b)
a
When mutual information is detected,
the intensity of pixel is linearly converted from 0
to 255. In this paper, the amount of the
movement of the target image is described by
optimization that uses multi-dimensional
direction set method (powell’s method) of which
the index is mutual information. This method can
be used to 3-D data. But, it needs more
computational time. Thus the calculation cost is
decreased by using 2-D data of target image and
reference image, 2-D data is obtained by the
method of SIP with weighted value.
The target image is the CT image and MRI is
the reference image. The reference image
compare to the target image we get the matched
image and rotated image shown in Fig 2, and Fig
3.
Proposed Algorithm
PROCEDURE: Image Registration that uses 2D data
Input: F and R
Output: Registration using Mutual information
Begin
Flag
0 and ni
0 (i=1,2,…..6)
I0
I(F,R).
/*The 2D data fj and rj are made from F and R.*/
if ni
0,j
i
/*The movement parameter mi*/
mj
fj * rj.
Fj
T(Fi,mj), Ij
I( Fj,R)
/*The maximum value of Imax*/
Imax
Ij and I0. .
If Imax =Ij .
K
j* Imax .
F=Fk , nk=1, I0 =Imax , Flag =0.
/*if Imax not included in Ij*/
If Flag=1,
Flag=0,I0=Imax and n i=0.
Here I shows the mutual information, F shows
target image, R shows reference image, i shows
the number of 2D data and T(F,m) shows the
image F is transformed by using the movement
parameter m.
3. EXPERIMENTAL RESULTS
The Proposed technique was applied to
the two different modalities images- CT and MR
images of five human head images and
4
©2010 International Journal of Computer Applications (0975 - 8887)
Volume 1 – No. 20
satisfactory results are achieved. The parameters
of CT and MR images are shown in table 1. In
this paper, the image that has 512*512 pixels in
each slice were converted into 256*256 pixels to
reduce the processing time.
The experimental results are shown in
Fig 2, Fig 3, table 2.shows the transformation
and matching of the image using mutual
information.
4. CONCLUSION
Multi-modal
medical
intensity
registration is an important capability for finding
the correct geometrical transformation that brings
one intensity pixel in precise spatial
correspondence with another intensity pixel.
Multi-modal intensity registration is needed to
benefit from the complementing information in
medical images of different modalities. The
result demonstrates that our registration
technique allows fast, accurate, robust and
completely
automatic
registration
of
multimodality medical images. From the image
the value of mutual information in MRI and CT
are compared with existing method and shown in
Table 2. Compared with the existing method, our
experimental results are better with using the
intensity of two images.
Figure 2: Original image and the
corresponding matched MRI image
Table 1.Image parameter
CT image
MRI image
Size[pixels]
512*512
512*512
Pixel spacing
[mm]
Slice
thickness[mm]
0.638
0.638
2
2
Figure 3: Original image and the
corresponding rotated MRI image
5
©2010 International Journal of Computer Applications (0975 - 8887)
Volume 1 – No. 20
Table 2: Experimental results
Number
Existing method
Proposed method
Mutual information
Mutual information
1
0.4473
0.5658
2
0.5174
2.5083
3
0.4667
2.5396
4
0.5639
2.9598
5. REFERENCES
[1] L. Ding, A. Goshtasby, M. Satter, "Volume
image registration by template matching", Image
and Vision Computing, Vol.19, No.12, pp.821832,2001.
[6] S. Klinski, C. Derz, D. Weese et al., "Model
based image processing using snakes and mutual
information", Medical Imaging: Image
Processing,Proc. SPIE, pp. 1053-1064, 2000.
[2] J. M. Fitzpatrick, D. L. G. Hill, Y. Shyr et
al.,"Visual Assessment of the Accuracy of
Retrospective Registration of MR and CT
Images of the Brain", IEEE Trans. on Medical
Image., Vol.17, No.4, pp.571-585, 1998.
[7] D.Mattes, D.R.Haynor, H.Vesselle, T.K.
Lewellen, and W.Eubank, “Non-rigid
multimodality image registration,” Medical
Imaging 2001: Image Processing, pp 1609-1620,
2001.
[3] Friston KJ, Ashbuner J, Frith CD et al.,
"Spatial registration and normalization of
images", HumanBrain Mapping, Vol. 3, pp.165189, 1995.
[8] F. Maes, A. Collignon, D. Vandermeulen et
al,"Multimodality Image Registration by
Maximization of Mutual Information", IEEE
Trans.on Medical Image., Vol.16, No.2, pp.187198, 1997.
[4] T. Gaens, F. Maes, D. Vandermeulen, P.
Suetens,"Non-rigid Multimodal Image
Registration Using Mutual Information", Intl.
Conf on Medical Image Computing and
Computer-Assisted Intervention,
Pp.1099-1106, 2003..
[9] J. Maintz, E. Meijering, M. Viergever,
"General multimodal elastic registration based on
mutual information", Medical Imaging: Image
Processing, Proc. SPIE, pp. 144-154, 1998.
[5] B. Kim, J. Boes, K. Frey et al., "Mutual
information for automated unwarping of rat brain
autoradiographs", Neuro Image, Vol. 5, No. 1,
pp. 31-40, 2004.
[10] J.P.W.Pluim, J.B.A. Maintz, and
M.A.Viergever, “Image Registration by
Maximization of Combined Mutual Information
and Gradient Information,” IEEE Transactions
on Medical Imaging
6
©2010 International Journal of Computer Applications (0975 - 8887)
Volume 1 – No. 20
[11] J.P.W.Pluim, J.B.A. Maintz, and
M.A.Viergever, “Mutual Information-based
registration of medical images:a survey,
“Medical Imaging. IEEE Transactions on,
vol.22,pp.986-1004, 2003.
[12] A. Roche, G.Malandain, X. Pennec et al.,
"The Correlation Ratio as a New Similarity
Measure for Multimodal Image Registration",
Intl. Conf on Medical Image Computing and
Computer-Assisted Intervention, pp.1115-1124,
1998.
[13] Wu.A,Hartov.A,Paulsen.D,Roberts.W,
”Multimodal Image Registration via Mutual
information to account for intilal tissue motion
during image-guided neurosurgery”. Intl. Conf
of the IEEE EMBS Pp.1675-1678, 2005.
[14] B.Zitovaa and J.Flusser, “Image
Registration Methods: a survey,” Image and
Vision Computing, vol.21, pp.977-1000, 2003.
7