Collaborators/Support
C-arm Flat Panel CT:
Image Quality Considerations
Rebecca Fahrig
Department of Radiology, Stanford University
Introduction
o C-arm CT for visualization in 3D of
vasculature and other high-contrast
structures has become commonplace
o The transition from XRIIs to digital flat
panels opened the doors to the
possibility of low-contrast and high
resolution 3D CT imaging in the
interventional suite
o The Physics Gang:: ‘Bob’ R.L.Dixon, ‘Tom’ J.
Payne, ‘Rick’ R.L. Morin
o A. Ganguly, N. Strobel and T. Moore
o L. Hoffman, N. Kothary, D. Sze, M. Marks, H. Do
o Technical support : A. White, N.R. Bennett, J.
Kneebone, M. Lozada-Parks, W. Baumgardner
o Siemens Medical Solutions
o Varian GTC
o NIH R01 EB003524
o Lucas Foundation
C-arm System :: CT System
half scan, area detector, ‘thin’ CsI converter
vs.
full scan, narrow detector, ‘thick’ Gd2O2S
1
Creating 3D Images in the
Interventional Lab
1) Rotational Angiography Run
4) In-room Display
Topics of Discussion
o Dose measurement
o Signal and Noise
o Scatter
3) Reconstruction and
Visualization
2) Image transfer
Dose Measurement
o small 0.6cc Ion Chamber, measuring dose max.
o CTDI phantom (16cm diameter, 15cm long)
o Dose measured at center and eight peripheral
positions for :
(30x40) cm detector format
based on 543 views
o Beam Size (iso-center):
Width: 26.67cm
16 cm
Height: 20.00cm
32 cm
Dose Measurement
o small 0.6cc Ion Chamber, measuring dose max.
o averaging with a 10-cm pencil chamber doesn’t
make sense since the slab thickness is variable
o CTDI phantom (16cm diameter, 15cm long)
o Dose measured at center
and eight peripheral positions for :
(30x40) cm detector format
based on 543 views
16 cm
o Beam Size (iso-center):
Width: 26.67cm
32 cm
Height: 20.00cm
2
Dose Measurements: 81kVp
Dose Measurements: 81kVp
D (0) = (1 / 3) D0 + ( 2 / 3) D p
2/3
mean
peripheral
+
1/3 central
source side
2/3
mean
peripheral
+
1/3 central
detector side
Peak
Dose
(mGy)
86
Center
Dose
(mGy)
34
“CTDIw”
(mGy)
1167
Detector
dose
(uGy/view)
0.46
81
608
0.44
63
28
37
109
310
0.70
66
31
40
125
260
0.92
76
38
46
kVp
Total
mAs
70
48
Variation in CTDIw in spite of AEC.
The EU guidelines for routine head CT scans specify a CTDIw
of 60mGy.
Dose in Abdomen Phantom
Dose
Slab width @
mA ms
mGy
det/iso
22.2/17 cm
348 4.9
34
18.4/14 cm
357 4.9
32
14/11 cm
367 4.9
27
8.4/6.5 cm
383 4.9
21
28cm
26cm
16cm
Measured Doses :
‘Medium-High’ dose requested
37cm
Circumference = 104cm
system uses automatic
exposure control : mA
ramped up to max at
thickest part of phantom
3
Which statement is true
about dose in C-arm CT...
20%
1.
2.
20%
20%
3.
20%
4.
20%
5.
Answer
Collimation does not affect the measured
dose
CTDIw is defined as ½ the dose measured at
the center of the phantom + ½ the average
dose measured at the periphery of the
phantom
The peripheral dose is higher on the x-ray
tube side of the C-arm sweep
The average dose measured in the periphery
of a 16-cm phantom is lower than the dose
measured at the center of the phantom
CTDI100 represents the angular average dose
at z=0 for a scan length of L=100 cm
1.
Collimation does not affect the measured
dose
FALSE
2.
CTDIw is defined as ½ the dose measured at
the center of the phantom + ½ the average
dose measured at the periphery of the
phantom FALSE : 1/3 center + 2/3 periphery
3.
The peripheral dose is higher higher on the xray tube side of the C-arm sweep
TRUE
The average dose measured in the periphery
of a 16-cm phantom is lower than the dose
measured at the center of the phantom
FALSE
CTDI100 represents the angular average dose
at z=0 for a scan length of L=100 cm
FALSE : gotcha question … L=10 cm
4.
5.
10
Visibility vs. kVp
Visibility of Low-Contrast Objects?
120
Nominal Contrast
U)
(3H
15mm
9mm
0%
6mm
Visibility [%]
100
0.5% (5HU)
%
0 .3
1.
