Optimizing Image Quality and Dose in Digital
Mammography: X-ray Spectrum and Exposure
Parameter Selection
Mark B. Williams1, Priya Raghunathan1, Mitali
J. More1 Martin J. Yaffe2, Aili Bloomquist2,
Gordon Mawdsley2, Joseph Lo3, Ehsan
Samei3, Nicole Ranger3, J. Anthony Seibert4,
Alex Kwan4, Laurie Fajardo5, Allen McGruder5,
Sandra Maxwell5, Andrew D.A. Maidment6
Objective
Examine the effect on image
quality and radiation dose of the
choice of beam quality parameters
(target, filter, kVp) for current
commercial FFDM systems
of Virginia, 2University of Toronto
University, 4University of CaliforniaDavis, 5University of Iowa, 6University of
Pennsylvania
1University
3Duke
What is being optimized?
• Screen-film mammography maximization of contrast within the
constraint of acceptable film
darkening
• FFDM - separation of the processes
of acquisition and display permits the
displayed contrast of individual
structures to be adjusted when the
image is viewed
What is being optimized?
• However, display contrast in FFDM is limited by
the inherent image signal to noise ratio (SNR),
because as the signal contrast is increased, so
is the visibility of noise.
• Thus beam optimization in digital mammography
requires maximization of the image SNR.
• But since the SNR can be improved almost
arbitrarily by increasing the number of detected
x-ray photons, exposure parameter
optimization must balance increased image
SNR with increased patient radiation dose
1
Previous Study
FFDM systems evaluated:
• Fischer
• GE
• Trex
Williams MB, More MJ, Venkatakrishnan V, Niklason L,
Yaffe MJ, Mawdsley G, Bloomquist A, Maidment ADA,
Chakraborty D, Kimme-Smith C, and Fajardo LL. “Beam
optimization for digital mammography”. IWDM 2000: 5th
International Workshop on Digital Mammography. (Martin J.
Yaffe, Editor). Medical Physics Publishing. Madison,
Wisconsin, 2001; 108-119.
FFDM Systems Evaluated
Manufacturer
Model
Test Site
GE Healthcare
Senographe
2000D
University of Virginia
University of
Pennsylvania
Duke University
Siemens
Hologic
Mammomat
Novation DR
Selenia
University of Iowa
Fischer
Senoscan
University of Toronto
Fuji
5000MA
University of California
Davis
Current Study
•
•
•
•
•
Fischer SenoScan
GE Healthcare Senographe 2000D
Hologic Selenia
Siemens Mammomat Novation DR
Fuji 5000 MA (Lorad M-IV)
GE Senographe 2000D: amorphous silicon
photodiodes with overlying x-ray converter
Columnar
CsI(Tl)
Photodiode –
TFT
array
CsI(Tl)
p+
n+
a-Si
2
Hologic Selenia and Siemens Mammomat
Novation DR: photoconducting amorphous
selenium layer read out by a-Si thin film
transistor array
bias electrode
Fischer Senoscan:
Senoscan: CCDCCD-based scanned
slot system
Front View
drain
Fan Beam
Collimator
Swing
Arm
gate
Compressed
Breast
Phosphor
Fiber Optic
CCD
Pb
Shield
charge collection electrode
dielectric
A/D A/D
Direction
of Scan
DMA
Computer
storage capacitor
Dual Sided Imaging Plate Readout
Display
X-ray
Tube
amorphous selenium
source
Side View
Focal Spot
Acquisition Parameters
FFDM
system
Fischer
Target
Filter
kVp range
W
Al
28 – 37
GE
Mo, Rh
Mo, Rh
23 – 39
Hologic
Mo
Mo, Rh
23 – 39
Siemens
Mo, W
Mo, Rh
23 – 35
Fuji (Lorad
M-IV)
Mo
Mo, Rh
22 – 34
Diagram courtesy of FUJIFILM
3
Methods
• Common set of breast equivalent phantoms
circulated among all test sites
• Phantoms simulated nine breast types (3
compressed thicknesses x 3
fibroglandular/adipose compositions)
• For each phantom, for each possible target/filter
combination, images were obtained over a
range of kVps, manual mode, average pixel
value approximately constant
• SNR and mean glandular dose (MGD) were
calculated for each
Number of Images Processed
Example: Senographe 2000D at UVa
•
•
•
•
9 phantom types
3 target/filter combinations
8 kVps per target/filter combination
2 images per target/filter/kVp/phantom
combination
• Total of 432 images
Phantom
SNR Analysis
Thickness
3 cm, 5 cm, 7 cm
Mass stepwedge
1 cm
Glandular/adipose mass fraction
30/70, 50/50, 70/30
Skins are 100% adipose
Two images obtained for
each phantom/target/filter/kVp
combination.
