Transcript transparencies

6th International Workshop on Radiation Imaging Detectors, Glasgow 25 - 29th July, 2004
Performances of different
digital mammography
imaging systems: evaluation
and comparison
Maria Giuseppina Bisogni
Dipartimento di Fisica “E. Fermi”, Universita’ di Pisa and
Sezione INFN Pisa
Outline
 Introduction
• Mammography
• Digital mammographic systems
 The MEDIPIX pixel detector for digital mammography
• The assembly
• Si and GaAs Detector development
 Imaging performance of the MEDIPIX pixel detector
• Experimental set - up
• Mammographic Phantom
• Beam optimization
 Comparison with commercial digital systems
 MTF (Modulation Transfer Function) evaluation
The mammography challenge
Breast cancer
Glandular tissue
Adipose Tissue
Comparison between the linear
attenuation coefficient of carcinoma,
fat tissue and glandular tissue
•The problem for an early diagnosis: the X-rays attenuation of a pathological tissue is
very similar to that of breast tissue.
•Typical lesions: masses (of the order of a cm and with low contrast) and clusters of
microcalcifications (high contrast but small dimensions, of the order of hundreds of mm).
Digital Mammography (DM)
 Main features
• Linear response with X-ray exposure
• Wide dynamic range (104 – 105)
• Mammography of dense breast
• Reduced radiation dose
• Exposure determined as a function of the Signal to Noise Ratio
(SNR) not of the Optical Density of the film (OD)
•
•
• Dose reduction from 20 to 80 %
Image processing
Time required for the examination (texp<1s; Tproc~minutes)
 Limits (?)
• Spatial resolution
• Film-screen systems  20 lp/mm
• Digital systems  10 lp/mm
• Costs
Digital systems
 Indirect integration detection systems
• the X-rays are absorbed in a phosphor layer and produce light photons
which convert into electric charges in a photo detector and then to an
electric signal
• Imaging Plates based on photostimulable phosphors
• CsI(Tl) – aSi Flat Panels
• Phosphor screen +OF taper+ CCD array (slot scanning)
 Direct integration detection systems
• the X-ray photons directly convert into charges (electron-hole pairs) and
thus to an electric signal in a photoconductor
• aSe Flat Panels
 Direct photon counting systems
• single photons are counted, i.e. the number of photons directly represent
the intensity level in a pixel
• Sectra Microdose Mammography (MDM) based on Si detectors
• Xcounter based on gas avalanche photodiodes
The MEDIPIX1 Pixel detector for DM
http://medipix.web.cern.ch/MEDIPIX/
Semiconductor Detector
Detector
• Si, GaAs
• pixel 170 x 170 mm2
• channels 64 x 64
• area 1.2 cm2
From upper
pixel
Read-Out chip
Shutter
Sel
Mask
(1 bit)
Cfb
Input
Threshold
Adjust
(3 bits)
0
Shift
Reg
Clk
1
Latched
Comparator
Clkout
Analog
Reset
Test
Input
Sel
Mux
Rst
Test
(1 bit)
0
Mux
Data
Pulse
shaper
Preamp
Ctest
1
To lower
pixel
Photon Counting Chip (PPC)
CERN SACMOS 1 mm
FASELEC Zurigo
Pixel 170 x 170 mm2
Channels 64 x 64
Area 1.7 cm2
Threshold adjust 3-bit
Pseudo-random counter 15bits
The GaAs pixel detector: material characterization
100
GaAs 200 mm
Counts
10000
2
current density ( mA/cm )
10
VHV=350 V
109
Cd
E=22 keV
CCE = 90 %
8000
1
6000
SUM 082
SUM 051
DOWA 077
DOWA 079
0,1
0,01
CCE=100 %
4000
2000
1E-3
0
100
200
300
400
500
reverse voltage (V)
0
100
0
10
15
20
25
30
35
40
GaAs “bulk”
LEC (Liquid Encapsulated Chochralski),
Semi-insulating (SI)
Sumitomo
Resistivity = 108 Ohm cm
Contacts
Schottky : multilayer Ti/Pd/Au
(AMS Roma)
Thickness
200 mm
E(keV)
90
80
70
c.c.e. (%)
5
60
50
40
30
20
10
0
0
100
200
300
reverse voltage (Volt)
400
500
The Si pixel detector: geometry simulation
with multiguardrings
without multiguardrings
with multiguardrings
without multiguardrings
30000
100
electric field (Volt/cm)
electric potential (Volt)
25000
80
60
40
20
20000
15000
10000
5000
0
0
0
200
400
600
800
1000
1200
1400
0
200
width (mm)
width
Simulation performed with the package ISE-TCAD
400
600
800
1000
1200
1400
width (mm)
300 mm
 Geometry
The pixel detectors
Pixel 150 mm x 150 mm
• Pixel 150 mm, interpixel 20 mm
