Clinical Radiology (2008) 63, 1073e1085

REVIEW

The design and imaging characteristics ofdynamic, solid-state, flat-panel x-ray imagedetectors for digital fluoroscopy and fluorography

A.R. Cowena,*, A.G. Daviesa, M.U. Sivananthanb

aLXi_Research, The University of Leeds, Division of Medical Physics, and bDepartment of Cardiology,Yorkshire Heart Centre, The General Infirmary, Leeds, West Yorkshire, UK

Received 28 March 2008; accepted 4 June 2008

Dynamic, flat-panel, solid-state, x-ray image detectors for use in digital fluoroscopy and fluorography emerged at the

turn of the millennium. This new generation of dynamic detectors utilize a thin layer of x-ray absorptive materialsuperimposed upon an electronic active matrix array fabricated in a film of hydrogenated amorphous silicon(a-Si:H). Dynamic solid-state detectors come in two basic designs, the indirect-conversion (x-ray scintillator based)and the direct-conversion (x-ray photoconductor based). This review explains the underlying principles and enablingtechnologies associated with these detector designs, and evaluates their physical imaging characteristics, comparingtheir performance against the long established x-ray image intensifier television (TV) system. Solid-state detectorsafford a number of physical imaging benefits compared with the latter. These include zero geometrical distortionand vignetting, immunity from blooming at exposure highlights and negligible contrast loss (due to internal scatter).They also exhibit a wider dynamic range and maintain higher spatial resolution when imaging over larger fields ofview. The detective quantum efficiency of indirect-conversion, dynamic, solid-state detectors is superior to that ofboth x-ray image intensifier TV systems and direct-conversion detectors. Dynamic solid-state detectors are playinga burgeoning role in fluoroscopy-guided diagnosis and intervention, leading to the displacement of x-ray image inten-sifier TV-based systems. Future trends in dynamic, solid-state, digital fluoroscopy detectors are also briefly consid-ered. These include the growth in associated three-dimensional (3D) visualization techniques and potentialimprovements in dynamic detector design.ª 2008 The Royal College of Radiologists. Published by Elsevier Ltd. All rights reserved.

Introduction

Diagnostic and interventional radiology have a con-tinuing requirement for dose-efficient x-ray-basedmodes of imaging to visualize moving anatomicalstructures, organs and/or clinical devices (e.g.,guide-wires, catheters, stents, pacemakers,etc).1 Historically, real-time dynamic imaging us-ing x-rays (for the purpose of procedural guidance)

* Guarantor and correspondent: A.R. Cowen, LXi_Research,The University of Leeds, Division of Medical Physics, WorsleyBuilding, Clarendon Way, Leeds LS2 9JT, West Yorkshire, UK.Tel.: þ44 0113 3438312.

E-mail address: a.r.cowen@leeds.ac.uk (A.R. Cowen).

0009-9260/$ - see front matter ª 2008 The Royal College of Radiolodoi:10.1016/j.crad.2008.06.002

has been referred to as ‘‘fluoroscopy’’. The serialacquisition of x-ray images of higher quality foruse in diagnosis and documentation is known as‘‘fluorography’’. Such imaging techniques are com-monly associated with contrast medium aided ex-aminations of the gastrointestinal (GI) tract, thecardiovascular system, and various other soft-tissue organs and structures. During the secondhalf of the 20th century fluoroscopy has been sup-ported by the electronic imaging device known asthe x-ray image intensifier television (IITV) system.Such a system comprises a chain of electron-optical imaging components including an x-rayimage intensifier tube, suitable coupling lensesplus a high specification closed-circuit television

gists. Published by Elsevier Ltd. All rights reserved.

1074 A.R. Cowen et al.

channel combined with a suitable electronic dis-play.2 The TV image is either recorded by an elec-tronic camera tube (e.g., a Plumbicon, Saticon,Chalnicon, etc) or a semiconductor charge-coupleddevice (CCD) sensor.3 Modern IITV fluoroscopysystems are capable of producing good-quality,dynamic x-ray images with economical use of radi-ation dose.

The emergence of digital subtraction angiogra-phy (DSA) imaging equipment circa 1980 pioneeredthe integration of computerized video processorsand magnetic storage discs with x-ray IITV sys-tems.4 The success of DSA fuelled a seminal era inthe development of digital fluoroscopy/fluorogra-phy equipment, which continued through the1990s. The availability of user-friendly, high-performance, digital x-ray IITV systems led to aradical shift in clinical imaging practice. This in-cluded the replacement of spot-film-based record-ing of clinical results by digital (computerized)fluorography in the screening room. This made itpossible to access and replay sequences of fluoro-graphic images on-line, and to view them in a digi-tally enhanced form. At the same time there was anenthusiastic adoption of radiation dose-saving mea-sures, such as digital recursive filtering (to amelio-rate noise), last image hold and two-dimensional(2D) road-mapping, further increasing the clinicalusefulness of digital fluoroscopy.5 Digital x-rayIITV systems proved flexible and effective plat-forms for expanding the range of dynamic imageacquisition protocols. Digital x-ray IITV systemshave been a crucial (albeit largely unsung)enabling technology in modern radiology. Notablythey have underpinned the growth in x-ray image-guided interventional radiology. At the turn of thenew millennium, the digital x-ray IITV system wasthe dominant image receptor not only for routinefluoroscopy, but also dynamic x-ray image acquisi-tion in general. Around this time, however, a newgeneration of dynamic image detectors first ap-peared, which has subsequently gone on tothreaten the established role of digital x-ray IITVsystems.

Solid-state, flat-panel detectors were originallydesigned for use in standard projection radiogra-phy; the basic physical and technical characteris-tics of these devices were described in thepreceding review.6 Solid-state digital radiography(DR) detectors provide on-line access to the elec-tronic signal data, so the radiographic images areavailable to view in a matter of seconds after theexposure, (rather than after delays of several min-utes more typical of conventional and computedradiography). Significantly, researchers foundthat with suitable technical optimization these

solid-state detectors can be used equally well torecord and read-out images at rates high enoughto support fluoroscopy.7,8 Prototype clinical dy-namic solid-state detector systems started to ap-pear toward the end of the 1990s.9e15 The firstcommercial solid-state detector-based digital fluo-roscopy products became available in 2001; thesedetectors were designed specifically for cardiacimaging.16e18 In recent years most new cardiaccatheterization laboratories have utilized solid-state detectors, ousting digital x-ray IITV fromone of its most celebrated clinical roles. With therecent introduction of dynamic, solid-state detec-tors of larger area, a similar shift away from digitalx-ray IITV systems is now occurring in radiographyand fluoroscopy and vascular imaging.19e25 Theaim of this review is to describe the physical designand imaging characteristics of the dynamic, solid-state, flat-panel x-ray image detectors that aredriving this trend.

