Philips MicroDose

Photon counting technology

+100V

holes
 Index
Requirements for detectors. . . . . . . . . . . . . . . . . . . . . . . . . 3

Available detector technologies. . . . . . . . . . . . . . . . . . . . . . 4

Photon counting technology. . . . . . . . . . . . . . . . . . . . . . . . . 5

Detector design. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5

Readout electronics. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6

Noise rejection. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7

Weighting of X-ray photons with different energies. . . . . . 8

Anti-scatter grids. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9

Scanning technology and scatter rejection. . . . . . . . . . . . . 10

System DQE. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11

Summary of characteristics for Philips MicroDose. . . . . . 12

References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12

2
Requirements
for detectors
Of all medical X-ray applications, mammography puts
the highest demands on the imaging system. As a
result, mammography is the last X-ray-based imaging
technology to convert from film-screen to digital
detector techniques.

For the image quality to be optimized and the radiation

dose to be reduced, all the links in the imaging chain
have to be considered; no imaging system is stronger
than its weakest link. For a digital mammography system,
the following requirements have to be met:

High image quality

• High-spatial resolution is needed to visualize and
characterize the small structures in breast tissue. The
detector element size and spacing define the maximum
spatial resolution in an image. In mammography, a
spatial resolution of 50 µm to 100 µm is required. This
results in a limiting spatial resolution of 10 line pairs
per mm to 5 line pairs per mm respectively.
• High-contrast resolution is needed for differentiation
of breast tissues that have very similar X-ray
absorption.
• Wide dynamic range is important, as the image has
to include information from both dense regions of
glandular tissue, regions of fatty tissue, and all the way
out to the skin line.
• Low noise facilitates accurate diagnosis. Noise reduces
the contrast resolution.
• Absence of artifacts. Artifacts can for example
be caused by electronic noise in the detector, or
“ghost images.”
• No dead pixels. Dead pixels can be interpolated from
neighboring pixels; this will improve the appearance of
the image but not retrieve the lost information.

Low radiation dose

• High quantum efficiency allows the detector to
record every incident X-ray photon, maximizing dose
efficiency of the system.
• Efficient rejection of scattered radiation. Scatter
adds noise and reduces the contrast resolution in the
image, which has to be compensated by increasing
the exposure.

3
 Available detector
technologies
Conceptually, what differentiates the different detector
technologies is the number of intermediate conversion
steps before a digital signal is formed. This is shown
schematically in Figure 1, which indicates the stages at
which the signals from individual photons are added
(enclosed in green boxes). Each conversion step has
the potential to deteriorate image quality. For systems
that use techniques other than photon counting, the
initial digital signal from the X-rays is converted into an
analog signal, and after the exposure one attempts to
retrieve the original digital signal through an analog-to-
digital converter.

Image acquisition cycle

Decrea
sing nu
mber o
f noise
source
1985 s
Time fo
r intro
duction
1999

CR plates 2002
• X-ray excitation of electron
states a-Si / CCD 2003
• Accumulation of trapped charge • X-ray conversion to light
• Conversion to light with laser • Light conversion to electrical a-Se
beam charge • X-ray conversion to electrical
• Light conversion into electrical • Charge accumulation in pixel charge Photon counting
analog signal
• Readout of total charge in pixel • Charge accumulation in pixel • X-ray conversion to electrical
• Readout of total charge in pixel pulse

• A/D conversion of signal • A/D conversion of signal • A/D conversion of signal • Incrementation of counter if
charge above threshold

Digital Digital Digital Digital

output output output output

Figure 1 Schematic representation of different digital detector

technologies for mammography. Analog steps are above the red
line; digital steps are below the red line. The green boxes indicate
the stages at which the signals from individual photons are added.

4
Photon counting
technology
X-rays are often perceived as a continuous fluid of
energy. In reality this fluid consists of individual particles,
X-ray photon
so called X-ray photons, which react one-by-one in an
X-ray detector. The X-rays are thus digital by nature.
The Philips MicroDose detector is built to match
this digital nature of X-rays. The Philips MicroDose
detector increments the photon counter directly as
Photon
each individual X-ray is detected. This means that there
counter
are no intermediate steps that can add noise from
the detection of the photon to the pixel value in the
mammogram as indicated in Figure 2. Since there is a
risk in every conversion step of adding noise and losing
signal, significant effort by several companies has been
MicroDose detector
focused on reducing the number of conversion steps.

