a literature review of electronic portal imaging for radiotherapy

Radiotherapy and Oncology 88 (2008) 289–309www.thegreenjournal.com

Systematic review

A literature review of electronic portal imagingfor radiotherapy dosimetry

Wouter van Elmpta,*,1, Leah McDermottb,1, Sebastiaan Nijstena,Markus Wendlingb, Philippe Lambina, Ben Mijnheera,b

aDepartment of Radiation Oncology (MAASTRO), GROW, University Hospital Maastricht, The Netherlands, bDepartment of RadiationOncology, The Netherlands Cancer Institute-Antoni van Leeuwenhoek Hospital, Amsterdam, The Netherlands

Abstract

Electronic portal imaging devices (EPIDs) have been the preferred tools for verification of patient positioning forradiotherapy in recent decades. Since EPID images contain dose information, many groups have investigated their use forradiotherapy dose measurement. With the introduction of the amorphous-silicon EPIDs, the interest in EPID dosimetryhas been accelerated because of the favourable characteristics such as fast image acquisition, high resolution, digitalformat, and potential for in vivo measurements and 3D dose verification. As a result, the number of publications dealingwith EPID dosimetry has increased considerably over the past �15 years. The purpose of this paper was to review theinformation provided in these publications. Information available in the literature included dosimetric characteristicsand calibration procedures of various types of EPIDs, strategies to use EPIDs for dose verification, clinical approaches toEPID dosimetry, ranging from point dose to full 3D dose distribution verification, and current clinical experience. Qualitycontrol of a linear accelerator, pre-treatment dose verification and in vivo dosimetry using EPIDs are now routinely usedin a growing number of clinics. The use of EPIDs for dosimetry purposes has matured and is now a reliable and accuratedose verification method that can be used in a large number of situations. Methods to integrate 3D in vivo dosimetry andimage-guided radiotherapy (IGRT) procedures, such as the use of kV or MV cone-beam CT, are under development. It hasbeen shown that EPID dosimetry can play an integral role in the total chain of verification procedures that areimplemented in a radiotherapy department. It provides a safety net for simple to advanced treatments, as well as a fullaccount of the dose delivered. Despite these favourable characteristics and the vast range of publications on the subject,there is still a lack of commercially available solutions for EPID dosimetry. As strategies evolve and commercial productsbecome available, EPID dosimetry has the potential to become an accurate and efficient means of large-scale patient-specific IMRT dose verification for any radiotherapy department.

�c 2008 Elsevier Ireland Ltd. All rights reserved. Radiotherapy and Oncology 88 (2008) 289–309.

Keywords: EPID dosimetry; Pre-treatment dose verification; In vivo dosimetry; IMRT verification; Portal dosimetry

IntroductionMotivation

Advances in external beam radiotherapy have increasedthe accuracy requirements for dose delivery to patients.Quality control procedures before and during treatmentcan potentially ensure a high level of accuracy necessaryfor treatments designed to achieve adequate tumour con-trol and reduction of normal tissue complications. As a re-sult of the high number of patients being treated inradiotherapy departments world-wide, there has to be atrade-off between the extent of verification required to at-tain a certain level of accuracy and the work-load requiredto carry out such procedures.

1 Both authors contributed equally to this work.

0167-8140/$ - see front matter �c 2008 Elsevier Ireland Ltd. All rights re

In general, there are several aspects that should be incor-porated in radiotherapy verification procedures. These in-clude procedural verification using well-documentedprotocols and workflow, set-up verification measurementsof the patient position and also verification of the dosedelivered to the patient. Documentation and workflow arehighly dependent on the size of the department and a myr-iad other department-specific factors. Set-up verification isgaining significant ground in the field of external-beamradiotherapy with the recent proliferation of 3D and 4D(in-room) imaging tools. Detailed reviews on the clinicaluse of electronic portal imaging devices (EPIDs) for 2D por-tal imaging have been provided by Herman et al. [1–3],Langmack [4] and Antonuk [5]. Standards in dose verifica-tion also need to keep up with the pace, since high levels

served. doi:10.1016/j.radonc.2008.07.008

290 EPID dosimetry review

of geometrical accuracy and other QA procedures becomemeaningless when the dose cannot be delivered as planned.

With the increased use of intensity-modulated radiother-apy (IMRT), higher dose prescriptions and changes to elec-tronic recording and transfer of patient treatment data,the demand for patient-specific verification has increased.The complexity of treatment plans is also increasing, mak-ing it more difficult to discover possible errors becausethe dose levels and field shapes are no longer intuitive forspecific treatment sites, and are therefore more difficultto detect by experience. During the treatment planning pro-cess, deviations from the planned dose can be present ifcharacteristics of the linear accelerator are not modelledaccurately in the treatment planning system (TPS), e.g.parameters such as the tongue-and-groove effect or multi-leaf collimator (MLC) transmission. Errors can also occurdue to data transfer problems, especially when convertingfrom paper to electronic systems, or from manual to auto-matic systems. Regardless of whether the treatment tech-nique is conventional or advanced, the department islarge or small, or the treatment makes use of the latesttechnology or not, errors can, and often do, occur duringthe delivery process. Recommended QA procedures mightno longer be sufficient and patient-specific verification pro-cedures are therefore necessary to detect possible errorsfor these complex plans. Furthermore, the legal require-ments for radiotherapy are increasing. National or interna-tional legislation (e.g. the EC Directive 97/43) imposerequirements on these types of verification procedures.

Not long after their clinical introduction for set-up verifi-cation, it was realised that EPID images contain dose infor-mation. Consequently, several groups investigated thedosimetric characteristics of various types of EPIDs. In somecases, EPIDs have gone on to replace traditional dosimetrydevices in the clinic for plan verification. Interest in EPIDdosimetry has been accelerated by the advantages of fastimage acquisition, high resolution, digital format, and po-tential for in vivo measurements and 3D dose verification.Fig. 1 illustrates this increased interest through the numberof publications in peer-reviewed journals relating to EPIDdosimetry over a period of more than a decade since itsinception.

Fig. 1. Number of publications on EPID dosimetry (as found onPubmed www.pubmed.com).

The aim of this study was to review how EPIDs have con-tributed to the verification of the dose delivered eitherprior to or during the treatment of a patient. The dosimetriccharacteristics published for different types of EPIDs arefirst summarised, followed by the various strategies toimplement EPID dosimetry in the clinic. Finally, a reviewof current clinical practice is presented with special atten-tion to the acceptance and rejection criteria used in variousinstitutes.

Search strategyA comprehensive search strategy through the indexed

database ‘Pubmed’ was performed using search terms thatincluded ‘‘EPID’’, ‘‘portal dosimetry’’, ‘‘portal imaging’’,‘‘dose verification’’, ‘‘quality control’’, ‘‘quality assur-ance’’, ‘‘treatment verification’’, ‘‘pre-treatment verifica-tion’’ and ‘‘in vivo dosimetry’’. Relevant papers, includingrelevant cited references in these papers, published in Eng-lish between 1985 and January 2008 were included.

DevicesDose verification devices

This review will focus primarily on the use of EPIDs fordosimetry, other devices are also used for verifying the doseeither before or during treatment. Devices such as pointdetectors, film and gel have been widely reported. More de-tailed information about the various systems used for verifi-cation of IMRT can be found elsewhere, e.g. [6,7].

Like EPIDs, various matrix detectors have been producedfor measuring energy fluence or absorbed dose in twodimensions [7,8]. Commercial options are available basedon different measurement techniques. These include two-dimensional detectors consisting of a large number of ioni-sation chambers [9,10] or diodes [11,12] placed in a regu-larly spaced array or at specific points in a phantom. Theuse of fluorescent screens in combination with an opticalcamera system has also been used for dose measurements[13–15]. Matrix detectors can be attached to the gantryof the linear accelerator or placed on the treatment couch.A quick verification of the beam is performed by comparingthe output of these devices with a dose distribution pre-dicted by the TPS. A possible drawback of these devices isthat they have relatively few measuring points and there-fore a low spatial resolution. This potentially limits the pro-cedures for verification of highly intensity-modulated fields.A possible solution has been proposed whereby the detectoris shifted between multiple irradiations of the same plan,which extends the verification time [10].

An advantage of using EPIDs for dosimetry is that mostdepartments wishing to deliver, and hence verify, morecomplicated treatments usually already have one attachedto their linac. Using this device for dosimetry usually re-quires additional software. The various EPIDs that are com-mercially available can be categorized according to theirtechnical design. The first commercially available EPIDswere the liquid-filled ionisation chamber EPID and the cam-era-based EPID. More recently, the amorphous-silicon EPIDhas become the most commonly available portal imager.For a detailed overview of the historical perspective and

http://www.pubmed.com

W. van Elmpt et al. / Radiotherapy and Oncology 88 (2008) 289–309 291

the development of EPID technology, as well as more indepth information on the various types of EPIDs that havebeen produced, see the review articles by Boyer et al.[16], Antonuk [5] and more recently by Kirby and Glendin-ning [17].

EPIDs used for dosimetryOne of the first publications that proposed to use EPIDs

for dose measurement involved the scanning liquid-filledionisation chamber EPID (Li-Fi EPID, also referred to as‘SLIC EPID’ in the literature) [18]. It was developed inthe late 1980s/early 1990s at the Netherlands CancerInstitute, Amsterdam by Meertens and van Herk [19–21]and became commercially available as PortalVision (Var-ian, Palo Alto, CA, USA). Original versions consisted of amatrix of 256 · 256 ionisation chambers filled with an or-ganic liquid (isooctane) over an area of 32 · 32 cm2. Usingthe standard acquisition mode, the sampling time for onerow was 20 ms with an image scan time of 5.6 s [22]. Dueto the relatively long read-out time, this device cannotmeasure dose directly, but is suitable for measuringdose-rate. The measured dose-rate can be converted toabsolute dose by recording a continuous readout of themonitor chamber signal of the linac during image acquisi-tion and the number of MUs delivered for the measureddose image. Dose–response characteristics of the Li-FiEPID have been investigated by several groups [18,22–36]. These characteristics can be summarised as follows:(i) the stability of the response is within 1% up to 2 years,as long as corrections for room temperature dependenceand radiation damage are incorporated; (ii) the dose-rate/response relationship can be accurately describedby an equation with a term proportional to the square-root of the dose-rate and another term linear to thedose-rate, or a power-law; (iii) if a mini-phantom is usedto calibrate the system, additional corrections are re-quired to account for lateral scatter within the EPID,which can be done via a de-convolution filter; (iv) a rela-tive sensitivity map is needed to account for variation inresponse for each ionisation chamber–electrometer com-bination. The amount of build-up material required forelectron equilibrium was investigated by Boellaard et al.[26]. They found that above the detector layer, the EPIDhad a water-equivalent thickness of 8 mm. An additional28 mm of polystyrene was required for a 25 MV beam;however, the disadvantage was that 4.5 kg was added tothe mass of the EPID system, and some deterioration inimage quality was reported.

