Image
processing: 3D reconstruction
Principle
To apply one of the
reconstruction techniques, a preliminary step is windowing of the 2D slices. It is
the process of using the measured radiographic densities, i.e. Hounsfield units,
(HU, see CT scan (dual
energy)) to make an
image. The various HU amplitudes are mapped to 256 gray-shades. They are distributed
over a wide range of HU values when an overview of structures is needed. They
can also be distributed over a narrow range (called a narrow window)
centered over the average HU value of a particular structure in order to discern
subtle structural details. This image processing technique is known as contrast compression.
There exist various 3D
reconstruction techniques. Briefly some are mentioned. Of some of them, details
can be found in. Important are
Because contemporary CT scanners offer (nearly) isotropic
(the property of being independent of direction) resolution in space, a
software program can 'stack' the individual slices one on top of each other to
construct a volumetric or 3D image. The image is obtained by taking slices
through the volume in a different plane (usually orthogonal). The procedure is
called multiplanar reconstruction (MPR).
Fig. 1 Typical
screen layout for diagnostic software, showing a 3D (left, torso) and a MPR (right,
vertebrate column) view.
A special projection method maximum-intensity projection (MIP) or minimum-intensity
projection (mIP) can be used to build the reconstructed slices. With MIP a
transparent 3D view is visualized, which can be rotated.
With rendering
techniques boundaries or volumes can be highlighted.
Segmentation is the technique to separate kinds of tissues, for instance with head MRI
skull, liquor and brain from each other.
Application
3D-reconstruction
techniques are basic tools for all imaging techniques in medicine,
so it is not limited to CT (see CT scan (dual energy)), but also used
for PET, SPECT,
MRI, fMRI
and optical imaging techniques such as thermography, diaphanography and optical mammography. They all rely on windowing as the preliminary step.
To explain how windowing works in practice some examples are given.
Windowing
liver and bones
For example, to evaluate the abdomen in order to find
subtle masses in the liver, a good liver window is 70 HU as an average with the
shades of gray distributed over a narrow window of ± 85 HU. Any HU value below
-15 would be pure black, and any HU value above 155 HU would be pure white in
this example. Using this same logic, bone windows would use a wide window since bone contains dense cortical
bone as well as the low dense fatty marrow. The process will take between some
minutes and one hour.
Visualization
of spines
Axial images through the spine will only show one
vertebral body at a time and cannot reliably show the intervertebral discs. By
reformatting the volume with MPR it is possible to visualize the position of
one vertebral body in relation to the others.
More
Info
A maximum
intensity projection (MIP) is a computer visualization method for 3D
data that projects in the visualization plane the voxels (a voxel is the 3D
analog of the 2D pixel) with maximum intensity that fall in the way of parallel
rays traced from the viewpoint to the plane of projection. This implies that
two MIP renderings from opposite viewpoints are symmetrical images.
This technique is computationally fast, but the 2D
results do not provide a good sense of depth of the original data. To improve
the sense of 3D, animations are usually rendered of several MIP frames in which
the viewpoint is slightly changed from one to the other, thus creating the
illusion of rotation. This helps the viewer's perception to find the relative
3D positions of the object components. However, since the projection is orthographic the viewer cannot distinguish
between left or right, front or back and even if the object is rotating
clockwise or anti-clockwise.
Retrieved from "http://en.wikipedia.org/wiki/Maximum_intensity_projection"
where also an animation can be found.
Modern software allows reconstruction in non-orthogonal
(oblique) planes so that the optimal plane can be chosen to display an
anatomical structure (e.g. for bronchi as these do not lie orthogonal to the
direction of the scan).
For vascular imaging, curved-plane
reconstruction can be performed. This allows bends in a vessel to be
'straightened' so that the entire length can be visualized on one image. After
'straightening', measurements of length and diameter can be made, e.g. for
planning surgery.
Since MIP reconstructions enhance areas of high
radiodensity, they are useful for angiography.
mIP reconstructions enhance air spaces, so they are useful for assessing lung structure.
3D rendering techniques
Surface rendering A threshold value of radiodensity is chosen by
the operator (e.g. for bone). A threshold is set, using edge detection image
processing algorithms. In edge detection, small groups of pixels (e.g. forming
a line piece) with strong intensity differences from neighboring ones are marked.
In this way, finally 3D-surface is extracted and a 3D model with contour is
constructed. (Actually, complicated mathematics is used to realize edge
detection.) Multiple models can be constructed from various different
thresholds, allowing different colors to represent each anatomical component
such as bone, muscle, and cartilage. Surface rendering only displays surfaces
meeting the threshold, and only displays the surface most close to the
imaginary viewer. Therefore, the interior structure of each element is not
visible.
Volume rendering In volume rendering, transparency and colors
are used to allow a better representation of the various structures within a volume,
e.g. the bones could be displayed as semi-transparent, so that even at an
oblique angle, one part of the image does not conceal another.
Fig.2 Brain
vessels reconstructed in 3D after bone has been removed by segmentation
Where different structures have similar radiodensity, it is
impossible to separate them with volume rendering. The solution is called
segmentation, a manual or automatic procedure that can remove the unwanted
structures from the image.
After using a
segmentation tool to remove the bone of the skull, now the previously concealed
vessels can be demonstrated.