Dynamic x-ray detector design

Currently, the majority of dynamic solid-statedetectors in clinical use are based upon the so-called ‘‘indirect conversion principle’’.6 These de-tectors exploit the conversion of x-ray energy tolight photons in a layer of thallium-activatedcaesium iodide (CsI:Tl). The emitted light is thenconverted to an electronic signal in a 2D array oflight-sensitive elements (i.e., photodiodes), fabri-cated in a thin layer of hydrogenated amorphoussilicon (a-Si:H). CsI:Tl is a very similar scintillatorto the CsI:Na used in x-ray image intensifier tubes.Caesium and iodine have comparatively highatomic numbers [Z¼ 55 and 53, respectively],and as such have good x-ray absorption properties.Additionally they exhibit a boost in x-ray absorp-tion at photon energies exceeding their k-edges,at 36 and 33 keV for caesium and iodine, respec-tively. This ensures efficient absorption of x-rayphotons over the energy range that is most rele-vant to fluoroscopy and fluorography. The CsI:Tllayer has a columnar (pillar-like) crystal micro-structure. Consequently, this phosphor has a highpacking density (w90%), again helping to maximizex-ray absorption. The channelled (fibre-optic like)micro-structure of CsI:Tl helps minimize scatter ofthe fluorescent light emission. As a result, compar-atively thick layers of scintillator can be employedbefore spatial resolution is degraded significantly.For dynamic x-ray imaging applications a relativelythick CsI:Tl layer (typically 550e650 mm) is used tomaximize detector sensitivity (and, therefore,minimize patient dose). The CsI:Tl layer absorbs

The design and imaging characteristics of dynamic x-ray image detectors 1075

80e90% of the incident x-ray photons. Each x-rayphoton absorbed in CsI:Tl yields w3� 103 lightphotons, predominantly in the green portion ofthe optical spectrum. Of these light photons, ap-proximately half are recorded by the 2D array ofphotodiodes and contribute to the electronic sig-nal. The scintillator is grown or mounted (depend-ing on the detector design) on top of the a-Si:Hactive matrix array. A light reflector may becoated on the top surface of the CsI:Tl to minimizethe loss of fluorescent light and to maximize signalgain (albeit at the cost of reduced spatialresolution).

Solid-state, digital fluoroscopy systems utilizepulsed-mode x-ray exposure. In this mode thedetector is exposed to a sequence of moderateto high intensity x-ray pulses of short duration(w5e20 ms depending upon the type of examina-tion and patient size). Pulsed-mode fluoroscopyrequires a powerful grid-controlled x-ray source.This enables the pulses of radiation to be deliveredrapidly and precisely, and as a result, motion-linked artefacts and unsharpness (blur) are keptto a minimum.26 Each succeeding x-ray pulse pro-duces fluorescent light that illuminates the photo-diode array releasing electrical charge carriersthat are (temporarily) stored in the matrix array.The magnitude of the charge packet stored ata particular pixel is proportional to the dose ab-sorbed at that location. Each pixel in the activematrix array comprises a photodiode plus an asso-ciated thin film transistor (TFT) switch. The 2D ar-ray of TFT switches are addressed sequentially viaa set of (horizontal) gate control lines. The signalinformation is, therefore, read out line-by-line tothe external circuitry via a set of (vertical) datatransfer lines. Fluoroscopic image frames are typ-ically acquired at rates of up to 30 frames s�1.Where clinical circumstances permit the framerate can be reduced to 15 or 7.5 frames s�1 (oreven lower) to moderate patient dose. The outputsignal is then amplified prior to digitization andtransfer to the system computer. Dynamic se-quences of images are finally viewed on a suitabledisplay device in the procedure room, or at a sepa-rate viewing station.

Alternatively, dynamic, solid-state detectorscan be designed to directly convert x-ray energyto electronic charge. This detector utilizes a layerof a-Se x-ray photo-conductor superimposed uponthe a-Si:H active matrix.6 The latter comprisesa 2D array of charge-sensing electrodes, storagecapacitors, and TFT switches. Although the densityof a-Se is similar to that of CsI:Tl, selenium hasa much lower atomic number (Z¼ 34), and thek-absorption edge lies at 13 keV (a very low

energy). For the beam energies used in fluoroscopythe x-ray absorption efficiency of a-Se is signifi-cantly lower than that of an equivalent thicknessof CsI:Tl. In addition, the x-ray absorption effi-ciency of a-Se falls more rapidly with increasingbeam energy than it does with CsI:Tl. Conse-quently, when used in dynamic detectors the a-Selayer thickness is increased to 1000 mm, from the500 mm normally used in radiographic versions.Absorption of x-ray photons in the a-Se layer re-leases electronic charge carriers directly. A biasvoltage is applied across the a-Se layer to transferthe charge carriers to the appropriate signal elec-trode. Due to the high strength of the resultingelectric field charge carries rapidly cross the a-Selayer with negligible lateral diffusion. In theorynegligible loss of spatial resolution during imagecapture should benefit image quality; however, inpractice fluoroscopic image quality will be de-graded due to the effects of noise aliasing.6 The2D array of stored charge packets are read outusing a mechanism similar to that used in indirectconversion detectors (as described above).

Dynamic, solid-state detectors have a compact,flat-panel construction. A sectional view througha dynamic, solid-state, flat-panel detector (indi-rect conversion type) is shown in Fig. 1. Compo-nents of the detector identified in this figureinclude the surface light reflector, the CsI:Tl layer,the a-SiH active matrix array, glass substrate plusthe refresh light (whose function is explainedlater). The detector depicted is the Pixium 4800(Trixell SA, France); this detector was specificallydesigned for use in cardiac imaging.16 A workingcardiac catheterization laboratory employing thisdetector is shown in Fig. 2. Recently imaging sys-tems incorporating dynamic, solid-state detectorswith larger fields of view (up to 40 cm� 40 cm)have become available. This has broadened therange of clinical applications that can be supportedby solid-state detectors to include radiography andfluoroscopy and vascular examinations.19e25 Thevascular imaging system depicted in Fig. 3 incorpo-rates the large-field Trixell Pixium 4700 dynamicdetector, (which again exploits indirect conver-sion). This design of dynamic detectors is incorpo-rated in medical x-ray imaging devicesmanufactured by Philips Healthcare and SiemensMedical Solutions (in Europe). Other leading manu-facturers of dynamic solid-state detector systemsinclude GE Healthcare and Varian Medical Systems(in the USA) and Toshiba Medical Systems andShimadzu Medical Systems (in Japan). The two for-mer companies utilize indirect-conversion dynamicdetectors, whereas the latter two companies usedirect-conversion detectors.