The scintillation crystal-photodiode detector, ‘RTIM-AGE’, was also developed in the beginning of the 1990sand its use for dose measurements has been described invery early EPID dosimetry publications. It is a linear scan-ning array imager, developed at the Royal Marsden Hospitalin London by Morton et al. [37]. A similar prototype was de-scribed by Symonds-Taylor et al. [38]. While not developedcommercially, Hansen et al.[39] used the RTIMAGE to per-form transit dosimetry, and demonstrated the linear rela-tionship between the response and dose, pulse height andpulse frequency. Mosleh-Shirazi et al. [40] described a 2Dcrystal array mirror system that was used to measure trans-mission dose images.

The camera-based EPID was developed into a commercialproduct two decades ago. This device consists of a fluores-cent phosphor-screen with a metal plate on top that con-verts high-energy photons into visible photons. Thesephotons are imaged with a video camera (mostly CCD-basedcameras are used) by means of mirrors and a lens. A largeportion of the field can be imaged quickly due to the fastread-out of the camera and it has a high spatial resolution.The various camera-based systems for which the dosimetricproperties are described are SRI-100 [41–43] (Philips Medi-cal Systems, Best, The Netherlands), iView [44] (ElektaOncology Systems, Crawley, UK), TheraView NT [45–48](Cablon, Leusden, The Netherlands). The measured greyscale value is approximately linear with dose and does notdepend on the absorber thickness that is placed in thebeam. Some groups have introduced a correction functionto correct for the nonlinearity of the EPID electronics[43,45]. There is, however, a large field size dependencecaused by scattered visible photons inside the optical sys-tem. This dependence is usually removed via deconvolutionwith a cross-talk kernel that describes the spread of thephotons inside the optical system [42,43], or using posi-tion-specific cross-talk contributions [48,49], while alsoanti-scatter grids have been proposed [50,51]. In addition,a sensitivity matrix is used to remove some of the residualcross-talk and to correct for possible non-uniformities inthe phosphor-layer. The short-term variation in the re-sponse is smaller than 1% (1 SD) and its long-term stabilityis within 1–2% (1 SD) for periods up to one year[42,45,47,52,53]. Once the camera-based EPID is cali-brated, mainly by removing the influence of the opticalscatter effects, it is a suitable device for performing portaldosimetry with deviations around 1–2% (1 SD).

The most common type of EPID available today is theamorphous-silicon EPID (a-Si EPID) or flat-panel imager. Itwas first described by Antonuk et al. [54,55] from the Uni-versity of Michigan Medical Center, Ann Arbor, USA. The pa-nel consists of an X-ray converter, light detector, and anelectronic acquisition system for receiving and processingthe resulting digital image. The dose–response behaviourof the three commercially available a-Si EPIDs has been de-scribed, including the Elekta iView GT system [56–60], theSiemens OptiVue [61,62] and the Varian a-Si PortalVisionaS500/aS1000 [63–70]. Stability of the response has beenreported as 0.5% (1 SD) over 2 years for panels fixed to linacswith lower energy beams (4 and 6 MV photon beams) [71].The dose–response relationship is independent of thedose-rate and approximately linear with integrated dose[54,56,72].

Over-sensitivity to low energies has also been reported,where the signal is influenced by the scatter within the bulklayers of the imager [15,67,70,73,74]. The EPID response istherefore dependent on the off-axis position of a specificpixel (beam softening) as well as the thickness of the phan-tom or patient in the beam, which causes hardening of thephoton beam, as reported by several groups[58,62,63,66,75]. Various studies have used models to re-duce the field size dependence from 5% to 8% (1 SD) to val-ues less than 1–2% (1 SD) [64,69,76]. Yeboah et al.identified the effect of a copper plate on the attenuationof low-energy scattered photons, while producing additional


annihilation, bremsstrahlung and scattered photons. Conse-quently, several groups have added an additional Cu plateon the detector (usually fixed within the housing underthe cover) [56,60,66,73,77]. Additional build-up materialfor portal dosimetry serves to attenuate scattered radiationfrom the patient, to which the a-Si EPID is over-sensitive,and to ensure measurements are made beyond the positionof dose maximum.

An additional issue specific to amorphous-silicon EPIDs isthat of ghosting and image lag. Both are due to the chargetrapped within the photodiode layer resulting in both gainand offset effects. Ghosting is a gain effect whereby thetrapped charge alters the electric field strength of the bulkand surface layers, lasting minutes and affecting subsequentimages. Image lag is an offset effect as a result of incom-plete charge transfer from the pixel capacitor. This trappedcharge is read-out in subsequent frames, lasting up to 30 s.Under-dosages from 4% to 10% at short irradiation time (2–5 MUs) relative to that of 1000 MUs have been reported forall commercially available a-Si EPIDs [78]. Ghosting wasfound to depend on irradiation time and the dose per frame,not on dose or dose-rate directly [56,73,78–80].

Changes in EPID technology over recent decades have notbeen driven by the technical demands of dosimetry, but bythe need for improved imaging standards for patient posi-tioning, such as enhancement in image quality and fasterread-out. Coincidentally, they have also aided in improvingthe EPID’s capacity as a dosimeter. Faster read-out time hasmeant images can be used to measure integrated dose, in-stead of dose-rate (as with the Li-Fi EPID). Higher resolu-tion, especially of the a-Si EPID, has improved theresolution to which dose can be verified, along with the im-age quality as intended. Changes arising from EPID dosime-try requirements involve software changes, such as thePortalVision system by Varian. The earlier version(v6.1.03) used a reset every 64 frames to move the framebuffer content to the CPU, creating a dead time of 0.28 s,or loss of one to two image frames (depending on the framerate). This was a problem for dosimetry and corrected insubsequent versions of the software [65,78]. As mentionedearlier, the addition of build-up material (e.g. a copperplate) is important for dosimetry and may be incorporatedin future EPID designs. At this stage both imaging and dosim-etry functions would have to be considered by manufactur-ers, and as in the past, whichever ends up becoming theprimary function will drive technological advances.

An overview of the various types of EPIDs and key refer-ences to the literature are given in Table 1.

EPID calibration: general approachesTwo approaches have been adopted for the calibration of

an EPID for dose measurements: conversion of the greyscale pixel value to a dose value or simulation (or predic-tion) of the grey scale pixel value.

The first approach applies (semi-) empirical models toconvert the measured grey-scale image of the EPID into aportal dose image, e.g. [42,61,73,81,82]. The EPID signalis converted to dose by using a calibrated detector, usuallyan ionisation chamber inside a water- or a mini-phantom, orfilm [35,76]. The advantage of this approach is that the

results can be verified independently with an ionisationchamber or film measurement. In other words, the interpre-tation of portal dose images is now simplified to the inter-pretation of the measurement of a conventionaldosimeter. A possible drawback is that empirical or mea-surement-based models must be validated also outside thereference situation; especially the robustness of these cali-bration models under various different clinical situationsneeds to be tested.

The second approach simulates or models the detector re-sponse, usually by Monte Carlo (MC) simulation [67,68,83] orby empirical models [64]. For accurate simulations, a de-tailed model of the EPID is necessary; however, accuratetechnical details are not always available, and often assump-tions with respect to the construction and materials of thevarious layers are made. If the detector can be modelledaccurately, an advantage of simulating the detector re-sponse is that the model can be directly verified by EPIDmeasurements. A disadvantage in simulation methods is thatfor treatment verification, it is difficult to interpret patientdose differences directly from predicted and measured re-sponse differences. In principle, it is possible to determinethe fluence that resulted in the response (measured or pre-dicted), and from this the corresponding dose to the patient,or detector, can be calculated. However, different sourcesof error can produce similar erroneous response measure-ments. The simulation approach is useful for quality controlof the linac and treatment verification by means of a consis-tency check with the patient treatment plan.

The two approaches, either conversion or simulationmethod, have been used for different purposes. The conver-sion method usually applied a model-based approach that issimpler and faster than a Monte Carlo simulation of the EPIDresponse, and therefore more suited for implementation inclinical routine. The drawback is that these calibrationmodels can be too simple to cover all treatment techniquesand irradiation configurations. Simulation of the detectorresponse is a useful tool to derive, test and validate assump-tions concerning the dose–response characteristics of theEPID. However, detailed information on the technical designof the EPID and long calculation times are needed. Accuratesimulations could be used to develop models that convertthe measured portal image to a portal dose distribution thatopens various possibilities for (pre-) treatment dosimetryverification, as described in the later sections.

Methods of EPID dosimetryThere is no clear consensus in the literature on the defi-

nition of various procedures and methods related to EPIDdosimetry. Here are definitions of various terms as used inthis review, and further illustrated in Fig. 2.

Verification procedures can be classified according towhether they are performed during treatment time (i.e.with the patient) or outside of treatment time (i.e. withoutthe patient).

• Pre-treatment verification: a procedure comparing thewhole or part of the intended treatment plan with mea-surements of corresponding radiation beams delivered by

Table 1List of key references that describe dosimetric properties and calibration methods for various EPIDs

Type of EPID Key references Investigated quantity or characteristics

Liquid-filled (Li-Fi) EPID Meertens [19,20], van Herk [18,21] Physical characteristics of Li-Fi EPIDsEssers [22,24], Yin [23], Zhu [25], Boellaard [26],Keller [27], Parsaei[28], Van Esch [30], Chang [31]

Dosimetric properties of Li-Fi EPIDs

Louwe [33] Short- and long-term stability of Li-FiEPIDs

Non-commercial EPIDs Morton [37] Characteristics of the ZnWO4 scintillationcrystal-photodiode detector (‘‘RTImage’’)

Symonds-Taylor [38] Characteristics of the CsI scintillationcrystal-photodiode detector

Mosleh-Shirazi [40] Characteristics of the CsI scintillationarray imaged with a CCD camera

Camera-based EPID Althof [41] Physical characteristics of camera-basedEPIDs

Heijmen [42], Pasma [43], Franken [47,48],de Boer [49], Glendinning [45], Partridge [50]

Dosimetric properties of camera-basedEPIDs

Heijmen [42], Franken [47], Glendinning [45], Dirkx [53] Short- and long-term stability of camera-based EPIDs

Amorphous silicon (a-Si) EPID Antonuk [54,55] Physical properties of a-Si EPIDsMcDermott [56,78], Siewerdsen [79], Partridge [73],Winkler [59]

Ghosting and image lag for a-Si EPIDs

Chen [61], Nijsten [62] Dosimetric properties of Siemens OptiVueEPIDs

McDermott [56], Winkler [57,59], Parent [58,60], Louwe [71] Dosimetric properties of Elekta iView GTEPIDs

Greer [63,65,75], Kirkby [66,70], Van Esch [64], Grein [69] Dosimetric properties of VarianPortalVision aS500/aS1000 EPIDs

Fig. 2. Three arrangements for EPID dosimetry, each with the possibility to verify a dose distribution at the EPID level or inside the patient orphantom.


the linear accelerator outside patient treatment time,i.e. with open fields or a phantom. This comparison canfocus on different aspects of the planned treatment:e.g. predicted and measured leaf positions, dose deliv-ered to the detector or phantom, or incident energy flu-ence extracted from measurements.

