ArticlePDF Available

Abstract and Figures

Medical imaging domain is a field of interest which concentrates great efforts to offer software tools for the assisted diagnosis. Segmentation, 3D reconstruction and visualization are techniques that allow physicians to observe possible pathological structure inside of the human body. Computing the volume of such a structure offers critical information in the evaluation of the disease gravity. In this paper we present a method for 3D reconstruction of the pathological zone and propose a method for volume computing. We consider this as a start for developing a complex software tool for Deep Brain Stimulation procedure in the neurosurgery domain. I. INTRODUCTION Medical imaging is a set of techniques and methodologies used to process images of the human body for clinical purposes. Some branches may be used for surgical planning, monitorization and navigation of different diseases. Every year, millions of images of patients are taken, in different dimensions and sizes. Most of them are 3D images of patients in order to assist in therapy and diagnosis. To process this large set of data, fast and accurate visualization methods are needed. For this, we need to provide dedicated computer support and it is necessary to have an in-depth understanding of the computer graphics field and visualization. In Webster's Ninth New Collegiate Dictionary the term visualization is defined as the act or process of interpreting in visual terms or of putting into visual form. This implies vision, which is the primary human sensory apparatus, and the power to process of the human mind. Three terms are defined in this way, [1]: Scientific visualization is a field of computer science that includes the user interface, data representation and processing algorithms. Data visualization refers to representation of data sources in different domains. Information visualization is specific for computer vision and deals with abstract information that exists in a computer. As we can see, the base of the visualization process is the data. First, the data is acquired from some source, then it is transformed using various methods, and then mapped to a form appropriate for presentation to the user. Finally, the data is rendered or displayed, completing the process. This process can be repeated and new models can be developed. Medical image data are usually represented as a stack
Content may be subject to copyright.
Volume 52, Number 3, 2011 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
_______________________________________________________________________________
Manuscript received September 2, 2011; revised October 21, 2011
18
3D RECONSTRUCTION AND VOLUME COMPUTING IN MEDICAL
IMAGING
Ligia-Domnica CHIOREAN, Teodora SZASZ, Mircea-Florin VAIDA, Alin VOINA
Technical University of Cluj-Napoca, Baritiu Street no. 26-28, 400027, Cluj-Napoca, Romania,
Phone: +40264401810, Fax: +40264591689
Chiorean.Ligia@com.utcluj.ro, dora.szasz@yahoo.com, Mircea.Vaida@com.utcluj.ro, Alin.Voina@com.utcluj.ro
Abstract: Medical imaging domain is a field of interest which concentrates great efforts to offer software tools for the assisted
diagnosis. Segmentation, 3D reconstruction and visualization are techniques that allow physicians to observe possible
pathological structure inside of the human body. Computing the volume of such a structure offers critical information in the
evaluation of the disease gravity. In this paper we present a method for 3D reconstruction of the pathological zone and
propose a method for volume computing. We consider this as a start for developing a complex software tool for Deep Brain
Stimulation procedure in the neurosurgery domain.
Keywords: volume/ surface rendering, multi-planar reformatting, DICOM files, segmentation
I. INTRODUCTION
Medical imaging is a set of techniques and methodologies
used to process images of the human body for clinical
purposes. Some branches may be used for surgical
planning, monitorization and navigation of different
diseases.
Every year, millions of images of patients are taken, in
different dimensions and sizes. Most of them are 3D
images of patients in order to assist in therapy and
diagnosis. To process this large set of data, fast and
accurate visualization methods are needed. For this, we
need to provide dedicated computer support and it is
necessary to have an in-depth understanding of the
computer graphics field and visualization.
In Webster’s Ninth New Collegiate Dictionary the
term visualization is defined as the act or process of
interpreting in visual terms or of putting into visual form.
This implies vision, which is the primary human sensory
apparatus, and the power to process of the human mind.
Three terms are defined in this way, [1]:
Scientific visualization is a field of computer science
that includes the user interface, data representation and
processing algorithms.
Data visualization refers to representation of data
sources in different domains.
Information visualization is specific for computer
vision and deals with abstract information that exists in a
computer.
As we can see, the base of the visualization process is
the data. First, the data is acquired from some source,
then it is transformed using various methods, and then
mapped to a form appropriate for presentation to the user.
Finally, the data is rendered or displayed, completing the
process. This process can be repeated and new models can
be developed.
Medical image data are usually represented as a stack
of individual images, stored in many medical formats, the
common one being the DICOM format. When we develop
an application it is important to study how this format is
processed by the tools that application is based, because
in this way we can guarantee the accuracy of the data.
Many tools can store the DICOM format slices in other
formats to be viewed, and in this case the files can be
compressed, so we can lose important information.
Using the best suited format for the files, we can
render the graphical data into images and if we want, to
reconstruct a 3D volume from these slices. In the context
of medical volumetric data we can use either direct or
indirect volume rendering. Starting from the rendered
image data, we can outline regions of interest from each
slice. Using these selections, a 3D model may be
obtained. It is important to numerically calculate the
surface of the region of interest (ROI) on each slice and
the volume of the obtained 3D structure, to allow a
quantitative evaluation of the structure, which can be a
tumor.