Raw (For Processing) images
used.
1 cm
12 cm
(
Signal ≡ P2 − P1 − P4 − P3
SKINS
MASS STEPWEDGE
Calc stepwedge
10 cm
CIRS, Inc., Norfolk, VA
Noise ≡
FOM ≡
σ 5, difference
5
3
4
1
2
)
2
2
SNR
MGD
4
Dose Calculation
Limitations of the FOM
• Boone, J. (1999). Glandular breast dose for
monoenergetic and high-energy x-ray beams:
Monte Carlo assessment. Radiology 213, 2337.
• Sobol, WT and Wu, X. (1997) Parameterization
of mammography normalized average glandular
dose tables. Medical Physics 24(4), 547-555.
• Stanton, L., Villafana, T., Day, J., and Lightfoot,
D. (1984). Dosage evaluation in mammography.
Radiology 150, 577-584.
Independence of Shape of Signal vs. kVp
on Step Height
• SNR did not include spatial frequency
dependence
– Systems with higher MTFs and thus less smoothing of
Poisson x-ray noise may be penalized
– Inter-system comparisons can be misleading
• Exposure time was not taken into account
• The primary sources of uncertainty in the data
were:
– Noise in the measured HVL data
– Fluctuating system noise; this varied greatly among
manufacturers
Independence of Shape of Signal vs. kVp
on Step Height
Normalized Signal vs kVp, Novation DR, W/Rh
7 cm 50/50
Signal v s kVp, Nov ation DR, W /Rh, 7 cm 50/50
180
180
160
160
140
signal (ADUs)
signal (ADUs)
140
120
100
80
60
Step
Step
Step
Step
40
20
0
1
2
3
120
100
80
Step 0
Step 1
Step 2
Step 3
60
40
20
0
0
20
25
30
kVp
35
40
20
22
24
26
28
30
32
34
36
38
40
kVp
5
Independence of Optimum
Target/Filter on Step Height
Independence of Optimum
Target/Filter on Step Height
Normalized Signal vs. kVp, Senographe 2000D, All Target/Filters
3 cm 30/70
Signal vs. kVp, Senographe 2000D, All Target/Filters
3 cm 30/70
1600
Normalized Signal (ADUs)
Step 1 MoMo
Step 2 MoMo
Signal (ADUs)
1200
Step 3 MoMo
1000
Step 0 MoRh
Step 1 MoRh
800
Step 2 MoRh
600
Step 3 MoRh
Step 0 RhRh
400
Step 1 RhRh
200
Step 2 RhRh
0
25
30
35
Step 1 MoMo
1400
Step 2 MoMo
1200
Step 3 MoMo
Step 0 MoRh
1000
Step 2 MoRh
600
Mo/Mo
400
Step 3 MoRh
Mo/Rh
Step 0 RhRh
Step 1 RhRh
200
Step 2 RhRh
0
Step 3 RhRh
20
40
25
30
35
40
kVp
kVp
Contrast vs kVp, 5 cm 70/30, best
performing target/filter combinations
(Noise/ADU)2 vs. kVp, 5 cm 30/70, best
performing target/filter combinations
0.0005
0.40
Siemens W/Rh
0.25
Hologic Mo/Rh
0.20
0.15
Fuji Mo/Mo
signal
avg. bkgnd . pixel value
0.0004
2
0.30
GE - Rh/Rh
0.0005
contrast =
(Noise/ADU)
GE Rh/Rh
0.35
Contrast
Step 1 MoRh
Rh/Rh
800
Step 3 RhRh
20
Step 0 MoMo
1600
Step 0 MoMo
1400
Siemens
W/Rh
0.0004
0.0003
Hologic
Mo/Rh
0.0003
0.0002
Fuji Mo/Mo
0.0002
0.10
Fischer
0.05
0.0001
Fischer
0.0001
0.0000
0.00
22
25
28
31
kVp
34
37
40
20
23
26
29
32
kVp
35
38
6
SNR versus kVp, 5 cm 50/50, best
performing target/filter combinations
30
DgN versus kVp
Example: Fuji, 50/50 composition
7
Siemens
W/Rh
3 cm Mo/Mo
25
6
Hologic Mo/Rh
Fuji Mo/Mo
15
10
Fischer W/Al
5
3 cm Mo/Rh
DgN (mRad/R)
SNR
20
5
4
5 cm Mo/Mo
