Dual-energy (DE) imaging and digital tomosynthesis (DT) have been around for a few decades, but recent advancements in digital detectors have made this technology increasingly promising in clinical use. For more information about Carestream's imaging portfolio, visit www.carestream.com/medical or http://www.carestream.com/blog/2016/03/15/dual-energy-imaging-and-digital-tomosynthesis/

1. Dual-Energy imaging and Digital Tomosynthesis:
Innovative X-ray based imaging technologies
S Sajja, F Ursani, A Ursani, and N S Paul
TRIIO Core Laboratory, Departments of Medical Imaging,
University Health Network, University of Toronto
S Richard, N Packard, X Wang, and L Vogelsang
Carestream Health, Rochester

2. 2
Outline
•Chest Radiography – Limitations of 2D imaging
•Anatomical clutter
•Dual-Energy (DE) X-ray
•Anatomical clutter reduction via tissue discrimination
•Factors that impact DE imaging performance
•Clinical examples
•Digital Tomosynthesis (DT) X-ray
•Anatomical clutter reduction via spatial discrimination
•Factors that impact DT imaging performance
•Clinical examples

3. 3
Chest Radiography (CXR)
• Chest Radiography is the most common technique for 2D chest imaging
•Limitation
• A chest radiograph is a 2-dimensional projection of a complex 3-
dimensional volume in which several tissues overlay each other.
• Overlapping anatomy – ribs, lungs and vessels are background
structures which may confound detection of a lung nodule on a single
projection.
Conspicuity (detectability) = Contrast of the object
Complexity of the background

4. 4
Limitation of Chest Radiography (CXR)
3-d object
2-d projection of
the objectsCylinders made of material 1
Different objects made of
material 2

5. Overcoming anatomical noise via tissue discrimination
•Removing overlapping ribs via tissue removal using dual-energy imaging.
Material 1 only image2-d projection
Tissue discrimination using x-ray
imaging results in reduced
background anatomical noise
and improved feature
conspicuity.
DR - Digital Radiography
DE - Dual Energy X-ray

6. Tissue Discrimination – Dual-Energy (DE) imaging
6
High energy image
Low energy image
Bone image
WbWs
Soft-tissue image
X-ray tube
Anti-scatter bucky grid
flat-panel
digital detector

7.  Energy selection: Choice of
high and low energy values
 Dose allocation:
 Filtration:
 Delay between two
acquisitions:
7
Factors that affect DE imaging performance
• DE soft tissue images of polyethylene nodule
• High energy = 130 kVp (fixed)
• Low energy = 60-90 kVp
Patient (cm) equivalent phantom [acrylic] thicknesses:
“slim patient” (18 cm) [7.5cm]
“average patient” (24 cm) [10cm]
“large patient” (30 cm) [12cm]
Nodule contrast is highest at low kVp (60 kVp) and in a
slim patient
Nick Schumat, Masters Thesis, University of Toronto, 2008

8.  Energy selection:
 Dose allocation: Ratio of dose
between high and low energies
 Filtration:
 Delay between two
acquisitions:
8
Factors that affect DE imaging performance
Aε = 0.06 Aε = 0.29
Aε = 0.72 Aε = 0.91
• DE bone-only images at different dose
allocations (AƐ)
• Noticeable increase in image noise at very
low and very high allocations
Nick Schumat, Masters Thesis, University of Toronto, 2008

9. 9
Factors that affect DE imaging performance
 Energy selection:
 Dose allocation:
 Filtration: helps in increasing the
energy separation between the high
and low energy images
Fixed filtration: Same filtration for both
high and low energy acquisitions.
 Delay between two
acquisitions:
4
0 50 100 150
0
2
4
6
8
10
12
x 10
Intensity
4
keV
0 50 100 150
0
2
4
6
8
10
12
x 10
Intensity
60 kVp
60 kVp
1 mm Cu
keV
No Filtration – high unused radiation
Fixed Filtration – lower unused radiation

10. 10
Factors that affect DE imaging performance
Differential Filtration – very low unused
radiation
 Energy selection:
 Dose allocation:
 Filtration: helps in increasing the
energy separation between the high
and low energy images
Differential filtration: Filtration is
changed between high and low
images to separate spectra.
 Delay between two
acquisitions:
4
keV
0 50 100 150
0
2
4
6
8
10
12
x 10
Intensity
4
keV
0 50 100 150
0
2
4
6
8
10
12
x 10
Intensity
60 kVp
1 mm Cu
120 kVp
1mm Cu
70 kVp
0.1 mm Cu
120 kVp
0.5 Ag
Fixed Filtration – lower unused radiation

