380 PIERS Proceedings, Kuala Lumpur, MALAYSIA, March 27 30, 2012 MR Perfusion Visualization in 3D Image M. Cap 1, E. Gescheidtova 1, P. Marcon 1, K. Bartusek 2, and E. Kroutilova 1 1 Department of Theoretical and Experimental Electrical Engineering Brno University of Technology, Kolejni 2906/4, 612 00 Brno, Czech Republic 2 Institute of Scientific Instruments, Academy of Sciences of the Czech Republic Kralovopolska 147, 612 64 Brno, Czech Republic Abstract MRI is a constantly developing region of medicine, which is suitable for the study of soft tissues. The current methodologies for obtaining images weighted by relaxation times give only an idea of the distribution of soft tissues. Differential diagnosis of a high-grade glioms and solitary metastases is in some cases inconclusive. Investigators in several studies have demonstrated that in perfusion MRI (magnetic resonance imaging) of high-grade glioms and solitary metastases are differences. Analysis of the peritumoral region could be more useful than the analysis of the tumor itself. Precise evaluation of mentioned differences in peritumoral region gives a hopeful chance for tumor diagnosis. This article describes image processing and fusion of the MR images. T 2 weighted (T2W) images and perfusion weighted images are processed for creating a 3 dimensional image. System is designed for calculating slice from 3D image in any direction and position defined by user. Consequently, T2W and PWI are used for image fusion and to image perfusion in structural image in defined direction and position. 1. INTRODUCTION This article deals with fusion of the perfusion weighted images with T 1, T 2 weighted image and its imaging in 3D images. The aim of this work is to create system for tumor and peritumoral region detection, image registration and fusion of the T 2 and perfusion weighted images. The importance of the perfusion imaging method lies in its ability to describe anatomy and physiology of the tumor and peritumoral region microvasculature [1]. Several studies have demonstrated that in perfusion MRI of high-grade glioms and solitary metastases are differences [2 4]. In this article, we are focused on the differences in perfusion weighted images in the peritumoral region. Three sets of images were processed. T 1 and T 2 weighted images are used for detection of the tumor and peritumoral region size and position. These two image sets contain 22 images. Resolution of the T 1 weighted images is 256 256 px (0.9 mm 0.9 mm pixel size), resolution of the T 2 weighted images is 512 512 px (0.45 mm 0.45 mm pixel size). Fig. 1 gives as examples of the processed images. Perfusion images are acquired each 1.22 s with space resolution 64 64 px (pixel size 3.4 mm 3.4 mm). Image fusion requires precise image registration. In an effort to obtain the most accurate diagnosis is the image registration the key part of the image fusion algorithm. Hence, the PWI images are processed wit accuracy of ±1 pixel. With the aim of image fusion the structural image is processed for tumor and peritumoral region detection. Figure 1: T 1 weighted image, T 2 weighted image and perfusion weighted image of the same slice.
Progress In Electromagnetics Research Symposium Proceedings, KL, MALAYSIA, March 27 30, 2012 381 2. METHODS 2.1. Image Registration Image sets are acquired in different resolutions and slice positions. DICOM standard tags used for image positioning in 3D space are shown in Table 1. Precise position detection in both images is necessary for right perfusion evaluation in peritumoral region. T1W, T2W and PWI images has the same Image orientation (patient) tag so all images are parallel. Slice thickness and spacing between slices tags for structural and PWI images are different. Image registration and consequent image fusion is provided in T2W image and in PWI image which are in the same volume. Detection of the size and position of the patient s brain is based on the interhemispheric fissure position as can be seen in Fig. 2. Position of the interhemispheric fissure is detected from the local minimum in the frontal and occipital lobe. Each value is calculated as an average of 10 positions of the local minimum in surrounded rows. For size adjustment the bilinear method is used. Pixel is calculated as a weighted average of pixels in the nearest 2-by-2 neighborhood. 2.2. Segmentation Most important part of the work is the segmentation of the regions of interest. T2W image was chosen for higher contrast in peritumoral region as you can see in Fig. 1. The goal of segmentation is to find position of the tumor and peritumoral region. Segmentation method is based on the detection of the high intensity of the tumor region [6]. At first, the high contrast area of the tumor Table 1: Used tags for image positioning in 3D space. Tag Atribute name 0018/0050 Slice Thickness 0018/0050 Spacin Between Slices 0020/0032 Image Position (IP) 0020/0037 Image Orientation (Patient) (IOP) 0028/0030 Pixel Spacing Figure 2: Results of the interhemispheric fissure detection. T2W image left, PWI image right. Figure 3: On the top the detected tumor, on bottom the detected peritumoral region in T2W image.
