Transcript Silicon Drift Detectors - Wayne State University

Silicon Drift Detectors:
A Novel Technology for Medical Imaging
Rene Bellwied, Wayne State University






Applications in medical imaging
Why semiconductor detectors ?
Which semiconductor ?
Which Silicon technology ?
What is a SDD ? Applications
Proposal for future research
Why Semiconductors ?



Presently most medical devices are based on photo-imaging
(film) => excellent resolution but low sensitivity and lack of
image uniformity => long exposure times in b- and x-ray
imaging and development time.
Lately, digital imaging based on integrating devices (i.e.
MWPC (multi-wire proportional chambers, gas devices).
Sensitivity better than film but resolution poorer (~400 mm)
New idea: single particle counting using semiconductor
detectors has the following advantages:





High sensitivity (low exposure time)
High dynamic range and excellent linearity
Energy discrimination of particles
Direct digital imaging and online image display
Very good resolution (< 50 mm)
Medical Applications
X-ray applications
 Digital mammography <E> = 20 keV
 Dental X-ray tube <E>= 35 keV
 Fast frame medical diagnostics
Nuclear medicine
 Thyroid measurements, <E> = 60-140 keV
photons
 DNA probe array, b-emitter (32P, 14C, 35S), <E>
= 50-700 keV
X-ray Mammography statistics

Breast cancer





Most frequent neoplastic pathologies in women in the world
One of the first causes of death. U.S. has one of the highest rates in the world.
Last year 180,000 women in the U.S. were diagnosed, 44,000 women died of breast cancer
During last 50 years incidence rate (probability of woman infected during life cycle) is increasing

In 1950 the probability was one in fifty (0.2%)

Last year the rate was one in eight (12.5%)
Mammography



85-90% malignancies are visible on mammography
Early diagnosis

5-year survival rate after treatment is over 95% for localised breast cancer (dimensions less
than 15 mm)

5-year survival rate after treatment is less than 20% for metastatic cancer

Life quality improved: conservative surgery
Mammographic screening

Effective tool in reducing breast cancer mortality for women aged 50 years and over

Benefits for women younger than 50? Presently not applied because of dose.
Dose Considerations

Defining factors





Young vs. old women




Mammography relies on photon attenuation coefficient differences between
tumours and normal breast tissue
The highest contrast between micro-calcification and glandular tissue is at low
photon energies (generally below 30 keV)
But low energies have small photon transmission coefficients as a function of
tissue thickness.
Breast compression leads to large improvements in dose rates.
single shot doses are low (less than 1 mGy), but dose integration over several
years increases risk of associated carcinogenesis.
Young women have denser breasts, which strongly reduces the contrast to
cancerous spots. Higher exposures with higher doses have to be used.
Breast cancer is most dangerous in young women as it tends to develop faster that
in older patients.
The main proposed way to reduce the dose is to significantly increase the
image quality
Contrast and Transmission
Requirements for radiation detectors
Definitions for device comparisons


Optical Density (O.D.)
 The optical density of a film system can be directly compared to the
count response for a digital system.
 The optical density is a measure of the ‘blackening’ of a
mammographic film
Spatial frequency
 Spatial frequency n is measured in lp/mm (line pairs per mm)
 Definition: n = 1 / (2 spatial resolution), e.g. 1 / 0.2 = 5 lp/mm for
100 micron resolution. In a pixel system the pixel size should be at
least twice as small as the highest detectable spatial frequency.
 Small objects require sharp images i.e. high spatial frequencies
 Highly granular systems have a more stable MTF with increasing
spatial frequency.
 MTF is being measured at a fixed spatial frequency for comparison
MTF and DQE measurements

Modulation Transfer Function (MTF):

MTF(n )   LSF(x)  e2inx dx




Spatial resolution properties of the imaging system
LSF(x) Line Spread Function: image of an “infinitely thin” line object
Detective Quantum Efficiency (DQE):
 SNR(out)
DQE  

 SNR(in) 

2
Noise transfer properties through the different stages of the imaging
system
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)
Limitations (?)

Spatial resolution
 Film-screen systems  20 lp/mm
 Digital systems  5 lp/mm (spatial frequency is smaller, but image is sharper)

Monitor resolution
 Monitor 2000x2500 pixels, resolution 0.1 mm
Development of digital imaging

Need to be superior to standard X-ray film
cassette:





Better resolution
Lower radiation dose, shorter exposure time
Need to match film detector size (18 x 24 cm2)
Problem: quantum efficiency of 300 mm Si wafer for
medical x-rays is at best 25 %. Use Silicon for
readout but not for signal generation.
Development based on amorphous Si (a-Si):
detector based on converter with a-Si active matrix.