• Treatment verification: a comparison of all or part of theplanned and the delivered dose distribution based onmeasurements acquired during radiotherapy of thepatient. These measurements can be used to determinethe dose delivered to the detector or patient, or incidentenergy fluence obtained from measurements.

Dosimetry methods, independent of the type of detec-tor used, can be grouped according to whether or notbeams have passed through an attenuating medium, orwhether the dose is reconstructed inside a phantom orpatient.

• Non-transmission (or non-transit) dosimetry: determi-nation of the dose in the detector, patient or phantom,or determination of the incident energy fluence, basedon measurements without an attenuating mediumbetween the source and the detector, i.e. phantomor patient.


• Transmission (or transit) dosimetry: determination ofthe dose at the position of the detector, patient or phan-tom, or determination of the incident energy fluence,based on radiation transmitted through the patient orphantom.

• In-phantom dosimetry: measurement or determinationof the dose inside a phantom (rarely performed withEPIDs but included for completeness). This can be thedose at points, lines, planes or volumes within thephantom.

• In vivo dosimetry: measurement or determination of thedose inside the patient. Measurements performed duringtreatment can be performed invasively, i.e. inside thepatient, or non-invasively, i.e. on or at some distancefrom the patient, whereby the in vivo dose at the pointof interest is obtained by extrapolation. EPID dosimetryand the use of dosimeters positioned on the skin of apatient belong to the second category of in vivodosimetry.

The dose can be measured and determined at differentlocations with different configurations of the EPID, phantomor patient, as described in Fig. 2. Abbreviations: E: level ofthe EPID, P: at the patient level/inside the patient.

• (a) Non-transmission pre-treatment dosimetry: acquireimage for each field without a patient or phantom inthe beam.

– E: compare image (raw or converted to dose-to-water) with predicted EPID response or dose-to-waterat the level of the imager, also referred to as ‘portaldosimetry’.– P: compare dose reconstructed inside patient/phantom CT scan (convert image to energy fluence,use as input for dose calculation algorithm) with plancalculated with patient/phantom CT scan.

• (b) Non-transmission treatment dosimetry: acquireimage for each field with the detector located betweensource and patient (during treatment), usually attachedto the linac-head.

– E: compare image (raw or converted to dose-to-water) with predicted EPID response or dose-to-waterat the level of the imager during treatment time, alsoreferred to as ‘portal dosimetry’.– P: compare dose reconstructed inside patient/phantom CT scan (convert treatment image to energyfluence, use as input for dose calculation algorithm)with plan calculated with patient/phantom CT scan.

• (c) Transmission treatment dosimetry: acquire image foreach field with the detector located behind patient orphantom.

– E: compare image (raw or converted to dose-to-water) with predicted EPID response or dose-to-waterat the level of the imager, behind the patient/phan-tom, also referred to as ‘portal dosimetry’.– P: compare reconstructed dose inside patient CTscan either back-project primary signal (simplecorrection-based algorithm) or convert image toenergy fluence, use as input for a dose calcula-tion algorithm with plan calculated with patientCT scan.

Non-transmission EPID dosimetryWhen considering EPID dosimetry, it is important to

understand the capabilities and limitations of various pos-sible methods to determine precisely what can be veri-fied. In this chapter we describe non-transmissiondosimetry methods, including 2D verification of the deliv-ered energy fluence or dose distribution at the level ofthe EPID, and 3D verification, where the dose is recon-structed in a phantom or patient CT scan based on 2Dimages. For non-transmission dosimetry, both 2D and 3Dverification methods are based on EPID measurements ac-quired without a phantom or patient between the radia-tion source and the detector. See Table 2 for anoverview of methods and key references presented inthe literature.

Non-transmission images can be useful for performingquality control of treatment parameters independent ofthe patient, related to dosimetric and geometric character-istics of the linear accelerator. Test fields have been de-signed to check the field flatness or symmetry of thebeam, absolute output of the linear accelerator and radia-tion-light field coincidence [52,53,84–86]. Furthermore, alarge number of papers have been published on the use ofEPIDs for the geometrical verification of MLC leaf positionsor leaf trajectories during dynamic multileaf collimation[52,53,84,86–97]. Most of these applications are based onusing a grey-scale image not converted into a portal doseimage, although generally these methods can also be usedwith portal dose images. These methods report in generalan accuracy of about 1 mm.

Furthermore, Vieira et al. [98] proposed a simple leaftrajectory check for dynamic treatments. They used a slid-ing field with a constant gap, and measurements with acamera-based EPID were performed. Differences in gapwidth as small as 0.2 mm could be detected making it anaccurate QA procedure. The same group also described aQA procedure for the verification of MLC calibration andthe delivery of static segmented MLC fields combined witha verification procedure for the absolute output verifica-tion of low dose segments [86]. Differences smaller than1% (1 SD) in output compared to ionisation chamber mea-surements were obtained. Partridge et al. [99] compareddifferent methods for the verification of dynamic MLCtreatments based on various EPID types: an in-house devel-oped imager [40], a method based on the liquid-filled EPID[100] that was later extended for a-Si EPIDs [101], and amethod for a camera-based EPID [102]. With all methods,it was possible to verify the position of the leaves duringdynamic delivery with an accuracy of approximately1.0 mm.

2D verification based on non-transmission imagesVerification of the energy fluence or dose by means of an

EPID is a common approach to ensure accurate delivery ofradiotherapy beams. Two strategies have been adoptedfor verification of the planned beams: a comparison per-formed at the level of the detector or a method that recon-structs the dose distribution inside the patient or phantombased on non-transmission portal images. The fluence orthe portal dose is either exported from the TPS, or calcu-

Table 2List of key references on non-transmission based dose verification methods

Verification procedure Type of verification Key references Objective of verification or subject of thestudy

QA of treatment machine QA Prisciandaro [93] Radiation-light field congruenceQA Dirkx [52,53], Budgell [80,87] Linac output, beam profile flatness and

symmetryQA Baker [84], Yang [95], Samant [96],

Parent [97]MLC leaf position for step-and-shoot fields

QA Vieira [86] MLC leaf position and absolute output forlow MU segmented fields

QA Vieira [98], Partridge [99],Chang [88]

MLC leaf position during dynamictreatment

2D fluence or dose verificationat EPID level

Pre-treatment Chang [82,106] Dose profilesPre-treatment Siebers [68], Parent [58] Measured EPID response and Monte Carlo

simulationPre-treatment Van Esch [30,64] 2D energy fluence verification using Li-Fi

and a-Si EPIDsPre-treatment Nijsten [44] Absolute dose at centre of field, field

shapes and wedge angles

2D dose reconstructionin phantom

Pre-treatment Warkentin [109] Reconstruction of the dose in a plane in aphantom based on non-transit portal doseimages

Cozzi [110], Moran [111],Nicolini [112]

Dose distribution measured with an EPIDpositioned inside a phantom

3D dose recalculation inphantom or planning CT

Pre-treatment van Elmpt [116,117] Dose distribution reconstructed inside(in)homogeneous phantoms

Pre-treatment Ansbacher [115] Dose distribution reconstructed inside avirtual phantom

Pre-treatment van Zijtveld [121], McNutt [123,124],Steciw [120]

Calculated dose distribution (TPS) using thepatient planning CT scan based onmeasured non-transit portal dose images

Pre-treatment van Elmpt [118] Calculated dose distribution (Monte Carlo)using the patient planning CT scan based onmeasured non-transit portal dose images

Pre-treatment Renner[119] Calculated dose distribution (independentdose engine) using the patient planning CTscan based on measured non-transit portaldose images


lated independently of the TPS from planned treatmentparameters. The energy fluence is used to predict eitherthe EPID response or the portal dose; i.e. the EPID signalis generally converted to a distribution of dose-to-waterdeposited at the level of the EPID.

Kirkby et al. [66] and Li et al. [103] investigated theimportance and the consequences of spectral response ofa-Si EPIDs for dosimetric calibration showing that on onehand additional copper shielding could reduce the over-re-sponse of detectors to low-energy components, while onthe other hand, when convolution/superposition methodsare used to predict EPID images, the incorporation of abeam-hardening dependence of the EPID kernel would im-prove dosimetric accuracy. Chytyk et al. [104] investigatedthe effect of tilted dose kernels for portal dose predictionand found that if one assumes parallel dose kernels, thedosimetric effect is less than 1.8% in extreme cases and lessthan 1.0% in most clinically relevant situations. Yeboah etal. [105] used Monte Carlo simulations to model the EPID re-sponse and found that the air gap and field size are roughly

complementary and suggested that an equivalent field sizeconcept may be used to account for intensity and spectralchanges arising from air gap variations.

Several other groups developed models to predict theEPID response to compare with non-transit EPID images oftreatment fields [30,31,44,51,58,64,68,106–108]. Parentet al. [58] also used a Monte Carlo model, and predictedthe image of a test IMRT field delivered on the beam axisand with an offset. A variation of EPID response up to 28%was found between the on- and off-axis delivery. Chang etal. [31,106] compared measured and planned relative pro-files and central axis dose of 25 prostate fields. After theircorrection methods were applied, very good agreementwas obtained, with a difference of 1.9% (± 0.5%, 1 SD).The agreement was better than 2.0% for the central axisdose. Nijsten et al. [44] reported a method to verify theabsolute point dose and the 2D geometrical parameters ofthe treatment fields at the level of the EPID, to check themanual treatment parameter transfer from TPS toaccelerator.


Besides predicting the dose at the imager level or theEPID response, another approach is to convert the energyfluence (measured without a phantom) to a 2D dose distri-bution inside a phantom at a specific depth. This method,first proposed by Warkentin et al. [109], involves deconvolu-tion of the raw EPID images to two-dimensional distributionsof primary fluence using Monte Carlo generated scatter ker-nels. The phantom dose-distribution that would result fromthe energy fluence derived from measurements is thenreconstructed. The calculated dose distribution inside thisphantom at a specific depth, e.g. 10 cm, can then be com-pared to treatment planning dose plane at that depth. Inthis way a direct comparison between the dose distributioncalculated by the TPS and the delivered in-phantom dosedistribution can be made.

A method that is more difficult to classify involves pre-treatment verification by inserting the EPID inside a phan-tom, or between attenuating material, at the position ofthe patient [110–112]. The material above the detectoreffectively acts as build-up layer, without an air-gap, sotherefore it can also be considered as a non-transmissionEPID dosimetry approach. Agreement with film has been ob-tained within 4%, 3 mm for phantom depths of smaller than4 cm [112].

