Image Processing Reference
In-Depth Information
24.3 Pre-processing
3D ultrasound is often employed in clinical diagnostic imaging. If a dedicated 3D
probe is unavailable, 3D volumes can be acquired using freehand ultrasound systems;
a 2D probe with an attached tracker which places and orients the 2D images in 3D
space. Volume compounding consists of two levels: acquisition and reconstruction.
Precise reconstruction requires calibration of the tracking system and correction of
pressure-induced artifacts from the probe onto the skin.
Ultrasound allows for extracting more information, such as tissue strain. Strain is
a tissue-deformation property and can be used to detect functional deficiencies, e.g.,
frommyocardial infarction. Strain determination via tissue tracking is a complex task
and can be done by using tissue Doppler [
27
]. Deprez et al. advanced in 3D strain
estimation by providing a better out-of-plane motion estimation [
12
]. Visualization
of strain has, however, stagnated compared to the development of technology and is
mostly depicted by elementary color coding.
For freehand ultrasound systems, it is necessary to calibrate the position and ori-
entation of the 2D image with respect to the tracking sensor. Wein and Khamene
proposed to make two perpendicular sweeps through tissue containing well-visible
structures [
79
]. They used an optimization strategy to maximize the similarity be-
tween two volumes reconstructed from each sweep.
To achieve the best possible quality of scans, the clinician presses the probe against
the body. However, the human factor causes a non-constant pressure and different
deformations of underlying structures in the body. Prager et al. correlated images
in the sequence and used a rigid translation in the
x
and
y
directions followed by a
non-rigid shift in depth
z
[
58
].
Ultrasound acquisition takes place in polar coordinates
(φ,
R
)
for 2D or
(φ, ψ,
R
)
for 3D. The angles
correspond to the azimuth and elevation angles of the
beam and
R
is the depth of the tissue boundary which has reflected the echo. In order
to use off-the-shelf 3D volume rendering techniques, the grid must be scan-converted
to a Cartesian lattice. This can be done as a preprocessing step or on-the-fly directly
at the rendering stage.
This section is dedicated to selected methods for
volume reconstruction
from
scan-converted freehand ultrasound and for
data enhancement
tailored for ultrasound
volumes, which in the pipeline typically follow the reconstruction stage.
φ
and
ψ
24.3.1 Reconstruction
Volume reconstruction from a set of 2D images needs to solve several important
problems. Each image must be inserted precisely into the right spatial context.
Space-filling between individual images is also crucial and the high framerate of
2D ultrasound implies speed requirements.
Search WWH ::
Custom Search