11.  Energy selection:
 Dose allocation:
 Filtration:
 Delay between two
acquisitions: This has an impact
on the motion artifacts observed in
resultant DE image.
11
Factors that affect DE imaging performance
ECG gating can be used to reduce motion artifacts.
Timing diagram displaying the (ECG) trace,
plethysmogram, and digital trigger.
DE “soft-tissue” images acquired (a) with (b) without
cardiac motion – custom built insert for motion
simulation. Some motion blur is observed.
Nick Schumat, Masters Thesis, University of Toronto, 2008

12. 12
Modelling nodule conspicuity
DR
Objective function for system optimization - combines:
MTF – System resolution
GNPS- Noise and anatomical clutter
Wtask- description of task i.e., nodule detection
d‘ – Detectability index – surrogate for nodule conspicuity

13. 13
Nodule detectability – fixed versus differential filtration
DR
DR DE fixed filtration
DE differential
filtration
• Figure shows the detectability index values for
DE fixed and differential filtration.
• The results are normalized such that d’ = 1 for
DR.
• d’ = 1.2 and 1.3 for fixed and differential filtration
respectively.
• d’norm refers to detectability normalized by dose.
• d’norm = 0.7 and 1.1 for fixed and differential
filtration indicating fixed filtration less dose
efficient than differential filtration.
Richard S. et al., Diagnostic Imaging, 2015

14. 14
DE Clinical patient imaging
CXR of a patient with lung nodules:
limited conspicuity due to
overlapping structures
DE-soft tissue of a patient with lung
nodules: Improved conspicuity due
to subtracted ribs.

15. 15
DE – Cadaver Images
Cadaveric chest x-rays were performed using conventional Digital
Radiography (DR) (a) followed by DE projections decomposed into Bone
only (b) and Soft tissue only (c) images to demonstrate improved
conspicuity of bone and soft tissue details from DE projections compared
to DR chest x-ray.

16. How to overcome anatomical clutter via spatial
discrimination
•Separation of overlapping structures. Commonly used techniques –
Computed tomography (CT) and Digital Tomosynthesis (DT) .
Illustration of CT
Illustration of DT
Illustration of spatial discrimination in x-ray medical
imaging resulting in reduced overlying clutter and
improved feature conspicuity.

17. 17
DT (Spatial discrimination) – Imaging
Detector
Detector
X-ray
tube
Table top
Wall Stand
(a) Working of the tomosynthesis in wallstand position (b) Working of the tomosynthesis in table top
position
Phantom
Phantom

18. Factors that affect DT imaging performance
 Angular range (Ɵ), number of
projections (N) : A smaller range for
image acquisition results in a lower tomo
angle and reduced x-ray dose; but the
depth resolution is also reduced.
 Detector binning: The process of
selecting regions of pixel and finding the
mean value. Improves signal-to-noise
ratio, speeds but results in loss of spatial
resolution.
 Scan time: The time taken to acquire
the set of projections. This has an impact
on motion artifacts due to cardiac and
respiratory motion.
(a) N=15, Ɵ=30o (b) N=60, Ɵ=30o
(a) N=15, Ɵ=7.5o (b) N=60, Ɵ=30o
Söderman C. et al., Medical Physics, 2015

19. Factors that affect DT imaging performance
 Angular range (Ɵ), number of
projections: A smaller range for
image acquisition results in a lower tomo
angle and reduced x-ray dose; but the
depth resolution is also reduced.
 Detector binning: The process of
selecting regions of pixel (NxN, N=
number of pixels) and finding the mean
value. Improves signal-to-noise ratio,
speeds but results in loss of spatial
resolution.
 Scan time: The time taken to acquire
the set of projections. This has an impact
on motion artifacts due to cardiac and
respiratory motion.
Pacemaker generator: Greater detail is seen
with 1X1 binning

20. Factors that affect DT imaging performance
 Angular range (Ɵ), number of
projections: A smaller range for
image acquisition results in a lower tomo
angle and reduced x-ray dose; but the
depth resolution is also reduced.
 Detector binning: The process of
selecting regions of pixel (NxN, N=
number of pixels) and finding the mean
value. Improves signal-to-noise ratio,
speeds but results in loss of spatial
resolution. .
 Scan time: The time taken to acquire
the set of projections. This has an impact
on motion artifacts due to cardiac and
respiratory motion.
DT phantom images acquired
(a) without motion
(b) with respiratory (breathing) motion
(c) with cardiac motion using a custom built
insert for motion simulation.

