Digital Tomosynthesis: Theory of Operation

White Paper | DRX-Evolution Plus Digital Radiography System
Technical Brief Series
Digital Tomosynthesis – Theory of Operation
Pending FDA 510(k) clearance
Author: Xiahui (Ed) Wang
Digital Tomosynthesis (DT) is a new radiographic imaging
technique that is revived from the nearly century-old traditional
film-screen tomography. This rejuvenation is all made possible
by the recent advances in high frame-rate, high-sensitivity flat-
panel digital radiographic detector, rapid pulsed-exposure
sequence-capable high-frequency x-ray generator, the widely
available and low-cost computer GPU processing power, and
the precision motion controls built in the digital radiography
system hardware.
Figure 1: An example of DT acquisition. Both wall stand and
table positions are supported by the Carestream DT option.
Carestream Health implements the DT functionality as an
upgradable option on the standard DRX-Evolution Plus Digital
Radiography System (Figure 1). The portable DRXPlus detector
can be placed inside either the wall bucky or the table bucky to
accommodate different exam types and positions.
The major benefit of DT over the traditional film-screen
tomography is that DT greatly simplifies the operator’s
workflow by separating the process of DT exposure acquisition
from image volume formation; it needs only a single sweep of
x-ray exposures and then relies on computer software in post-
processing to generate as many DT slices as needed at any slice
thickness and slice interval.
Background in Traditional Tomography
The traditional tomography acquisition uses synchronized,
reciprocal mechanical motion of the x-ray source and the film
cassette combined with a single, long-duration, continuous x-
ray exposure. Many scan patterns have been developed for the
x-ray source and cassette motion, such as linear, circular,
elliptical, spiral, hypocycloidal, etc., but linear motion is the
most frequently used in practice due to its simplicity.
Figure 2 illustrates how linear tomography works. Before a
tomography procedure, the operator visually estimates the
midline of the anatomical area of interest, determines the slice
thickness requirement for the exam, and then sets the initial
pivot level (fulcral plane) and the sweep angle for the
tomography scan accordingly.
For each tomographic scan the anatomical details in the fulcral
plane (e.g., object O in Figure 2) will be preserved while those
anatomical clutters above or below the fulcral plane will be
blurred or smeared (e.g., object X in Figure 2). The larger the
sweep angle, the more blurred the anatomical clutters
become. As anatomy clutter is the primary source of noise for
radiography, tomography isolates the anatomy of interest in
the fulcral plane from the unwanted clutters from the rest of
the anatomy, and therefore improves contrast-to-noise ratio of
the image details.
Each tomography acquisition uses one single continuous x-ray
exposure, and many acquisitions are frequently needed in
order to cover the anatomical area of interest fully. This would
definitely increase the overall exam time and the total radiation
dose by the total number of scans. As the overall exam time
and the total radiation dose must be limited, the number of

2
tomographic scans is limited, which may limit the smallest
tomographic slice interval (slice pitch) and the spatial resolution
in the depth direction of the tomography volume.
Digital Tomosynthesis
Figure 3(a) illustrated how DT acquisition works. DT operates
based on the same principle as tomography except the
detector can either be mechanically motorized in
synchronization with the x-ray source or simply be stationary
during the exam. The DT implemented by Carestream chooses
to use stationary x-ray detector for the purpose of reducing
motion blur and improving spatial resolution. In addition, DT
uses multiple, low-power, pulsed x-ray exposures instead of a
single continuous exposure. Correspondingly the x-ray detector
in a DT exam will capture each short x-ray pulse as individual
images from different projection angles of the x-ray source.
These projections are used to reconstruct a stack of DT slices
for viewing, of which each slice is equivalent to a traditional
tomogram with fulcral plane set at the center of the DT slice.
Because the image acquisition is separated from the DT image
volume formation process, sophisticated reconstruction
processes are used to maximize the anatomy contrast-to-noise
ratio in the DT slices. Figure 3(b) and (c) shows the “Shift-and-
Add” method used in some of the early DT systems.
Individually captured images are shifted differently and added
Figure 3: Schematic view of DT acquisition (a) and reconstructions of slices at different depth (b) (c).
O
X
X-Ray Source
Motion
O X
X O
X
O
X X X
O
O
O
O X
X O
X
O
O O
X
X
X
O
(a) (b) (c)
Stationary
Detector
Plane A
Plane B
Plane A
Reconstruction
Plane B
Reconstruction
Figure 2: In traditional tomographic imaging, object structure
details (e.g., Object O) in the fulcral plane stays focused
(stationary in the film plane) while objects outside the focal
plane (e.g., object X) become blurred or smeared due to
motion.
O
X
O X X O
O
X
Focal Plane
X-Ray Source
Motion
Synchronized
Detector Motion

