Transcript L. Ramello

CINVESTAV 2005 Advanced
Summer School
Medical Imaging
with Semiconductor
Detectors
L. Ramello – Dip. Scienze e Tecnologie Avanzate,
Univ. Piemonte Orientale, ALESSANDRIA (Italy)
13-15 July 2005
L. Ramello
1
Topics



Basic properties of semiconductor detectors
Image quality: contrast, SNR, MTF, DQE
Recent detector developments:
–
–
–


MEDIPIX (2D pixels)
SYRMEP (Synchrotron Light Source)
High Z semiconductors
Dual Energy Mammography
Dual Energy Angiography
13-15 July 2005
L. Ramello
2
Basic properties of
semiconductor detectors
13-15 July 2005
L. Ramello
3
Why semiconductor detectors ?
• Advantages for medical imaging with x-rays:
–
High spatial resolution (down to ~50 micron)
– High detection efficiency, especially in the low energy
range (mammography)
– Combine x-ray conversion and electrical signal generation
– Decrease radiation dose and/or improve image quality
• Semiconductor imaging system concepts:
– Digital radiography with scintillator + amorphous silicon
(commercially available)
– Digital radiography with direct conversion in
semiconductor material (R & D)
– PET and SPECT with high Z semiconductors (R & D)
13-15 July 2005
L. Ramello
4
Semiconductor materials


Material
Z
density,
g/cm3
Eg,
eV
W,
eV
Si
14
2.33
1.12
3.6
GaAs
31,33
5.31
1.42
4.2
Ge
32
5.32
0.73
2.9
Se
34
4.3
1.71-1.75
5.6
CdTe
48,52
6.20
1.52
5.0
HgI2
80,53
6.36
2.13
6.7
PbI2
82,53
6.2
2.31
7.2
Atomic number Z, density and thickness  probability of
x-ray photon conversion
Average energy loss to create electron-hone pair, W
(roughly proportional to Eg)  energy resolution
13-15 July 2005
L. Ramello
5
Energy resolution
Radiation ionization energy (W):
determines the number of primary
ionization events
Band gap energy (Eg):
lower value  easier thermal
generation of e-h pairs
(kT = 26 meV for T = 300 K)
13-15 July 2005
L. Ramello
6
Semiconductor detectors

To fully exploit these attractive semiconductor
detector features:
– High electric field is needed to collect signal
– Dedicated, low noise electronics is needed (usually
the first element is a charge amplifier)
– For silicon, a p-n junction is needed to reduce dark
current (operation at room temperature is OK)
– For germanium, cryogenic operation (liq. N2
temperature) is needed

Multichannel systems require special care for
power density, connection technique, cross-talk
13-15 July 2005
L. Ramello
7
The p-n junction (1)
Abrupt junction approximation
Net charge density vs. distance
Electric field vs. distance
Electrostatic potential vs. distance
Valence and Conduction band
energies vs. distance
13-15 July 2005
L. Ramello
8
The p-n junction (2)



In reverse polarization (positive voltage to n-side):
the diode current density saturates at a low value Js
the depletion layer thickness (d) increases with
increasing voltage, so does the active volume
d = (2VB/eeND)1/2
e = ere0  12 e0 (Si)
ND = donors/cm3 (n-Si)
qV/kT = ratio between potential
energy and thermal energy
13-15 July 2005
L. Ramello
9
The microstrip detector
SIGNAL = number of electronhole pairs:
ne-h = DE/W,
where W=3.62 eV for silicon
C
e
d
13-15 July 2005
1/ 2
 eeN D 

C  
 2V B 
REVERSE POLARIZED DIODE
• Depletion region => free from charge
carriers: e-h pairs may be detected
• Reverse Bias voltage (VB) =>
controls diode depletion thickness, i.e.
active volume
• p-n junction capacitance per unit
area C:
1/C2 grows linearly with VB =>
C-V measurement determines full
depletion voltage VFD
L. Ramello
10
A microstrip detector
• AC coupling: Bias
Line and resistors
to bias each strip,
without shorting
adjacent strips
• Guard ring(s) are
essential to collect
surface currents
– This introduces
a dead layer for
edge-on
geometry
13-15 July 2005
DC contact (to p+ implant)
guard
ring
bias line first strip (AC contact)
L. Ramello
11
A readout chain
vthp
IN
calib in