In neurosurgery domain, dependent to medical
visualization, the problem of predicting behavior from the
brain activity has attracted considerable attention over the
past decades. If we are referring especially to brain
activity, two main types of signals are most considered:
the spikes (extracted from a single neural cell) and the
local field potentials (LFPs) (viewed as a sum of multiple
neurons activity).
LFPs refer to the low frequency component of the
recorded neural activity, and have been shown to carry a
lot of useful information. These signals are extracted
within human brain using Deep Brain Stimulation (DBS),
an invasive neurosurgical procedure used in the treatment
of severe forms of epilepsy, Parkinson’s disease,
obsessive-compulsive disorder and depression.
Combining medical visualization with signal
Volume 52, Number 3, 2011 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
_______________________________________________________________________________
19
processing is a domain which concentrates special efforts
from researchers.
II. 3D VISUALIZATION SYSTEMS
In 3D medical image modeling and visualization there are
many systems and software tools which can be used.
Visualization Toolkit (VTK) and the Insight
Segmentation and Registration Toolkit (ITK) are
frequently used in 3D medical image visualization, they
being also used as a solid foundation for many other
systems. These two software tools are Kitware company
products, and besides these, they also have some good
visualization tools that include powerful interfaces:
ActiViz, VolView, MIDAS, CDashPro, ParaView and
IGSTK. Other tools, which are not belonging to Kitware,
and need to be mentioned, are: MIPAS, VolumePro100,
Vizard II, SepINRIA, VisTools, SciRun, Amira,
MeVisLab, MITK, VolV, BioImageSuite and 3D Slicer.
The most of these tools include powerful interface and
interactive systems for volume visualization and are
designed to be flexible for different purposes. Some of
them are briefly described in this chapter.
The MIPAS system produces a photo realistic face skin
surface model by using texture mapping. Thus, it is
possible to see the whole human head on 3D surface
models completely with outer face and inner brain tissues,
[2].
Another system, VolumePro1000 is an accelerator
based system that uses its hardware-implemented ray-
casting algorithm. The system is based on a typical
graphics accelerator that uses “volume texture” feature
and pixels shaders.
Vizard II is a volume rendering system based on a
reconfigurable ray casting implementation. The high
image quality is guaranteed by enabling the use of
different gradient filters, pre-integration to avoid common
slicing artifacts, and high precision color and opacity
accumulation, [3].
SepINRIA is a free software dedicated to Multiple
Sclerosis (MS) patient brain MRI analysis. The aim of the
product is to provide to clinicians tools that allow
analyzing MS (Multiple Sclerosis) brain images, [4].
VisTools is a suite of cross-platform interoperable
toolkits for analyzing, rendering, visualizing, and
interacting with very large data sets from a variety of
scientific disciplines, [5].
3D Slicer is a tool widely used in clinical research,
especially for image-guided therapy, because of its ability
to visualize collections of MRI data. It has also some
segmentation options.
ParaView – is an instrument built using VTK libraries.
The main feature is that it is able to analyze extremely
large datasets due to its power to distribute memory
computing resources. If it is used on supercomputers it
can analyze datasets of terascale, [5].
BioImageSuite – it belongs to National Institutes of
Health (NIH) from Yale and it comes with a great support
for the beginners in medical visualization field using VTK
and Tcl. It offers image analysis in different domains and
supports manual segmentation and registration, [6].
Most of the presented software instruments were
developed for general 3D visualization. They also include
the most used image processing techniques in the field of
medical visualization, like: zooming, padding, contour
extraction, and the interaction with the volume. Many of
these software tools are developed under the VTK
(Visualization Toolkit) software package, able to
implement visualization data structures, algorithms and
image processing, [5].
III. SEGMENTATION OF SLICE IMAGES
Segmentation of medical images is used in all medical
domains, being a great tool for identifying and outlining
relevant anatomical structures.
The main task of the segmentation is to compute a
map, called label volume, which classifies pixels in an
image or voxels in a volume. Based on this purpose,
different algorithms and methods for segmentation have
been implemented in recent years. Even if we are talking
about manual-interactive or semi-automated segmentation
methods, there exist different techniques used to improve
the segmentation process.
The most general and robust method for image
segmentation is the manual selection of the relevant
structure, on each slice of the radiological data. This type
of segmentation is commonly used when we deal with
low contrast images and hard-to-select shapes. Its
drawbacks: time-consumption and non-reproducibility
make semi-automatic segmentation methods to be
preferred, but sometimes tumor segmentation is preferred
to be done manually.
A very common approach of image segmentation is
based on region growing algorithms, [7]. The principle is
to select seed pixels from the target structure, and then
aggregate successively neighboring pixels to obtain a
connected volumetric region. We can include pixels in the
volumetric region by verifying if the intensity values of
the pixels remain in a specific threshold interval. The
most common application of region growing is the
segmentation of the vascular structures, having the
capability for improved vessel segmentation. There were
developed advanced region growing methods, by
considering a second criterion for obtaining the
volumetric region, or by developing adaptive threshold
intervals.
Threshold-based segmentation, [3] is a simplified
method of region growing, implying the generation of a
binary image based on a threshold or an interval of a
lower and upper threshold applied to the image intensity.
The most typical application of this type of segmentation
is the bones extraction in the CT data, but for an accurate
detection of the bones, the method must be improved
using a connected component analysis (CCA) approach.