0
(1
80
60
40
Average (70kVp)
Average (81kVp)
20
Average (109kVp)
)
HU
o Catphan Module CTP515
used as image quality
phantom (20cm housing)
o Acquired 543 views over
20sec at various dose and
kVp settings, Zoom 0
o Reconstructed soft tissue
segment (smooth kernel,
10mm slice width)
o Analyzed visibility of (outer)
5HU insets
Detail Diameters [mm] (2, …, 9, 15)
Average (125kVp)
0
0
2
4
6
8
10
12
14
16
Normalized Diameter [mm]
Scoring Question: What size “5HU” objects can you see?
4
Nominal vs. Measured Contrast
Values for a 0.5% contrast object
70 kVp
0.49
81 kVp
0.47
109 kVp
0.43
o CsI vs. Gd2O2S of thickness 1.4 mm
With16cmwater
o Measured contrast
depends on kVp
o Contrast decreases
as kVp increases
1.1
Absorbedenergyfra
kVp
Measured
Contrast
Detector Efficiency?
1
0.9
0.8
0.7
Flat Panel Detector
0.6
CTDetector
0.5
125 kVp
60
0.36
MTF: 100 µm steel wire
o Take advantage of the high resolution of
the flat panel : ~150 µm native at panel
70
80
90
100
kVp
110
120
130
140
High Resolution Imaging
In-vivo Stent in the Superficial Femoral Artery
displayed at constant win/lev 1000/900
150 µm no binning
300 µm 2x2 binning
600 µm 4x4 binning
1.0
Sharp
0.8
Normal
MTF
Smooth
0.6
0.4
0.2
0.0
0.0
0.5
1.0
1.5
2.0
Spatial Frequency (cycles/mm)
2.5
var{µ ( x, y )} ∝
1
∆R 3 S
5
Visibility vs. Dose
CNR ∝
Which statement is true about
signal and noise in C-arm CT?
1
Dose
Visibility Chart (81kVp, 543view s)
20%
100
80
20%
Visibility (%)
70
60
20%
50
4.
CTDIw = 19.93mGy
CTDIw = 26.31mGy
20
20%
CTDIw = 37.01mGy
CTDIw = 54.12mGy
10
0
0
3
6
9
12
15
Detail Diameter [mm]
Answer
2.
3.
4.
5.
3.
40
30
1.
1.
2.
90
Inherent object contrast increases with
increasing kVp
TRUE
In C-arm CT head scans, better detectability
occurs for kVps above 100 because the detector
has been optimized for visualization of iodinefilled vessels
FALSE : detector optimized for 70-80 kVp
The variance in a projection image is proportional
to the slice width
FALSE : inversely proportional to slice width
The variance in a reconstructed C-arm CT slice is
inversely proportional to the square of the inslice reconstructed pixel size
FALSE : inversely proportional to cube of pixel size
The contrast to noise ratio is proportional to
1/dose FALSE : inversely proportional to 1/sqrt(dose)
20%
5.
Inherent object contrast increases with
increasing kVp
In C-arm CT head scans, better detectability
occurs for kVps above 100 because the detector
has been optimized for visualization of small
iodine-filled vessels
The variance in a projection image is proportional
to the slice width
The variance in a reconstructed C-arm CT slice is
inversely proportional to the square of the inslice reconstructed pixel size
The contrast to noise ratio is proportional to
1/dose
10
Intracranial Imaging
C-arm CT
Clinical CT
In vivo pig model,
Autologous blood,
NO iodine contrast
ARTIFACT:
Beam hardening
Scatter
Conebeam
M. Marks, H. Do et al.
6
Body Applications : TACE
Scatter is Significant
o depends on such geometric factors as
ARTIFACTS:
Scatter
Truncation
•
•
•
•
air gap between object and detector
focal-spot-to-object distance
focal-spot-to-detector distance
weakly dependent on spectrum energy (kVp)
over the range of diagnostic energies of
interest
o The magnitude of the scatter-to-primary
ratio (SPR) depends on irradiated volume
L. Hoffman et al.
SPR is increasing as Cone Angle
(Imaged Volume) Increases
A Rough Rule-of-Thumb
(no collimation!)
o Cone angles on clinical scanners are getting
larger
• 64-slice conventional scanners have a cone angle of
~ 4°°
• 256-slice scanners have a cone angle of ~13°°
o CT imaging using large-area flat panels
• C-arm CT in interventional suites
have a cone angle of ~ 12°°
• On-board kV CT on radiation therapy
machines have a cone angle of ~ 15°°
RESULT: Scatter-to-primary ratios
are on the rise…
modified from Fox et al. Proceedings of SPIE Vol. 4320 (2001)
7
For a 16-cm diameter
head-sized object
64-slice scanner
For a 16-cm diameter
head-sized object
Conebeam C-arm CT
30-cm z-extent @ detector
For a 16-cm diameter
head-sized object
256-slice scanner
For a 16-cm diameter
head-sized object
Conebeam Rad Therapy
41-cm z-extent @ detector
8
For a 32-cm diameter
body-sized object
64 slice Rad Therapy
256-slice
C-arm CT
Image Quality vs. SPR
A Simple Phantom Study
o scatter field is shown as a lowfrequency background signal that
does not vary much from view to
view
o the primary varies significantly
depending on the angle
o signal (with scatter) in view 1
behind the two rods is significantly
over-estimated compared to signal
(with scatter) behind a single rod in
view 2
A Simple Phantom Study
As SPR increases:
o overall cupping increases, and therefore mean
offset increases
o variation due
to streaking
reaches a
max of ~ 200 HU
9
How can image quality be
maintained?