OUT
charge
preamplifier
shaper
vthn
discriminator
This is just one possibility, the binary readout scheme –
another one is to put an ADC instead of the discriminator,
preserving the full analog information
13-15 July 2005
L. Ramello
12
The RX64 ASIC
RX64 - Krakow UMM design - (28006500 m2) consists of:
- 64 front-end channels (preamplifier, shaper, discriminator),
- 64 pseudo-random counters (20-bit),
- internal DACs: one 8-bit threshold setting and and two 5-bit for bias,
- internal calibration circuit (square wave 1mV-30 mV),
- control logic,
13-15
July
2005 to external bus).
- I/O
circuit
(interface
L. Ramello
13
Conversion efficiency (1)
} 300 μm (standard thickness)
} 10-20 mm μm (edge-on)
Si (300 μm): efficiency drops to 50 % at 15 keV
(Al window limits efficiency at low energies)
 Recover efficiency with edge-on orientation
13-15 July 2005
L. Ramello
14
Conversion efficiency (2)
} 300 μm GaAs (Z ~32)
} 10-20 mm μm Si (Z =14)
GaAs (300 μm): efficiency drops to 50 % at 48 keV
 Material of choice for mammography, E ~ 22 keV
13-15 July 2005
L. Ramello
15
Image quality:
contrast, SNR, MTF, DQE
13-15 July 2005
L. Ramello
16
•
•
X-ray beams (1)
X-rays are generated by bremsstrahlung of electrons
emitted from cathode, accelerated by an applied
voltage and impinging on the anode
The energy spectrum of x-rays is determined by:
– Peak kilovoltage (kVp)
– Anode material (concerning peaks at characteristic
energies)
– Intrinsec and added filtration
Effect on an 80 kVp x-ray beam of added
filtration with a light material (Al) and with
a rare earth material (La, K-edge @ 39 keV)
13-15 July 2005
L. Ramello
17
X-ray beams (2)
•
•
•
•
Most common anode materials:
–
–
–
W (Z=74) for general radiographu (chest, whole body, …)
Mo (Z=42) & Rh (Z=45) for mammography
Cu (Z=29) for diffractometry
Energy emitted as x-rays is only 0.5-1% of input
energy, the remaining part must be dissipated as
heat
X-ray tubes with moderate power are with fixed
anode, high power ones have a rotating anode to
avoid melting
Typical currents are 1-5 mA for prolonged
exposure (fluoroscopy) and 50-1000 mA for short
exposures; exposure is measured in mAs
13-15 July 2005
L. Ramello
18
X-ray imaging techniques
•
•
•
•
•
•
Film: sensitivity is very low, it would require too
high a dose to the patient
Film + screen: conventional radiography
Image intensifier (I.I.): fluoroscopy
Photosensitive phosphor (computed radiography)
Indirect digital radiography (I.I. or
photoconductor coupled to a semiconductor)
Direct digital radiography (semiconductor)
13-15 July 2005
L. Ramello
19
Film + screen (1)
X-rays transmitted through patient
first screen
double coated emulsion / AgBr
second screen
•
•
•
•
About 50% of the photons convert in the film-screen, mostly
(95%) in the two screens
The film exposure is mainly due to the blue-green light emitted
by the phosphorescent screens (CaWO4, Gd2O2S:Tb, etc.)
Film-screen systems are classified according to their speed,
with faster systems requiring less incident radiation to obtain
same optical density
The standard speed is = 100, slower (50) and faster (200, 400,
600) speed film-screen systems are commonly used
13-15 July 2005
L. Ramello
20
Film + screen (2)

X-ray absorption vs.
energy by different
screens

Spectrum of primary and
scattered x-rays from a
tube operated at 80 kVp,
with a Perspex (clear
acrylic resin) phantom
 usefulness of Gd
screen to suppress
scattered x-rays
13-15 July 2005
L. Ramello
21
Exposure and optical density (1)
Radiographic film blackening
radiografico (mostly due to visible
light emitted by screens) may be
quantified by optical density (D):
D = -log(T)
where T is the transmission:
T = I1/I0
Useful optical density goes from 0.2 to 2.5-3.0
Exposure X quantifies the number of incoming x-rays
13-15 July 2005
L. Ramello
22
Exposure and optical density (2)
Relation between optical
density D and exposure X:
1) Film-screen:
D = cX
highly non linear, constants
depend on film speed
Transm. T
D = -log(T)
1.000
0.0
100% transm.
0.741
0.13
base
0.100
1.0
good exposure
0.010
2.0
lung
0.001
3.0
very dark
0.0003
3.5
maximum
darkness
2) Electronic detector (e.g. phosphor + photodiode):
D =kX
linear (image may be subsequently processed to
“emulate” film of any given speed)
13-15 July 2005
L. Ramello
23
X-ray film: dynamical range
0.5 mAs
2 mAs
4 mAs
8 mAs
16 mAs
underexposed
32 mAs
63 mAs
overexposed
M. Overdick (PHILIPS), 11/09/2002, IWORID 2002, Amsterdam
13-15 July 2005
L. Ramello
24
Flat panel detector: dynamical range
typical usage
Digital Diagnost (PHILIPS) 43 cm x 43 cm, 143μm x 143 μm
M. Overdick (PHILIPS), 11/09/2002, IWORID 2002, Amsterdam
13-15 July 2005
L. Ramello
25
Image quality