3
5 cm Mo/Rh
2
7 cm Mo/Mo
GE Mo/Rh
1
7 cm Mo/Rh
0
20
23
26
29
32
35
0
38
kVp
20
23
26
29
32
35
kVp
Dose versus kVp
Example: Fuji, 50/50 composition
FOM versus kVp, fixed composition
Hologic, 50/50
2.50
10
3 cm Mo/Mo
3 cm Mo/Mo
3 cm Mo/Rh
3 cm Mo/Rh
5 cm Mo/Mo
1.50
5 cm Mo/Mo
1.00
5 cm Mo/Rh
5 cm Mo/Rh
FOM
MGD/ADU
2.00
1
7 cm Mo/Mo
22
26
30
7 cm Mo/Mo
34
7 cm Mo/Rh
AF - 3 cm
0.50
AF - 5 cm
7 cm Mo/Rh
AF - 7 cm
0.00
0.1
20
25
30
kVp
35
kVp
7
FOM versus kVp, fixed composition
GE, 50/50
100
FOM versus kVp, fixed composition
Siemens, 50/50
3 cm Mo/Mo
10
3 cm Mo/Rh
3 cm Mo/Mo
3 cm Rh/Rh
FOM
10
5 cm Mo/Mo
3 cm Mo/Rh
5 cm Mo/Rh
3 cm W/Rh
5 cm Rh/Rh
5 cm Mo/Mo
5 cm Mo/Rh
7 cm Mo/Rh
7 cm Rh/Rh
FOM
7 cm Mo/Mo
22
AEC - C - 3 cm
1
22
25
28
31
34
37
40
5 cm W/Rh
1
24
26
28
30
32
34
36
7 cm W/Rh
AEC - D - 3 cm
AEC - 3 cm
AEC - C - 5 cm
AEC - 5 cm
AEC - SD - 5 cm
0.1
AEC - C - 7 cm
kVp
AEC - 7 cm
0.1
AEC - SD - 7 cm
FOM versus kVp, fixed composition
Fischer, 50/50
7 cm Mo/Mo
7 cm Mo/Rh
AEC - S - 3 cm
kVp
FOM versus kVp, fixed composition
Fuji, 50/50
10
10
3 cm Mo/Mo
3 cm W/Al
3 cm Mo/Rh
5 cm Mo/Mo
5 cm W/Al
FOM
FOM
5 cm Mo/Rh
1
7 cm Mo/Mo
20
7 cm W/Al
22
24
26
28
30
32
34
7 cm Mo/Rh
AEC - 3 cm
AEC - 5 cm
1
26
29
32
35
kVp
38
41
44
AEC - 7 cm
0
kVp
8
Techniques producing maximum FOM
GE
3 cm
5 cm
7 cm
30/70
Mo / Rh / 27
Mo / Rh / 27
Mo / Rh / 27
50/50
Mo / Rh / 27
Mo / Rh / 27
Rh / Rh / 31
70/30
Mo / Rh / 27
Mo / Rh / 27
Rh / Rh / 33
3 cm
5 cm
7 cm
W / Rh / 27
W / Rh / 27
W / Rh / 27
W / Rh / 27
W / Rh / 27
W / Rh / 29
W / Rh / 29
W / Rh / 29
W / Rh / 29
3 cm
5 cm
7 cm
Mo / Rh / 27
Mo / Rh / 27
Mo / Rh / 27
Mo / Rh / 27
Mo / Rh / 27
Mo / Rh / 28
Mo / Rh / 27
Mo / Rh / 27
Mo / Rh / 28
HVL at peak FOM, 3 cm
WAl
Fischer
MoRh
MoMo
Siemens
Hologic
Fischer
3 cm
5 cm
7 cm
W / Al / 27
W / Al / 29
W / Al / 35
W / Al / 27
W / Al / 30
W / Al / 35
Fuji
RhRh
GE
MoRh
Hologic
MoMo
MoRh
Siemens
MoMo
30/70
WRh
W / Al / 27
W / Al / 30
W / Al / 35
50/50
MoRh
MoMo
Fuji
3 cm
Mo / Mo / 24
Mo / Mo / 24
Mo / Mo / 26
5 cm
7 cm
Mo / Mo / 24
Mo / Rh / 28
Mo / Mo / 24
Mo / Rh / 28
Mo / Mo / 26
Mo / Rh / 28
70/30
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70
HVL (mm Al)
HVL at peak FOM, 5 cm
HVL at peak FOM, 7 cm
12
W Al
WAl
Fischer
MoRh
MoRh
10
Fischer
MoMo
MoMo
Fuji
MoRh
GE
Hologic
MoMo
MoRh
RhRh
8
RhRh
MoMo
6
MoMo
WRh
MoRh
MoMo
30/70
Hologic
MoRh
Siemens
4
Fuji
GE
MoRh
Siemens
MoMo
30/70
WRh
50/50
50/50
MoRh
2
70/30
70/30
MoMo
0
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70
HV L (m m Al)
HVL (mm Al)
9
FOM versus kVp, 3 cm 30/70, best
FOM versus kVp, 3 cm 50/50, best
performing target/filter combinations
performing target/filter combinations
14
Fischer W/Al
14
Fischer W/Al
Fuji Mo/Mo
Fuji Mo/Mo
12
12
Fuji AEC
Fuji AEC
GE Mo/Rh
GE Mo/Rh
10
10
GE AOP - C
8
GE AOP - S
GE APO - D
6
GE AOP - S
FOM
FOM
GE AOP - C
8
4
GE AOP - D
Hologic Mo/Rh
Hologic Mo/Rh
6