Another important region-based segmentation method
is watershed segmentation, [3] based on the idea of
considering an image as a landscape topology with water
flowing into low collecting basins. As the water fills these
low basins, they will merge together into larger basins
(called catchment basins), which form the regions.
Watersheds are the border lines that separate basins from
each other. A watershed transform decomposes the entire
image and assign each pixel either to a region or a
watershed. The drawback of this method is
oversegmentation, encountered in the noisy medical
image data, but there exist different techniques (criteria
for merging regions) for avoiding this problem.
A special segmentation method is livewire method,
because instead of directly generating a target region (as
in the region growing methods), it performs edge
detection. This method extracts the contours of interest
Volume 52, Number 3, 2011 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
_______________________________________________________________________________
20
from each slice of the medical dataset. Livewire uses
Dijkstra’s graph search algorithm to compute the minimal
cost paths and based on these paths, it selects the edges of
the region of interest. In order to apply the search
algorithm, each slice must be represented as a graph,
where the vertices represent the image pixels and the
edges represent the costs between neighboring pixels.
Because of the graph representation of each slice, the
livewire method is not suited for larger 3D dataset, being
time-consuming even for a minimum cost function. Then,
it can be imprecise in defining the contours in each slice.
In order to eliminate these drawbacks, some interpolation
methods were developed for this segmentation algorithm.
Besides traditional segmentation methods, there exist
also methods used for specific segmentation tasks, called
model-based segmentation methods. In order to be
implemented, we need to have information about the size
and the shape of the region to be segmented. One
approach of the model-based segmentation is the active
contour models, being a deformable model, where the
initial contours are deformed towards edges in the image,
[3]. The most common application is to approximate the
boundary shape of a region of interest, under the
assumption that the boundaries are smooth.
We must pay attention that in segmentation not only
the development of strong and fast algorithms must be
considered. The interaction aspects, referring to the
accuracy and the simplicity in using a segmentation
application, must be taken into account. They take into
account the strength and the weaknesses of human beings
and allow human-computer collaboration for a given
application. Interaction methods were developed for every
segmentation method, being a support for developing
advanced segmentation applications. Then, we must
consider also some postprocessing methods in order to
correct the small errors that may appear and enhance the
visualization of the result.
In order to perform segmentation, there exist several
open source software tools. The ITK (Insight ToolKit)
includes a basic set of algorithms implemented in C++
programming language that can be used to develop a
segmentation application, based on its segmentation
filters; MITK used for manual segmentation; ITK-
SNAP – use level-set methods to combine manual and
semi-automatic segmentation, [7].
IV. VISUALIZATION IN MEDICAL IMAGING
Medical visualization is primarily based on 3D volume
data. The interaction facilities must be focused on 3D
interaction techniques which allow the immediate
exploration of 3D data.
The volumetric visualization methods can be divided
in two categories: surface rendering and volume
rendering. If we consider surface rendering, the most
common method is marching cubes. This method
produces polygonal approximations of the real edges,
known as isosurface. On the other hand, the volume
rendering techniques display the volumetric data as a
colored semi-transparent fog, [9].
Volume rendering visualization methods make
possible obtaining of a 3D good perception, by integrating
the 2D slices in a volume. That volume can be seen from
any angle and with shadows effects, with small variations
in density and opacity. There exists a variety of modes for
presentation of the volumetric data on the display which
will be discussed in the following.
The first step in medical imaging is the visualization of
the data. This step is very important in obtaining the
results of an experiment and in testing different methods
that can be beneficial for a patient. The volumetric data is
composed of a very large number of individual voxels.
There exist two main methods of volume rendering in
medical visualization, [8]: indirect volume rendering and
direct volume rendering.
1. Indirect volume visualization
Two methods are common used in indirect volume
visualization:
- Plane-based volume rendering: we can easily
visualize a plan aligned with the cuboid of the volume,
and we obtain the three views: axial, coronal and sagittal.
- Surface-based volume rendering: to represent a
material surface of the object. The resulting surface is
called an isosurface and the value used for extraction of
this surface is called isovalue. The common methods used
for contour extraction are the contour tracing and
marching cubes.
2. Direct volume visualization
The method consists in the visualization of the volume
using all the voxels. Despite there exist many volume
rendering algorithms, classified as image space
approaches and object-space approaches, four techniques
are the most common ones, [8]:
- The classic direct volume rendering technique is
ray casting, in which the rays are casted from the eye or
viewpoint through the image-plane and through the
dataset. The ray samples the volumetric dataset along its
path and depending on the specified properties, makes
parts of the volume visible or not.
- In splatting, a voxel can be discarded if the voxels
surrounding it in the direction of the viewing ray have
“high” opacity. Splatting methods for volume rendering
are substantially faster than ray-casting.
- Shear warp is an optimized version of the ray
casting algorithm, developed by Cameron and Undrill. It
simplifies the volume sampling to reduce the memory
access costs, by transforming the viewing transformation
such that the nearest face of the volume becomes axis
aligned with an off-screen image buffer with a fixed scale
of voxels to pixels. The volume is than rendered into this
buffer using the far more favorable memory alignment,
fixed scaling and blending factors. After the slices are
rendered, the buffer is warped into the desired orientation
and scaled in the displayed image.