Another possible arrangement for EPID dosimetry is ‘non-transit imaging during treatment’, by mounting the EPID tolinac head (Fig. 2b). Commercial development of this strat-egy using an EPID is, however, still at an early stage [113].There are various other dosimetry devices such as ionisationchamber matrices (i.e. they are not EPIDs) with the sameconfiguration [114]. The idea is to image beams when deliv-ered to the patient through the detector. The signal ac-quired during treatment is converted to the ‘deliveredenergy fluence’, and may be compared with the ‘plannedenergy fluence’. The advantage of using EPIDs is that theyhave a much higher resolution than the current matrixdetectors.

3D dose reconstruction based on non-transmissionimages

Another approach to verification based on non-transmis-sion dosimetry is the reconstruction of the 3D dose distribu-tion based on measured portal (dose) images. The idea isthat with both the 2D energy fluence delivered by the linearaccelerator (derived from portal images of each beam) and adose calculation algorithm, one can reconstruct a dose dis-tribution in 3D based on measurements. Images can be ac-quired pre-treatment (Fig. 2a) or during treatment (Fig.2b). The reference dose distribution can be calculated witha CT-scan of either a phantom or the patient, as appropriate.

For 3D in-phantom dose reconstruction, a method wasdescribed by Ansbacher [115]. The delivered segments ofan IMRT treatment were measured and reconstructed in apre-defined cylindrical ‘virtual phantom’ of fixed diameterusing a set of attenuation and phantom scatter correctionfactors. Absolute point dose values and relative dose pro-files were both reconstructed with an error smaller than2% (1 SD), compared to ionisation chamber and film mea-surements for a group of 20 head-and-neck cancer IMRT pa-tients analysed retrospectively.

An alternative algorithm was described by van Elmpt etal. [116–118], who used the energy fluence extracted froma measured portal dose image as input for a Monte Carlodose calculation in an object of choice. Reconstructionwas based on either a CT-scan of a phantom or the planningCT-scan of a patient. For both homogeneous and inhomoge-neous phantoms, agreement within 3% of the dose measuredusing film or ionisation chamber dosimetry was achieved. Inprinciple any dose calculation engine could be used, how-ever, a Monte Carlo-based approach provides greater accu-racy, especially for inhomogeneous geometries. Renner etal. [119] extracted the energy fluence from EPID measure-ments and applied an in-house developed dose calculationalgorithm using the planning CT-scan. A similar approachwas used by Steciw et al. [120], but these authors appliedthe same TPS as used for the original dose calculation.The advantage of these methods is that the transfer anddelivery of the plan, the creation of the initial energy flu-ence file and the modelling of the MLC by the TPS are allchecked. However, certain aspects of the dose calculation,occurring after the creation of the fluence distribution, re-main unchecked if the same TPS is used to calculate boththe planned and reconstructed dose distributions. van Zijt-veld et al. [121,122] used a method developed previously byMcNutt et al. [123,124] and applied it to non-transmissionmeasurements. In this method, the incident primary energyfluence of the plan (exiting the treatment head) is adjusteduntil it converges to the measured portal image. The flu-ence is then used to re-calculate the 3D dose distributionusing a cone-beam CT scan acquired just prior to treatment.

The main advantage of 3D verification methods usingplanning CT scans is that differences between planned anddelivered beams may be assessed in terms of dose in the tar-get volume or dose in normal tissue, resulting in a pre-treat-ment verification method that is more intuitive thanconventional methods based on measurements at points orin planes. Errors in the dose calculation of the TPS can bedetected along with the plan deliverability, such as correctdata transfer and mechanical or dosimetric functioning ofthe linear accelerator. Furthermore, with pre-treatmentverification all errors due to patient anatomy changes, pa-tient set-up or treatment machine malfunctioning that oc-cur during treatment will be missed. Dose reconstructionmethods that do not include patient information at the timeof treatment are potentially misleading, so users should beaware of their limitations (for more detail, see Clinical ap-proaches to EPID dosimetry).

Transmission EPID dosimetryThe ultimate aim of radiotherapy verification is to verify

that the patient is receiving the correct dose during treat-ment. To this end, it is becoming increasingly advantageousto determine the dose from transmission images, wherebythe beam passes through the patient, as described in Fig.2c. In this way the dose can be verified either at the levelof the EPID or by reconstructing the dose inside a digitalrepresentation of the patient. While dosimetry with or with-out a phantom (pre-treatment transit or non-transit dosim-


etry) provides similar information, the reason for doingtransit dosimetry behind a phantom is to be able to extendsuch a procedure to patient treatment verification. Modelsbased on homogeneous phantoms need to be adapted to ac-count for inhomogeneities inside the patient. For verifica-tion during treatment, the closer this representationresembles the patient at the time of treatment (e.g. by

Table 3List of key references on transmission based dose verification methods

Verification procedure Position of verification:EPID or patient/phantom

Key re

Point doseverification

Patient Pasma

EPID and patient (in vivo) Nijste

Patient/phantom(pre-treatment and in-vivo)

Chang[130,1

2D transit doseverification at EPID level

EPID Pasmavan El

EPID KroonEPID Dahlgr

EPID Spezi

Extraction of entrance fluence HanseVieira

EPID McCur

EPID McNutMoham

2D transit doseverification at patient level

Patient/phantom (in vivo) EssersBoella

Patient/phantom (in vivo) Boella

Phantom (pre-treatment) Wend

Patient/phantom(pre-treatment and in-vivo)

McDer

Phantom (pre-treatment) Talam

3D dose verification (dosecalculated using a CT scanacquired from the planningstage or acquired ‘in room’)

Patient (planning CT) Hanse

Patient (planning CT) Jarry

Patient (planning CT) McNut

Patient (planning CT) Louwe

Patient (in room CT) McDer

Patient (in room CT) Partri

using a set of (cone-beam) CT scans), the more accurate thisvalidation will be. EPID dosimetry based on transmissionimages is described in this section, and classified by point-dose verification, 2D portal dose prediction, 2D back-pro-jection models and finally 3D dose reconstruction models.Various methods and references for transmission dosimetryare summarised in Table 3.

at EPID level and patient/phantom level, as indicated in Fig. 2

ferences Objective of verification or subjectof the study

[127] Measured point dose distributionback-projected to the dose in apatient at 5 cm depth

n [128] Measured point dose verified at EPIDlevel and back-projected to the dosein a patient at 5 cm depth

[82], Piermattei31]

Measured point dose back-projectedto the dose in a patient at theisocentre

[132,141],mpt [135]

Predicted transit portal dosedistribution using measured beamdata and the planning CT scan

wijk [142] Clinical results for prostateen [139,140] Predicted transit portal dose using

the collapsed cone superpositionmethod

[145] Portal dose image prediction usingMonte Carlo calculations

n [150], Fielding [144],[143], Spies [148]

Extract entrance energy fluence

dy [67,136–138] Predicted portal dose using MonteCarlo-based scatter kernels

t [152], Reich [34],madi [35,154],

Predicted portal dose using a TPS byadding the EPID to the planning CTscan (‘‘extended phantom’’)

[22,126], Kirby [156,157],ard [81,158]

Reconstructed exit dose (3D CRT)

ard [159,160] Reconstructed 2D midplane dose (3DCRT)

ling [76] Reconstructed 2D midplane dose(IMRT)

mott [161,162] Clinical results for prostate IMRT

onti [77] 2D verification (IMRT)

n [39] Back-projected dose based ontransmission EPID images andplanning CT scan

[163] Back-projected energy fluence basedon EPID images and a Monte Carlocalculation using the planning CT scan

t [123,124] Combined EPID transit dosemeasurement and planning CT scan inan ‘‘extendedphantom’’ using theTPS

[166] Reconstructed 3D dose distribution(breast)

mott [167] Reconstructed in vivo dose for rectalcancer patients using cone-beam CTscan

dge [164] Back-projected dose based ontransmission EPID images and MVcone-beam CT


Point-dose verification based on transmissionimages

The motivation for point dose verification based on trans-mission EPID images was to replace diode in vivo dosimetry[125]. In vivo dosimetry at a point has been useful fordetecting large errors in the delivery of conformal fieldsre-occurring at successive treatment fractions. Usually invivo dosimetry is performed using diodes positioned at theentrance or exit side (or both) of the patient. The time re-quired to position the diodes on the patient, to maintain anaccurate calibration and to develop a ‘trust-worthy’ systemare rather laborious and therefore a disadvantage. Further-more the shadowing effect of the entrance diode, having arelatively thick build-up cap, may compromise its use. Pointdose verification based on EPID measurements, therefore,has been an approach initiated by various groups from theonset of EPID dosimetry. Point dose measurements madebehind the patient have been compared at the position ofthe EPID [126], extrapolated (back-projected) to a pointwithin the patient (or phantom), at 5 cm depth [127,128]and at the isocentre [82,129–131].

Pasma et al. [127] and Nijsten et al. [128] correlated thedose measured with the EPID on the central beam axis withdose values at 5 cm depth using a (semi-) empirical relation-ship between these two quantities. Chang et al. [82] andPiermattei et al. [130] described similar methods to recon-struct the absolute dose at the isocentre of a phantom, bycalibrating the EPID under the same scatter conditions asthe actual measurements, i.e. using a set of correlation fac-tors between the EPID signal and the dose at the isocentre.

2D verification based on transmission images(verification at the EPID level)

A popular approach to transmission dosimetry involvespredicting the portal dose at the level of the imager behinda phantom or patient [22,43,49,132–141].

At the Erasmus Medical Center – Daniel den Hoed CancerCenter in Rotterdam (The Netherlands), the portal imagewas predicted behind patient CT scans. Pasma et al.[43,132] began by comparing corrected EPID images againstionisation chamber measurements, including modulatedfields, obtaining agreement up to 1%. De Boer et al. [49]compared EPID portal dose images and ionisation chambermeasurements (at the detector plane) behind an absorberfor various field shapes. The average difference in absolutedose was 0.1 ± 0.5% (1 SD) over the entire irradiation field,with no deviation larger than 2%. This strategy has also beenused during treatment to investigate organ motion [142].The various research and clinical approaches used by thisgroup have been summarised in a report by Dirkx et al.[134]. These authors and other groups have also proposedmethods to separate the modulated beam component fromthe anatomical information to check the entrance fluenceindependently [143,144].

Spezi and Lewis [145] developed a planar calibration ofthe Li-Fi EPID response using a full Monte Carlo simulationof a radiotherapy treatment facility including an MLC andEPID. The calibrated EPID was used as a dosimetric systemto validate an MC model for the MLC. Computations andmeasurements agreed within 2% of dose difference (or

2 mm in regions of high dose gradient). Spies et al.[146,147] developed a scatter model for portal images usingMonte Carlo generated data, followed by a method toreconstruct incident beam profiles from images acquiredbehind a phantom [148]. Their method was based on an iter-ative algorithm that compensated for beam attenuation andsubtracting phantom scatter reaching the detector, yieldingan accuracy for the incident beam distribution of betterthan 3% (relative difference). Researchers at the RoyalMarsden Hospital, London (UK) presented Monte Carlo simu-lations and a simple physical model for scatter correction[149]. Hansen et al. [150] extracted the primary signal fromtransmission measurements, accurate to within 1.5% (1 SD)for scatter-to-primary ratios up to 25%. Later Partridge andEvans [151] described the practical range for which themodel was valid; for centre-of-mass to detector distanceslarger than 60 cm and field areas up to 625 cm2.