The Philips MicroDose detector is a spin-off from

high-energy physics research. It is an integrated result
from thousands of researchers working from 1960
to the discovery of the top quark in 1996.1 In high-
energy physics, e.g. the search for the top quark,
the requirements have been very high for detection
efficiency and for extracting maximum information from Digital signal
every detected particle, such as X-ray photons.

Figure 2 Philips MicroDose photon counting detector.

Detector Design
The detector system consists of a large number of
crystalline silicon strip detectors, with the strips tapered
to point back to the X-ray source. Silicon has excellent X-ray
X-ray photons
electronic properties and very good energy resolution. photon
A drawback is the relatively weak attenuation due to the
low atomic number (Z=14). This is overcome by placing
the strip detectors in edge-on geometry with their long
axis parallel to the direction of the X-ray beam. In this
way, a sufficient absorption length is created, resulting
in high quantum efficiency for the full energy spectrum
used in mammography.

Figure 3 Crystalline Silicon strip detectors with edge-on

geometry.

5
 The X-ray interaction in the Si transfers energy to
either a photo-electron or Compton electron that X-ray photons
subsequently excites electron hole pairs. Since only 3.6 +100V
eV is required to create an electron hole pair, several
thousand pairs are created for one incident X-ray
photon and there is no need for further amplification.
A bias voltage is applied across the Si-detector, which electrons
generates a bias field. This electrical field causes the holes
charge cloud to be swept towards electrodes with
positive voltage and the charge cloud of holes to be
swept to electrodes on the opposite side. This drift Figure 4 X-ray generates a cloud of electron-hole pairs that
induces a short electrical signal on the electrodes. The create a short electrical signal.
signal is transferred from the aluminum strip to the
pulse-shaping electronics through wire-bonds.

Due to the large market for microelectronics that has

emerged during the last 50 years, crystalline silicon
is by far the purest detector material available. The
existing production facilities are huge and standard batch
processing of excellent quality and high yield are routine
in a reliable production process that has been optimized
over the last 4-5 decades.

Readout Electronics
The custom-designed ASIC (Application Specific
Integrated Circuit) contains pre-amplifier, shaper,
comparator and a digital counter. All channels in the
ASIC work in parallel. The shaper creates a signal
around a few hundred nanoseconds in duration. The
comparator discriminates pulses above a certain
threshold level from the detector noise and counts
these pulses. The counter has 15 bits to enable wide
dynamic range. Each ASIC read-out channel can handle
count rates up to 2 MHz, allowing acquisition of images
with sufficient statistics in a few seconds. The number
of counted photons in each channel is accumulated
during a single sampling interval before being sent to the
workstation for image reconstruction and display.

6
 Noise rejection
The pulses created in the Philips MicroDose detector by For film-screen systems and scintillator-based digital
the incident photons are easily detected, since the pulse systems, Swank noise can be a serious cause of image
height (5000 electrons) is an order of magnitude higher degradation. This noise is introduced by the statistical
than the electronic noise in the detector (200 electrons variation of released light after the conversion of the
rms). Thus the detector counts the number of peak photon.2 Hence, two photons of the same energy
pulses and nothing else. With no conversion steps from can have different responses in the detector, which
X-ray photons to a digital signal; Swank noise, electronic translates to an overall uncertainty in the integrated
noise and noise from the quantification of the electrical signal. This effect is to some extent present in
signal from the detector can be removed almost entirely. integrating detectors such as amorphous silicon
detectors.
If all external noise sources are eliminated, the system
is said to work at the quantum limit. This is the The absence of Swank noise and the significant removal
ultimate limit for information provided by the laws of of electronic noise is a major advantage of the photon
physics. Ideally the X-ray photons in all pixels should counting technology.
be distributed according to the mathematical Poisson
distribution. These fluctuations are sometimes referred
to as quantum mottle or (wrongly) as quantum noise.
Energy

Lowest photon energy

Threshold

Electronic noise

X-ray X-ray X-ray Time

photon photon photon

Figure 5 The threshold is set to an optimum level for the

energy of the X-ray photons used in mammography, which
results in a quantum efficiency of more than 90%.

