Principles of Digital Radiography with Large-Area, Electronically ...

Harrell G. Chotas, MS
James T. Dobbins III, PhD
Carl E. Ravin, MD
Index terms:
Radiography, digital, * .121 2
Radiography, direction conversion
Radiography, indirect conversion
Radiography, selenium detector,
* .121
Radiography, technology
Radiology 1999; 210:595–599
Abbreviation:
CCD charged-coupled device
1 From the Departments of Radiology,
Digital Imaging Research Division
(H.G.C., J.T.D., C.E.R.), and Biomedical
Engineering (J.T.D.), Duke University
Medical Center, Box 3302, Rm 139,
Bryan Research Bldg, Research Dr,
Durham, NC 27710. Received February
24, 1998; revision requested April
2; final revision received July 7; accepted
September 11. Address reprint
requests to H.G.C.
2
* . Multiple body systems
RSNA, 1999
Review
Principles of Digital
Radiography with Large-Area,
Electronically Readable
Detectors: A Review of
the Basics 1
The practice of digital radiographic imaging is poised to undergo dramatic change in the
very near future owing to a rapid proliferation of electronically readable x-ray detectors.
Although self-scanning, direct-readout digital detectors have been in use since the
introduction of the charged-coupled device (CCD) almost 30 years ago, recent advances in
manufacturing technology have made possible a new generation of large-area, flat-panel
detectors with integrated, thin-film transistor readout mechanisms. The excitement
surrounding this new technology is based on two factors—the promise of very rapid access
to digital images wherever radiography with stationary x-ray equipment is performed and
the anticipation of image quality that exceeds that of both screen-film receptors and
photostimulable storage phosphor computed radiographic systems because of improvements
in x-ray detector technology.
As digital radiography continues this rapid evolution, it is likely that radiologists will be
inundated with information concerning a wide variety of large-area, flat-panel electronic
detectors. Unfortunately, it is also likely that there will be a tendency to think of these
devices as equivalent, interchangeable commodities because they will be similar in physical
size, appearance, and targeted applications. It must be emphasized, however, that
important differences exist among these detectors, and differences in digital image quality
among the various systems are inevitable and may be quite large. To minimize confusion,
therefore, it is important that radiologists have a working understanding of this emerging
technology.
In this report, we provide an overview of digital electronic x-ray detectors, including two
broad classes of detectors based on thin-film transistor arrays and the older, CCD-based
designs. Computed radiographic systems based on photostimulable storage phosphors are
omitted from this report because they do not contain integrated readout mechanisms (1).
Our goal is to provide a brief review of the basic methods, designs, and materials used in
direct-readout radiographic systems and to emphasize important characteristics that may
affect system performance and image quality. The advantages and disadvantages of the
different detectors, as well as the important factors that should be considered when
performing a critical analysis of these new digital imaging systems, are discussed.
DIRECT VERSUS INDIRECT CONVERSION
Electronic x-ray detectors can be divided into two classes—those in which direct methods
are used to convert x rays into an electric charge and those in which indirect methods are
used (Fig 1). (‘‘Direct conversion’’ should not be confused with ‘‘direct readout,’’ which is a
capability of all electronic detectors.) Direct-conversion detectors have an x-ray photoconductor,
such as amorphous selenium, that directly converts x-ray photons into an electric
charge. Indirect-conversion detectors, on the other hand, have a two-step process for x-ray
detection; a scintillator is the primary material for x-ray interaction. When x rays strike the
scintillator, the x-ray energy is converted into visible light, and that light is then converted
into an electric charge by means of photodetectors such as amorphous silicon photodiode
arrays or CCDs. In both direct- and indirect-conversion detectors, the electric charge
pattern that remains after x-ray exposure is sensed by an electronic readout mechanism,
and analog-to-digital conversion is performed to produce the digital image.