Several groups have used the TPS to calculate the trans-mitted dose map at the EPID level [34,35,152–154], as wellas relative fluence maps for tomotherapy systems [155].Phantom thickness changes of �1 mm have been detected,corresponding to dose changes on the central beam axissmaller than 0.6% [34]. In a later study, these authors re-ported that both for homogeneous and inhomogeneousphantoms, more than 90% agreement was achieved usingdose-difference and distance-to-agreement criteria of 2%,3 mm or 3%, 2.5 mm [153].

2D back-projection models based on transmissionimages (verification at the patient or phantomlevel)

In order to verify the patient dose in vivo, it is necessaryto make a measurement at the same time that the dose isbeing delivered to the patient and use that measurementto determine the dose inside the patient. To do this withEPID dosimetry, two things are required: transmission mea-surements and an algorithm to reconstruct the dose in thepatient from the EPID signal.

One of the first back-projection methods described in theliterature deals with exit dose reconstruction. The EPID hasbeen used for this purpose by several groups [81,126,156–158]. The main objective was to replace diode exit dosime-try during treatment, with an accuracy reported to be with-in 3% compared with diode measurements behind phantoms[156]. Parsaei et al. [28] used the Li-Fi EPID to measure rel-ative exit dose profiles behind a phantom. Open and wedgeddose profiles measured with the Li-Fi EPID agreed with ioni-sation chamber measured profiles to within 3.5%.

Zhu et al.[25] used a Li-Fi EPID to back-project pixel val-ues to a 2D dose map of an ionisation chamber matrix mea-sured at the depth of dose maximum (Dmax) of 1.5 cm inwater. Boellaard et al. [159,160] subsequently introduceda back-projection midplane-dose dose reconstruction modelfor the Li-Fi EPID. The calibration of the EPID for back-pro-jection dosimetry consisted of two parts. First, a dosimetriccalibration established the dose–response relationship byrelating EPID pixel values to dose values at the position ofthe imager. Secondly, the parameters for the back-projec-tion algorithm, i.e. attenuation factors and scatter withinthe EPID and patient, and from patient to EPID, were deter-


mined to enable the conversion from the dose at the EPIDposition to the dose inside the patient or phantom. Thismodel was extended to the a-Si EPID by Wendling et al.[76], who reported close agreement between EPID and filmmeasurements using gamma evaluation with 2%, 2 mm dose-difference and distance-to-agreement criteria. Clinicalapplications were reported by McDermott et al. [161,162],where back-projected 2D EPID dosimetry replaced verifica-tion with film, and subsequently in vivo verification replacedpre-treatment checks for prostate treatments.

Spies et al.[148] also investigated reconstructing the 2Dincidence fluence based on a thickness map of the phantom,or transmission measurements. Pre-calculated Monte Carloscatter kernels were used to estimate scatter. The methodwas tested for a homogeneous water-equivalent slab phan-tom for simple rectangular and complex multileaf colli-mated fields. They reported an accuracy for the incidentbeam distribution of better than 3% (relative difference).

An alternative calibration method was described by Tala-monti et al. [77] for pre-treatment verification in a phan-tom. They used an EPID to measure exit doses, andrelated pixel values to film dose distributions measured ina phantom at the patient level. Measurements were madefor various phantom thickness, air gap distances, field sizesand measurement position in the phantom. These authorswanted to obtain a universal calibration factor, regardlessof field size. To this end, a copper plate was added to atten-uate low-energy phantom scattered photons which largelydepend on the field size of the irradiated volume. Thismethod was used for the measurement of the dose distribu-tions for 15 clinical IMRT fields. On average, 97.6% and98.3% of the points agreed (using gamma criteria of 3%,3 mm) for EPID versus TPS and for EPID versus film,respectively.

3D dose reconstruction based on transmissionimages

Three-dimensional dose reconstruction begins with a por-tal image taken behind a patient or phantom. Based on theinformation in this image, in combination with a 3D dosereconstruction model, the dose distribution for this segmentor field can be reconstructed in 3D within the patient or aphantom. Finally, the 3D dose distributions of all fieldsmay be summed to yield the total 3D dose distribution. Ingeneral, 3D dose reconstruction based on transmissiondosimetry requires (a) modelling of the EPID response,including a correction for the scatter from the patient orphantom to the measured EPID image, (b) a 3D ‘‘densityrepresentation’’ of the patient or phantom, preferablyincluding position and anatomy of various types of tissueand organs, and (c) a dose calculation algorithm.

Hansen et al. [39] demonstrated this concept on ahumanoid phantom. First, the EPID image was correctedfor scatter from the phantom to the detector yielding thetransmitted primary fluence. This fluence was then back-projected through the planning CT dataset yielding the pri-mary fluence within the phantom. By convolving this distri-bution with energy deposition kernels (taken from theliterature) the dose in the phantom was calculated. Thereconstructed dose distribution agreed with the dose distri-

bution from the TPS within 2% in the pelvic region of ahumanoid phantom. Jarry and Verhaegen [163] investigatedthe use of Monte Carlo calculations for this step and reachedan agreement within 5% for IMRT treatments. The planningCT-scan was used for this method and is only reliable foranatomy that is not changing during treatment.

Partridge et al. [164] extended the model of Hansen etal. by incorporating an on-line imaging modality and per-formed a proof-of-principle experiment on an anthropomor-phic phantom. The portal image was converted to primaryfluence using an iterative phantom-scatter estimation.Next, the energy fluence was back-projected through a‘‘treatment-time’’ MV (cone-beam) CT model of the phan-tom. For dose reconstruction, the same convolution/super-position algorithm as for the original planning dosedistribution was used. Recently, Chen et al. [165] also dem-onstrated the use of a MV cone-beam CT dataset, acquiredjust prior to treatment, for dose reconstruction within a pa-tient. In their approach, no correction for the scatter fromthe patient to the EPID was included; hence reconstructeddose values were 10% higher than planned values.

Another concept was developed by the group from theNetherlands Cancer Institute – Antoni van LeeuwenhoekHospital (NKI-AVL) in Amsterdam (The Netherlands) basedon their back-projection algorithm for 2D dose reconstruc-tion. The model was extended to multiple parallel planesperpendicular to each beam direction. In this way it waspossible to get a full 3D dose reconstruction as demon-strated for breast cancer treatments using EPID images incombination with planning CT data. This information wasused to determine the dose distribution in irradiated anthro-pomorphic breast phantoms [166]. The scattered compo-nents of the dose within the EPID, from the phantom tothe EPID, and within the phantom were taken into accountby using empirical kernels. For the attenuation correction,the external contours of the phantom were used, whichwere determined from the (planning) CT-scan. The averagedeviation of the 3D reconstructed dose from the planneddose was 1.4% with a SD of 5.4%. A later study combinedthe use of an a-Si EPID-based 3D in vivo dosimetry andcone-beam CT (CBCT) to obtain a complete account of theentire treatment for a selected patient group. For that pur-pose, nine hypo-fractionated rectum cancer IMRT patientplans were investigated [167]. Planned and measured in vivo3D dose distributions showed excellent agreement, based on3D gamma evaluation and dose-volume histograms (DVHs).Over-dosage of up to 4.5% was measured only in one fractiondue to the presence of gas pockets.

McNutt et al. [123,124] included the EPIDwith the CT data-set into a so-called ‘‘extended phantom’’. With an iterativeconvolution/superposition algorithm the primary energy flu-ence at the EPID was adjusted until it converged to a fluencethat would have produced the measured portal image. Theconverged primary energy fluence was back-projected andconvolved with dose-deposition kernels. The dose distribu-tion agreed within 3% with the planned dose distribution.

For the tomotherapy treatment machine, a procedurewas developed by Kapatoes et al. [168,169]. A reconstruc-tion of the incident energy fluence was made using the mea-sured transit signal in the detector and a measured transferfunction. This function was calculated from a database of


calibration data combined with the on-line MV CT-scan forthe extraction of the radiological path lengths and pa-tient–detector distances. The reconstructed energy fluencewas then used as input for a dose calculation inside the MVCT-scan. A phantom study [169] showed that positional dif-ferences down to 2 mm or 2� could be detected, but hadnegligible impact on the 3D dose distribution. Acceleratoroutput differences or stuck leaves of the MLC could be de-tected and their influence on the 3D dose distribution beestimated.

Clinical approaches to EPID dosimetryIn the past, treatments were fairly standardised, so expe-

rienced radiographers, physicists and radiation oncologistsknew what to expect in terms of sizes and shapes of treat-ment fields, and number of monitor units delivered, for a gi-ven type of treatment. With the introduction of advancedradiotherapy techniques and higher dose prescriptions, awide range of parameters is possible, so deviations are lessobvious and it is much more difficult to know when some-thing is wrong. With the introduction of new data manage-ments systems, change of TPSs and protocols, the risk oferror also increases, even for a simple treatment technique.It is therefore even more surprising that the commercialdevelopment of EPID dosimetry is lagging far behind otheradvances such as patient positioning, i.e. image-guidedradiotherapy (IGRT). To the best of our knowledge, thereare only two commercial products available. The first oneis produced by Varian, who released a dosimetric modulein ‘PortalVision’ based on a publication by Van Esch et al.[64]. Their method allows verification of pre-treatment,non-transmission fields, at the level of the EPID (Fig. 2a).The second commercial product currently available is basedon the method described by Renner et al. [119]. It uses apre-treatment 3D dose verification method as described inthe section on 3D non-transmission dosimetry. Despite thelack of commercial availability, several clinics have intro-duced in-house solutions and produced data describing thepossibilities of EPID dosimetry in clinical practice. Beforediscussing the clinical experience using EPID dosimetry, anoverview is presented on the type of errors that can bedetected.

Error detection potential using EPID dosimetryDescriptions of the various ways an EPID may be used

to verify dose distributions in 2D/3D are summarised inTable 4. The methods are divided into two categories:pre-treatment verification (image measured outside oftreatment time, behind or without a phantom) and treat-ment verification (image measured before or behind thepatient, dose determined inside the patient, in vivo, orat the detector level). Under ‘potential errors’, variouspossible sources of error in radiotherapy are listed andthe type of EPID verification that is able to detect this er-ror is also indicated.