Image quality has a decisive impact on the
radiologist’s ability to detect pathologies (other
factors: visualization conditions, radiologist’s
experience)
Most important aspects of image quality:
–
–
–

Contrast
Noise (hence signal/noise ratio, SNR)
Spatial resolution (sharpness)
Then of course the dose to the patient must be
minimized
13-15 July 2005
L. Ramello
26
Contrast (1)
The radiographic contrast C between two areas
A (signal) and B (background) of an image
may be defined in terms of optical densities:
C = DA-DB
 The radiographic contrast depends from both
subject contrast Cs and detection method (filmscreen, digital detector, etc.)
 The subject contrast Cs depends on the
radiation-subject interaction, in the case of xrays it depends on the linear attenuation
coefficient μ and on the thickness x of areas A
and B
 In electronic imaging systems the contrast can
be manipulated in a second time

13-15 July 2005
L. Ramello
27
Contrast (2)
Transmission of monochromatic
photons of several energies vs. soft
tissue thickness:
T = exp[-μx]
Subject contrast Cs:
Cs = (I1-I2)/I1 =ΔI/I1
with I1, I2 representing absorbed energy per
unit area of photoreceptor:
I0 = NE
I1,2 = N E ε exp[-∫μdz] (1+R)
con N = number of primary photons per unit
area, ε = detection efficiency,
R = ratio secondary/primary photons
13-15 July 2005
L. Ramello
I0
I0
t
μ1
μ2
x
I1
I2
28
Contrast and Signal
Subject contrast Cs:
Cs = ΔI/I1= {1-exp[-(μ2-μ1)x]}/(1+R)
 depends on the thickness x of the detail under study
(but not on the background tissue tickness t)
 depends on the difference between linear
attenuation coefficients μ1 and μ2
 decreases as diffused radiation (by Compton effect)
impinging on the detector increases: this can be
countered by antiscatter grids or exploiting the
lesser energy of diffused photons
The signal relative to a certain area A may be defined
as ΔI·A, and must be compared with fluctuations of
the background I1·A (same area)
13-15 July 2005
L. Ramello
29
Noise and signal-to-noise ratio




Fluctuations are due both to quantum noise (fluctuation in
the number of converted photons) and to properties of the
photoreceptor and of the imaging system
Quantum noise in our case follows Poisson statistics:
noise = E(I1A/E)1/2 = E[NεAexp(-μ1t)(1+R)]1/2
Taking the ratio of signal:
ΔI·A = I1CA = CANεEexp(-μ1t)(1+R)
to noise we get the signal-to-noise ratio:
SNR = {1-exp[-(μ2-μ1)x]}[NεAexp(-μ1t)/(1+R)]1/2
Setting a minimum SNR (Rose criterion: SNR > 5) one
can compute the number N of incident photons per unit
area necessary to detect a detail of thickness x and
transverse area A
13-15 July 2005
L. Ramello
30
Spatial resolution (1)


Every imaging system has intrinsic resolution
limits which define the smallest detectable detail
For example, in the case of film-screen systems,
several factors contribute to the spatial resolution:
–
–
–

finite dimensions of the focal spot and
magnification value
possible motion of the patient (breathing, hearth
beat) during exposure
resolution loss in the photoreceptor, due e.g. to
diffusion of light in screens (or in image
intensifiers)
Many test objects and procedures have been
developed to measure spatial resolution of
imaging systems
13-15 July 2005
L. Ramello
31
Spatial resolution (2)


An objective measure of spatial resolution is given by the
MTF (Modulation Transfer Function), which quantifies the
ratio between output and input contrast vs. spatial frequency
The MTF may be measured by taking an image of a lead
object having a series of slits with given spatial frequency
(lp/mm, line pairs per mm), or an image of a sharp edge
13-15 July 2005
L. Ramello
32
Spatial resolution (3)
Radiographic image of a test object
with an array of 3 x 7 groups of slits
with different spatial frequencies
Optical density profiles of the top-left 3 rows
by 4 columns of the test object.
The resolution limit (*) corresponds to a
spatial frequency of 1.5 cycles/mm
13-15 July 2005
L. Ramello
33
Detective Quantum Efficiency