595

LARGE-AREA, THIN-FILM
TRANSISTOR ARRAYS
Recent advances in photolithography and
electronic microfabrication techniques
have enabled the development of largearea
x-ray detectors with integrated readout
mechanisms based on arrays of thinfilm
transistors. Unlike older, CCD-based
detectors that require optical coupling
and image demagnification (discussed
later), thin-film transistor–based, flatpanel
systems are constructed such that
the pixel charge collection and readout
electronics for each pixel are immediately
adjacent to the site of the x-ray interactions.
Several investigators (2,3) have
reported prototypic thin-film transistor–
based systems with x-ray–sensitive elements
that achieve detective quantum
efficiencies of 65% or more; this greatly
exceeds the performance of photostimulable
storage phosphor systems, which
have efficiencies of 20%–35% (4), and
of screen-film systems for chest radiography,
which have nominal efficiencies
of 25% (5), and is comparable to the
performance of a commercial seleniumdrum
digital chest radiographic system
(6).
Thin-film transistor arrays are used as
the active electronic elements in both
indirect- and direct-conversion, flat-panel
detectors. Thin-film transistor arrays are
typically deposited onto a glass substrate
in multiple layers, beginning with readout
electronics at the lowest level and
followed by charge collector arrays at
higher levels. Then, depending on the
type of detector being constructed, x-ray
elements, light-sensitive elements, or
both, are deposited to form the top layer
of this ‘‘electronic sandwich,’’ and the
entire assembly is encased in a protective
enclosure with external cabling for computer
connection.
THIN-FILM TRANSISTOR-DIRECT
CONVERSION
Direct-conversion systems based on thinfilm
transistor arrays are constructed by
adding an x-ray photoconductor as the
top layer of the electronic thin-film transistor
sandwich (Fig 2). Typically, amorphous
selenium is used as the photoconductor
material because of its excellent
x-ray detection properties (6–8) and extremely
high intrinsic spatial resolution
( 20 line pairs per millimeter [lp/mm] at
100 keV) (9). Before exposure, an electric
field is applied across the amorphous
Figure 1. Direct-readout electronic x-ray detectors use either a direct technique or an indirect
technique for converting x rays into an electric charge. Direct-conversion detectors have an x-ray
photoconductor, such as amorphous selenium, that converts x-ray photons into an electric charge
directly, with no intermediate stage. Indirect-conversion devices have a scintillator that first
converts x rays into visible light. That light is then converted into an electric charge by using an
amorphous silicon photodiode array or a CCD. Thin-film transistor (TFT) arrays may be used in
both direct- and indirect-conversion detectors.
Figure 2. Direct-conversion thin-film transistor detectors use a uniform layer of evaporated
amorphous selenium to convert x rays into electron-hole pairs. Electric fields that are established
within the selenium by means of bias voltages channel the charge to the nearest collector,
which preserves spatial resolution. Very high fill factors are achievable with appropriate electrode
design.
selenium layer through a bias electrode
on the top surface of the selenium. As x
rays are absorbed in the detector, electrons
and holes are released within the
selenium, and owing to the electric field
within the selenium, electric charges are
drawn directly to the charge-collecting
electrodes. Pixels are effectively separated
by means of field shaping within the
selenium layer, and the entire selenium
surface is available for x-ray charge conversion.
Thus, with properly designed charge-
collection electrodes, effective fill factors approaching
100% are achievable (3,10).
Amorphous selenium is well developed
technologically; it has been used for decades
as a photoconductor in photocopiers
and in an x-ray imaging technique
known as xeroradiography (11,12).
Because selenium is used in its amorphous
form, selenium plates can be
made by means of evaporation and thus
can be made relatively easily and inexpensively.
596 • Radiology • March 1999 Chotas et al

Figure 3. In indirect-conversion detectors, the x-ray scintillator can be structured or unstructured.
Structured scintillators, which typically are crystalline cesium iodide, reduce the spread of
visible light; this improves spatial resolution and permits the use of thicker scintillator materials
for improved quantum detection.
THIN-FILM
TRANSISTOR-INDIRECT
CONVERSION
Indirect-conversion systems based on
thin-film transistor arrays are constructed
by adding amorphous silicon photodiode
circuitry and a scintillator as the top
layers of the thin-film transistor sandwich.