• 64 x 64 pixels
• Sensitive area 1.2 cm2
 Si detectors
• Produced by ITC-IRST
(Trento, Italy)
• Thickness 300, 525, 800 mm
• p+-n junction
• Bump bonding by VTT (Finland)
guardring
multiguardrings
 GaAs detectors
•
•
•
•
Produced by AMS (Italy)
Thickness 200 mm
Schottky contacts
Bump bonding by AMS
Detection efficiency @ 20 keV
•
Si 300 mm 26 %
•
Si 525 mm 40 %
•
GaAs 200 mm 98 %
150 mm
500 mm
guardring
Experimental setup
Source: standard mammographic tube
(Instrumentarium Imaging Products Diamond)
Molibdenum target
Mo (0.03 mm) and Be (1 mm) filters
Distance between the beam focus and the detector = 64 cm
Mammographic facility of the Istituto di Radiologia,
Ospedale S. Chiara, Pisa, Italy
3500
3000
mammographic
spectrum (28 kVp)
After 4 cm Lucite
Simulated with the Program
IPEM Report 78
2500
2000
1500
1000
500
0
0
5
10
15
20
25
30
E(keV)
Mammographic Phantom: RMI 156
RMI 156 phantom is recommended for quality checks in mammography
(as suggested by American College of Radiology (ACR) Phantom)
dimension : 8 x 8 cm2
a Lucite block 3.3 cm thick
a wax block 0.6 cm thick
a 0.3 cm thick cover
It
simulates
a
compressed breast
(4.2 cm)
16 test objects embedded in wax:
5 groups of simulated micro-calcifications of different
diameter
(0.54 mm, 0.40 mm, 0.32 mm, 0.24 mm and 0.16 mm),
5 different thickness tumor-like masses
(2.00 mm, 1.00 mm, 0.75 mm, 0.50 mm and 0.25 mm)
6 different size nylon fibers that simulate fibrous structures
(1.56 mm, 1.12 mm, 0.89 mm, 0.75 mm, 0.54 mm and 0.40 mm)
Acquisition Geometry
focus
scan direction
105 mm
3 mm
Pb collimator
440 mm
42 mm
phantom
50 mm
detector
 Move and tile technique
 72 different acquisitions
 scanning performed with stepper
motors controlled by the PC
Exposure conditions optimization
The effect of the collimator has been evaluated by measuring the contrast of a
W edge with and without a collimator
N1
N2
With one
collimator
26 kV 250 mAs
25000
0 coll
1 coll
counts
20000
15000
10000
contrast
(65.0 + 0.7) %
(95.11 + 0.03) %
5000
0
0
200
400
600
800
1000
•The use of a collimator close to the beam focus allows a significant improvement in
the detected contrast and therefore in the image quality.
•The system equipped with 2 collimators shows a slight contrast improvement (~ 0.6%).
MEDIPIX I –Film Images Comparison
 kodak trimatic film
• kodak min-r 2190 screen
• Digitized 12 bits, 85 mm
pitch
Tube settings
25 kV 80 mAs
 Medipix 1
• Detector Si
• 525 mm thick
MedipixI-Film SNR comparison
6.0
film
MedipixI
5.5
5.0
4.5
SNR
4.0
3.5
3.0
2.5
2.0
1.5
1.0
12
13
14
15
Particular #
 The Signal to Noise Ratio (SNR) of the details imaged with the
MEDIPIXI system is systematically higher than the SNR of the
same details imaged by the film
 The SNR of the detail 15 (0.24 mm thick) is not measurable on
film
The commercial digital systems
•GE Senographe 2000D based on CsI(Tl)-aSi
•Diagnostic Centre Prevenia in Torino (Italy)
•Ge Senographe 2000D is a full field digital mammography system:
•19 x 23 cm2 amorphous silicon detector coupled to CsI (Tl) (flat panel) with a
pixel size of 100 mm
•Fuji FCR 5000MA Computed Radiography (CR) based on imaging plates with a
photostimulable phosphor screen
•C.S.P.O. in Firenze (Italy)
•Image Plate with a photostimulable phosphor screen
•Senographe 800T X-ray unit:
•18 x 24 cm2 with a pixel size of 50 mm
•Giotto Image MD Internazionale Medico Scientifica Srl (I.M.S. Bologna, Italy) based
on aSe
•Istituto di Radiologia APGD in Udine (Italy)
•Amorphous selenium technology (flat panel)
•active area 17.4 x 23.9 cm2 with a pixel size of 85 mm
•Mo and Rh filter
SNR comparison
•SNR as a function of the mAs for the detail 12 (2 mm thick tumor mass)
of the RMI 156 phantom, obtained with the different acquisition systems
Si 525 mm
GE Senographe 2000D
FCR 5000 MA
Giotto Image MD
10
9
8
25 kVolt
25 kVolt
7
SNR
6
5
4
3
2
1
0
0
50
100
150
mAs
200
250
GaAs- Si detectors SNR comparison
•SNR as a function of the mAs for the detail 12 (2 mm thick tumor mass) of
the RMI 156 phantom with GaAs and Si 525 mm thick detectors.