Figure 1 Cross-section through a dynamic, solid-state, flat-panel detector, identifying the surface reflector, CsI:Tllayer, a-SiH active matrix array, and refresh light (Trixell SA Pixium 4800 detector). Reproduced with the permission ofMedicamundi.

1076 A.R. Cowen et al.

Physical imaging characteristics

The physical image quality of dynamic digital x-rayimage detectors can be evaluated using a toolkit ofparameters such as: dynamic range; geometricaldistortion, vignetting, and veiling glare; spatialresolution; temporal resolution (lag and memoryeffect); and detective quantum efficiency (DQE).These parameters are transportable across differ-ent designs of x-ray image detector, and can beused to compare imaging system performance onan objective basis.

Figure 2 A modern cardiac catheterization laboratoryincorporating an Allura Xper FD10 solid-state, cardiac,flat-panel detector in the Yorkshire Heart Centre, Leeds.Reproduced with the permission of Medicamundi.

Dynamic range

The dynamic range of a digital x-ray image de-tector describes the maximum range of entrancedoses over which substantive image information isrecorded. In basic terms dynamic range is de-scribed by the ratio of the maximum to theminimum detector usable operating dose levels;(specifically the former is defined by the maximumsignal capability and the latter by noise). Dynamicx-ray image detectors require a wider dynamicrange than DR detectors, as they are multi-

Figure 3 A modern neurovascular intervention labora-tory incorporating a Philips Allura Xper FD20 large-field,dynamic, solid-state detector. Reproduced with the per-mission of Philips Healthcare.

Figure 4 A double-contrast barium enema imageacquired with a large-field dynamic solid-state flat-paneldetector. Figure courtesy of Sue Rimes.

The design and imaging characteristics of dynamic x-ray image detectors 1077

functional image receptors operating over a widerrange of dose levels.27 For example, modern fluo-roscopy demands effective image recording downto dose levels as low as 10 nGy per frame. Wherehigher quality dynamic images are required expo-sures are made at between w100 nGy and 1 mGyper frame, depending upon the image acquisitionmode. The former (lower) value is typical of thedose per frame used in digital cine fluorography,here images are usually acquired at 15 or 30frames s�1. The latter (higher) dose value is typicalof that used in more general applications of digitalfluorography. In DSA the dose per frame can ap-proach (or even exceed) radiographic dose levels,viz. 10 mGy or greater. In order to accommodateexposure highlights and minimize blooming arte-facts, (e.g., over sections of the GI tract contain-ing gas or between the lower limbs duringperipheral angiography), detectors must havea maximum dose capability of w50e100 mGy.Solid-state detectors have a linear signal responseacross the dynamic range. The linearity is suffi-ciently accurate that detectors can support thelogarithmic subtraction algorithm used in DSA im-aging over a wide dose range.25 Image signals arenormally digitized to 14-bits of grey-scale resolu-tion, (corresponding to 16,384 levels). Solid-statefluoroscopy detectors reportedly offer a dynamicrange some 10 times greater than x-ray IITV fluo-roscopy systems.28 The wide dynamic range andexcellent contrast rendition of indirect conversiondynamic solid-state detectors can be gauged fromthe double-contrast barium enema study shown inFig. 4.

Geometry, vignetting, and veiling glare

X-ray IITV channels are susceptible to a number offield-dependent defects that can degrade the cos-metic quality of fluoroscopic and fluorographicimages. Geometrical distortion arises from theelectron-opticaldesignof thex-ray image-intensifiertube. The mapping of electrons from a concavephotocathode onto a planar output screen resultsin a form of geometrical distortion known as pin-cushion distortion. The degree of pincushion distor-tion exhibited by a large-field x-ray image intensifieris illustrated in Fig. 5a. Pincushion distortion reflectsa progressive increase in geometrical magnificationtowards the periphery of the image field. At thesame time the luminance of the image field falls,producing a non-uniform distribution in brightness(or vignetting). Image intensifiers are also subjectto a second form of geometrical distortion, due toextraneous magnetic fields, known as S-distortion.The impact of S-distortion on image quality is

compounded by the fact that it varies with changingangulation of the C-arm; this is felt most strongly inrotational angiography and related reconstructiveimaging applications. Solid-state detectors are sub-ject to none of these forms of geometrical distor-tion, and therefore, consistently record imageswith excellent field homogeneity as shown inFig. 5b. As solid-state detectors are insensitive tomagnetic fields they can be successfully imple-mented in hybrid x-ray/magnetic resonance (MR)imaging laboratories29 or in conjunction with mag-netic catheter navigation equipment.27

Images produced by x-ray IITV systems aresubject to a loss of contrast due to the large-area scatter of x-rays, electrons, and light, whichoccurs at the various stages of image conversion.The overall effect of such scatter mechanisms onthe recorded image contrast is known as veilingglare. Veiling glare is often quantified in terms ofthe low frequency drop (LFD), which is derivedfrom the modulation transfer function (MTF). MTFis the concept often used to describe the spatialresolution properties of x-ray imaging systems.6

LFD measures the deterioration in MTF directly at-tributable to large-area scatter mechanisms. Thegreater the value of LFD the poorer the reproduc-tion of large-area contrast will be, and vice-versa.Solid-state detectors exhibit a much smaller LFDthan x-ray IITV systems, typically by a factor of

Figure 5 Comparison of the geometrical distortion exhibited by a large-field x-ray image intensifier (a), comparedwith the distortion-free image of a large-field, dynamic, solid-state detector (b). Figure courtesy of Pat Turner.

1078 A.R. Cowen et al.

between five and 10.21 Consequently, dynamic,solid-state detectors produce images with superiorcontrast and wider dynamic range than x-ray IITVsystems; this is reflected in the excellent repro-duction of the high-contrast structures depictedin Fig. 4.