These layers replace the x-ray photoconductor
layer that is used in directconversion
devices. When x rays strike
the scintillator, visible light is emitted
proportional to the incident x-ray energy.
Visible light photons are then converted
into an electric charge by the photodiode
array, and the charge collected at each
photodiode is converted into a digital
value by using the underlying readout
electronics.
The scintillators used in indirect-conversion
detectors can be either structured
or unstructured (Fig 3). With an unstructured
scintillator, such as conventional
fluorescent screens, visible light that is
emitted in the material can spread to
adjacent pixels and thereby reduce spatial
resolution. To reduce this problem, some
manufacturers now use a structured scintillator
that consists of cesium iodide
crystals that are grown on the detector.
The crystalline structure, which consists
of discrete and parallel ‘‘needles’’ approximately
5–10 µm wide, behaves similar to
a bundle of light pipes and channels most
of the signal directly to the photodiode
layer. Because light spreading is greatly
reduced in a structured scintillator, thicker
layers of this material can be used in the
detector; this increases the number of
x-ray photon interactions and thus the
available visible light. Although there are
trade-offs and practical limitations to this
approach (13), use of thicker layers of
structured scintillators greatly increases
the potential detective quantum efficiency
that is achievable by using the
detector, especially at higher imaging frequencies.
For this reason and because of
its high x-ray absorption efficiency, most
manufacturers of indirect detector systems
appear to be moving toward the use
of structured cesium iodide scintillator
materials.
CHARGED-COUPLED DEVICES
The basic CCD consists of a series of
metal-oxide-semiconductor capacitors
that are fabricated very close together on
a semiconductor surface. Originally invented
by Bell Laboratories in 1969 in an
effort to create new computer memory
devices, CCDs were quickly adapted for
use as photodetectors when their sensitivity
to visible light was recognized. Today,
CCDs are used in a wide variety of indirect-conversion
x-ray imaging devices, including
large-area radiographic imaging
systems and the familiar image-intensifier
television system that is used in fluoroscopy.
The single most salient characteristic of
CCDs with regard to digital radiography
is that they are physically small—typically
2–4 cm 2 , which is much smaller
than typical projected x-ray areas. Because
of this, cost-effective CCD-based
radiographic systems must include some
means of optical coupling to reduce the
size of the projected visible light image
and transfer the image to the face of one
or more CCDs. Some CCD-based systems
have an image intensifier that reduces the
large x-ray field to the size of one CCD.
Other systems are based on an array of
CCD cameras, each of which is coupled to
a scintillator by a lens or a fiberoptic
taper. The lens system substantially reduces
the number of photons that reach
the CCD; this increases the appearance of
image noise (ie, quantum mottle) and
degrades image quality. In addition, optical
coupling with lenses can introduce
geometric distortions, optical scatter, and
reduced spatial resolution. Fiberoptic
tapers can be used in place of optical lens
systems to mitigate light loss and optical
scatter problems to some degree, but imperfections
in the optical fiber bundles
can introduce structure artifacts on the
image. Finally, thermal noise within the
CCD itself also can degrade image quality,
although this is less of a factor with
modern, cooled CCDs.
SYSTEM CONSIDERATIONS
If recent professional radiology conferences
are any indication, numerous digital
radiographic systems based on directreadout,
flat-panel x-ray detectors will
soon be available commercially. Because
all of these detectors are compact, electronically
readable devices that offer rapid
image capture and direct connection to
computers in an imaging network, it may
appear to many radiologists that they are
equivalent, interchangeable commodities.
As indicated previously, this is not
the case.
Generally speaking, it is our opinion
that digital radiographic systems based
on thin-film transistor arrays will ultimately
provide superior image quality
relative to CCD-based detectors, particularly
when large-area detectors are required.
This is owing principally to the
requirement of CCD systems for either
image demagnification or a more costly
design with many tiled CCDs. In comparing
systems based on thin-film transistor
arrays, however, it is not yet clear whether
direct-conversion or indirect-conversion
designs will ultimately provide the preferred
solution. It may well be that optimum
system selection will depend on the
intended application (eg, static imaging
or fluoroscopy), although it is premature
to make such a determination at this
time.