Several types of errors can be distinguished. Machine er-rors are related to hardware faults, either random or sys-tematic (occurring at every treatment fraction). Pre-treatment measurements using EPIDs reveal systematic er-

rors and may be used to estimate the probability of randommachine errors that might occur during treatment. How-ever, they cannot be used to detect random errors occurringduring actual patient treatment, such as a change in linacoutput. If such a random error occurs during pre-treatmentverification, it can either be corrected or the measurementcan be repeated yielding information about its average va-lue and standard deviation. The detection of a wrong gantryangle, due to a mechanical failure, may be difficult withEPID dosimetry. Whether or not EPID dosimetry is sensitiveto gantry angle errors is highly dependent on the treatmenttechnique. A case where it might be possible to detect awrong gantry angle is when transit measurements are madebehind a non-homogeneous phantom or patient. Wrong gan-try angles may then result in different transmission imagesand therefore in a wrong dose distribution at the patient le-vel. A problem arises when the transmission images do notchange much, and the unplanned angle results in the ‘cor-rect dose’ delivered to the wrong location. Such a situationemphasises the need for dose reconstruction in combinationwith patient anatomy and set-up verification at the time oftreatment, for instance by including an analysis of anatom-ical structures visible in EPID images. For transit or non-transit dosimetry in 3D, the contribution of dose from awrong gantry is likely to render a different total dose distri-bution in certain cases.

Plan errors include dose calculation errors, wrong TPScommissioning data (such as beam fit errors) and data trans-fer errors, which include the selection of a wrong patientplan. If the error is related to the heterogeneity of the pa-tient anatomy, the error will not be detected by re-calculat-ing the dose in a homogeneous phantom. EPID-based (3D)dose reconstruction algorithms with an independent dosecalculation that take into account inhomogeneities (e.g.Monte Carlo, convolution/superposition or collapsed-conealgorithms) are necessary for this type of verification. Errorsdue to poor modelling of the MLC, however, are likely to bedetected by using 2D EPID verification models. As with gan-try angle errors, selection of the wrong plan will most likelybe detected by EPID transit dosimetry, but only if the trans-mission results in a different dose distribution. The errorwill not be detected if the measured and planned dose dis-tributions are similar, even though the medium (i.e. the pa-tient) is different.

Patient errors are specifically related to errors due tochanges in the patient’s positioning or anatomy fromthe situation at planning. Verification of the dose basedon EPID measurements prior to treatment cannot detectany of these types of errors. Furthermore, at least onevendor is developing a system, whereby EPID images areacquired with the detector between the beam sourceand the patient (Fig. 2b). The detection of patient-relatederrors will be limited to situations where additional imag-ing information is available (usually acquired before and/or after treatment) and the patient did not change be-tween imaging and treatment time. 2D EPID verificationmodels during treatment are used to provide a consis-tency check for the dose delivered for treatment sites,where no large differences in anatomy are expected.These methods are either based on the portal dose imageor use a back-projection algorithm to reconstruct a dose-

Table 4Overview of the various errors that can be detected with EPID dosimetry

Potential errors Pre-treatment verification Treatment verification

2D/3D 2D 3D 2D 3D

No phantom Behindphantom

Insidephantom

Insidephantom

Beforepatient

Behindpatient

Insidepatient

Insidepatient

MachineWedge presence anddirection

Yes (systematic errors) Yes (systematic and random errors)

Presence of segment Yes (systematic errors) Yes (systematic and random errors)MLC leaf position/speed Yes (systematic errors) Yes (systematic and random errors)Leaf sequencing Yes (systematic errors) Yes (systematic and random errors)Collimator angle Yes (systematic errors) Yes (systematic and random errors)Beam flatness andsymmetry

Yes (systematic errors) Yes (systematic and random errors)

Linac output duringtreatment

No Yes

Gantry angle No Possible Possible Possible No Possible Possible Possible

PlanTransmission throughleaves

Yes Yes Yes Yes Yes Yes Yes Yes

Steep dose gradients Yes Yes Yes Yes Yes Yes Yes YesTPS modelling parametersfor MLC

Yes Yes Yes Yes Yes Yes Yes Yes

Delivery of wrong patientplan

Yes (if same plan is used for verification andtreatment)

Yes Yes Yes Yes

Dose calculation inphantom or patient

No No Yes Yes No No Yes Yes

PatientTable arm obstruction No No No No No Yes Yes YesObstructions fromimmobilisation devices

No No No No No Yes Yes Yes

Anatomical changes inpatient sinceplanning CT


Anatomical movementsduring treatment


Wrong patient duringtreatment


Under/over-dose tovolumes of interest

No No No No No No Single plane Yes

Dose distribution inpatient during treatment

No No No No No No Single plane Yes


plane inside the patient. For treatment sites where anat-omy changes are common, it is difficult to say anythingmeaningful about the dose delivered to the region-of-interest (i.e. target volumes or organs at risk) or thetotal dose distribution. This will only be possible if thedose is determined in 3D and based on accurate 3D imag-ing data (e.g. cone-beam CT) acquired close to the timeof treatment.

For machine, plan or patient-related errors, pre-treat-ment and treatment EPID dosimetry raises an alarm beforeor from the first fraction (after other procedures and checksfail), alerting clinical staff to what often turns out to be asimple error with drastic consequences. Ideally, when EPIDtreatment images can be acquired for every fraction andanalysed automatically, a check and record of the entiretreatment will be guaranteed without any cost to patienttreatment time or clinical workflow.

Acceptance and rejection criteriaQuick and adequate interpretation of a verification result

can be hampered by the fact that 2D and 3D images rendervast amounts of information for the user. For dose distribu-tions with high dose gradients, simple dose difference mapsare usually insufficient to deduce clinically relevant differ-ences between planned and measured dose distributions.Dose-difference and distance-to-agreement (DTA) analysesare extremely sensitive in regions of high- and low-dose gra-dient, respectively. The most common method today tocompare two images is the gamma evaluation method[170,171], although other methods have also been proposed[172,173]. The gamma evaluation takes both the dose dif-ferences and the DTA into account. The dose criterionshould be stated relative to a reference value. Usually thisis the maximum or reference dose at the dose specificationpoint (i.e. global), but also the local dose difference could


be used (i.e. local). Gamma criteria are indicated as x%,y mm. However, in clinical practice easy to use criteriaare needed to quickly evaluate differences between mea-sured and planned dose distributions. Acceptance andrejection criteria can facilitate this process. There is, how-ever, no consensus on which criteria should be used for spe-cific treatment sites or equipment used.

Various strategies are proposed in the literature withmost 2D metrics and criteria based on film dosimetry usedto verify the total dose in a plane in a phantom. Film dosim-etry is commonly used to detect plan errors, and sometimesalso machine errors. Childress et al. [174] suggested usingcriteria of 5% for local dose difference and 3 mm for DTA.They described the type of errors that could be detected,e.g. beam energy change, wrong collimator rotation or gan-try angle. Budgell et al. [175] used 3% of the maximum doseand 3 mm DTA. They focused on the region within the 80%isodose surface, but switched to 4%, 4 mm criteria if largedeviations were visible. The ESTRO QUASIMODO project[176] used 4%, 3 mm and looked at median and 95% percen-tile. Another example of pass/fail criteria for IMRT pre-treatment verification using film dosimetry was presentedby Stock et al. [177], based on the percentage of points witha gamma value larger than a threshold value and the aver-age values of the gamma points in a specific region, usinga 3% dose difference to a reference point and a DTA of3 mm. If the acceptance criteria were not met, they recom-mended looking at dose profiles and dose difference maps orat additional parameters such as the gamma angle. Thisparameter describes whether the dose or the DTA-criterionis dominant. Spezi et al. [178] described the use of gamma-area and gamma-volume histograms to analyse 2D and mul-tiple 2D gamma evaluations, respectively, using 3%, 3 mmcriteria. For all of these studies, the final interpretation,if the criteria were not met, was still made by an expert,who decided to accept or reject the dose distribution.Therefore the gamma index has been used more to alertthe user to discrepancies, rather than as a strict set ofpass/fail criteria.

McDermott et al. [162] presented criteria and accep-tance levels for field-by-field comparisons of multiple frac-tions. Composite images for each field were made byselecting the minimum, median and ‘low’ – half-way be-tween minimum and median – gamma value per pixel, overmulitiple fractions (1–5), in order to filter out random er-rors. They also introduced four levels of ‘alert criteria’,from very easy to easy, medium and strict based on theaverage, maximum gamma value and the percentage ofpoints exceeding unity. The various alert criteria, and thenumber of fractions combined, were tested to develop adecision tree to distinguish between systematic and randomerrors, and minimise false positive and false negative re-sults. The optimum result required a medium level alert cri-terion, based on combining ‘low’ gamma values formeasurements of three fractions for prostate IMRTtreatments.

Such decision trees can facilitate the interpretation ofdifferences between dose distributions and the implemen-tation in automatic analysis. van Zijtveld et al. [179] imple-mented such a decision tree for the automatic evaluation ofgamma images, both 3%, 3 mm global and 3%, 3 mm local,

acquired for IMRT pre-treatment verification of the treat-ment fields delivered and measured using the EPID. They in-cluded not only simple metrics such as the number of pointsfailing the gamma evaluation but also incorporated moreadvanced features of the image such as the size of the areawhere differences are visible. They suggested that forbeam-by-beam analysis the local dose difference criterionshould be used to prevent differences in low dose areasfrom contributing to unacceptable discrepancies in the totalplan.

A single solution to what criteria should be used in clini-cal practice is difficult. Most centres use a gamma evalua-tion with a dose criterion of 3% of the maximum orreference dose combined with a DTA criterion of 3 mm,and pose restrictions on metrics based on gamma statisticsover a specified area or volume, such as the average gammavalue, the maximum gamma value or the proportion ofpoints with a gamma value above a certain threshold value.However, acceptance criteria for the percentage of pointsnot fulfilling these restrictions greatly differ between vari-ous centres, treatment techniques and treatment sites.The main reason for these differences is that the level ofaccuracy that must be achieved should be carefully bal-anced between the rate of false positive errors (increasingworkload unnecessarily) and false negative errors (failingto detect clinically relevant discrepancies).

With the emerging methods that can reconstruct dose in3D, another interesting point is how these dose distributionsneed to be compared. Two strategies can be performed.First, a ‘standard’ verification procedure using the gammaevaluation or dose-difference maps can be performed: averification of the two datasets at a selected slice, multiple2D slices or even a gamma evaluation for the entire volume[180]. A second strategy is to analyse the dose distributionin terms of anatomical structures or a region-of–interest-based comparison: dose-area histograms for 2D verificationor dose-volume histograms for 3D verification methods. Thelatter strategy might be preferable because a calculation ofthe parameters in terms of patient-related characteristicscan be performed: isodose display on top of the structuresof interest, mean dose to the tumour or in the normal tis-sue, or eventually, a reconstructed dose-volume histogramto get a quick idea of the total treatment plan. This wayof verification would ease the process of interpretation ofpossible differences to the judgement of the verified planin terms of a ‘normal’ treatment plan evaluation.