Spatial resolution

The spatial resolution of a solid-state detector isaffected by a number of physical and technicalfactors. These include light scatter in the x-rayabsorption layer (in the case of indirect-conversiondevices), the detector pixel sampling interval andaperture size, and the bandwidth of the readoutelectronics. Dynamic, solid-state, flat-panel de-tectors come in a variety of sizes, form factors andpixel resolutions, matched to their target clinicalapplication(s). Typical values of spatial imagingcharacteristics for indirect-conversion dynamicdetectors designed for cardiac, vascular, andradiography and fluoroscopy applications are listed

Table 1 Typical spatial imaging characteristics of dynamic, solidcompared with a digital IITV system (results are quoted for the la

Cardiac detector Vascula

Maximum field ofview (cm)

Square field 24.8� 24.8 Rectan38.2�

Pixel samplinginterval (mm)

184 154

Maximum pixel array 956� 954 2480�Nyquist frequency

(lp/mm)2.72 3.25

in Table 1 (Readers should note that relevant char-acteristics, including pixel sampling interval andNyquist frequency, were defined in a previous re-view.6). Equivalent values for a large-field digitalx-ray IITV system are included for reference. Themaximum spatial resolution of a large-field, dy-namic, solid-state detector exceeds that of thedigital x-ray IITV system by a factor of over two.It should be noted that the spatial resolution ofdigital x-ray IITV systems deteriorates toward theperiphery of the image field. The spatial resolutionof solid-state detectors is maintained throughoutthe whole field of view. Dynamic, solid-state de-tectors offer multiple (sometimes up to five) ancil-lary zoom-field selections. These zoom-fields areused to magnify the presented image and, there-fore, improve the resolution of fine-detail struc-tures (albeit for a reduced field of view). Thespatial resolution of solid-state detectors remainsessentially constant, independent of the fieldsize selected. The frame rate of a dynamic solid-state detector can be increased, say from 15 to

-state detectors designed for three clinical application areas,rgest field selection in each case)

r detector Radiography andfluoroscopy detector

Digital IITV

gular field29.4

Square field42.6� 42.6

Circular field35 cm diameter

148 341

1910 2880� 2881 1024� 10243.38 1.46

The design and imaging characteristics of dynamic x-ray image detectors 1079

30 or from 30 to 60 frames s�1 by sacrificing spatialresolution and/or field coverage. In a detector’slargest field mode this is usually achieved by bin-ning (averaging) data over blocks of [2� 2] or[4� 4] pixels. For a digital x-ray IITV system thespatial resolution falls markedly as the field ofview is increased. The spatial resolution ofa large-field, solid-state detector can be gaugedfrom the abdominal DSA image of the superiormesenteric artery presented in Fig. 6. The vascularbed is depicted with excellent detail resolutionacross the whole image field.

Lag and ghosting

In a fluoroscopy examination of the GI tractstructures can move with a velocity of between10 and 30 mm s�1 due to peristalsis.30 In cardiacangiography the mean velocity of a coronary ves-sel31 is typically w50 mm s�1, while the peak ve-locity can exceed 100 mm s�1. Fluoroscopy

Figure 6 Abdominal DSA image of superior mesenteric arttector. Figure courtesy of Anne Allington, and Drs Raman Ub

detectors, therefore, must be able to record im-ages with sufficient temporal resolution to meetthe needs of their target examination(s). The tem-poral resolution of a dynamic solid-state image de-tector is defined by two physical mechanisms32:memory effect (or ghosting) and lag. These effectsoccur concurrently; under a given set of circum-stances a detailed experimental analysis is re-quired to distinguish and quantify their individualcontributions.

Memory effect refers to the production ofa spurious frozen pattern (a so-called ghost im-age), which mirrors the image content producedby the preceding x-ray exposure. This phenomenoncan persist for some time (even several minutes)particularly after an intense x-ray exposure ofa high-contrast structure is made.33 Both indirectand direct-conversion detectors are susceptibleto memory effect, but they occur due to differingphysical mechanisms. In general, the latter designof detector is more susceptible to ghosting

ery acquired with a large-field, dynamic, solid-state de-eroi and Phil Boardman.

1080 A.R. Cowen et al.

artefacts.32 Memory effect reflects a non-uniformvariation in detector response depending uponthe exposure history. With regard to indirect-con-version detectors, memory effect represents anincrease in CsI:Tl light emission and a-Si:H photo-diode gain, following the x-ray exposure. Theseeffects manifest themselves as a spurious increasein detector conversion efficiency, producing aso-called bright-burn artefact in images acquiredsubsequently. Conversely, in direct-conversiondetectors memory effect reflects a reduction(fatigue) in photoconductor response. Memory ef-fect can be a particular problem in mixed-modeimaging applications, specifically where low-dose fluoroscopy might follow immediately afteran image is acquired at a high fluorographic doselevel, e.g., as occurs in DSA.27 The x-ray dose perframe during fluoroscopy can be as low as onethousandth of that used during serial image acqui-sition; therefore, even a modest degree of memoryeffect may intrude upon the fluoroscopic imagesthat follow.

Lag is the property that quantifies the ability ofan image detector to accurately record time-varying changes in image content; the larger thelag, the poorer the temporal response and viceversa. Lag results from the carry-over of a pro-portion of recorded signal content into succeedingframes in the sequence. In the case of indirect-conversion detectors a small contribution of lagarises from afterglow in the CsI:Tl layer, (but this israrely significant in routine fluoroscopy). In prac-tice, lag largely results from the relatively slowtemporal response of a-Si:H. More specifically lagarises from the trapping and subsequent slowrelease (de-trapping) of charge carriers in thephotodiode array.34 In direct-conversion detectorslag is compounded by charge trapping/de-trappingmechanisms in the a-Se photoconductor.32 Withoutcorrection lag causes unacceptable unsharpness(smearing) of rapidly moving and time-varyingimage structures.