Evaluation and selection of a digital
radiographic system should involve a
thorough analysis of the complete imaging
system, including the x-ray detector
itself and the environment in which the
Volume 210 • Number 3 Review of Principles of Digital Radiography • 597

system will be used. The following considerations
are offered to facilitate the evaluation
of electronic digital radiographic
systems.
Detector Size
The physical dimensions of the x-ray
detector have an obvious influence on
the radiographic examinations performed,
and it is critical that the detector
be both sufficiently large to capture the
desired anatomic views and sufficiently
compact for the desired applications. Electronic
detectors for standing radiographic
examinations, for example, would ideally
have an active detector that is at least
43 43 cm (ie, 17 17 inches) in size to
allow both vertical (35 43-cm) and
horizontal (43 35-cm) imaging orientations
without detector rotation. With
good mechanical design, a rectangular
35 43-cm (ie, 14 17-inch) detector
that can be easily rotated may perform as
well as a larger detector.
Spatial Resolution
Image spatial resolution, often expressed
as the modulation transfer function,
can vary substantially, depending
on physical detector characteristics. The
intrinsic spatial resolution of amorphous
selenium (with direct conversion) and of
structured cesium iodide (with indirect
conversion) is higher than that of unstructured
scintillators, and the intrinsic resolution
of selenium is much higher. Digital
images can be processed to alter apparent
image sharpness; however, excessive processing
can lead to an increase in perceived
noise. Thus, the inherent detector
modulation transfer function, when expressed
as a function of spatial frequency,
is not as useful as a measure of overall
system performance as is the detective
quantum efficiency. (See the Image Quality
section that follows.)
Image Quality
Although observer studies are the most
conclusive methods of determining overall
system performance, the difficulty in
performing such studies for all clinical
detection tasks makes it desirable to find
a more suitable physical measurement
that gives an indication of overall system
performance. Although there is no physical
measurement that correlates perfectly
with perceived diagnostic quality, and
there are unique difficulties associated
with physical image assessment in digital
systems (14), detective quantum efficiency
is generally accepted as the best
single objective indicator of image fidelity.
Detective quantum efficiency combines
spatial resolution (ie, modulation
transfer function) and image noise (ie,
noise power spectrum) to provide a measure
of the signal-to-noise ratio of the
various frequency components of the image.
Higher detective quantum efficiency
values suggest greater image quality, and
the results should be evaluated at all
frequencies to estimate the ability of the
image to depict both small and large
image structures.
Pixel Size and Matrix Size
The maximum spatial resolution on an
image is defined by pixel size and spacing
(ie, pitch). It should be emphasized, however,
that more pixels do not always
mean higher spatial resolution, because
image blurring can result from x-ray scatter,
light scatter, or both, in the detector.
For chest radiography, some investigators
(15,16) have concluded that 0.2-mm pixel
spacing (2.5 lp/mm, approximately 2,000
pixels per row) is adequate for most diagnostic
tasks, and in one commercial selenium-drum
digital chest radiographic system,
these 2K images have been reported
to offer superior depiction of anatomic
structures and an image appearance that
is preferred over that of conventional
screen-film radiographs. For mammography,
smaller pixel sizes—probably in the
range of 50–100 µm per pixel—are required.
Although some detectors that offer
substantially higher pixel pitch will
soon be available, it should be remembered
that the advantages that may be
gained by using larger image matrices
also have implications in practical implementation,
including increased network
traffic, increased archiving requirements,
and the likely need for image reduction
when images are displayed on video
monitors.
Monolithic Panels versus Tiled
Arrays
Because fabrication of full-size thinfilm
transistor detector panels is technically
challenging and can have relatively
low yields, many manufacturers seek to
reduce costs by constructing detectors
that consist of two or more smaller panels
in a tiled configuration. In these detectors,
digital image processing is used to
‘‘stitch together’’ the image sections to
eliminate the appearance of the tile junc-
tions. Although attractive in concept, this
approach can be problematic in that it is
difficult to seamlessly reconstruct a complete
image without compromising image
quality (17). In addition, the effects of
physical and thermal stress on the longterm
structural integrity of tiled detectors
has not yet been determined.