- Texture mapping method is based on the texture-
mapping support of computer graphics hardware. The
dataset is loaded into texture memory and is resampled by
the texture-mapping functions of the graphics hardware
and mapped to a rasterized polygon on the screen. The
main disadvantage of this technique is that if the dataset
does not fit into main memory, need to be swapped. There
exist different solutions to this problem, which are based
on considering the reorganization of the volume dataset.
The main technologies used for 3D rendering are:
OpenGL, Direct3D and Java3D. There are software tools
and packages developed to manipulate and visualize the
image data, such as: VTK (Visualization Toolkit), H3D
API, 3D-Doctor, 3DVIEWNIX, and ANALYZE.
Volume 52, Number 3, 2011 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
_______________________________________________________________________________
21
V. 3D RECONSTRUCTION AND VOLUME
COMPUTING
Using Dicom images we propose a method for 3D
reconstruction of the investigated part of body and the
representation of the pathological volume inside of it. A
volume computing method is also proposed, in order to
estimate the dimension of the pathological anatomic
structure. We implemented some software modules for
improved visualization of dicom images, ROI selection
and 3d reconstruction.
3D visualization of medical images brings a plus of
information in diagnosis process and could be useful
when a pathological structure has to be extracted. For this
purpose we propose a method for 3D reconstruction from
Dicom images combined with a 3D reconstruction from
the segmented images and the visualization of the both
volumes in the same scene. For 3D reconstruction and
visualization we implemented an application which uses
vtk library (Visualization Toolkit).
The first step is the visualization of the 2D images.
There exist different methods for displaying Dicom
images. There are software applications that allow
transforming these images in a common format. Different
toolkits may also be used to create a Dicom viewer.
For the visualization of 2D images there are some
parameters that can be established, such as: scale,
window, level. In the left side of the image in Figure 1 we
can visualize the parameters that can be adjusted in our
implemented application for both 2D and 3D visualization
of the dataset.
Figure 1. 2D visualization of a CT axial slice.
Scale parameter is used for zooming. A filter
implemented in vtk is used to realize the magnification,
based on linear interpolation.
The Window/Level values determine the brightness
and contrast of the displayed slice. A high value for the
Window parameter and a low one for Level parameter
lead to a flat distribution of the grey values and even the
boundary of the slice cannot be detected. In this case it is
very hard to separate the bone from tissue or air.
In contrast, if a low value for Window and a high
value for Level are used, we lose the information
regarding the tissue and the bones are easy to be
distinguished.
It is very important to set the Level and Window
values of a slice in the most appropriate way, depending
on the region of interest. Otherwise, wrong interpretations
may arise, this being critical in analyzing a medical
image.
In the header of each Dicom file there exist
information about the Window and the Level values and
the range of these values. These parameters may be
obtained using the GetPointData(), GetScalars() and
GetRange() methods of the vtkImageData class.
In order to reconstruct the pathological zone, on every
2D images must be selected the region of interest. An
interactive method is proposed, because that region has to
be selected as accurate as possible. This method realizes
the marking of some contour points. A line is traced
between two adjacent points and after the last point was
selected, the closed natural cubic spline that interpolates
those points is calculated and displayed. Our module
application allows to display every slice and to realize the
selection with the mentioned method. Such a selection is
shown in Figure 2, where we can see the bright outlined
region in the mouth position of the human head.
Figure 2. Region of interest selection.
The 3D reconstruction is realized using surface-based
and plane-based volume rendering. Surface rendering
technique allows visualizing the obtained structure with a
given opacity. To observe the internal structure
represented inside, the opacity of the external volume has
to have a low value. For surface rendering we used
marching cubes mechanism, represented in vtk by some
classes: vtkContourFilter, vtkMarchingCubes,
vtkImageMarchingCubes. The instances of these classes
represent contour filters which can lead to isosurfaces.
We used the same filter for the structure reconstructed
from selections as well as for the entire investigated part
of the human body. In order to obtain a smoother
structure from the selection, a surface smoothing may be
applied before filtration. The vtkSrink3D class of the vtk
toolkit allows smoothing by a subsampling on a uniform
grid. Because the two surfaces are displayed with 2
actors, is very important to place them correctly in the
scene.
A method for 3D visualization is Marching Cubes,
which is based on isocontouring techniques. These
techniques are used to extract the skin and bone surfaces
and to display orthogonal cross-sections to put the
Volume 52, Number 3, 2011 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
_______________________________________________________________________________
22
isosurface in context. In [6] is defined that a density value
of 500 will define the air/skin boundary, and a value of
1150 define the soft tissue/bone boundary.
The main steps in representing the isocontour are:
- Read the input;
- For each anatomical feature of interest, create an
isosurface;
- Transform the models from patient space to world
space;
- Render the models.
For reading medical images, Dicom files or .vtk files
may be used, as well as other image common formats. In
order to transform Dicom files into other image format,
different software application or tools may be used. For
instance, we used dicom2 application. This is a free
command-line driven program which allows us to convert
medical images and DICOM files to various other formats
(.bmp, .png, .raw), while performing some rudimentary
image processing tasks. The Figures 3 and 4 show the 3D
reconstruction results obtained using Dicom files and .raw
files.
