Biomedical Engineering Reference
In-Depth Information
Figure 8.7
3D image processing. (a) Computed tomography setup. The X-ray source and the
detector rotate around the axis, which facilitated the collection of images at different angles. (b)
Reconstruction of 3D images using multiple slices.
emitted sources in the original image. A problem in the back projection is blurring
that occurs in other parts of the reconstructed image. A data filter (a mathematical
function) is used to eliminate blurring. As 2D filtering is computationally intensive,
a better approach is to reverse the order of the filtering and the back projection.
The multiple transmitted beams are registered and back-projected to generate an
“image slice” of high resolution and contrast in digital form. CT systems are usu-
ally configured to take many views of the object, often more than 100 to generate
reconstructed images of good quality, with excellent density discrimination.
All the digital images are combined in a computer using different algorithms to
build a 3D image of any part of the body. A rendering technique is the computer
algorithm used to transform serially acquired axial CT image data into 3D images
made of voxels (volume pixel), similar to pixels. The volume of the voxels is a
product of pixel size and the distance between slices. The brightness of each voxel
is dependent on the intensity of signal from the corresponding location in the ob-
ject. Each voxel has a corresponding pixel pattern in the display panel. A typical
MRI 3D image consists of 256
512 voxels. The voxel is usually con-
siderably larger in the third dimension, as the slice thickness is typically 1-5 mm.
This means that the in-plane resolution is typically 0.5-2.0 mm. One could then
instruct the computer to display 2D slices (each measurement lasts only a fraction
of a second and represents an “image slice” of tissue) from any angle.
There are a number of different rendering techniques, but most can be divided
into two classes: surface-based (binary) techniques and volume-based (continuous)
techniques. Rendering with either technique consists of three steps: volume forma-
tion, classification, and image projection. Volume formation involves the acquisi-
tion of the image data, the stacking of the resultant data to form a volume, and
preprocessing, which varies according to the rendering technique used. Typical pre-
processing includes resizing (by interpolation or resampling) of each voxel, image
smoothing, and data editing (e.g., deletion of the CT table on which the patient was
positioned). The classification step consists of determining the types of tissue (or
×
256 or 512
×