Hologic AEC
Hologic AEC
2
Siemens W/Rh
4
Siemens W/Rh
Siemens AEC
Siemens AEC
0
20
22
24
26
28
30
kVp
32
34
36
38
2
40
20
22
24
26
28
30 32
kVp
34
36
38
40
FOM versus kVp, 3 cm 70/30, best
FOM versus kVp, 5 cm 30/70, best
performing target/filter combinations
performing target/filter combinations
5
12
Fuji Mo/Mo
Fuji Mo/Mo
10
4
Fuji AEC
Fuji AEC
GE - Mo/Rh
GE Mo/Rh
8
GE AOP - C
GE AOP - S
6
GE AOP - D
Hologic Mo/Rh
4
GE AOP - C
3
FOM
FOM
Fischer W/Al
Fischer W/Al
GE AOP - S
GE AOP - D
2
Hologic Mo/Rh
Hologic AEC
Siemens W/Rh
2
Hologic AF
1
Siemens W/Rh
Siemens AEC
Siemens AEC
0
0
20
22
24
26
28
30 32
kVp
34
36
38
40
20
23
26
29
kVp
32
35
38
10
FOM versus kVp, 5 cm 50/50, best
FOM versus kVp, 5 cm 70/30, best
performing target/filter combinations
performing target/filter combinations
4
4
Fischer W/Al
Fischer W/Al
Fuji Mo/Mo
Fuji Mo/Mo
Fuji AEC
3
Fuji AEC
3
GE - Mo/Rh
GE Mo/Rh
GE AOP - C
GE AOP - D
GE AOP - C
FOM
FOM
GE AOP - S
2
GE AOP - S
2
GE AOP - D
Hologic Mo/Rh
Hologic Mo/Rh
Hologic AF
Hologic AF
1
Siemens W/Rh
1
Siemens W/Rh
Siemens AEC
Siemens AEC
0
0
20
23
26
29
32
35
38
22
25
28
31
kVp
kVp
34
37
40
FOM versus kVp, 7 cm 30/70, best
FOM versus kVp, 7 cm 50/50, best
performing target/filter combinations
performing target/filter combinations
2
1.5
Fischer W/Al
Fischer W/Al
Fuji Mo/Rh
Fuji Mo/Rh
Fuji AEC
Fuji AEC
1.5
GE Mo/Rh
1
GE AOP - D
Hologic Mo/Rh
0.5
GE AOP - C
FOM
FOM
GE AOP - S
GE Rh/Rh
1
GE AOP - C
GE AOP - S
GE AOP - D
Hologic Mo/Rh
0.5
Hologic AF
Hologic AF
Siemens W/Rh
Siemens W/Rh
Siemens AEC
Siemens AEC
0
0
22
26
30
34
kVp
38
42
22
25
28
31
34
kVp
37
40
43
11
FOM versus kVp, 7 cm 70/30, best
performing target/filter combinations
Mean Glandular Dose Using AEC
Selections (max FOM settings for Fischer)
350
1.2
Fischer W/Al
300
Fuji AEC
GE Rh/Rh
0.8
FOM
GE AOP - C
GE AOP - S
GE AOP - D
Hologic Mo/Rh
0.4
Hologic AF
Siemens W/Rh
Siemens AEC
Dose Using AEC (mrads)
Fuji MoRh
GE
Siemens
Hologic
Fuji
Fischer
250
200
150
100
50
0
22
25
28
31
34
kVp
37
40
43
0
3 cm
30/70
3 cm
50/50
3 cm
70/30
5 cm
30/70
5 cm
50/50
5 cm
70/30
7 cm
30/70
7 cm
50/50
7 cm
70/30
Conclusions – Detector Dependence
Conclusions – kVp Dependence
• Optimum exposure parameters for a
specific breast type are system (detector)
specific – there are no universal values
• The kVp dependence of the FOM and the
relative performances of different
target/filter combinations are independent
of signal (step) size and step composition
• Peaks in FOM versus kVp curves are
broad, when they exist
• The choice of target/filter is much more
crucial than the choice of kVp
12
Conclusions – kVp Dependence
• In all cases the FOM is a decreasing
function of kVp at the upper end of the kVp
range tested here
• Thus at least for the target/filter
combinations currently included in these
FFDM systems, there is no advantage in
expanding the available voltage range to
even higher kVp.