As can be observe in Figure 3 and Figure 4, there
exists a visible difference between the two methods, even
if we choose the same value of the density and the same
properties for the isosurface. This difference arises
because if we read Dicom files using vtkDicomReader the
slices are saved as JPEG 16 bit compressed, which cuts
the higher frequencies in the images. The conversion
from Dicom to .raw let the image uncompressed, which
leads to better resolution. An improvement for using
Dicom files can be achieved by performing histogram
normalization. We observe that there is a great difference
between these two types of visualization and it is crucial
to be taken into account when we want to develop a
strong reconstruction instrument tool, this problem being
usually neglected by the software developers.
Figure 3. Skin isocontour extraction from a CT dataset
saved as Dicom files.
Figure 4. Skin isocontour extraction from a CT dataset
saved as .raw files.
Usually we can provide more context of the volume if
we are able to see all the orthogonal planes (axial, coronal
and sagittal), and to visualize the volume and the tumor in
correlation with the slices used to obtain these volumes.
In this case we can visualize just a part of the volume, the
one generated between the selected slices.
We implemented this possibility by using plane-based
rendering technique, available in vtk by using the
vtkImageActor class that is able to combine a
quadrilateral, polygon plane with a texture map. In the
Figure 5 we can visualize the result of applying the plane-
based method.
Figure 5. Visualization of the orthogonal planes and the
skin isosurface.
The pathological structure and the investigated
anatomical part have to be displayed in the same 3D
scene. The volume of the former offers important
information for the physicians.
We propose a method for calculating the volume of a
3D structure. Different image’s attributes may be read
from Dicom file. The PixelSpacing attribute give us the
distance between pixels. This attribute offers the distances
on rows and columns between adjacent pixels’ centers.
For every 2D image the area of the selected region may
be calculated based on the number of the pixels in that zone,
using relation (1):
Volume 52, Number 3, 2011 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
_______________________________________________________________________________
23
=
=
1
0
)1__(**
n
i
ik linepixelsnryxA (1)
where:
Ak – area of the selected zone on the k slice
x, y distances between adjacent pixels’ centers on
rows and columns
nr_pixels_linei - the number of pixels on line “i”
n – the number of pixels’ rows on the selection
The thickness of a slice may be obtained from
SliceThickness tag. For the distance between slices there
exist an attribute named SpacingBetweenSlices, but it
may be absent in Dicom file and appears only for MRI
images. Moreover, some vendors do not establish in this
tag the distance between slice’s centers, but the distance
between their margins. We propose to establish the
distances d1 and d2 from two successive slices to the
same origin, on the plane normal direction. The distance
between slices is the difference between these two
distances. Using this information the volume is calculated
with relation (2):
2
*
2
** 1
1
1
g
A
g
AAhV m
m
k
k++=
=
(2)
where:
V – the structure volume
h – the distance between two successive slice’s centers
m - the number of slices
Ak - the area of slice “k”
A1 – the first slice’s area
Am - the last slice’s area
g – the thickness of a slice
The proposed methods are integrated in our software
module developed for ROI selection.
In the software component that we implemented for 3d
reconstruction and visualization, the volume can be
displayed with transparency, so the tumor can be
observed. In Figure 6 the dark-grey structure located
inside the rendered volume of the head represents the
pathological volume. It is the volume reconstructed from
selections. The entire volume can be rotated, to see the
tissues and the tumor from different angles and
perspectives.
Another important aspect in 3D medical image
visualizing is the multiplanar reformatting, [10]. This
technique allows visualizing the structure on the main
planes used in medicine: axial, coronal and sagittal. To
extract a slice in the desired orientation vtkImageReslice
class may be used, from VTK toolkit.
In order to visualize the 2D projections of the volume
on the three main planes, we have placed three cutter
planes which intersect the volume in those directions. By
moving a plane, in the correspondent right window is
displayed the image reformed by intersection of that plane
with the volume.
Figure 6. Visualization of the pathological structure and
the investigated anatomical part in the same scene
For a more natural visualization mode of the two
volumes (human part of the body and the tumor), direct
volume visualization methods were implemented by using
vtk classes such as vtkVolumeRayCastCompositeFunction
and vtkVolumeRayCastMapper, together with some
properties that were added to the volume and some
cropping methods. An example of this kind of
visualization is presented in Figure 7. The surgeon is able
to explore into the volume and to extract some
information that can be used in taking decisions regarding
the treatment of the affected area.
Figure 7. Direct visualization of the volume
VI. CONCLUSION AND FUTURE WORK
Medical images are very important in assisted diagnosis
and for the surgery domain. Useful information may be
provided by visualization in 2D space, as well as in the
tridimensional one.
We offer a method to represent in 3D space a
pathological structure, related with the surrounding
tissues. For this purpose a selection has to be made on
every image slice. The proposed method for ROI
selecting may be time consuming, but assures a good
accuracy.
Volume 52, Number 3, 2011 ACTA TECHNICA NAPOCENSIS
Electronics and Telecommunications
_______________________________________________________________________________
24
The volume of the structure being an important factor
in medical evaluation, we offer a method for volume
computing.
The multi-planar reformatting may be useful for
observing how the tissue is affected in different
directions.