3
together in the software to generate DT slices focused at
different depths of the anatomy. The Carestream DT
reconstruction software leverages the state-of-the-art
algorithms that are the same in principle to those applied in
the computed tomography (CT), such as filtered back
projection or iterative reconstruction. As a consequence, the
anatomical details are greatly improved with DT when
compared to tomogram. Figure 4 shows two DT slices of the
pelvis region at different depth locations.
The exam time is greatly reduced with DT than with traditional
tomography, and the operator’s trial-and-error process in
locating the anatomical region of interest is eliminated.
Multiple projections are acquired during a single DT acquisition
sweep. Afterward the anatomical features at different depth
locations are reconstructed mathematically. The total number
of slices and the slice interval can be made arbitrarily large or
small in the software as needed; and they are limited by only
the computer memory capacity, network bandwidth, PACS
storage space, and most importantly the diagnostic needs on
the image resolution.
DT Scout
Each DT acquisition sequence starts with a scout view. The
scout serves two purposes for DT: 1) collimation and anatomy
positioning, and 2) exposure control. The scout view is
captured at the same technique settings (kVp, mAs, etc.) as
with a standard radiography except the projection angle is
always straight on to the detector center and the SID is fixed
(110 cm for table and 180 cm for wall stand). The scout is
processed as a radiographic image and then presented to the
operator to verify the collimation for anatomy coverage as well
as the anatomy positioning. The scout image can be captured
using either automatic exposure control (AEC) or manual
technique settings. The operator would decide on the
appropriateness of the image quality of the scout view for
diagnosis or retake the exposure if adjustment is needed.
DT Sequence Exposure Control
Once the DT scout is captured and accepted by the operator,
its exposure technique will be used as the basis to derive the
total exposure for the DT sequence. This is an important
Figure 4: DT slice examples of the pelvis region at different depth locations.

4
operation step for automatic exposure control of the DT
sequence.
The total exposure level, mAs in particular, of a DT sequence is
set as the multiplication of the DT scout exposure level (mAs)
and the DT Technique Scale Factor that is preconfigured in the
Preference Editor. This is regardless of the total number of
projections and the x-ray tube sweep angle specified. The total
DT mAs is allocated equally to all the DT projections. The other
exposure technique settings between the scout and the DT
sequence, such as kVp and beam filtration, are kept the same.
DT Acquisition Modes and Reconstruction
Resolution
DT projection images can be acquired in either high-resolution
(HR) mode or high-speed (HS) mode. In the HR acquisition
mode, the detector works in its native resolution (i.e., no pixel
binning) and acquires images at 4.5 frames per second (fps); in
the HS acquisition mode the detector uses 2x1 pixel binning
and acquires images at 9.0 fps due to 50% reduced data rate.
The primary usage of the HR mode is for musculoskeletal
exams (hand, foot, knee, etc.), when the patient motion
artifact due to the longer acquisition time is limited and it
would not be a major concern to the DT image quality. The
improved frame rate in the HS acquisition mode is directly
translated into 50% reduction in total DT image acquisition
time. Because the HS acquisition mode is 2X faster, it will be
helpful to reduce the patient motion artifact. There is a
potential image resolution loss due to 2x1 pixel binning;
however, the impact is minimized in the system design by
binning the pixels in the x-ray tube motion direction as well as
applying sharpness enhancement filtering operation to
compensate for the minor resolution loss. A comparative
image quality example is shown in Figure 5.
Regardless of the acquisition mode, DT slices can be
reconstructed and presented in two spatial resolutions during
post processing: the native 1x1 pixel resolution (HR-res) or the
2x2 pixel average resolution (STD-res). The size of an STD-res
DT volumetric image will be 1/8 (½ x ½ x ½) of that of the
corresponding HR-res image.
In summary, the DT volumetric images can be generated in
four combinations (see Table 1 for the recommended
applications). In theory, the HR/HR-res combination will offer
the best spatial resolution in the reconstructed DT slice image
because both the detector and the reconstruction work in the
native 1x1 resolution.
Figure 5: Image quality comparison of a hand phantom acquired in HR mode (left) and HS mode (right). Both images are
reconstructed in HR-res using 0.3 cm slice thickness and 0.05 cm slice interval. Sharpness enhancement is applied automatically in
the software to bring up the anatomical details in the image acquired in HS mode.