Dynamic solid-state detectors incorporate mea-sures to minimize lag and memory effect. Manymodern dynamic detectors achieve this using a so-called refresh (or reset) light, which reconditionsthe detector prior to each new image acquisitioncycle.9,20,28,34 The refresh light usually takes theform of an array of light-emitting diodes (seeFig. 1), which floods the detector with light pho-tons, saturating charge-trapping sites in thea-Si:H prior to each x-ray exposure. As a result,lag (and memory effect) is reduced to an accept-ably low level ensuring a suitably fast detector re-sponse. Signal retention due to lag in a modernindirect conversion detector is reportedly as low

as 0.3% at a time 1 s after termination of thex-ray exposure; after 10 s the lag reduces by a fur-ther order of magnitude.27 This ensures that thetemporal resolution is adequate for high-speedimaging applications, such as paediatric cardiacfluoroscopy. Equivalent lag figures for direct con-version detectors are reportedly higher.27 In someclinical applications a moderate degree of lag canbe tolerated, and is used to improve fluoroscopicimage quality by time-averaging (smoothing) noisefluctuations. Depending upon the type of clinicalapplication, a suitable degree of lag is normallysynthesized using digital recursive filtering.

DQE

DQE is the most effectual physical parameter usedto quantify and compare the performance of differ-ent x-ray image detectors objectively.35 To simplifythe discussion, here it is assumed that the fluoro-scopic image detector exhibits zero lag, (or any lagthat does exist is fully corrected). The DQE of thedetector can then be defined by the ratio,6,35

DQEdetector ¼ SNR2recorder=SNR2

input

Where SNRinput2 is the square of the signal-to-

noise ratio at the input of the image detector.This is defined by the fluence of x-ray photons(number per unit area) contributing to an individ-ual frame in the fluoroscopic image sequence.SNRrecorded

2 is the square of the signal-to-noise ratiorecorded by the image detector. The value ofSNRrecorded

2 can be computed from the outputdata. In terms of counting statistics SNRrecorded

2 isan estimate of the fluence of information carriersthat the recorded image frame is actually worth.

The information content of a recorded imageframe can never exceed that delivered to thedetector in the incident x-ray beam, therefore,

0 � DQEdetector � 1

A DQE of unity implies that the recording of x-rayimage information by the detector is perfect. At theother extreme a DQE of zero implies that no in-formation at all is recorded. Real-world x-ray imagedetectors obviously offer a DQE value falling some-where between these two extremes. The deterio-ration in recorded information is for two principalreasons. First, no detector can absorb all the in-cident x-ray photons with 100% efficiency. Inevita-bly some x-ray photons pass straight through thex-ray absorber, while others that are absorbed maythen be re-emitted and escape the detector. Thisloss in primary information is compounded by anynoise sources arising in the detector itself (e.g.,

DSD

IITV

Fluoroscopy DigitalCine

Digital Fluorography

DigitalAngiography

0.1

1.0D

QE

(X

)

1 10 100001000100

Dose per frame [nGy]

Figure 7 Variation of DQE as a function of the detector entrance dose-per-frame, comparing the performance ofa (indirect conversion) dynamic, solid-state detector with a digital IITV system.

The design and imaging characteristics of dynamic x-ray image detectors 1081

electronic noise arising in the a-Si:H matrix arrayand the readout circuitry). The DQE of a modern,indirect-conversion, dynamic, solid-state detectorfalls in the range 0.7e0.75.16,19,36 In the case ofdirect-conversion detectors noise-aliasing playsa significant part in degrading DQE.6,37 The DQE ofa direct conversion dynamic solid-state detectortypically lies in the range 0.5e0.6.38

The influence of electronic noise on detectorDQE performance is strongly dependent on signallevel (and, therefore, detector entrance dose perframe); and this is an important characteristic ofdynamic detector performance. This can be ana-lysed by measuring how DQE varies as a function ofdetector entrance dose per frame. In early designsof dynamic solid-state fluoroscopy detectors theimage quality fell below that of digital x-ray IITVsystems; this was due to the comparatively largecontribution of electronic noise at thattime.7,10,11,39 Typical DQE performance for a mod-ern indirect-conversion, solid-state, dynamic de-tector is shown in Fig. 7. Equivalent results fora modern digital x-ray IITV system have been in-cluded for reference. The dose ranges used instandard fluoroscopy, digital cine acquisition (ascommonly used in cardiac angiography), digitalfluorography (serial imaging) and (non-subtractive)

vascular imaging (where the dose-per-frame mayreach those used in radiography) are delineatedfor reader orientation. The high DQE of the indi-rect conversion detector is maintained across themajority of the dynamic range. The DQE perfor-mance is 10e15% greater than that of an IITV-based digital fluoroscopy system.40,41 This suggeststhat an improvement in image quality, and/or sav-ing in patient/staff dose, should be feasible if thisdesign of solid-state detector is used. This proposi-tion has been verified in cardio-angiography42 andcardiac electrophysiology imaging, respectively.43

In modern indirect-conversion detectors goodDQE performance can be maintained during fluo-roscopy down to dose levels of less than 10 nGyper frame.16,19,36 At exceptionally low dose levels(say well below 5 nGy per frame) digital x-rayIITV systems still offer better imaging perfor-mance; however, such dose levels would rarelybe used in clinical routine. At higher dose levels,say 1 mGy per frame and above, a marked deterio-ration in x-ray IITV system DQE occurs, and the su-periority of the dynamic, solid-state detectors isapparent. This fall in DQE results from the growinginfluence of the (noisy) granular structure of the IIinput and output phosphor screens. This acts asa residual fixed-pattern noise source, which

1082 A.R. Cowen et al.

becomes more pronounced as the x-ray quantumnoise diminishes with increasing dose. In solid-state detectors fixed-pattern noise is eliminatedduring system calibration, and this holds acrossthe dynamic range. Overall the graphs presentedin Fig. 7 confirm that modern, indirect-conversion,dynamic, solid-state detectors are dose-efficientimaging devices, which can support the full spec-trum of clinical applications previously underwrit-ten by digital x-ray IITV systems.