Bad Pixels and Rows
Perfect thin-section transistor arrays are
difficult to make, so practical limits with
regard to the number of defects allowed
in these devices must be established by
each manufacturer. Defects may manifest
as individual bad pixels or entire bad
pixel columns or rows. Image corrections
made by using system calibration masks
and additional image processing can ameliorate
some of these defects; however, as
with detector tiling, one must exercise
caution so that the image processing does
not alter the image in ways that adversely
affect the diagnosis.
Image Acquisition Time and Rate
The image quality with electronic detectors
is influenced by the details of circuit
design for charge collection, the charge
measurement technique, the analog-todigital
conversion rates during readout,
and in some detectors, the device recharging
between exposures. Generally, electronic
noise is the predominant noise
source in these detectors and the factor
that limits image quality, particularly at
the low-dose rates associated with fluoroscopy.
Whether one is performing fluoroscopy
or static radiographic imaging, the
precision of a pixel value (ie, the number
of gray levels) is related to the readout
time, with longer times generally resulting
in more precise image values. Because
image readout involves the active flow of
the current to and from each pixel element,
an incomplete charge transfer can
lead to inaccurate pixel values later if a
subsequent exposure is used without taking
the residual charge into account; the
result is an electronic memory artifact.
Similarly, rapid image acquisition can also
lead to image artifacts owing to x-ray–
induced charge trapping (18,19); however,
compensation for this effect may be
designed into the readout electronics of
the detector (20,21). For these reasons,
not all thin-film transistor system designs
are practical for use where high exposure
levels or high-speed image acquisition is
required (eg, in fluoroscopy); this factor
should be considered when selecting a
digital detector and how it will be used.
598 • Radiology • March 1999 Chotas et al

Dynamic Range
One of the principal advantages of
digital radiography is that image acquisition
and image display are decoupled;
this allows the detector to have a wide
dynamic range and linear response to
x-ray exposure. Radiologists should be
aware, however, that for a given analog-todigital
converter (ie, a given pixel depth),
there can be a trade-off between the detector’s
sensitivity range (ie, latitude) and
contrast resolution (ie, bits per pixel), and
some manufacturers may choose to limit
the dynamic range of the detector to
maximize the number of gray values on
the digital image. If used with an automatic
exposure control mechanism (ie,
phototimer), however, even detector
systems with limited latitude (ie, two
to three decades) should have no operational
problems in clinical use, because
the photodetector will ensure that the
exposure is appropriate for the available
dynamic range. If no phototimer is incorporated
into the system, it is recommended
that the detector have a minimum of four
decades of latitude, equivalent to the latitude
of photostimulable storage phosphor
systems, to avoid data loss owing to
overexposure or underexposure.
Antiscatter Grid
Any digital imaging system that includes
an antiscatter grid has the potential
for interaction between the grid lines
(ie, lead septa) and the rows of pixels that
constitute the digital image. For example,
grid line artifacts often manifest in a
‘‘corduroy pattern’’ of lines on the image
owing to image aliasing; however, some
high-strip-count grids do not leave apparent
grid lines on digital images. Artifacts
that can be produced by grid interaction
are not always obvious and can degrade
image quality, so caution and proper selection
of the grid is advised.
CONCLUSION
Electronically readable, large-area x-ray
detectors that promise rapid access to the
image for diagnosis, improved image quality
relative to that of screen-film and
storage phosphor–based radiography, reduced
patient examination time, a reduction
of consumable agents, and possibilities
for reduced patient exposure will
soon be commercially available for digital
radiography. While these devices will certainly
create many new opportunities in
diagnostic imaging, it is critical that radiologists
recognize that these new selfscanning
digital radiographic systems are
not interchangeable commodities and
can, in fact, produce images with very
different image quality.