o Scatter decreases Contrast-to-Noise ratio
CNRs CNR0*(1-SPR/2)
(an approximation in the ‘small scatter, small contrast’
regime)
o To recover CNR, the dose can be increased,
but
CNR ∝ 1/sqrt(dose)
so the dose penalty is high
• e.g. for an SPR of 1, need 4x dose to get the same
CNR as a scatter-free image
Scatter Correction
o removes mean scatter
signal from the
recorded signal,
thereby reducing
artifact, but does not
remove the noise
o can lead to increased
artifact if the estimate
of SPR is incorrect
Pencil beam is attenuated
and produces a scatter
“impulse response” at
detector.
Transmitted
Primary
Scatter PointSpread Function
Detector
1. L.A. Love and R.A. Kruger, Med Phys 1987 14(2) 1978-85. 2. B. Ohnesorge, T Flohr, K Klingenbeck-Regn, Eur
Radiol. 1999; 9(3) 563-9 3. J. Maltz et. al. Med Phys. 2006 33(6), 2280. 4. Star-Lack et. al. AAPM07, EPI2k8
Grids and/or Collimators
o Anti-scatter grid
• removes scatter from the beam,
but also removes primary
• depending on the efficiency of
the grid, the SPR and the
resolution, it may or
may not be dose efficient
o Siewerdsen et al. (Med Phys 31, 2006)
have shown that grids used
with fluoroscopic systems
(e.g. Ptrans = 75%, Strans = 25%)
are not efficient unless SPR is
high (>1) , and resolution is low
(vox size > 1 mm)
o Endo et al. (Med Phys 33, 2006) showed that a 30:1 Mo
collimator was dose efficient for a 256-slice CT scanner
Direct Measurement :
Modulator Technique
primary
o “code” the
primary
without
affecting the
scatter
high-frequency
semi-transparent grid
o Decoding
(primary modulator)
provides
scatter-free estimate
of the primary
o no additional dose
required
same window for both = [-50, 100] HU
scatter
uncorrected
corrected
10
Impact of Scatter on Image
Quality
o Images of a 16-cm ‘medium contrast’
insert in the abdomen phantom
o Objects are -20, -25, -30 and -45 HU
relative to background
o 32, 16, 8, 4 and 2 mm diameter
Which statement is true
about scatter in C-arm CT?
20% 1.
20%
2.
20% 3.
4.
20%
20%
5.
Scatter-to-primary ratio (SPR) increases with
decreasing cone angle
SPR decreases approximately linearly (at small
volumes) as a function of increasing irradiated volume
Scatter decreases contrast-to-noise ratio by (1-SPR/2)
For a SPR of 1, an increase in dose of a factor of 2 is
required to get the same contrast-to-noise ratio as a
scatter free image
For best image quality in C-arm CT, the collimators
should always be as wide open as possible
10
Answer
1.
2.
Scatter-to-primary ratio increases with decreasing
cone angle
FALSE : decreases
SPR decreases approximately linearly (at small
volumes) as a function of increasing irradiated
volume
FALSE : increases
3.
Scatter decreases contrast-to-noise ratio by (1SPR/2)
TRUE
4.
For a SPR of 1, an increase in dose of a factor of 2
is required to get the same contrast-to-noise ratio
as a scatter free image
FALSE : need an increase in dose of a factor of 4
For best image quality in C-arm CT, the collimators
should always be as wide open as possible
FALSE
5.
Conclusions
o For C-arm CT with current detector
technologies, operate at ‘low’ kVp
o Present reconstructed data using the largest
pixel size appropriate for the detection task
o expose no more than you need to see since
SPR increases as the irradiated volume
increases
o Consider grid/collimator properties to
determine if contrast improvement outweighs
noise increase for the particular task
11
References
o Fahrig R, Dixon R, Payne T, Morin RL, Ganguly A, and
Strobel S. “Dose and Image Quality for a Cone-Beam Carm CT System.” Medical Physics 33(12), 4541-4550,
2006.
o Riederer, SJ, Pelc, NJ, and Chesler DA, „The Noise
Power Spectrum in Computed X-ray Tomography“Phys
Med Biol 23 (3), 446-454 (1978)
o Faulkner, K and Moores, BM. “Noise and contrast
detection in computed tomography images” Phys Med
Biol 29 (4), 329-339 (1984).
o Endo M, Mori S, Tsunoo T, Miyazaki H. “Magnitude and
effects of x-ray scatter in a 256-slice CT scanner” Med
Phys 33(9), 3359-3368 (2006).
12