New directions in digital fluoroscopy

3D-enhanced fluoroscopy

The 1990s saw a growth in the use of digital x-rayIITV systems in 3D reconstruction imaging, basedupon a rotating C-arm imaging geometry.44 Beforeclinically acceptable reconstructions can be com-puted, extensive data processing is required tocorrect for defects such as the changing geometri-cal distortion (which occurs as the image intensi-fier rotates around the patient).45 For reasonsexplained above dynamic, solid-state detectors es-sentially produce distortion-free image data. Con-sequently, these new detectors yield 3D imagereconstructions with greater detail resolution46,47

and fewer artefacts.48,49 3D reconstruction imag-ing is typically used to improve the visualization

Figure 8 3D roadmap image of the iliac arteries withthe catheter in situ acquired using a dynamic, solid-state detector. Reproduced with the permission ofPhilips Healthcare.

of complex bone structures during orthopaedicsurgery50 or a tortuous network of blood vesselsin endovascular procedures.47 Reportedly the lat-ter can aid the clinician in navigating and deploy-ing interventional devices, thereby reducingprocedure times and patient/staff radiationdose.47,51 The availability of dynamic solid-statedetectors now makes it possible to reconstruct3D (and 2D sectional) images of not only high con-trast details, but also soft-tissue structures ofcomparatively low subject contrast52; (digitalx-ray IITV systems lack the contrast resolutionand dynamic range required to reliably achievethe latter). To illustrate the quality of 3D recon-structive imaging achievable with a solid-state de-tector let us focus on the technique known as‘‘dynamic 3D road-mapping’’.53,54 This visualiza-tion tool makes it possible to project (and automat-ically register) the live 2D fluoroscopy image upona 3D reconstruction of relevant vasculature, (andwhen useful, a CT-like sectional slice through thesurrounding soft-tissue). A 3D roadmap compositionof the iliac arteries (with a catheter in situ) ac-quired using a dynamic solid-state detector is pre-sented in Fig. 8. 3D-enhanced digital fluoroscopyis set to proliferate and increase in clinical utility,for example, incorporating real-time interven-tional procedure evaluation and device tracking.55

Such advances will facilitate the increasinglysophisticated and precise interventions that willbe realized in the future.

Increasing detector sensitivity

Further innovations in basic dynamic solid-statedetector design are anticipated. These possiblyinclude increases in detector sensitivity, by boost-ing signal gain and/or reducing electronic noise.Solutions considered include modifying the archi-tecture of the readout array to maximize the pixelfill-factor56 and implementing signal amplificationat a pixel-level.57 In the future signal digitizationis likely to be increased to 16 bit grey-scale resolu-tion (and in time possibly higher) to improve de-tector performance in 3D reconstructive imagingapplications. Automatic switching of the amplifiergain setting can also be used to extend detectordynamic range in these applications.58 It is con-ceivable that direct-conversion dynamic detectorsmight mature to the point where they can chal-lenge, or even out-perform indirect-conversion de-tectors. This could follow the adoption of moreefficient x-ray photoconductive converter mate-rials than a-Se, such as poly-crystalline HgI2, PbI2or PbO.59,60 Reportedly, however, significant

The design and imaging characteristics of dynamic x-ray image detectors 1083

refinement of the physical properties of thesematerials would be required before this is likelyto happen.61

Alternative readout arrays

Despite the success of a-Si:H-based dynamic de-tector arrays, alternative forms of electronicreadout have been mooted. For example, severalauthors have speculated that readout electronicsmight be better fabricated from crystalline siliconwafers, typically in the form of C-MOS (comple-mentary metal oxide semi-conductor) tech-nology.27,61e63 Such readout arrays could befabricated in semiconductor plants set up to man-ufacture C-MOS wafers for commodity products,possibly leading to reductions in detectormanufacturing costs. In addition, such dynamic de-tectors might afford performance enhancements,including improved spatial and temporal resolutionplus higher image acquisition rates.62 C-MOS is alsowell-suited to the implementation of complexcomponent structures in the detector array itselfrather than as external circuitry. Resulting pixel-level processing functions might include automaticgain control, array timing, adaptive digital imageenhancement, or more exotic concepts, such asx-ray photon counting (to maximize DQE) orenergy-selective (viz, colour) x-ray imaging.27,61

Consequently, adoption of C-MOS might conceiv-ably lead to more economic detector designs,potentially combined with enhanced dynamic,functional, and 3D imaging capabilities.

Conclusions

Dynamic, solid-state image detectors have reachedfull technological maturity; early deficiencies,such as moderate image quality at low dose rates,excessive dark current and artefacts due to lag andmemory effect having been resolved. An x-ray IITVsystem comprises a complex chain of electron-optical components, which are subject to drifts andvariations in adjustment over time, which candegrade clinical performance. Solid-state digitaldetectors are inherently more stable image acqui-sition platforms, which (in principle) require min-imal quality assurance monitoring. The cosmeticquality of dynamic solid-state detectors is excel-lent with zero geometrical distortion and vignett-ing, immunity from blooming (at exposurehighlights) and negligible contrast loss due tointernal scatter mechanisms. These detectors ex-hibit a wider dynamic range and retain high spatial

resolution when imaging over larger fields of view.They are also insensitive to magnetic fields andcan, therefore, be successfully implemented inmixed x-ray/MR imaging laboratories or wheremagnetic catheter-navigation equipment is used.The image quality of an x-ray IITV system can varymarkedly across the field of view. Indeed optimumimaging performance only really occurs withina quality area of limited extent. For solid-statedetectors the quality area encompasses the wholefield of view. Solid-state, flat-panel detectors arelighter and less bulky than x-ray image intensifierTV systems, affording better accessibility to thepatient and greater anatomical coverage. The DQEof modern indirect-conversion detectors is greaterthan that of both x-ray IITV systems and direct-conversion detectors, and offers high-quality fluo-roscopic imaging down to commendably low doserates. Dynamic, solid-state detectors can supportthe full range of fluoroscopy-guided procedures,including cardiac and vascular (DSA) imaging,radiography and fluoroscopy procedures and mo-bile surgical imaging. The transition from IITV tosolid-state detector-based digital fluoroscopy nowlooks irrevocable. Dynamic solid-state detectorsare extremely versatile x-ray image acquisitiondevices supporting fluoroscopy, serial image ac-quisition, and 3D reconstructive imaging; thelatter benefiting from the greater geometricalintegrity of the acquired data. Combining 3Dvisualization with fluoroscopy will prove increas-ingly influential in interventional radiology. Inves-tigations into ways of improving the technicalperformance of solid-state detectors are stillon-going. The longer term might conceivably seethe migration from a-Si:H to C-MOS based readoutelectronics. Such a re-alignment in dynamic, solid-state detector design might ease manufacturingcosts, while encouraging further innovations indynamic x-ray imaging.