Several examples are given in the literature of errors indose delivery observed by using EPID measurements in com-bination with a 3D dose reconstruction method. van Zijtveldet al. [121] described the impact of a malfunctioning leaf(machine error), picked up during pre-treatment verifica-tion, on the 3D dose distribution inside the patient. Malfunc-tioning of one of the leaves caused a difference betweenmeasured and predicted portal dose image of about 10% inone patient. Due to the combination of the different treat-ment fields and the large size of the tumour, the dose dif-ferences reduced to 5% in the reconstructed 3D dosedistribution. They therefore concluded that for clinicalevaluation, one cannot rely on analysing only the DVH. Prop-er local dose evaluation is important as well, for example,by using automatic slice-by-slice comparison. Another


example is given by Chen et al. [165] describing the in vivoverification of the dose delivered to a patient. The authorsmeasured the dose behind the patient with an EPID and con-verted the image data to dose values inside the patientusing a set of megavoltage cone-beam CT scans. They founddifferences due to the changed patient anatomy during thecourse of treatment.

Clinical experienceRoutine clinical use of EPIDs for dosimetry has been re-

stricted to a very few centres until recently. This was partlydue to the absence of commercial solutions, the limited useof EPIDs for set-up verification and the lack of demand forpatient-specific dose verification. With the introduction ofIMRT, these aspects have received much more attentionthan in the past. The geometric accuracy of patient posi-tioning is very important for IMRT techniques with steepdose gradients between target volumes and organs at risk,thus accelerating the introduction in the clinic of efficientand accurate use of EPIDs to replace film for portal imaging.Furthermore, the need for patient-specific IMRT dose verifi-cation also stimulated many groups to investigate the use ofEPID dosimetry. Finally many groups realised that EPIDscould be used to perform in vivo dosimetry in a relativelysimple way during IMRT treatments, without much addi-tional time at the accelerator. In this section we will sum-marise the results of clinical studies applying the methodsoutlined in the previous sections.

The group at the Erasmus Medical Centre – Daniel denHoed Cancer Center in Rotterdam (The Netherlands) hasextensive experience with clinical use of camera-basedEPIDs for dosimetric verification of patient treatments,as outlined in the previous sections. In their first paperon clinical use, they reported their experience with in vivodosimetry based on EPID images acquired during treatmentof 10 prostate cancer patients [142]. EPID images were ac-quired for all fields during each fraction. The resultingmeasured portal dose images were compared with the por-tal dose images predicted with their model, which uses theplanning CT scan of the patient. The difference betweenthe average measured and predicted on-axis portal dosewas 0.3 ± 2.1% (1 SD) for the lateral fields and 0.7 ± 3.4%(1 SD) for the anterior fields. Off-axis differences betweenmeasured and predicted dose values were often much lar-ger, up to 15%, and were caused by frequently occurringgas pockets in the rectum during treatment or CT scan-ning. In a larger group of 115 prostate patients, the doseat 5 cm depth in the patient, D5, was derived from EPIDmeasurements for each treatment beam and comparedwith the dose predicted by the TPS using the empiricalrelationship between D5 and the portal dose on the beam’scentral axis [127]. The mean differences between mea-sured and predicted D5 values were similar to those ob-served when using their portal dose image predictionmodel. The difference between measured and predictedD5 values was larger than 5% for seven patients. All thesepatients had relatively large gas pockets (3–3.5 cm diame-ter in the anterior–posterior direction) in the rectum dur-ing acquisition of the planning CT scan, which were notpresent during most of the treatments. Pre-treatment ver-

ification of dynamic IMRT delivery is performed in Rotter-dam using camera-based EPIDs in combination with thefirst model developed by the Rotterdam group [181], whichcompares measured with predicted portal dose images. Re-cently, van Zijtveld et al. [179] analysed the results of 270patients treated with dynamic MLC and verified using thismethod, in combination with a gamma evaluation. For thislarge patient group they intercepted four severe errors; inone case a wrong treatment plan was sent to the acceler-ator, and three times one of the leaves were malfunction-ing. The authors also proposed an automated imageanalysis scheme, as discussed in the former section, bywhich visual inspection of images could be avoided in 2/3 of the cases.

The group at the Netherlands Cancer Institute – Antonivan Leeuwenhoek Hospital (NKI-AVL) in Amsterdam (TheNetherlands) has developed a back-projection model basedon EPID measurements mainly with the idea for use as aroutine in vivo dosimetry tool. In a pilot study Boellaardet al. [160] demonstrated the usefulness of their approachto verify the midplane dose during radiotherapy of larynx,breast, lung, and prostate cancer patients. Midplane dosedata derived from portal dose measurements were com-pared with those calculated with a 3D TPS and generallyshowed good agreement. The reproducibility was within1% (1 SD) and agreed within 2.5% with the treatment plan-ning data for most parts of the field. For a few of theprostate and lung treatments, larger deviations werefound due to the differences between the actual patientanatomy and the planning CT data, as a result of variablegas filling in the rectum and anatomical changes in thelung. Clinical results for prostate treatments were laterreported, where a-Si EPID dosimetry replaced all filmand ionisation chamber measurements for IMRT pre-treat-ment verification from February 2005 [161]. For 20 patientplans, pre-treatment phantom verification was comparedwith an EPID and an ionisation chamber yielding an aver-age ratio of 1.00 ± 0.01 (1 SD) at the isocentre. The same20 plans were also verified in 2D, with only 1.5% of thepoints yielding gamma values larger than unity (3% ofDmax, 3 mm DTA). Results for 7 out of 20 plans revealedunder-dosage in individual fields ranging from 5% to 16%,occurring along the junction of abutting segments (ton-gue-and-groove side). The agreement was improved afteradjusting an incorrectly set tongue-and-groove widthparameter in the TPS. Subsequently, a report detailedthe replacement of pre-treatment with in vivo EPID dosim-etry for IMRT prostate radiotherapy [162]. Dose distribu-tions were reconstructed from EPID images, inside thephantom and patient. Planned and EPID isocentre dosevalues were, on average, within 1% (0.01 SD) for bothpre-treatment and in vivo verification of the total plan.Multiple in vivo fractions were combined by assessing gam-ma images, and showed that information obtained in threefractions rendered similar results as pre-treatment verifi-cation. Additional time for verification was about 2.5 hper plan for pre-treatment checks and 15 min + 10 min/fraction for in vivo dosimetry.

Researchers at the Royal Marsden Hospital in London (UK)have developed a model for using EPID measurements of thetangential fields to construct a thickness model of the


patient’s breast. Using this information, IMRT beams weredesigned to improve dose homogeneity in the breast byreducing the high-dose region [182]. This model was anextension of an earlier developed method [39] to designphysical compensators to deliver IMRT to breast cancer pa-tients. Their EPID dosimetry model was the basis of a ran-domized clinical trial in which 300 patients wererandomized to evaluate the results of improved dose unifor-mity on treatment cosmetic outcome [183].

At the MAASTRO Clinic in Maastricht (The Netherlands),pre-treatment verification and in vivo dosimetry usingEPIDs has been performed since 2004 for all patients trea-ted with curative intent with photon beams. They ac-quired and analysed approximately 32,500 images ofabout 2500 patients for a period of 24 months. Radiationtherapy technologists perform both the measurementsand the analysis by using in-house developed software.Calibrated camera-based EPIDs were used to measure thecentral field dose, which is compared with a dose predic-tion at the EPID level [44,62]. For almost 9000 treatmentfields, they found a small mean dose difference of0.1 ± 2.4% (1 SD) and several errors such as wrong manualentering of treatment parameters, the omission of ashielding block and an error in the machine output. Fortreatment dosimetry, transit dose data were calculatedusing patient transmission and scatter, and compared withmeasured values. Furthermore, measured transit dose dataare backprojected to a dose value at 5 cm depth in water(D5) and directly compared with D5 calculated by the TPS.Patient-related delivery errors reported were often due toset-up errors and organ motion, initiating further investi-gation of error sources. Imaging during treatment com-bined with set-up correction protocols is necessary todecrease dose differences, particularly in the case ofbreast, lung, and head and neck treatments.

In Bellinzona (Switzerland), results for pre-treatmentEPID verification of 44 static IMRT patient plans were re-ported [110]. With 9.9% of points within the measured fieldsbeyond a gamma value of unity (3% of Dmax, 3 mm DTA),they concluded that the technique was feasible, dosimetri-cally accurate and recommended for large-scaleimplementation.

A group in Rome (Italy) [130] used an a-Si EPID for verify-ing the dose at the isocentre for a group of 10 patients trea-ted with conformal techniques in the pelvic region. Adosimetric module of a commercial software tool was usedto convert the transmitted signal of 25 pixels around thecentre of the EPID into dose values, which was then usedto determine the dose at the isocentre using a model basedon a series of phantom measurements. The authors reportthat the use of EPIDs strongly reduced the workload of theirin vivo dosimetry procedure compared to their previousmethod in which a small ionisation chamber, positioned onthe EPID, was used.

Discussion and conclusionsThe use of EPIDs for dosimetry has matured, and cur-

rently a variety of reliable and accurate dose verificationmethods are used in clinical practice, but this is not yet

widespread. Quality control of a linear accelerator, pre-treatment and treatment dose verification, as well as invivo dosimetry methods using EPIDs is now routinely inuse in a growing number of clinics. First it has been shownby several groups that for IMRT verification, EPID dosime-try can replace the more cumbersome film and ionisationchamber measurements. From these studies it can be con-cluded that one set of EPID measurements is sufficient tocheck an IMRT plan, making it a useful, efficient and accu-rate dosimetric tool. Furthermore, there is a growing de-mand to verify that the planned dose distribution isindeed correctly delivered to the patient. For that pur-pose, treatment dosimetry is an essential tool, and EPIDdosimetry has shown to be a better alternative to othermethods that are limited to point dose information, partic-ularly during IMRT treatments. Once the EPID is calibratedfor dosimetry, treatment dosimetry performed with EPIDsis a fast procedure compared to pre-treatment verificationbecause the images are acquired during treatment time(for some methods an additional set of images may be re-quired). The only extra time is needed for analysis, whichcan be automated. Since portal imaging is generally per-formed by radiation therapy technologists at the treatmentmachine, IMRT dose verification based on EPID images ac-quired during treatment only requires additional input ofphysicists or radiation oncologists if criteria for acceptanceare not met, as for the routine use of portal imaging forpatient set-up verification.

As radiotherapy becomes more complicated, both invivo dosimetry and geometry verification in three dimen-sions using image-guided radiotherapy (IGRT) methods willbecome necessary. Also over longer treatment times, ana-tomical changes such as weight loss, flexing of the neckand variation in lung density, occur frequently, resultingin dose delivery variation over a series of fractions. Meth-ods to integrate 3D in vivo dosimetry and IGRT procedures,such as the use of kV or MV cone-beam CT, are thereforeunder development. IGRT procedures using a cone-beamscan made on the treatment day for dose reconstruction,as well as for re-contouring volumes of interest, would im-prove the accuracy of in vivo dosimetry for many treat-ment sites. A challenge will be to implement theseprocedures in clinical routine; i.e. finding answers forquestions such as when is a deviation in dose clinicallyunacceptable, when should a patient be re-planned, andhow should dose variations in deformed volumes be takeninto account in an efficient way.