The Detective Quantum Efficiency (DQE) measures the noise
added by the imaging system: DQE(f) = SNR2out(f) / SNR2in(f)
Comparison of DQE among four different
imaging systems:
• Film-screen (speed 400)
• Computed Radiography
• Indirect digital radiography (CsI + a-Si)
• Direct digital radiography (a-Se)
13-15 July 2005
L. Ramello
34
Recent detector developments
13-15 July 2005
L. Ramello
35
Medipix: Hybrid Pixel Detector
M. Campbell, V. Rosso, Rome IEEE NSS-MIC 2004 conference
13-15 July 2005
L. Ramello
36
Medipix detector - cross section
13-15 July 2005
L. Ramello
37
ITC-Irst Detector
– Si <111>
– 300-800 m thick
– pixel 170 x 170 m2
– p+ side 150x150 m2
– 64 x 64 chs
– 1.2 cm2 area
13-15 July 2005
MEDIPIX1 ASIC:
SACMOS 1 m technology
pixel: 170 x 170 m2
64 x 64 channels
area 1.7 cm2
threshold adjust 3-bit
15-bit counter
L. Ramello
VTT Bump-bonding
http://medipix.web.cern.ch/MEDIPIX/
Medipix1 ASIC with silicon pixel
detector
38
Medipix1 + Si: contrast measurement
X-ray (W-anode) settings : 40 kV, 25 mA, 500 ms
X-ray focus
Al thickness 75 m
Collimator
140 cm
Air
800m detector
1.5 cm
3.8
3.6
Al
Si detector
Contrast (%)
3.4
3.2
3.0
2.8
N air  N Al
C
N air
13-15 July 2005
2.6
2.4
8
10
12
14
16
18
20
22
24
Energy threshold (keV)
L. Ramello
39
Contrast
4 ,2
4 ,0
(  Al ( E )   air ( E )) ( E ) S ( E ) dE


x

(
E
)
S
(
E
)
dE

C  1 e
Al(E) and air(E) are the absorption
coefficients at the energy E
(E) is the detector efficiency at the
energy E
S(E) is the incident spectrum
3 ,6
3 ,4
3 ,2
3 ,0
2 ,8
8
40kV W
seen by 800 m
seen by 525 m
seen by 300 m
14000
13000
12000
11000
1 0
1 2
E n e rg y
1 4
1 6
1 8
th r e s h o ld
2 0
2 2
2 4
(k e V )
4.2
4.0
3.8
10000
2
Photons per (mA s mm ) at 750 mm
15000
C
(% )
3 ,8
3.6
9000
C (%)
8000
7000
6000
5000
3.4
3.2
300 m
525 m
800 m
3.0
4000
3000
2.8
2000
2.6
1000
0
0
5
10
15
20
25
30
35
40
45
2.4
8
Energy (keV)
13-15 July 2005
10
12
14
16
18
20
22
24
Energy threshold (keV)
L. Ramello
40
Medipix1 + Si: SNR
2 ,4
3 0 0
5 2 5
8 0 0
2 ,2
2 ,0
75 m Al
 m
 m
 m
S N R
1 ,8
air
1 ,6
1 ,4
1 ,2
1 ,0
SNR 
0 ,8
0 ,6
8
1 0
1 2
E n e rg y
1 4
1 6
1 8
t h r e s h o ld
2 0
2 2
2 4
2
2
 air
  Al
(k e V )
Thickness ratio
Calculated ratio
Experimental SNR ratio
525/300
1.24
1.25
800/300
1.41
1.42
800/525
1.14
1.14
13-15 July 2005
N air  N Al
L. Ramello
41
Medipix1 + Si: MTF
Energy threshold: 11 keV
300m
525m
800m
1,0
0,8
Nyquist Freq. (2.94 lp/mm)
MTF: 64 %
Evaluated aperture 168 m
Detector pitch 170 m
MTF
0,6
0,4
0,2
0,0
1
2
3
4
5
6
7
1,0
800 m detector
11 keV
15 keV
19 keV
23 keV
0,8
MTF
0,6
0,4
0,2
0,0
0
1
2
3
4
5
6
7
8
9
10
Spatial Frequency (lp/mm)
13-15 July 2005
11
800 m
aperture
(m)
168
15
161
19
155
23
146
Thr.
(keV)
Spatial Frequency (lp/mm)
L. Ramello
42
Medipix2: 55 μm x 55 μm pixels
55  m
55  m
13-15 July 2005
L. Ramello
43
Calculated x-ray spectrum and
energy thresholds used
Thresholds
9.1 keV
11.3 keV
12.8 keV
18.8 keV
Siefert FK-61-04x12 X-ray tube, W-target, 2.5 mm Al, Vpeak = 25 kV.
13-15 July 2005
L. Ramello
44
Medipix2: Measured MTF
@ various thresholds
13-15 July 2005
L. Ramello
45
Medipix2: DQE @ various
thresholds
13-15 July 2005
L. Ramello
46
SYnchrotron Radiation for
MEdical Physics