Acknowledgements

Figs. 1 and 2 have been reproduced with the per-mission of Medicamundi. The authors acknowledgethe help of Ruth Turner and Mary Bennett of PhilipsHealthcare (Reigate, UK) during preparation of thisreview, including supplying Fig. 3. Individualthanks are also due to the following: Sue Rimes(Superintendent Radiographer) of the RadiologyDepartment at Musgrove Park Hospital, Taunton,who kindly provided Fig. 4. Pat Turner (Superinten-dent Radiographer) of the Radiology Departmentat Derby City General Hospital helped to acquireFig. 5. Anne Allington (Superintendent

1084 A.R. Cowen et al.

Radiographer), and Drs Raman Uberoi and PhilBoardman (Consultant Radiologists) of the VascularAngiography Department at the John RadcliffeHospital, Oxford, who kindly provided Fig. 6. DrDrazenko Babic, Clinical Scientist at Philips Health-care (in the Netherlands), who kindly providedFig. 8.

References

1. Krohmer JS. Radiography and fluoroscopy, 1920 to the pres-ent. RadioGraphics 1989;9:1129e53.

2. Schueler BA. General overview of fluoroscopic imaging.RadioGraphics 2000;20:1115e26.

3. Van Lysel MS. Fluoroscopy: optical coupling and the videosystem. RadioGraphics 2000;20:1769e86.

4. Mistretta CA, Crummy AB. Diagnosis of cardiovascular dis-ease by digital subtraction angiography. Science 1981;214:761e5.

5. Pooley RA, McKinney JM, Miller DA. Digital fluoroscopy.RadioGraphics 2001;21:521e34.

6. Cowen AR, Kengyelics SM, Davies AG. Solid-state flat-paneldigital radiography detectors and their physical imagingcharacteristics. Clin Radiol 2008;63:487e98.

7. Schiebel U, Conrads N, Jung N, et al. Fluoroscopic x-ray im-aging with amorphous silicon thin-film arrays. SPIE Proc PhysMed Imaging 1994;2162:129e40.

8. Antonuk LE, Yorkston J, Huang W, et al. A real-time, flat-panel amorphous silicon digital x-ray imager. RadioGraphics1995;15:993e1000.

9. Chabbal J, Chaussat T, Ducourant T, et al. Amorphous sili-con x-ray image sensor. SPIE Proc Phys Med Imaging 1996;2708:499e510.

10. Colbeth RE, Allen MJ, Day DJ, et al. Characterisation of anamorphous silicon fluoroscopic imager. SPIE Proc Phys MedImaging 1997;3032:42e51.

11. Colbeth RE, Allen MJ, Day DJ, et al. Flat panel imaging sys-tem for fluoroscopy applications. SPIE Proc Phys Med Imag-ing 1998;3338:376e87.

12. Bruijns TJ, Alving PL, Baker EL, et al. Technical and clinicalresults of an experimental flat dynamic (digital) x-ray imagedetector (FDXD) systems with real-time correction. SPIEProc Phys Med Imaging 1998;3336:33e44.

13. Bury RF, Cowen AR, Davies AG, et al. Initial experienceswith an experimental solid-state universal digital x-raydetector. Clin Radiol 1998;53:923e8.

14. Bruijns AJC, Bury R, Busse F, et al. Technical and clinical as-sessments of an experimental flat dynamic x-ray image de-tector system. SPIE Proc Phys Med Imaging 1999;3659:324e35.

15. Jung N, Alving PL, Busse F, et al. Dynamic x-ray imagingbased on an amorphous silicon thin-film array. SPIE ProcPhys Med Imaging 1998;3336:974e85.

16. Busse F, Rutten W, Sandkamp B, et al. Design and perfor-mance of a high quality cardiac flat panel detector. SPIEProc Phys Med Imaging 2002;4682:819e27.

17. Granfors PR, Albagli D, Tkaczyk JE, et al. Performance ofa flat cardiac detector. SPIE Proc Phys Med Imaging 2001;4320:77e86.

18. Sivananthan UM, Moore J, Cowan JC, et al. A flat-detectorcardiac cath lab system in clinical practice. Medicamundi2004;48:4e12.

19. Granfors PR, Aufrichtig R, Possin GE, et al. Performance ofa 41 � 41 cm2 amorphous silicon flat panel x-ray detectordesigned for angiographic and R&F imaging applications.Med Phys 2003;30:2715e26.

20. Ducourant T, Michel M, Vieux G, et al. Optimization of keybuilding blocks for a large area radiographic and fluoroscopicdynamic x-ray detector based on a-Si:H/CsI:Tl flat paneltechnology. SPIE Proc Phys Med Imaging 2000;3977:14e25.

21. Bruijns AJC, Bastiaens R, Hoornaert B, et al. Image qualityof a large-area dynamic flat detector: comparison witha state-of-the-art IITV system. SPIE Proc Phys Med Imaging2002;4682:332e43.

22. Colbeth RE, Boyce S, Fong R, et al. 40 � 30 cm2 flat imagerfor angiography, R&FG and cone-beam CT applications. SPIEProc Phys Med Imaging 2001;4320:94e102.

23. Choquette M, Demers Y, Shukri Z, et al. Real time perfor-mance of a selenium based detector for fluoroscopy. SPIEProc Phys Med Imaging 2001;4320:501e8.

24. Tousignant O, Demers Y, Laperriere L, et al. Clinical perfor-mances of a 14’’�14’’ real time amorphous selenium flatpanel detector. SPIE Proc Phys Med Imaging 2003;5030:71e6.

25. Ducourant T, Couder D, Wirth T, et al. Image quality of dig-ital subtraction angiography using flat detector technology.SPIE Proc Phys Med Imaging 2003;5030:203e14.

26. Neitzel U. Recent technological developments and their in-fluence. Radiat Prot Dosimetry 2000;90:15e20.

27. Spahn M. Flat detectors and their clinical applications. EurRadiol 2005;15:1934e47.

28. Seibert JA. Flat-panel detectors: how much better are they?Pediatr Radiol 2006;36:173e81.

29. Fahrig R, Wen Z, Ganguly A, et al. Performance of a static-anode/flat-panel x-ray fluoroscopy system in a diagnosticstrength magnetic field: truly hybrid x-ray/MR imaging sys-tem. Med Phys 2005;32:1775e84.

30. Nguyen TC, Rowlands JA. A study of motion in gastro-intes-tinal x-ray fluoroscopy. Med Phys 1989;16:569e76.

31. Achenbach S, Ropers D, Holle J, et al. In-plane coronary ar-terial motion velocity: measurement with electron-beamCT. Radiology 2000;216:457e63.