•Tube for general radiography (W anode and an Al filter 2.5 mm)
•Tube settings: 40 kV and 16 - 125 mAs
4,0
GaAs
Si 525mm
3,5
• The performance of the GaAs detector
in terms of the SNR is superior to the
525 mm thick Si one
3,0
SNR
2,5
2,0
• The statistics on the images is in
agreement with the different detection
efficiencies of Si and GaAs at the
energies of the X ray beam (average = 28
keV)
1,5
1,0
0,5
0,0
0
20
40
60
80
mAs
100
120
140
MTF evaluation: edge method
•The MTF of the system has been evaluated using the image of
the W edge with the mammographic tube at 25 kV and 12 mAs.
1,0
with collimator, 25 kV 12 mAs
MTF Si 525 mm
0,8
MTF
0,6
0,4
0,2
0,0
0
2
4
6
8
10
12
lp/mm
ESF: from the image of a W edge
LSF: the ESF’ derivative
MTF: normalized Fourier Trasform of the LSF
At the Nyquist frequency (2.94 lp/mm) of
the MedipixI the MTF is 60 %
MTF comparison
•Comparison among the MTF of the different digital systems is presented
(the vertical lines represent the value of the Nyquist frequency for each
system).
GE Senographe 2000D
Si 525 mm
Fuji FCR 5000MA
Giotto Image MD
1.0
0.8
MTF
0.6
0.4
0.2
Nyquist frequency
MTF
Giotto Image MD
5.88 lp/mm
46 %
Single photon counting
2.94 lp/mm
60 %
GE Senographe 2000 D
5 lp/mm
20 %
FCR 5000MA
10 lp/mm
1%
0.0
0
2
4
6
8
lp/mm
10
12
14
These results are in agreement with data published in literature.
Bloomquist et al. “Acceptance Testing of Digital Mammography Units for
the ACRIN/DMIST Study”, Proceedings of the IWDM 2002, Bremen,
Germany, pp. 85-89
MTF evaluation: slit method
 Tantalum phantom, 1.5
mm of thickness, 10
mm width slit.
 Presampling MTF
• line slightly tilted with respect to the
perpendicular to the pixel lines.
• Composition of digital LSFs.
• Computation of the FFT for the MTF
Edge –slit methods comparison
 The two curves have the same
behavior up to the Nyquist
frequency
 For higher frequencies the slit
method is more accurate
MTF edge PPC
MTF slit PPC
1.0
0.6
MTF
• Good agreement with the theoretical
curve sinc(pfA) of an ideal imaging
system of sampling aperture A
0.8
0.4
 The first minimum position provides
the information on the sampling
aperture A
• For the PPC A = 0.17 mm which
corresponds to the pixel dimension
 The MTF does not depend on the
detector material and thickness
0.2
0.0
0
1
2
3
4
5
6
7
8
9
10
spatial frequency (lp/mm)
Sampling aperture
Conclusions
o Development of Si and GaAs pixel semiconductor detectors,
which have been bump bonded to single photon counting
chips (PCC-MedipixI system).
o These assemblies have shown very good imaging capabilities
in terms of:
SNR and MTF in comparison with commercial digital
systems
o The MTF measurements with the slit method for the
commercial systems have still to be done
o Assemblies based on Si pixel detectors and the new chip
Medipix2 (256 x 256 pixel, 55 mm in side), are currently
under test and an improvement in terms of SNR and MTF is
expected
Medipix I -Medipix II MTF comparison
MedipixI
MedipixII
fNy=2.9 lp/mm
1.0
fNy=9.1 lp/mm
0.8
MTF
0.6
0.4
0.2
0.0
0
5
10
15
20
25
Spatial Frequency (lp/mm)
Sampling aperture
a = 0.17 mm
 MTF measured with the slit method
 W anode tube, 40 kV, 125 mAs
Sampling aperture
a = 0.055 mm