7
 Weighting of X-ray
photons with
different energies
To date, all mammography systems available on the 5 cm thick breast
market, except for the Philips MicroDose system, have 0.8
been integrating systems, i.e. the signal is formed by
a sum of liberated electric charge or light from X-ray 0.7
photon interactions. The amount of charge, or light, is
0.6
proportional to the photon energy. Therefore, with an
integrating technique a higher energy photon will gain
0.5
a higher statistical weight compared to less energetic

Contrast
photons. However, the attenuation in the breast, as 0.4
well as the contrast, is strongly dependent on energy. 0.1 mm micro-calcification
It is the low-energy photons that carry more contrast 0.3
information and should be given a higher statistical
0.2
weight. This subject was first discussed two decades
ago3 but has not generated much attention. The contrast 0.1 5 mm mass
is proportional to the linear attenuation coefficient of
the object being imaged. 0
10 15 20 25 30 35
E (keV)
The subject contrast is roughly proportional to E-3
(photon energy), practically independent of tissue type Figure 6 Low energy photons carry more contrast information.
and the size of the abnormality being detected. For
integrating systems the signal is proportional to energy,
and hence the statistical weight function is incorrect by
four powers.

A photon counting system with the same weight for all

photons is an improvement over current integration,
but does not provide the level of contrast sensitivity
afforded by optimal energy weighting. The numerical
example to the right is exaggerated to demonstrate
the principle. Detailed calculations show that a photon
counting system improves the dose efficiency by more
than 10%.4

8
Anti-scatter grids
The conventional way of reducing the scatter-to-primary

Scattered radiation
(S/P) ratio is to insert a Bucky grid between the breast
and the detector. The anti-scatter grids are comprised of
high aspect-ratio lamellae that absorb the rays that have
Radiation scattered in the breast is a common
been scattered at angles beyond acceptance. Depending
source of image degradation in mammography.
on the geometry, discrimination between scattered
Scattered radiation will reduce the contrast and
photons and primary photons is more or less efficient.
increase the noise, which must be compensated
However, for mammography energies, the grids will also
by increasing the dose to the patient. This effect
remove a significant fraction of the direct photons and
is particularly pronounced for thick breasts. It is
are mechanically very difficult to manufacture.
quantified by the (S/P), which is ideally zero but
can be of the order of unity in practice.
The Bucky factor expresses the increase in patient
dosage that is needed to maintain the exposure to the
detector when a grid is inserted. It is typically 2-3, and
contrast improvements exceeding 80% can be achieved. 5
The grid usually moves during the exposure to prevent it
from appearing in the image, but there are also high strip
density, stationary grids. Grids are usually linear, unable
to reject photons scattered along the absorbing strips.
In recent years, high transmission cellular grids with
air interspaces have exhibited improved performance
and become more widespread. However, studies show
that the use of a grid in conventional mammography
provides far from ideal scatter rejection. To quantify, we
introduce the scatter DQE as the ratio of squared SNR’s
between actual scatter rejection and perfect scatter
rejection. This can be calculated through:

Tp
SRE =
1 +S / P

where Tp is the transmission of primary photons.6

Using values from Ref.,7 we find that SRE is in the range
0.5–0.6 for a standard grid.

9
 Scanning technology
and scatter rejection
Achieving efficient scatter rejection is much easier for
a scanning system in which a small detector is moved
across the object to acquire the full-field image. The
main principle is, the smaller the detector used, the X-ray tube
lower the risk that a scattered photon will interact. That
means no grid is needed in a scanning system and thus
a major mechanical simplification is attained since the
mechanical construction to achieve a linear translation
Collimator 1
of a slit is much simpler compared to the complicated
mechanics of a Bucky.
Object
Systems are divided into slot and slit scanning systems.
For the former, a 2D detector is used, while the latter Collimator 2
has a linear detector covering only one pixel row. A
multi-slit system has several of these linear slits to
Detector
create better photon economy, i.e. using a larger portion
of the output from the X-ray tube. In this way, the image
acquisition time and/or tube load can be reduced. Moving parts
Scan motion
The load for the X-ray tube will be higher for a scanning
system and this is compensated for by increasing the
diameter of the anode disc from 3 inches (standard Figure 7 Multi-slit scanning technology used in Philips MicroDose.
in mammography) to 4 inches, which is a minor
modification of the relatively simple X-ray tubes
used in mammography.

Philips MicroDose is of the multi-slit type. Photons

that are not initially directed towards the detector
are removed from the beam by a pre-collimator.
These photons are useless and would otherwise only
contribute to patient dose. The post-collimator absorbs
the photons scattered in the breast and not directed
towards the detector. During the continuous scan
motion, the detector is read every 2 ms. Besides the
cost advantages of using a smaller detector, the main
advantage of this technique is the efficient rejection
of scattered radiation. The S/P-ratio is only about 3%
without any loss of primary radiation.8

This inherent scatter rejection is another reason

why Philips MicroDose can reduce the radiation dose
significantly compared to that used on conventional
systems.