Indirect systems: scintillator as converter (e.g. CsI or P) , active matrix
records visible photons from scintillator
Direct systems: use heavy semi-conductor as converter (e.g. a-Se,
GaAs) with direct charge transfer to active matrix
Direct Digital Mammography (DMM)

Direct conversion of photons

Minimises image blurring and avoids an extra conversion stage from
X-rays into visible light

Higher SNR with respect to phosphor-based systems

Higher spatial resolution

Single Photon counting

More efficient noise discrimination

Higher SNR for polychromatic beams
Amorphous vs. crystalline
semiconductors

Advantage of amorphous semiconductors:
Can be produced into any size detectors (i.e. a-Si
can match film detector size)

Disadvantage of amorphous semiconductors:
Charge carrier lifetimes are orders of magnitude
lower than in crystalline semiconductors. High
voltages have to be applied to collect charges fast.
Flat Panel Detector based on a-Si
Famous recent commercial products

GE Senograph 2000D flat panel device:

Indirect system based on a Phosphor scintillator converter
and a-Si matrix for readout (first digital device with
FDA approval (2000)).

Fischer and Siemens slot scanning devices


Indirect systems based on Phosphor screen and CCD
array.
AGFA-Gevaert flat panel device:

Direct system based on a-Se converter with a-Si matrix for
readout.
Commercial systems for DM



Photostimulable phosphors (Imaging Plates)

Fujifilm HR-V 18 x 24 cm2

pixel 100 x 100 mm2

Double read-out
Flat Panels

GE Senographe 2000D

Revolution ™ Flat Panel Digital Detector

18x 24 cm2, pixel 100 x 100 mm2

11 years R&D and 130 M$ investment

I Digital Mammographic system approved by FDA
(2000)
Slot scanning systems

Fischer Imaging SenoScan

Phosphor screen +OF taper+ CCD array

22 x 30 cm2 in 4 s

Pixel 50 x 50 mm2

10 years R&D and 30 M$ investment

Approved by FDA (2001)
Device development by SIEMENS
The progress in image quality
Detection Quantum Efficiency
Photon Counting Chip (PCC)
From upper
pixel
Shutter
VLSI Circuit developed at CERN
in the frame of the MEDIPIX European collaboration
CERN, Universities and INFN Pisa and Napoli (IT),
Universities of Glasgow (UK) and Freiburg (GE)
http://medipix.web.cern.ch/MEDIPIX/
Input
Cfb
Threshold
Adjust
(3 bits)
Preamp
0
Test
(1 bit)
0
Mux
Shift
Reg
Clk
1
Clkout
Analog
Reset
12 mm
Sel
Mux
Latched
Comparator
Ctest
1
Data
Pulse
shaper
Rst
Test
Input
Sel
Mask
(1 bit)
To lower
pixel
EP CERN SACMOS 1 mm
FASELEC now Philips AG, Zürich
Pixel 170 x 170 mm2
channels 64 x 64
Area 1.7 cm2
Threshold adjust 3-bits
Pseudo-random counter 15-bits
Read-Out and Config. Freq. 10 MHz
Read-Out Speed 400 ms/image
Photon counting vs. optical density
Optical density as
a function of dose
for mammographic film
Photon counting as
a function of dose
for digital device
4.0
10000
1.0
2000
10 20 30 40 50 60 70 80
mAs
0.0
0.5
1.0
1.5
=10 mAs
2.0
2.5 3.0
log(E/E0)
=40 mAs
MTF Comparison (according to MEDIPIX study)
Direct detection of
the photons
GaAs PCC
Improved spatial
resolution
Mammographic system
Radiographic system
Flat panel
(a-Si+CsI)
Fluoroscopic system
DQE Comparison (according to MEDIPIX study)
DQE of the pixel
detector is higher and
more stable with the
frequency
Due to direct detection of
photons
and
Single photon counting
Improved noise transfer
properties
GaAs PCC
Trixell
Fluoroscopic system
Mammographic system
Radiographic system
Which Semiconductors ?




