International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Vol XXXV, Part B5. Istanbul 2004
sensor geometries. In many respects, current challenges in
medical imaging show remarkable similarities to usual
photogrammetric problems (Patias, 2002). However, there are
great differences between the terminologies of medical imaging
and photogrammetry. And this makes very difficult to be
motivated on applications that integrate photogrammetry and
medical imaging. This last determination is true for this paper's
introduction section too. As long as the cooperation between
photogrammetry and medical imaging grow up, we believe that
these problems would be solved.
With MIPAS, we also aimed to integrate photogrammetry and
medical imaging for various medical applications such as
radiology, neurosurgery, general surgery , plastic surgery,
dental applications etc. But up to now, we considered MIPAS
as it has two different modules; photogrammetry and medical
imaging and modelling. With the contribution of medical
doctors and biomedical engineers, we believe that two modules
would be synthesized.
For a system to be used for detection and measurement of the
pathological formations; segmentation, detection and
registration steps are very important tasks (Betke et.al, 2001). In
this paper, we briefly explain the methods which we used in
MIPAS for pre-processing, segmentation, detection, registration
and photorealistic 3D visualization. We present sample images
on finding tumour location. We have produced photo realistic
face skin surface model by using texture mapping. And thus, it
is possible to see the whole human head on 3D surface models
completely with outer face and inner brain tissues. This 3D
photo realistic face model may also be used for plastic surgery
after some modifications.
2. DATA ACQUSITION AND PREPROCESSING
OPERATIONS
For 3D reconstruction from CT ànd MR slices, spiral scanning
technique with parallel slices are preferred. During scanning
process, external markers could be used for registration
purposes. For CT scans, the best marker shape is sphere and the
best marker material is plastic. Radius of the sphere must be at
least two times greater than the slice thickness. Thus it would
be guaranteed that the same sphere will appear at least on two
successive slices (Dogan, 2003; Altan and Dogan, 2003).
Before and after scanning, digital images of patient are also
taken with digital photograph cameras. During the photograph
acquisition, for the precise photogrammetric evaluation, some
rules should be obeyed; base and camera-object distance ratios
should be selected carefully. Images must be convergent and
homologues bundle rays must intersect in the object space in an
adequate manner. For the fast and automatic digital image
acquisition, we are designing a camera network in our ongoing
projects with five digital cameras. But in the sample images
presented in this paper, we used only one camera for image
acquisition and we have taken images with one camera from
pre-planned camera stations. These images are used for
photorealistic visualization of the outer body surfaces. These
images can also be used for the 3D measurements of the outer
body surfaces of the patient. By using digital images, it is also
possible to reconstruct 3D point cloud model of the outer faces
(D'apuzzo, 2001).
After acquisition of CT and MR images, image data is recorded
in DICOM format. In DICOM files, detailed information on
scanning parameters is recorded too. For example, slice
thickness, pixel size on x and y directions, scanning type,
scanning direction etc. MIPAS can process DICOM files and
can convert the image contents of the file to TIFF image file
stacks and to individual BMP file sets. MIPAS can also open
slices recorded in TIFF, BMP and raw PGM formats.
After CT scanning, on the slice images, scanner table's images
are also appeared. And the densities of the table pixels are very
close to the densities of the soft tissues. So, during automatic
segmentation, table pixels might frequently be classified as soft
tissue pixels and this situation is undesired for accurate
segmentation and 3D reconstruction. In order to overcome this
problem, table images should be extracted and eliminated from
the slice images. In other words, only the volume of interest
(VOI) should be reside on slices. To select VOI from each slice
images, usually a binary VOI mask is designed and used in
medical applications. With this VOI mask, by using maximum
or minimum filters, only VOI regions of the slices can be
selected. To define binary VOI mask, we used logical image
operations. By applying union operation with "or", all of the
slices are combined on one image. This final image will contain
the all of the VOI regions together with undesired table images.
In this combined image, the largest VOI boundaries can be seen
easily. Some can draw the largest VOI boundary with mouse
and thus VOI mask is obtained. If the inner pixels of the
bounded VOI region are labelled with a unique density level,
binary VOI mask is obtained. Now this mask is compared to all
slices and maximum or minimum operation is performed to
discard table images.
During the pre-processing step, MIPAS provides manual image
coordinate measurement possibility. Thus it is possible to find
the 3D model coordinates of the each marker's center points.
3. SEGMENTATION OF SLICE IMAGES
Segmentation is the process of classifying pixels in an image or
volume. It is one of the most difficult task in the visualization
processes. For reconstruction of medical 3D surface and
volume, interest tissue boundaries should be distinguished from
others on all the image slices. After the boundaries have been
found, the pixels, which constitute the tissue, can be assigned to
a constant grey level value. This constant value represents only
this tissue.
Label values can be used as isocontour value for surface
rendering. For volumetric rendering, in spite of the surface
properties of the tissues, their inner properties are also
important. Because of this reason, we should find the opacity
values for the individual voxels. And near this, wee need
different colours to separate the volume the elements which
belong to different tissues. For these purposes, segmentation
results are used too. In MIPAS, we have used three different
segmentation approaches. These are; interactive histogram
thresholding, contour segmentation and manual segmentation.
3.1. Interactive Histogram Thresholding
The simplest way of image segmentation is thresholding. By
this technique, according to image histogram, possible threshold
values are found. The pixels, that have values above or below
this threshold is assigned to constant values. Thus a binary-
segmented image is obtained. One can choose more than one
thresholds. In this case, between values of these thresholds are
replaced with constant label values. In our software, in spite of
histogram analysis, we have also presented an interactive
Intern
thresh
thresh
synchi
segme
After
and al
make
By th
thresh
might
functi.
segme
operat
holes.
33. €
By th
with
contoi
this *
autom
After
are b
conto
segme
33.1
With
make
manu:
precis
best ı
intere:
editin:
segme
somet
segme
Regis
transf
in anc
the s
Regis
differ:
MR) (
In this
DR
Regis
that, €
know
image
image
cam «
define
image
inforr
be brc
fusior
same