5
HR Std. Res
Acquisition
HR (1x1)
Hand/Wrist
Foot/Ankle
Knee
Shoulder
Acquisition
HS (1x2)
Hand/Wrist
Foot/Ankle
Knee
Shoulder
Chest
Abdomen
Pelvis
C-Spine
Table 1: Recommended DT operation for different exam types.
HR-res reconstruction is optimal for musculoskeletal exams,
and the HS acquisition helps to reduce patient motion artifact.
DT Section Thickness and Sweep Angle α
As mentioned before, DT imaging follows the same principles
as traditional tomography. For any anatomical details that are
outside the fulcral plane, the greater the sweep angle, the
greater the blur effect. Because all the details in the fulcral
plane are unchanged, increasing the sweep angle effectively
reduces the clutter caused by anything that is outside the
fulcral plane, which makes the tomographic slice appear
“thinner.” In practice, because the anatomical details gradually
become blurrier as their distance from the fulcral plane
increases, Section Thickness is used to define the region of the
anatomy within which all the anatomical details are under a
permissible blur level. Section Thickness is a value that is
determined solely by the acquisition geometry and therefore is
inherent to the acquisition. It is different from and should not
be confused with the Slice Thickness used in the DT
reconstruction. Table 2 shows the pre-calculated section
thickness at different sweep angles α and 0.25 mm permissible
blur values.
Sweep Angle α
(Degrees)
Section
Thickness (mm)
10 2.86
15 1.90
20 1.42
25 1.13
30 0.93
35 0.79
40 0.69
Table 2: Look up table for selecting DT sweep angle to achieve
desired slice thickness, assuming the maximum permissible blur
is 0.25 mm.
DT Slice Thickness t and Slice Interval δ
The DT image is presented as a stack of reconstructed 2D
coronal slices. The center location of each slice corresponds to
the fulcral plane at which anatomical details are maximized.
The distance between the reconstruction slices is the slice
interval (δ). The smaller the δ value, the finer the DT
reconstruction resolution in the depth direction. The slice
interval can be set accordingly by the operator, depending on
the requirements on the anatomical features to be visualized
for diagnosis. For example, for small musculoskeletal exams
such as wrist, etc., the δ value can be set between 0.5 mm to
1.0 mm, and for chest exams between 1.0 mm to 2.0 mm.
Each DT slice covers a slab of anatomical region and its
thickness (t) can be made arbitrarily by the DT reconstruction
algorithm, which is independent of (i.e., decoupled from) the
Section Thickness or the slice interval (δ). In practice the
reconstruction slice thickness (t) should always be greater than
or equal to the slice interval such that no anatomical void is
created (see Figure 6), causing the slices to overlap each other
partially. A smaller t value more likely helps to resolve the
anatomical details better but at the potential expense of
greater noise appearance. On the other hand, greater t value
helps to reduce the image noise but potentially blends more

6
anatomical details from the region outside the fulcral plane
therefore defeating the purpose of DT. A default slice
thickness of 20.0 mm is a good starting point for all exam
types.
Figure 6: Thick slices can overlap with each other yet slice
interval can be much smaller than slice thickness.
DT Projection Image Number N
The number of discrete DT projection images acquired across
the sweep angle dictates the reconstruction rippling artifact.
Figure 7 shows a DT slice of an ankle in the arterial-posterior
view. The rippling artifact at the bottom of the image is caused
by the sharp transitions between the open-air region and the
foot, with each ripple corresponding to a DT projection. When
the sweep angle is fixed, more projections will help to reduce
the artifact but at the expense of increased exam time. The
ripple artifact will manifest itself more with thicker anatomy, in
which case increasing the DT projection number will help.
Figure 7: An example of ripple artifact in DT foot exam.
Summary
Compared to traditional tomography, DT significantly improves
the workflow, reduces the overall exam time, minimizes
exposure retakes, and offers improved image quality. Good
radiographic practice in anatomy positioning and x-ray
exposure technique selection should be followed for a DT
exam, just as in standard radiography.
Patient motion during a DT exam degrades the image quality
and therefore should be minimized as much as possible. In
addition to taking measures to restrain patient motion, a
better timing to start the DT sequence with respect to the
patient motion pattern would help. Finally, patient motion
artifact can be reduced by reducing the total exam time using
the HS acquisition mode.
Slice A Slice B
Slice
Increment

Digital Tomosynthesis: Theory of Operation

More Related Content

What's hot (20)

Similar to Digital Tomosynthesis: Theory of Operation (20)

More from Carestream (20)

Recently uploaded (20)

Digital Tomosynthesis: Theory of Operation