Considering the proposed methods as a start, a
neurosurgery navigation software tool can be developed
for different medical techniques. Using the manual
segmentation is well suited for such an application, to
select the affected structure of the brain. This can be done
in either Dicom files or on the specialized human brain
atlases that already exist. Then, it is important to visualize
both the entire brain volume and the pathological part,
where the electrode must be placed. After the
visualization of these two parts, a navigation system can
be developed to move the electrode in the target position.
Based on the LFP processing techniques, we can record
the neural LFP signal and determine the best position of
the electrode inside the brain.
As future work we will extend this application to make
possible the integration of signal processing techniques
needed for LFPs. This will be helpful in the evaluation of
the neural activity in a certain region of the brain.
Furthermore, we will develop a navigation system for
DBS, able to move the electrode in the affected structure
of the brain and to take some decisions regarding the
exactly location of the electrode.
REFERENCES
[1] W. Schroeder, K. Martin, B. Lorensen, The Visualization
Toolkit (3rd edition): An onject-Oriented Approach to 3D
Graphics, Kitware Inc., Pearson Education Inc., New Jersey,
2003.
[2] S. Dogan, “3d reconstruction and evaluation of tissues by
using ct, mr slices and digital images”, Proceedings of 20th
ISPRS Congress, Istanbul, Turkey, July 12-23, 2004, vol. 35, pp.
323-327.
[3] M. Meißner, U. Kanus, G. Wetekam, J. Hirche, A. Ehlert,
W. Straßery M. Doggettz, P. Forthmann, R. Proksa, “VIZARD
II: A Reconfigurable Interactive Volume Rendering System”,
Proceedings of the ACM SIGGRAPH/EUROGRAPHICS
conference on Graphics hardware (HWWS’02), Saarbrucken,
Germany, September 1-2, 2002, pp. 137 – 146.
[4] P. Fillard, J.-C. Souplet, N. Toussaint, “Medical Image
Navigation and Research Tool by INRIA (MedINRIA 1.9),
Tutorial v2.0”, INRIA Sophia Antipolis - Research Project
ASCLEPIOS, October 1, 2009.
[5] C. R. Johnson, C. D. Hansen, The Visualization Handbook,
Elsevier Inc., Burlington, MA, USA, 2005.
[6] X. Papademetris, A. Joshi, “An Introduction to Programming
for Medical Image Analysis with the Visualization Toolkit”,
2009, [Online],
[Available: http://www.bioimagesuite.org/vtkbook5/index.html
[Accessed: July 12, 2011].
[7] W. K. Pratt, Digital Image Processing 4nd Edition, John
Wiley & Sons, Inc., Los Altos, California, 2007.
[8] B. Preim, D. Bartz, Visualization in Medicine. Theory,
Algorithms, and Applications, Elsevier Inc., Burlington, MA,
USA, 2007.
[9] S. Somaskandan “Visualization in 3D Medical Imaging”,
2006, [Online], Available:
http://wenku.baidu.com/view/c57629acdd3383c4bb4cd212.html
[Accessed: July 30, 2011].
[10] L. Chiorean, M.-F. Vaida, “3D Rendering of Radiological
Images using Java Secure Technologies”, IFMBE Proceedings
(MediTech2009), Cluj-Napoca, Romania, September 27-29,
2009, vol. 26, pp. 257-260.
... This problem is important to the scientific community because of its numerous applications in diverse fields. For example, in the medical industry, 3D reconstruction may aid in diagnosis and treatment by reconstructing a patient's internal anatomy from 2D scans [Chi+11]. In robotics, depth perception and understanding of the three-dimensional structure of the scene may be necessary for a robot to navigate its environment successfully. ...
Thesis
Performing 3D reconstruction from a single 2D input is a challenging problem that is trending in literature. Until recently, it was an ill-posed optimization problem, but with the advent of learning-based methods, the performance of 3D reconstruction has also significantly improved. However, the state-of-the-art approaches mainly focus on datasets with highly textured images. Most of these methods are trained on datasets like ShapeNet, which contain rendered images of well-textured objects. However, in natural scenes, many objects are textureless and challenging to reconstruct. Unlike textured surfaces, reconstruction of textureless surfaces has not received as much attention mainly because of a lack of large-scale annotated datasets. Some recent works have also focused on textureless surfaces, many of which are trained on a small real-world dataset containing 26k images of 5 different textureless clothing items. Transparent surfaces have received even less attention from the deep learning community, with most works using traditional computer vision methods to reconstruct these surfaces. Most techniques depend on inferring the shape of the objects by how light is reflected off the surfaces. However, this may not be possible in the case of transparent surfaces as they allow some light to pass through them, and the algorithms now have to deal with light refraction and absorption in addition to reflections. To facilitate further research in this direction, we present a synthetic dataset generation strategy for images of both textureless and transparent objects and corresponding depth maps and surface normals map groundtruth. We also make available three new datasets: a large synthetic textureless dataset containing 364k samples and 2635 3D models, a small real-world textureless dataset containing 4k samples and six objects, and a large transparent object dataset containing 126k samples and ten 3D models. We also propose an autoencoder-based network for learning to reconstruct the depth maps and surface normal maps from a single image for textureless objects. Furthermore, we propose a novel architecture that combines a Vision Transformer with a residual autoencoder and uses an auxiliary silhouette output to find transparent objects in realistic scenes and reconstruct their depth maps and surface normal maps.