Generally available: Ge, Si, GaAs, CdTe
Ge needs liquid nitrogen cooling to yield good
resolution
GaAs and CdTe have higher X-ray absorption
efficiency than Si in relevant range (<E>=10-70 keV)
At 20 keV the detection efficiency for a 200 mm
thickness GaAs layer is 98%, which is four times
higher than the equivalent efficiency in Silicon.
GaAs is more advanced than CdTe but both
technologies are in their prototype stages compared
to Silicon.
Photon Absorption Efficiencies
Performance Comparison
Si (left) vs. GaAs (right)

C.Schwarz et al., NIM A 466, 87 (2001)
Relevant Literature







S. Webb, “The Physics Of Medical Imaging”, Institute of Physics
Publishing (1998)
M. Sandborg and G.A. Carlsson, “Influence of x-ray energy spectrum,
contrasting detail and detector on the SNR and DQE in projection
radiography”, Phys.Med.Biol. 37 (1992) 1245
H.J.Besch, “Radiation Detectors im medical and biological applications”,
NIM A419 (1998) 202
S.R.Amendolia et al, “Spectroscopic and imaging capabilities of a
pixellated photon counting system”, NIM A466 (2001) 74
C.Schwarz et al., “Measurements with Si and GaAs pixel detectors
bonded to photon counting readout chips”, NIM A466 (2001) 87
A.Castoldi et al., “The Controlled Drift Detector: a new detector for fast
frame readout X-ray imaging”, NIM A461 (2001) 405
A Castoldi et al., “The Controlled Drift Detector”, NIM A439 (2000) 519
Imaging of mammographic
Phantom
phantoms
Lucite cylinder
thickness 4 cm
Al disks diameter 4 mm
Embedded in wax cylinders diameter 12 mm
110 cm
diameter 10 cm
Detector
Image comparison
Si
300 mm
0
GaAs
200 mm
Film
12 bits
170 mm
255
125 mm 100 mm 75 mm 40 mm 25 mm
10.9 mm
Exposure = 32 mAs
Exposure Time ~ 1 s
Dose = 6 mGy
RMI 156 phantom

American College of Radiology
(ACR) Accreditation phantom

RMI 156 features

Details
 6 Nylon fibers
 Ø 1.56 – 0.4 mm
 5 microcalcification groups

Ø 0.54 – 0.16 mm
 5 tumour masses

2.00 – 0.25 mm

Dimensions 4.5 x 10.2 x 10.8 cm3
 Compressed breast 4.2 cm thick
.
...
..
.
. . .
. .
.
.
.
.
.
.
Images
255
6 cm
0




6 cm
Pixel detector
6x6 scanning
Pixel 170 mm
Entrance Dose/image
= 4 mGy



Mammographic film
Digitized
 100 mm pitch
 12 bits
Entrance Dose
= 4 mGy
Dose comparison

Al disk, 125 mm thick, embedded in 3 mm
thick wax cylinder, Lucite block 4 cm thick
0

Images


Dimension 1.2 cm2
Pixel 170 mm
255
Film 32 mAs
Dose ~ 6 mGy
GaAs 8 mAs
Dose ~ 1.5 mGy
Why Silicon Drift Detectors ?




Generally available: CCD, Si pixels, Si strip, Si drift
CCD (charge-coupled devices) are slow and not
very radiation resistant.
Silicon pixels are fast and have high resolution but
they are very expensive and the connection to the
electronics (bump bonded) is a difficult technical
process, but the main electronics development
(PCC) is optimized for bump-bonding
Silicon strip detectors have relatively poor resolution
and are not cost competitive to drift detectors.
SDD’s: 3-d measuring devices
Silicon Drift Detector Characteristics




Large units can be mass produced (presently up to
10 by 10 cm)
Detector operates well at room temperature or with
Peltier cell cooling
Electronics density is order of magnitude smaller
than resolution. (electronics pitch = 250 mm,
position resolution < 20 mm)
Three dimensional point determination with one
dimensional readout.
Present status of technology
STAR (detector at Relativistic Heavy Ion Collider (RHIC) at BNL on Long Island)



4in. NTD material, 3 kWcm, 280 mm thick, 6.3 by 6.3 cm area
250 mm readout pitch, 61,440 pixels per detector
SINTEF produced 250 good wafers (70% yield)
ALICE (future detector at Large Hadron Collider (LHC) at CERN in Geneva, Switzerland)


5in. NTD material, 2 kWcm, 280 mm thick, 280 mm pitch
CANBERRA produced around 100 prototypes, good yield
Future (potentially for detector at Next Linear Collider (NLC) in ?)