The main aim of the SYRMEP beamline is the investigation
and the development of innovative techniques for medical
imaging.
The challenge of mammography
– High image quality: Both high contrast and spatial resolution
– Very low delivered dose: Breast is very radiosensitive
– Very high social relevance

After successful feasibility studies on in vitro mammography,
the project for synchrotron radiation clinical mammography is
under development.
R. Longo, C. Venanzi, Rome IEEE NSS-MIC 2004 conference
13-15 July 2005
L. Ramello
47
SYRMEP Beamline
Conceptual Design
monochromator
Fast
shutter
Sample
Holder
5 d.o.f.
Detector
Holder
2/3 d.o.f.
filters
slit systems
13-15 July 2005
ionization
chamber
L. Ramello
48
SYRMEP silicon microstrip
detector

Silicon microstrip detector in edge-on geometry
 Single photon counting read-out electronics
 Active area matched with beam cross-section
 Pixel size 100x300 mm2
 Very high scattering rejection
 Maximum SNR
13-15 July 2005
L. Ramello
49
SYRMEP digital radiography
Laminar
beam
object
Si detector
SR digital image
Conventional
image
Energy 20 keV
100 m scan step
MGD 1.8 mGy
MGD 1.4 mGy
13-15 July 2005
L. Ramello
50
SYRMEP digital radiography
a) SR digital
image
Energy 17keV
Scan step
100 m
MGD 1 mGy
clinical
mammographic
unit
26 kVp
MGD 1 mGy
b) SR digital
image
Energy 20keV
Scan step 100
mm
MGD 0.33mGy
3 cm thick ‘in vitro’ human breast tissue
13-15 July 2005
L. Ramello
51
SYRMEP image acquisition
tomography
mammography
13-15 July 2005
L. Ramello
52
SYRMEP mammographic unit
Patient support
Detector and
Exposimeter
holder
Patient movement
stage
13-15 July 2005
L. Ramello
53
High Z semiconductors



Positron Emission Tomography (PET) and Single
Photon Emission Computed Tomography
(SPECT) make use of high energy photons up to
~500 keV
High Z semiconductors are being developed as a
replacement for scintillators (BGO, LSO, …)
currently used in commercially available systems
Due to present limits in the volume (and cost) of
semiconductors, the targeted applications are those
for small animals with a not too large field of view
13-15 July 2005
L. Ramello
54
PHILIPS prototype CZT for SPECT



Aim: improve
energy resolution
and spatial
resolution
Overall size:
20 cm x 48 cm,
pixel size 2.4
mm
Compare CZT (5
mm thick) and
NaI(Tl) (9.6 mm
thick)
with 3.5 mCi of
Tc-99m
M. Petrillo, Rome IEEE NSS-MIC 2004 conference
13-15 July 2005
L. Ramello
55
PHILIPS CZT imaging
performance

Sensitivity of CZT slightly inferior to NaI(Tl)
 Contrast and spatial resolution of CZT clearly
superior to NaI(Tl)
13-15 July 2005
L. Ramello
56
CdTe, CZT developments

Quite some progress in recent years for CdTe,
CdZnTe detectors concerning:
–
–
–
–
Crystal growth
Electrode design
Interconnect technology (bump bonding, …)
Hybrid and ASIC electronics
L. Verger, Rome IEEE
NSS-MIC 2004
conference
13-15 July 2005
L. Ramello
57
Single channel CZT 4x4x6 mm2

After bi-parametric correction (based on pulse rise
time – amplitude correlation) the efficiency at 122
keV rises from 30% to 75-80%
13-15 July 2005
L. Ramello
58
Single channel CZT 8x8x15 mm2

With bi-parametric correction the energy
resolution at 662 keV is improved and tailing is
drastically reduced
13-15 July 2005
L. Ramello
59
CZT-based micro-PET

Micro-PET system
with pixellated CZT
replacing LSO
scintillator:
–
improve spatial
resolution (2 mm
1 mm) with
depth-of-interaction
information
13-15 July 2005
L. Ramello
60