21. 21
Nodule detectability – Impact of dose in IQ
DT 100% DT 50% DT 30%
• Figure shows the detectability index values for
DT at different doses (30%, 50% and 100%).
• The results are normalized such that d’ = 1 for
DR.
• d’ = 11.8, 14.6 and 18.1 for 30%, 50% and
100% of the nominal doses respectively.
• d’norm refers to detectability normalized by dose.
• d’norm = 7.8, 7.7 and 5.9 for 30%, 50% and 100%
respectively since DT becomes anatomical noise
limited as opposed to quantum noise limited.Richard S. et al., Diagnostic Imaging, 2015

22. 22
DT – Patient imaging
Reconstructed DT patient image Corresponding coronal CT patient

23. 23
DT – Cadaver Images
Cadaveric DT through
the pulmonary hila
demonstrate high
resolution images of
the bifurcating left
main bronchus (inset)
using:
Slice thickness = 5 mm
Slice interval = 3 mm
Dose = full
Binning=1x1

24. 24
DT – Cadaver Images
Cadaveric DT
through the
pulmonary hila
demonstrate high
resolution images of
the branch
pulmonary arteries
and veins (inset)
using:
Slice thickness = 5 mm
Slice interval = 3 mm
Dose = full
Binning =1x1

25. 25
DT – Limitations
A nodule in a vessel branching point may be mistaken for
an enlarged vessel in tomosynthesis – DT (left), CT (right)
(Asplund S. , Acta Radiologica, 2011)
•DT involves a limited angle of
acquisition compared to CT.
•This results in limited sampling of
the signal in the frequency domain.
• Depth resolution is poorer in DT
than CT.
• In some cases this may result in
misinterpretation of structures as
seen in the image here.
•This can be alleviated by relating
the location of the ribs as compared
to the structure in question.
•Also for this particular example, the
potential use of contrast medium can
be explored.

26. 26
DT – Limitations
Nodule close to
pleura border –
With X-ray beam
tangential to ribs
Nodule close to
pleura border –
With X-ray beam
not tangential to
ribs
Nodule close to
pleura border –
With X-ray
beam not
tangential to
ribs
Nodule on vessel
branching point –
may be mistaken
for an enlarged
vessel on DT
DT may provide a false anatomical location of a lung nodule when it is located close to the
pleura; the effect depends on the angle of the incident X-ray beam relative to the nodule.
Asplund S. , Acta Radiologica, 2011

27. 27
DE-DT – Comparison Study – Low density objects
• Low density objects (cotton spheres) were inserted into the anthropomorphic phantom:
1 dry cotton sphere (green arrow) and 2 cotton spheres dipped in distilled water (blue
arrows)
• The phantom was imaged using (a) DR (b) DE (c) and DT and low dose CT (1mSv).
• CT images served as a reference standard (slide 28).
• DR and soft tissue DE images demonstrate the wet cotton spheres with a faint
projection of the dry cotton sphere. The DT images clearly demonstrate all of the
spheres.
• Lung pathologies (tumors) vary in their water content
• DE-DT may improve lesion characterization
Digital Radiograph Dual Energy Digital Tomosynthesis

28. 28
DE-DT – Comparison Study – Low density object
Coronal low dose CT images
Digital Radiograph Dual Energy Digital Tomosynthesis

29. 29
What is the future of DE and DT?
Future innovation: Qualitative to quantitative
Increased detector spatial resolution will facilitate extraction of quantitative image
data to provide more accurate diagnosis.
Application study: Volume estimation of thoracic water content
• In critically ill patients, clinical assessment of change
in thoracic fluid volume is necessary.
• A method based on temporal subtraction of CXR is
proposed to quantify the change in fluid volume
• Proof of concept testing was done using a chest
anthropomorphic phantom and solid water blocks.
• The estimated volume based on the technique was
compared with the actual volume.
•Dual energy (DE)- soft tissue only images had the
highest accuracy and correlated closely with the actual
volume with a root mean square (RMS)=.4.74 ml
Sajja et al., World Congress of Med. Phys. and Biomed. Eng. 2015

30. 30
What is the future of DE and DT?
DE- DT
An integrated DE-DT system would be beneficial for increasing the scope of
thoracic diseases which can be potentially diagnosed using chest radiography.
Illustration of DE-DT – material 1 Illustration of DE-DT – material 2

31. 31
Summary
• For DR CXR, overlapping of structures confounds the detection of
nodules. This results in anatomical noise.
• Anatomical noise can be overcome either via tissue discrimination or
spatial discrimination.
• Tissue discrimination is achieved via Dual-Energy (DE) which involves
acquisition of paired radiographs at 2 energies.
• Spatial discrimination is achieved via Digital Tomosynthesis (DT) which
involves acquisition of radiographs at different angles.
• It would be beneficial to combine the spatial and tissue discrimination
through an integrated DE-DT system.