References
1. Sonoda M, Takano M, Miyahara J, Kato H.
Computed radiography utilizing scanning
laser stimulated luminescence. Radiology
1983; 148:833–838.
2. Siewerdsen JH, Antonuk LE, el-Mohri Y, et
al. Empirical and theoretical investigation
of the noise performance of indirect detection,
active matrix flat-panel imagers (AM-
FPIs) for diagnostic radiology. Med Phys
1997; 24:71–89.
3. Zhao W, Rowlands JA. Digital radiology
using active matrix readout of amorphous
selenium: theoretical analysis of detective
quantum efficiency. Med Phys 1997; 24:
1819–1833.
4. Dobbins JT, Ergun DL, Rutz L, Hinshaw
DA, Blume H, Clark DC. DQE(f) of four
generations of computed radiography acquisition
devices. Med Phys 1995; 22:
1581–1593.
5. Bunch PC. Performance characteristics of
asymmetric zero-crossover screen-film systems.
Proc SPIE 1992; 1653:46–65.
6. Neitzel U, Maack I, Gunther-Kohfahl S.
Image quality of a digital chest radiography
system based on a selenium detector.
Med Phys 1994; 21:509–516.
7. Papin PJ, Huang HK. A prototype amorphous
selenium imaging plate system for
digital radiography. Med Phys 1987; 14:
322–329.
8. Fahrig R, Rowlands JA, Yaffe MJ. X-ray
imaging with amorphous selenium: detec-
tive quantum efficiency of photoconductive
receptors for digital mammography.
Med Phys 1995; 22:153–160.
9. Que W, Rowlands JA. X-ray imaging using
amorphous selenium: inherent spatial
resolution. Med Phys 1995; 22:365–74.
10. Rowlands JA, Zhao W, Blevis I, Pang G, Ji
WG, Germann S. Flat panel detector for
digital radiology using active matrix readout
of amorphous selenium. Proc SPIE
1997; 3032:97–108.
11. Boag JW. Xeroradiography. Phys Med Biol
1973; 18:3–37.
12. Paulus DD. Xeroradiography: an in-depth
review. Crit Rev Diagn Imaging 1980;
12:309–384.
13. Chabbal J, Chaussat C, Ducourant T, et al.
Amorphous silicon x-ray image sensor.
Proc SPIE 1996; 2708:499–510.
14. Dobbins JT. Effects of undersampling on
the proper interpretation of modulation
transfer function, noise power spectra,
and noise equivalent quanta of digital
imaging systems. Med Phys 1995; 22:171–
181.
15. Floyd C Jr, Baker JA, Chotas HG, Delong
DM, Ravin CE. Selenium-based digital
radiography of the chest: radiologists’ preference
compared with film-screen radiographs.
AJR 1995; 165:1353–1358.
16. Woodard PK, Slone RM, Gierada DS, Reiker
GG, Pilgram TK, Jost RG. Chest radiography:
depiction of normal anatomy
and pathologic structures with seleniumbased
digital radiography versus conventional
screen-film radiography. Radiology
1997; 203:197–201.
17. Slump CH, van Dijk PW, Laanstra G, et al.
Real-time diagnostic imaging with a novel
x-ray detector with multiple screen-CCD
sensors. Proc SPIE 1999; 3336:418–429.
18. Chotas HG, Floyd C Jr, Ravin CE. Memory
artifact related to selenium-based digital
radiography systems. Radiology 1997; 203:
881–883.
19. Kasap SO, Aiyah V. X-ray induced hole
trapping in electroradiographic plates. J
Appl Phys 1991; 69:7087–7096.
20. Zhao W, Blevis I, Germann S, Rowlands
JA, Waechter D, Huang Z. Digital radiology
using active matrix readout of amorphous
selenium: construction and evaluation
of a prototype real-time detector.
Med Phys 1997; 24:1834–1843.
21. Tsukamoto A, Yamada S, Tomisaki T, et al.
Development of a selenium-based flatpanel
detector for real-time radiography
and fluoroscopy. Proc SPIE 1999; 3336:
388–395.
Volume 210 • Number 3 Review of Principles of Digital Radiography • 599