6in. NTD, 150 micron thick, any pitch between 200-400 mm
10 by 10 cm wafer
low radiation length, low cost for large area
The STAR Detector at RHIC
STAR/SVT at RHIC (BNL)

Search for the quark-gluon plasma (QGP) and
investigate the behavior of strongly interacting
matter at high energy density.
Installed in February 2001, first beam in July 2001.
2500 charged particles/event
Radiation length: 1.4% per layer

0.3% silicon, 0.5% FEE (FrontEnd Electronics),
0.6% cooling and support. Beryllium support structure.
FEE placed beside wafers. Water cooling.

SVT costs $7M for 0.7m2 of silicon.


RHIC Au-Au Beam Collisions
Approach Collision
Particle Shower
Actual Collision in STAR (central)
The SVT in STAR
Construction
in progress
Connecting
components
STAR-SVT characteristics







216 wafers (bi-directional drift) = 432 hybrids
3 barrels, 103,680 channels, 13,271,040 pixels
6 by 6 cm active area = max. 3 cm drift
3 mm (inactive) guard area
max. HV = 1500 V
max. drift time = 5 ms, (TPC drift time = 50 ms)
anode pitch = 250 mm, cathode pitch = 150 mm
The SVT in STAR
The final device….
… and all its
connections
Wafers: Spatial resolution (measured)
R&D on SDD thickness
All recent silicon detectors were produced from
300 mm thickness wafers
Potential Problems
(noise, depletion voltage)
STAR-SDD Summary






Mature technology.
<10 micron resolution achievable with $’s
and R&D. Easy along one axis (anodes).
<0.5% radiation length/layer achievable if
FEE moved to edges.
Low number of channels translates to low
cost silicon detectors with good resolution.
Detector could be operated with air cooling
at room temperature
Technology is viable for position and
energy measurements and first tests show
excellent response to single photons.
Where do we go from here ?

Strong European efforts are already
underway, based on state of the art high
energy semi-conductor detectors



MEDIPIX collaboration at CERN is an EU
supported project
IMI collaboration is an Italian consortium of
research groups and industry
We have an agreement with MEDIPIX to
use their photon counting chip in
connection with our SDD devices. An R&D
program to convert bump-bonded chips to
wire-bonded chips has started.
Integrated Mammographic Imaging Project

The aim of the project is to build an innovative system for both morphologic and
functional mammography
The project has been financially supported by the Italian Government
The research fields are:






Monochromatic X ray tube for mammography
Digital mammographic system based on GaAs pixel detectors
Bump-Bonding technology for GaAs pixel detectors
High resolution mammoscintigraphic system
The project is managed by five Italian companies in collaboration with the
Universities of Ferrara, Napoli, Pisa, Roma “La Sapienza” and the “Istituto
Nazionale di Fisica Nucleare” (INFN)
3 years and 4 MEuro investment
Our short-term goals

Form an Interdisciplinary Medical Imaging Project at Wayne
State University (WSU-IMIP) between the Physics Department,
the Medical School, and the Karmanos Institute

Collaborate with MEDIPIX on the electronics and with
Brookhaven National Laboratory on the Silicon Drift Detectors
Find a U.S. based industrial partner (Photon Imaging, Inc. ?)
Obtain funding through the WSU Life Science Corridor Initiative


What’s next ?? One organ in one breathhold
Computed tomography (CT) data acquisition
CT history in the past 30 years
New Applications ?
Somatom by SIEMENS
Digital Tomosynthesis
Image quality in digital tomosynthesis
Digital device requirements
The SVT-SDD Characteristics
R. Bellwied, NURT 2001
Wafers: B and T dependence



Used at B=6T. B fields
parallel to drift increase the
resistance and slow the drift
velocity.
The detectors work well up
to 50oC but are also very Tdependent. T-variations of
0.10C cause a 10% drift
velocity variation
Detectors are operated at
room temperature in STAR.
We monitor these effect via
MOS charge injectors
6.1
6.0
5.9
Drift Velocity ( mm/ns)

5.8
5.7
5.6
5.5
5.4
5.3
5.2
0
1
2
3
4
5
Magnetic Field (T)
R. Bellwied, NURT 2001
6
The SVT Multi Chip Module (Hybrid)
R. Bellwied, NURT 2001