Conclusions – Noise Dependence
• There is a general inverse correlation
between the normalized noise and the
FOM, with higher noise systems tending to
have lower FOM values
• This suggests that the system noise is an
important determinant of FFDM system
performance, irrespective of signal and
dose performance.
Conclusions – Target/Filter Dependence
• For the Senographe 2000D, the maximum FOM
was always obtained with either Mo/Rh or Rh/Rh
target filter combinations, but never with Mo/Mo,
even for the thinnest and most fatty breasts
• The Selenia produced maximum FOM values for
nearly all breast types at an exposure parameter
setting of Mo/Rh, 27 kVp, with only a slightly
higher kVp (Mo/Rh, 28 kVp) for the two densest
7 cm breasts
• Contrary to the Senographe 2000D and the
Selenia, the maximum FOM was obtained for the
5000MA with a Mo/Mo combination for most (6 of
9) breast types, with only the 7 cm phantoms
benefiting more from Mo/Rh.
Conclusions – HVL for Optimum Settings
• The HVL of the technique producing the highest
FOM tended to fall at a location within the range
of available HVL values that is quite systemspecific.
• For the Senoscan and 5000MA systems, the
optimum techniques had HVL values that
increased with increasing breast thickness, but
were always in the lower half of the available
HVL ranges.
• For the Mammomat Novation DR, the optimum
technique factors were nearly identical for all
breast types, always utilized the W/Rh
combination, and corresponded to a HVL that
was near the top of the available range
13
Conclusions – AEC Dose
Conclusions – AEC Performance
• There is a substantial difference among FFDM
systems in the radiation dose delivered under
AEC operation.
• In general, the Fuji system delivers the highest
dose under its AEC settings of all systems
tested, while the Novation DR delivers the least
of all systems with AECs.
• The Senoscan, operated with kVp settings close
to those maximizing the FOM for that system,
delivers the lowest radiation dose of all systems
tested for 5 cm and 7 cm breasts .
• The AEC systems of current commercial
systems generally select technique factors
close to those producing the maximum
FOM value, but selection could be
improved in several cases
• Preset minimum kVp values may be too
high in some systems
Conclusions – AEC Performance
Conclusions – AEC Performance
• For the Senographe 2000D, the best FOM
values for all 3 cm phantoms were
obtained with the Mo/Rh target/filter
combination.
• However, the AEC selected Mo/Mo for all
3 cm phantoms in all three modes,
resulting in lower FOM values
• For the Selenia, the FOM for the Mo/Rh
combination is superior to that for the
Mo/Mo combination at virtually any voltage
for all breast types.
• Since the AEC (autofilter mode) selected
Mo/Mo for 3 cm breasts of all
compositions, dose performance could be
improved without loss of image quality by
programming the AEC for selection of the
rhodium filter for smaller breast thickness
14
Conclusions – AEC Performance
• The W/Rh target filter combination on the
Mammomat Novation DR outperformed Mo/Mo
and Mo/Rh for all breast types imaged.
• Novation DR kVp selection is based on
compressed thickness only, thus AEC selection
is insensitive to changing breast composition,
and the same target/filter/kVp is chosen for all
breast compositions of a given thickness.
• The FOM curves suggest that compared to the
current AEC selections, some performance
improvement may be possible through the
utilization of somewhat lower kVp settings for
thicker (i.e. 7 cm or thicker) breasts with only
minimal increase in MGD.
Acknowledgements
• Patty Goodale Judy, University of Virginia
• Gordon Axt, University of Iowa
15