... The first will have a trustworthy origin, whereas the second relies on the knowledge of the operator. The data from imaging methods allow automatic reconstruction algorithms, including volume and surface rendering (Chiorean et al. 2011) (described later in this chapter). Though software generates many of the features automatically, there is an important work to be done manually. ...
Chapter
Image-guidance has been the mainstay for most neurosurgical procedures to aid in accuracy and precision. Developments in visualization tools have brought into existence the current microscope and even sophisticated augmented reality devices providing a human–computer interface. The current microscope poses an ergonomic challenge particularly in scenarios like sitting position. Also, the cost associated with the present microscope hinders the accessibility of micro neurosurgery in most low-to-middle-income countries.
... This paper does not intend to survey the field; however, a few papers are relevant for this research (an excellent survey on available software packages is found in [5]). The papers can be grouped into three main categories: (1) visualization toolkits [3], [4], [6], (2) 3D reconstruction [1], [2], [7], [11], and (3) multidimensional views with possible overlays [10], [12]. The literature shows that all methods use images either for visualization or 3D reconstruction (voxel and surface-based). ...
... Medical image reconstruction is mainly based on 3D data and the workstation must support 3D interaction functions which allow the real-time exploration of 3D data [18]. According to segmented results, we obtained the image data of scalp and lesion. ...
Article
Full-text available
Precise craniotomy localization is essential in neurosurgical procedures, especially during the preoperative planning. The mainstream craniotomy localization method utilizing image-guided neurosurgery system (IGNS) or augmented reality (AR) navigation system, require experienced neurosurgeons to point out the lesion margin by probe and draw the craniotomy manually in patient’s head according to cranial anatomy. However, improper manual operation and dither from AR model will bring in errors about craniotomy localization. Additionally, there is no specific standard to evaluate the accuracy of craniotomy. This study attempts to propose a standardized interactive three-dimension (3D) method using orthogonal transformation to map the lesion onto the scalp model and generating a conformal virtual incision in real time. Considered clinical requirements, the incision can be amended by 3D interaction and margin modification. According to the IGNS and the virtual incision, an actual craniotomy will be located on patient’s head and the movement path of probe will be recorded and evaluated by an indicator, which is presented as an evaluated standard to measure the error between virtual and actual craniotomies. After experiment, an incision is drawn on a 3D printing phantom based on the generated virtual one. The results show that the proposed method can generate a lesion-consistent craniotomy according to the size of lesion and mapping angle, and delineate the incision on patient’s head precisely under the IGNS.
... Different approaches and methods of 3D-visualization of biological tissues are discussed depending on the scientific goals and objectives of practical applications [1,2,14]. A significant number of software complexes, libraries and specialized systems of scientific and technological visualization for three-dimensional digital biomedicine have been developed [7,9,12,24,25]. Approaches using medical X-ray tomography are very effective for 3D-reconstruction of tissues/organs [8]. We can also highlight some areas that are associated with the localization of implanted biomaterials [8,26], the colonoscopy, the ultrasonic sounding [9], the stereoscopic fluorescence imaging, the multispectral magnetic resonance image analysis [16], the single photon emission computed tomography (CT) [11], the electron tomography [6], the use of combined methods [9]. ...
Conference Paper
Full-text available
A software for processing sets of full-color images of biological tissue histological sections is developed. We used histological sections obtained by the method of high-precision layer-by-layer grinding of frozen biological tissues. The software allows restoring the image of the tissue for an arbitrary cross-section of the tissue sample. Thus, our method is designed to create a full-color 3D reconstruction of the biological tissue structure. The resolution of 3D reconstruction is determined by the quality of the initial histological sections. The newly developed technology available to us provides a resolution of up to 5 - 10 {\mu}m in three dimensions.
Chapter
Full-text available
Many fields have adopted 3D technologies, and medicine is no exception. Their use ranges from educational purposes to skill training and clinical applications. This chapter proposes a possible protocol related to obtaining 3D anatomical models from Computed Tomography Angiogram (CTA) data and its subsequent 3D printing. We describe relevant features of free software available for this process as an introductory guide to those who want to make their first steps. We briefly discuss some of the benefits and drawbacks of applying 3D anatomy in pedagogical and surgical areas.
Chapter
Hepatocellular Carcinoma (HCC) is the most frequent form of liver cancer, being the fourth leading cause of cancer-related death worldwide. The curative treatment in most cases is the tumor removal from the body (surgery), but more than 70% of HCC patients have advanced tumors and cannot be treated with such procedures. Alternative laparoscopic surgical treatments, such as high dosage radiation-brachytherapy (HDR-BT) or inside-tumor drug release (IDR), are currently researching for tumor size reduction. Our target is to develop computerized methods for assisting the medical robot used in such treatments, to make them safer and more efficient. We build an accurate 3D model of the HCC anatomical context, based on Computed Tomography (CT) images acquired before surgery, putting into evidence the HCC tumor, its position within the liver and the most important blood vessels connected to it. We also highlight, in real time, during surgery, the 2D slice corresponding to the transducer position. In this article, we describe the corresponding software system, focusing on the segmentation and 3D reconstruction techniques, assessed through specific experiments.