32. Zhao W, DeCresenzo G, Rowlands JA. Investigation of lagand ghosting in amorphous selenium flat-panel x-ray detec-tors. SPIE Proc Phys Med Imaging 2003;4682:9e20.

33. Siewerdsen JH. Jaffray. DA A ghost story: spatio-temporalresponse characteristics of an indirect-detection flat-panelimager. Med Phys 1999;26:1624e41.

34. Overdick M, Solf T, Wischmann H-A. Temporal artefacts inflat dynamic x-ray detectors. SPIE Proc Phys Med Imaging2001;4320:47e58.

35. Dainty JC, Shaw R. Image science. London: Academic Press;1975.

36. Tognina CA, Mollov I, Yu JM, et al. Design and performanceof a new a-Si flat panel imager for use in cardiovascular andmobile C-arm imaging systems. SPIE Proc Phys Med Imaging2004;5368:648e56.

37. Moy JP. Signal-to-noise ratio and spatial resolution in x-rayelectronic imagers: is the MTF a relevant parameter? MedPhys 2000;27:86e93.

38. Tousignant O, Demers Y, Laperriere L, et al. Spatial andtemporal image characteristics of a real-time large areaa-Se x-ray detector. SPIE Proc Phys Med Imaging 2005;5745:207e15.

39. Davies AG, Cowen AR, Kengyelics SM, et al. Threshold con-trast detail detectability measurement of the fluoroscopicimage quality of a dynamic solid-state digital x-ray imagedetector. Med Phys 2001;28:11e5.

The design and imaging characteristics of dynamic x-ray image detectors 1085

40. Spekowius G, Boerner H, Eckenbach W, et al. Simulation ofthe imaging performance of x-ray image intensifier YV cam-era chains. SPIE Proc Phys Med Imaging 1995;2432:12e23.

41. Baker EL, Cowen AR, Kemner R, et al. A physical evaluationof a CCD-based x-ray image intensifier fluorography systemfor cardiac applications. SPIE Proc Phys Med Imaging 1998;3336:430e41.

42. Vano E, Geiger B, Schreiner A, et al. Dynamic flat panel de-tector versus image intensifier in cardiac: dose and imagequality. Phys Med Biol 2005;50:5731e42.

43. Davies AG, Cowen AR, Kengyelics SM, et al. X-ray dose re-duction in fluoroscopically guided electrophysiology proce-dures. Pacing Clin Electrophysiol 2006;29:262e71.

44. Fahrig R, Fox AJ, Lownie S, et al. Use of a C-arm system togenerate true three-dimensional computed rotational an-giograms: preliminary in vitro and in vivo results. AJNRAm J Neuroradiol 1997;18:1507e14.

45. Fahrig R, Moreau M. Holdsworth. Three dimensional com-puted tomographic reconstruction using a C-arm mountedXRII: correction of image intensifier distortion. Med Phys1997;24:1097e106.

46. Baba R, Konno Y, Ueda K, et al. Comparison of flat-panel de-tector and image-intensifier detector for cone-beam CT.Comput Med Imaging Graph 2002;26:153e8.

47. Hirota S, Nakao N, Yamamoto S, et al. Cone-beam CT withflat-panel detector digital angiography system: early expe-riences in abdominal interventional procedures. CardiovascIntervent Radiol 2006;29:1034e8.

48. Hirai T, Korogi Y, Ono K, et al. Pseudostenosis phenomenonat volume-rendered three-dimensional digital angiographyof intracranial arteries: frequency, location and effect onimage evaluation. Radiology 2004;232:882e7.

49. Kakeda S, Korogi Y, Ohnari N, et al. 3D digital subtractionangiography of intracranial aneurysms: comparison of flatpanel detector with conventional image intensifier TV sys-tem using a vascular phantom. AJNR Am J Neuroradiol2007;28:839e43.

50. Akpek S, Brunner T, Benndorf G, et al. Three-dimensionalimaging and cone beam volume CT in C-arm angiographywith flat panel detector. Diagn Interv Radiol 2005;11:10e3.

51. Hatakeyama Y, Kakeda S, Korogi Y, et al. Intracranial 2Dand 3D DSA with flat panel detector of the direct con-version type: initial experience. Eur Radiol 2006;16:2594e602.

52. Heran NS, Song JK, Mamba K, et al. The utility of DynaCT inneurovascular procedures. AJNR Am J Neuroradiol 2006;27:330e2.

53. Soderman M, Babic D, Homan R, et al. 3D roadmap in neuro-angiography: technique and clinical interest. Neuroradiol-ogy 2005;47:735e40.

54. Wilhelm K, Babic D. 3D angiography in the interventionalclinical routine. Medicamundi 2006;50:24e31.

55. Racadio JM, Babic D, Homan R, et al. Live 3D guidance inthe interventional radiology suite. AJR Am J Roentgenol2007;189:357e64.

56. Weisfield RL, Yao W, Speaker T, et al. Performance analysisof a 127 micron pixel large-area TFT photo-diode array withboosted fill factor. SPIE Proc Phys Med Imaging 2004;5368:338e48.

57. Matsura N, Zhao W, Huang Z, et al. Digital radiology usingactive matrix readout: amplified pixel detector array forfluoroscopy. Med Phys 1999;26:672e81.

58. Roos PG, Colbeth RE, Mollov I, et al. Multiple-gain-rangingreadout method to extend the dynamic range of amorphoussilicon flat-panel imagers. SPIE Proc Phys Med Imaging 2004;5368:139e49.

59. Street RA, Ready SE, van Schuylenbergh K, et al. Compari-son of PbI2 and HgI2 for direct detection active matrixx-ray image sensors. J Appl Phys 2002;91:3345e55.

60. Kasap SO, Rowlands JA. Direct-conversion flat-panel x-rayimage sensors for digital radiography. Proc IEEE 2002;90:591e604.

61. Neitzel U. Status and prospects of digital detector tech-nology for CR and DR. Radiat Prot Dosimetry 2005;114:32e8.

62. Nakagawa K, Mizuno S, Aoki Y, et al. C-MOS flat-panelsensor for real-time x-ray imaging. Radiat Med 2000;18:349e53.

63. Graeve T, Weckler GP. High-resolution C-MOS imaging de-tector. SPIE Proc Phys Med Imaging 2001;4320:68e76.