382 PIERS Proceedings, Kuala Lumpur, MALAYSIA, March 27 30, 2012 is detected. Tumor area intensity exceeds a threshold which is adjusted according to whole image intensity. In the second step the incorrectly detected areas are removed or added. Results of the tumor and peritumoral region detection are in Fig. 3. Signal intensity in peritumoral region is not so significantly higher as is in tumor area. Algorithm detects higher signal intensity close to the tumor and signal loss on the peritumoral region boundaries. 2.3. Image fusion Detection of the tumor and peritumoral region gave as a position of these objects. These positions are used for a mask creating. By multiplying the mask with PWI image the perfusion in peritumoral region is obtained. Results of the image fusion you can see in Fig. 5. 3. RESULTS Results of the tumor and peritumoral region detection algorithms are masks shown in Fig. 3. Resulting mask is obtained as the peritumoral region mask reduced about the tumor mask in size of whole image. By multiplying this mask and PWI image is obtained perfusion only in peritumoral region. Position of the T2W images in the space of tomograph is in Fig. 4. Zero position is defined as Figure 4: Position of the T2W images in the space of tomograph. Figure 5: Fusion of the T1W and PWI image in peritumoral region. Figure 6: Perfusion in the transversal and coronal plane.
Progress In Electromagnetics Research Symposium Proceedings, KL, MALAYSIA, March 27 30, 2012 383 Figure 7: Perfusion visualization in structural 3D image. center of the tomograph. Directional cosines of the Image orientation tag for this images are (0, 9946; 0, 0831; 0, 0657; 0, 0935; 0, 971; 0, 2167). In this work designed system is based on fusion of perfusion images to the 2D structural data as is shown in Fig. 5. Gadolinium based contrast cause in image signal loss depending on the content of the contrast in volume of the voxel. Perfusion data are always imaged in peritumoral region and in always the same color scale from 30 to 180 units. Goal of this work is to realize system for visualizing perfusion in more than one slice and in 3D image. System use IP a IOP DICOM tags as a information about requested slice. Voxel value is used from the nearest voxel to the calculated position of the new slice voxel. Example for the perfusion on the transversal and coronal plane you can see in Fig. 6. Perfusion visualization in three dimensions could be useful for understanding of structure of tissue microvasculature, example of the cut throw the human brain with brain tumor you can see in Fig. 7. User is able to define in which plane (defined by IOP) will be made a cut throw the analyzed volume and how many cuts is necessary to image requested part of the analyzed tissue. 4. CONCLUSION The paper presents registration, segmentation and fusion of the T2W and PWI images and its visualization in two defined planes and in 3D structural image. Interhemispheric fissure were localized and used for image registration. All PWI images were processed for registration wit corresponding T2W images. Perfusion imaging is possible to provide in two or more by user defined planes. This is possible to provide in defined number of slices or in 3D structural image. Presented result hold possibility to be useful for more precise tumor diagnosis. ACKNOWLEDGMENT This work was supported within GACR 102/11/0318 and CZ.1.05/2.1.00/01.0017 (ED0017/01/01) and GA FEKT-S-11-5/1012. REFERENCES 1. Jackson, A., D. L. Buckley, and G. J. M. Parker, Dynamic contrast-enhanced marnetic, Resonance Imaging in Oncology, Vol. XII, 311, Elsevier, 2005. 2. Law, M., S. Cha, E. Knopp, G. Johnson, J. Arnett, and A. Litt, High-grade gliomas and solitary metastases: Differentiation by using perfusion and proton spectroscopic MR imaging, Radiology, 715 21, 2002. 3. Hakyemez, B., C. Erdogan, N. Bolca, N. Yildirim, G. Gokalp, and M. Parlak, Evaluation of different cerebral mass lesions by perfusion-weighted MR imaging, J. Magn. Reson. Imaging, 817 24, 2006.
384 PIERS Proceedings, Kuala Lumpur, MALAYSIA, March 27 30, 2012 4. Rollin, N., J. Guyotat, N. Streichenberger, J. Honnorat, V. T. Minh, and F. Cotton. Clinical relevance of diffusion and perfusion magnetic resonance imaging in assessing intra-axial brain tumors, Neuroradiology, 150 9, 2006. 5. Cianfoni, A., R. Calandrelli, P. De Bonis, A. Pompucci, L. Lauriola, and C. Colosimo, Nocardia brain abscess mimicking high-grade necrotic tumor on perfusion MRI, Journal of Clinical Neuroscience, Vol. 17, No. 8, 1080 1082, 2010. 6. Mikulka, J., E. Gescheidtova, and K. Bartusek, Modem edge-based and region-based segmentation methods, 32nd International Conference on Telecommunications and Signal Processing, 89 91, Dunakiliti, Hungary, 2009. 7. Matlab, Help, Sections: Visualizing MRI data: Volume Visualization Techniques (3-D Visualization); Image Processing Toolbox; Creating Graphical User Interface.