Article
Full-text available
In this project, it is provided to reconstruct 3D models of human body by using CT, MR slices and digital images and precisely finding locations of pathological formations such as tumours. For this purpose, within this project we developed a software, which we called as "Medical Image Processing and Analysis System (MIPAS)" and it is still under development for multipurpose medical applications. In this paper, we introduce the abilities of MIPAS briefly and also give a sample application on finding location and visualization of a brain tumour. MIPAS uses volume and surface rendering techniques for D modelling of the tissues and provides both volume and surface models at the same time on the screen when required. MIPAS gives many editing and analysis functionalities to medical doctors. For pre-processing of CT and MR images, there are wide range of image processing functionalities. It is possible to register CT and MR images by using both anatomical landmarks and artificial (external) markers. MIPAS also provides surface registration functions with both rigid body and non-rigid body transformation with the variants of ICP algorithm. For photo-realistic visualization of the external human body like face of the patient, it provides a photogrammetric module. This module consists of self calibration with bundle adjustment by using additional lens parameters, automatic image coordinate measurement with ALSM matching technique and texture mapping functions with both colinearity equations and 3D affine transformation. We are still studying to progress the photogrammetric module for plastic surgery. Up to now, we could not still test the MIPAS on the real patients. But we have just started a new project with medical doctors, in order to test the MIPAS on real patients. During this new project, we will see the actual reliability of the system.
Article
Full-text available
This paper presents a reconfigurable, hardware accelerated, volume rendering system for high quality perspective ray casting. The volume rendering accelerator performs ray casting by calculating the path of the ray through the volume using a programmable Xilinx Virtex FPGA which provides fast design changes and low cost development. Volume datasets are stored on the card in low profile DIMMs with standard connectors allowing both, large datasets up to 1 GByte with 32 bit per voxel, and easy upgrades to larger memory capacities. Per-sample Phong shading and post-classification is performed in hardware, giving immediate feedback to changes in the visualization of a dataset. Adding new features, such as pre-integrated classification, can be accomplished using the existing card without expensive and time consuming redesigns. The card can also be used for medical image reconstruction by reconfiguring the FPGA broadening its usefulness for end users. For the first time, users are able to generate high quality perspective images as required for applications such as virtual endoscopy and colonoscopy, and stereoscopic image generation.
Article
Volume visualization has been driven by the advances of medical image acquisition, since most of the modern imaging devices produce 3D image volumetric data. Usually, projection needs to be employed on the numerical description of a volumetric data, to get desired information. These volumetric data are hard to effectively visualize since 3D structures of the interior of a volume are difficult to derive from viewing individual slices. Volume visualization methods make it possible to give a better 3D perception by integrating these slices together in a volume that can be seen from any viewpoint and containing shading effects, as well as small variations in density and opacity. In this seminar, some significant visualization techniques for medical data will be presented.
Article
The Visualization Handbook provides an overview of the field of visualization by presenting the basic concepts, providing a snapshot of current visualization software systems, and examining research topics that are advancing the field. This text is intended for a broad audience, including not only the visualization expert seeking advanced methods to solve a particular problem, but also the novice looking for general background information on visualization topics. The largest collection of state-of-the-art visualization research yet gathered in a single volume, this book includes articles by a "who's who" of international scientific visualization researchers covering every aspect of the discipline, including: Virtual environments for visualization Basic visualization algorithms Large-scale data visualization Scalar data isosurface methods Visualization software and frameworks Scalar data volume rendering Perceptual issues in visualization Various application topics, including information visualization. * Edited by two of the best known people in the world on the subject; chapter authors are authoritative experts in their own fields; * Covers a wide range of topics, in 47 chapters, representing the state-of-the-art of scientific visualization.
Chapter
The paper presents an application implemented for 3D rendering in the medical domain, especially for radiological images. For deployments we used Java and VTK technologies. The application allows selecting a pathological area on every 2D image, to reconstruct the pathological volume and rendering it in the entire tissue volume. It also offers facilities for multi-planar reformatting. The application is intended to be used for maxillo-facial tumor rendering in 3D space, to evaluate the volume and to assist the diagnosis process. Keywords3D rendering-multi-planar reformatting-DICOM files-volume rendering-surface rendering
An Introduction to Programming for Medical Image Analysis with the Visualization Toolkit
  • X Papademetris
  • A Joshi
X. Papademetris, A. Joshi, "An Introduction to Programming for Medical Image Analysis with the Visualization Toolkit", 2009, [Online], [Available: http://www.bioimagesuite.org/vtkbook5/index.html [Accessed: July 12, 2011].
[9] S. Somaskandan – " Visualization in 3D Medical Imaging
  • B Preim
  • D Bartz
B. Preim, D. Bartz, Visualization in Medicine. Theory, Algorithms, and Applications, Elsevier Inc., Burlington, MA, USA, 2007. [9] S. Somaskandan – " Visualization in 3D Medical Imaging ", 2006, [Online], Available: http://wenku.baidu.com/view/c57629acdd3383c4bb4cd212.html [Accessed: July 30, 2011]. [10] L. Chiorean, M.-F. Vaida, " 3D Rendering of Radiological Images using Java Secure Technologies ", IFMBE Proceedings (MediTech2009), Cluj-Napoca, Romania, September 27-29, 2009, vol. 26, pp. 257-260.