1. Feature based Matching of CT & MRI Brain Images Ayush Dogra

Abstract -Multimodal image matching calls or demands for feature based or object based matching. Feature used in matching can be edges ,ridges, blobs, valleys etc. Feature based matching is considered as less tedious and low level image processing task whereas object based matching is consider as high level image processing task and complex.

2. Matching

hile integrating two multimodal and mono modal images, first step is matching (geometrical matching) and second step is fusion (combined display of data involved). In this paper we will concentrate on first step. Matching of CT and MRI images is useful in radiation therapy planning. This type matching is also used in skull base surgery and epilepsy surgery. And the generalization of a 3D images by lvv operator can be found in [4]. When applying lvv operator to the CT set the resultant image show skull ridge. Then this ridgeness image is superimposed on to the original

3. Registration Method

Now we need to register the feature volume. The technique used is cross correlation [5] since the two images have different physical realities. By using the grey values directly [1], avoid segmentation our feature images.

The correlation value c(t) of CT feature volume L1 and MR feature volume L2 over all rigid transformations t, where c(t) is defined Here next level is formed by maximizing the trough images or minimizing the ridge images. This method is also called multi resolution correlation [2,6]. The disadvantage of this approach is high computational effort required. "

4. Conclusion

The terms matching and registration are both used to denote the process of determining a transformation that relates the contents of two images in a meaningful way. The registration of MR and CT images was one of the first application of medical image registration. Further more the applicability of MR-CT registration is restricted primarily to the head. MR-CT registration is likely to become more widely used in future.

CAT And MRI Scans CAT scans are a specialized type of x-ray. The patient lies down on a couch which slides into large circular opening. The x-ray tube rotate around the patient and computer collects the results. These results are translated into images that look like a slice of a person. CT is very good for imaging bone structure. "Fig." 1 shows CT slice. — Figure 1.

Figure 1 : A CT slice MRI scan uses magnets and radio waves lo locate the images. The patient lies on the couch that looks very similar the ones used for CT. they are then placed in a very long cylinder. The machine will produced a lot of noise and examinations typically run — Figure 2. Figure 1 :

slice to construct the 3D volume. "Fig." 8 shows ridged CT image. Similar example can be done in MRI image. "Fig." 9 shows a MRI image. "Fig." 10 shows the ridged MRI image (also called troughed image as in MRI image the skull area is dark). — Figure 3.

Figure 5 : CT slice — Figure 4. Figure 5 :

Figure 7 : Blurred by Gaussian Convolution — Figure 5. Figure 7 :

Fig." 11 shows resultant image after CT and MR volume matching using ridgeness correlation. — Figure 6.

Figure 11 : CT and MRI matching using ridgeness correlation — Figure 7. Figure 11 :

Appendix A

Evaluation of ridge seeking operators for multimodality medical image matching. J B A Maintz , P A Van Den Elsen , M A Viergever . IEEE Transactions on pattern analysis and machine intelligence 1996. 18 (4) p. .
Automatic registration of CT and MR brain images using correlation of geometrical features. P A Van Den Elsen , J B A Maintz , E J D Pol , M A Viergever . IEEE Transactions on medical images
Image fusion using geometrical features. P A Van Den Elsen , J B A Maintz , E J D Pol , M A Viergever . Visualization in biomedical computing, volume 1808 of Proc. SPIE, R A Robb (ed.) (Bellingham, WA
) 1992. SPIE Press. p. .
Multimodality matching of brain images, P A Van Den Elsen . 1993. Netherlands. Utrecht University (PhD thesis)
Grey value correlation techniques used for automatic matching of CT and MR brain and spine images. P A Van Den Elsen , E J D Pol , T S Sumanaweera , P F Hemler , S Napel , J Adler . Visualization in Biomedical Computing, Proc. SPIE, (Bellingham, WA
) 1994. SPIE Press. (in press)
Automatic Registration of CT and MR Brain Images Using Correlation of Geometrical Features. P Van Den Elsen , J Maintz , E-J Pol , M Viergever . IEEE Trans. Medical Imaging June 1995. 14 (2) p. .

Table of contents

Appendix A