10
System DQE
Scatter DQE Detector DQE
Scatter DQE shows how well the system rejects Detector DQE is the conventional measure of how a
scattered radiation from the breast. Detection of regular pattern is imaged by the system and especially
scattered X-rays decreases the contrast and degrades shows how this is degraded by noise in the system.
image quality. Conventional Bucky grids are far from Degradation by scattered radiation is not included
perfect and have a scatter DQE of roughly 50%. The here. Neither does this measure show how contrasts
scanning multi-slit geometry of Philips MicroDose in actual breast tissue are imaged.9
provides an excellent scatter rejection and a scatter
DQE of 96%.8

Energy weighting
Energy weighting describes how the information in
the X-ray beam is captured by the detector. The low
energy photons contain more information than the
high energy ones, and should thus be weighted higher
by the detector. Energy-integrating systems do the
opposite, while photon counting detectors put equal
weight on each photon. The energy weighting factor
shows how much efficiency is lost compared to a
photon counting system.4

Traditionally, X-ray detectors have been compared

through the detective quantum efficiency (DQE).
With the advent of new technologies, such as photon
counting and multi-slit-scanning, detector DQE is an
inadequate measure of the complete imaging system’s
performance. A better figure-of-merit is the system
DQE, in which several additional properties have been
taken into account. The system DQE is composed of
three factors:
• Detector DQE
• Scatter DQE
• Energy weighting

11
 Summary of
characteristics for
Philips MicroDose
• Matches the digital nature of X-rays; each X-ray is
processed individually and external noise is almost
entirely eliminated High Quantum Efficiency

• Almost perfect scatter rejection

• 50 µm pixel size

• Field of view 24 x 26 cm

• Scan time 3-15 seconds

• Tungsten anode, 0.5 mm Al filter

• All radiation hits active detector area,

no radiation “spill”

• Maximum count rate 2 MHz per pixel and

no “ghost images”

• No dead pixels

References
1 Abe, F. et al., 1994. Evidence for Top Quark Production in pp Collisions 7 Dance, D.R., Persliden, J., Carlson, G.A., 1992. Calculation of dose and
at sqrt(s) = 1.8 TeV. Fermilab-PUB-94/116-E. May 16, pp.1-19 contrast for two mammography grids. Physics in Medicine and Biology,
2 Swank, R., 1973. Absorption and noise in X-ray phosphors. Journal of 37(1), pp.235-48.
Applied Physics, 44, pp.4199– 203. 8 Åslund, M., Cederström, B., Lundqvist, M., Danielsson, M., 2006.
3 Tapiovaara, M., Wagner, R., 1985. SNR and DQE analysis of broad Scatter rejection in multi-slit digital Mammography. Medical Physics,
spectrum X-ray imaging. Physics in Medicine and Biology, 30, pp.519–29. 33(4), pp.933-40.
4 Cahn, R.N., Cederström, B., Danielsson, M., Hall, A., Lundqvist, M., 9 Dobbins III, J.T., 2000. Image Quality Metrics for Digital Systems. In:
Nygrenn, D., 1999. Detective Quantum Efficiency dependence on X-ray J. Beutel, H.L. Kundel, R.L. van Metter, ed. 2000. Handbook of Medical
energy weighting in Mammography. Medical Physics, 26(12), pp.2680-3. Imaging, Volume 1. Physics and Psychophysics. Bellingham WA, USA: SPIE
5 Rezentes, P., de Almeida, A., Barnes, G., 1999. Mammography grid Press. pp.161-219
performance. Radiology, 210(1), pp. 227–32.
6 Wagner, R., Barnes, G.T., Askins, B.S., 1980. Effect of reduced
scatter on radiographic information content and patient exposure: A
quantitative demonstration. Medical Physics, 7(1), pp.13-8.

Please visit www.philips.com/microdose

© 2011 Koninklijke Philips Electronics N.V. Philips Healthcare is part of Royal Philips Electronics
All rights are reserved.
www.philips.com/healthcare
[email protected]
Philips Healthcare reserves the right to make changes in specifications and/
or to discontinue any product at any time without notice or obligation
and will not be liable for any consequences resulting from the use of this Printed in The Netherlands
publication. 4522 962 79741 * OCT 2011