3D Brain Volume and Surface Visualization with CT and MR Images
for Diagnosing Abnormal Structures:
Integration with Photogrammetric Techniques
(EXTENDED SUMMARY)
S. Dogan', M. O. Altan"
" Ondokuz Mayis University, Engineering Faculty, Dept. of Geodesy and Photogrammetry, 55139 Kurupelit, Samsun, Turkey —
sedatdo@samsun.omu.edu.tr
® Istanbul Technical University, Civil Engineering Faculty, Dept. of Geodesy and Photogrammetry, 80626 Maslak, Istanbul, Turkey
— oaltan@itu.edu.tr
Commission V, WG V/3
KEY WORDS: Volume Rendering, Surface Rendering, CT and MR, Medical Image Segmentation, Digital Photogrammetry,
Texture Mapping.
ABSTRACT
In this paper a study on 3D volume and surface reconstruction of human brain with CT and MR image slices is pre-
sented. It is planned to integrate these constructed 3D representations with the external facial surface of the patient in
the future works. When 3D model of the human brain is created, tumours or like other abnormalities may be detected
and measured on the reconstructed 3D model. In this study, both 3D volume and 3D surface representation of human
brain have been derived. 3D volume is a projection for visualization of volumetric data on a 2D scene. For volumetric
imaging, ray casting technique is used with different ray functions such as composite, maximum intensity, average ray
functions etc. On the other hand, surface model is constructed by using isosurface extraction techniques. In this study
we have also studied on 3D reconstruction of external face surface with digital photogrammetric techniques. Both for
inner brain 3D reconstruction and external face reconstruction we have written a medical imaging software with Bor-
land C++ Builder. But because of limitation of our hardware capacities we couldn’t still test the photogrammetric mod-
ule of the program for the external face surface. For medical module of the software we have used both VTK (Visuali-
zation Toolkit) C++ libraries and also our own implementations. By this program many options have been presented to
users. One can view the models from different angles with different scales. Some may cut or clip models from intended
points by using implicit functions like plane, sphere etc. It is possible to make interactive segmentation with histogram
thresholding, manual segmentation or contour segmentation techniques. Surface models can be smoothed with many
options. Smoothing errors may be visualized graphically or in the list form with numerical error values. Models can be
viewed with one or many light sources. Light properties can be set with different options like location, colour, illumina-
tion properties etc. Parallel, orthogonal and perspective views are possible with different viewing camera settings,
where coordinate transformations are also possible. In this paper we explain the techniques for inner brain structures
and give the results of 3D models of our software.
1. INTRODUCTION to communicate between radiologists and clinicians who are
responsible for treatment.
Various medical imaging techniques are used for diagnosing
and treatment of many diseases as well as for surgical opera-
tions. With medical imaging techniques, inner tissues of a
human body may be visualized and thus abnormal formations of
interested tissues may be defined. The mostly used medical
imaging techniques are CT (Computerized Tomography) and
MR (Magnetic Resonance) imaging. These two imaging meth-
ods use different hardwares and different mathematical and
physical models. With these two techniques, interest region of
the body is scanned with CT or MR scanners and thus 2D image
slices of the interest region of the body is obtained with recon-
struction of the projected rays. Nowadays, since there are no
full automatic systems, which can interpret these 2D images
slices, interpretation of the images mostly depends on the
knowledge and the experience of the radiologists. Even if the
radiologists diagnose the abnormalities of tissues, treatment is
held by the clinician doctors. By using 2D images, it is difficult
The data acquired from CT or MR scans is usually displayed in
2D planar format. Although these display techniques are gener-
ally quite adequate for diagnosis, it doesn’t optimally commu-
nicate the 3D nature of the involved anatomy or the full extent
of pathology. Radiologists become skilled at forming a mental
model of the 3D aspects of a case, but may be less successful in
communicating this mental model to the clinician. Three di-
mensional imaging presents complex anatomical findings in a
format easily understood without prior specialized training and
experience (Ney et al, 1990). Furthermore, by using three di-
mensional imaging, it is possible to measure geometric features
like distance, area and volume and thus it is possible to find the
exact locations of the abnormalities precisely.
—250—
(Se ON CCS tj e ana 6 NY
PN En
aa