EPID dosimetry is evolving now to an integrated part inthe total chain of verification procedures that are imple-mented in a radiotherapy department. It provides a safetynet for advanced treatments as well as a full account ofthe dose delivered to specific volumes, allowing adapta-tion of the treatment from the original plan if necessary.Finally it should be noted that large-scale clinical imple-mentation of EPID dosimetry is dependent upon commer-cial availability. Current models and software aredeveloped by research groups and are mainly limited toacademic centres. Commercial solutions should be easyto implement, simple, robust and efficient in their dailyuse, and be sufficiently accurate for the purpose theyare serving.


AcknowledgementTwo authors, L.N. McDermott and M. Wendling, acknowledge

the financial support of the Technology Foundation STW (ProjectNo. 7184). The authors are also grateful to the colleagues at MAAS-TRO and NKI-AVL for contribution to discussions about the clinicalrole of EPID dosimetry in a radiotherapy department.

* Corresponding author. Wouter van Elmpt, Department ofRadiation Oncology (MAASTRO), GROW, University Hospital Maas-tricht, Dr. Tanslaan 12, NL-6229 ET Maastricht, The Nether-lands. E-mail address: [email protected]

Received 9 April 2008; received in revised form 9 July 2008;accepted 12 July 2008; Available online 14 August 2008

References[1] Herman MG. Clinical use of electronic portal imaging. Semin

Radiat Oncol 2005;15:157–67.[2] Herman MG, Balter JM, Jaffray DA, et al. Clinical use of

electronic portal imaging: report of AAPM Radiation TherapyCommittee Task Group 58. Med Phys 2001;28:712–37.

[3] Herman MG, Kruse JJ, Hagness CR. Guide to clinical use ofelectronic portal imaging. J Appl Clin Med Phys2000;1:38–57.

[4] Langmack KA. Portal imaging. Br J Radiol 2001;74:789–804.[5] Antonuk LE. Electronic portal imaging devices: a review and

historical perspective of contemporary technologies andresearch. Phys Med Biol 2002;47:R31–65.

[6] De Wagter C. QA-QC of IMRT: European perspective. In:Bortfeld T, Schmidt-Ullrich R, De Neve W, Wazer DE,editors. Image-guided IMRT. Berlin-Heidelberg: Springer-Verlag; 2006.

[7] Moran JM. QA-QC of IMRT: American perspective. In: BortfeldT, Schmidt-Ullrich R, De Neve W, Wazer DE, editors. Image-guided IMRT. Berlin-Heidelberg: Springer-Verlag; 2006.

[8] Wiezorek T, Banz N, Schwedas M, et al. Dosimetric qualityassurance for intensity-modulated radiotherapy feasibilitystudy for a filmless approach. Strahlenther Onkol2005;181:468–74.

[9] Amerio S, Boriano A, Bourhaleb F, et al. Dosimetric charac-terization of a large area pixel-segmented ionization cham-ber. Med Phys 2004;31:414–20.

[10] Spezi E, Angelini AL, Romani F, Ferri A. Characterization of a2D ion chamber array for the verification of radiotherapytreatments. Phys Med Biol 2005;50:3361–73.

[11] Letourneau D, Gulam M, Yan D, Oldham M, Wong JW.Evaluation of a 2D diode array for IMRT quality assurance.Radiother Oncol 2004;70:199–206.

[12] Jursinic PA, Nelms BE. A 2-D diode array and analysis softwarefor verification of intensity modulated radiation therapydelivery. Med Phys 2003;30:870–9.

[13] Monti AF, Frigerio G. Dosimetric verification of 6 and 18 MVintensity modulated photon beams using a dedicated fluoro-scopic electronic portal imaging device (EPID). RadiotherOncol 2006;81:88–96.

[14] Hesse BM, Spies L, Groh BA. Tomotherapeutic portal imagingfor radiation treatment verification. Phys Med Biol1998;43:3607–16.

[15] Cremers F, Frenzel T, Kausch C, Albers D, Schonborn T,Schmidt R. Performance of electronic portal imaging devices(EPIDs) used in radiotherapy: image quality and dose mea-surements. Med Phys 2004;31:985–96.

[16] Boyer AL, Antonuk L, Fenster A, et al. A review of electronicportal imaging devices (EPIDs). Med Phys 1992;19:1–16.

[17] Kirby MC, Glendinning AG. Developments in electronic portalimaging systems. Br J Radiol 2006;79 Spec No. 1:S50–65.

[18] van Herk M. Physical aspects of a liquid-filled ionizationchamber with pulsed polarizing voltage. Med Phys1991;18:692–702.

[19] Meertens H, van Herk M, Bijhold J, Bartelink H. First clinicalexperience with a newly developed electronic portal imagingdevice. Int J Rad Oncol Biol Phys 1990;18:1173–81.

[20] Meertens H, van Herk M, Weeda J. A liquid ionisation detectorfor digital radiography of therapeutic megavoltage photonbeams. Phys Med Biol 1985;30:313–21.

[21] van Herk M, Meertens H. A matrix ionisation chamber imagingdevice for on-line patient setup verification during radiother-apy. Radiother Oncol 1988;11:369–78.

[22] Essers M, Boellaard R, van Herk M, Lanson H, Mijnheer B.Transmission dosimetry with a liquid-filled electronic portalimaging device. Int J Rad Oncol Biol Phys 1996;34:931–41.

[23] Yin FF, Schell MC, Rubin P. Input/output characteristics of amatrix ion-chamber electronic portal imaging device. MedPhys 1994;21:1447–54.

[24] Essers M, Hoogervorst BR, van Herk M, Lanson H, Mijnheer BJ.Dosimetric characteristics of a liquid-filled electronic portalimaging device. Int J Rad Oncol Biol Physics1995;33:1265–72.

[25] Zhu Y, Jiang XQ, Van Dyk J. Portal dosimetry using a liquid ionchamber matrix: dose response studies. Med Phys1995;22:1101–6.

[26] Boellaard R, van Herk M, Mijnheer BJ. The dose responserelationship of a liquid-filled electronic portal imagingdevice. Med Phys 1996;23:1601–11.

[27] Keller H, Fix M, Ruegsegger P. Calibration of a portal imagingdevice for high-precision dosimetry: a Monte Carlo study. MedPhys 1998;25:1891–902.

[28] Parsaei H, el-Khatib E, Rajapakshe R. The use of an electronicportal imaging system to measure portal dose and portal doseprofiles. Med Phys 1998;25:1903–9.

[29] Chang J, Mageras GS, Ling CC, Lutz W. An iterative EPIDcalibration procedure for dosimetric verification that con-siders the EPID scattering factor. Med Phys2001;28:2247–57.

[30] Van Esch A, Vanstraelen B, Verstraete J, Kutcher G, HuyskensD. Pre-treatment dosimetric verification by means of a liquid-filled electronic portal imaging device during dynamic deliv-ery of intensity modulated treatment fields. Radiother Oncol2001;60:181–90.

[31] Chang J, Mageras GS, Ling CC. Evaluation of rapid dose mapacquisition of a scanning liquid-filled ionization chamberelectronic portal imaging device. Int J Rad Oncol Biol Phys2003;55:1432–45.

[32] Chin PW, Lewis DG, Spezi E. Correction for dose–responsevariations in a scanning liquid ion chamber EPID as a functionof linac gantry angle. Phys Med Biol 2004;49:N93–103.

[33] Louwe RJ, Tielenburg R, van Ingen KM, Mijnheer BJ, van HerkMB. The stability of liquid-filled matrix ionization chamberelectronic portal imaging devices for dosimetry purposes.Med Phys 2004;31:819–27.

[34] Reich P, Bezak E, Mohammadi M, Fog L. The prediction oftransmitted dose distributions using a 3D treatment planningsystem. Australas Phys Eng Sci Med 2006;29:18–29.

[35] Mohammadi M, Bezak E. Two-dimensional transmitted dosemeasurements using a scanning liquid ionization chamberEPID. Phys Med Biol 2006;51:2971–85.

[36] Tateoka K, Ouchi A, Waka M, et al. Dosimetric properties ofthe liquid ionization chamber electronic portal imagingdevice (EPID). Igaku Butsuri 2006;26:28–38.

[37] Morton EJ, Swindell W, Lewis DG, Evans PM. A linear array,scintillation crystal-photodiode detector for megavoltageimaging. Med Phys 1991;18:681–91.

mailto:[email protected]


[38] Symonds-Tayler JR, Partridge M, Evans PM. An electronicportal imaging device for transit dosimetry. Phys Med Biol1997;42:2273–83.

[39] Hansen VN, Evans PM, Swindell W. The application of transitdosimetry to precision radiotherapy. Med Phys1996;23:713–21.

[40] Mosleh-Shirazi MA, Evans PM, Swindell W, Symonds-Tayler JR,Webb S, Partridge M. Rapid portal imaging with a high-efficiency, large field-of-view detector. Med Phys1998;25:2333–46.

[41] Althof VG, de Boer JC, Huizenga H, Stroom JC, Visser AG,Swanenburg BN. Physical characteristics of a commercialelectronic portal imaging device. Med Phys 1996;23:1845–55.

[42] Heijmen BJ, Pasma KL, Kroonwijk M, et al. Portal dosemeasurement in radiotherapy using an electronic portalimaging device (EPID). Phys Med Biol 1995;40:1943–55.

[43] Pasma KL, Kroonwijk M, de Boer JC, Visser AG, Heijmen BJ.Accurate portal dose measurement with a fluoroscopicelectronic portal imaging device (EPID) for open and wedgedbeams and dynamic multileaf collimation. Phys Med Biol1998;43:2047–60.

[44] Nijsten SM, Minken AW, Lambin P, Bruinvis IA. Verification oftreatment parameter transfer by means of electronic portaldosimetry. Med Phys 2004;31:341–7.

[45] Glendinning AG, Bonnett DE. Dosimetric properties of theTheraview fluoroscopic electronic portal imaging device. Br JRadiol 2000;73:517–30.

[46] Glendinning AG, Hunt SG, Bonnett DE. Measurement of theresponse of Gd2O2S:Tb phosphor to 6 MV X-rays. Phys MedBiol 2001;46:517–30.

[47] Franken EM, de Boer JC, Barnhoorn JC, Heijmen BJ. Charac-teristics relevant to portal dosimetry of a cooled CCDcamera-based EPID. Med Phys 2004;31:2549–51.

[48] Franken EM, de Boer JC, Heijmen BJ. A novel approach toaccurate portal dosimetry using CCD-camera based EPIDs.Med Phys 2006;33:888–903.

[49] de Boer JC, Heijmen BJ, Pasma KL, Visser AG. Characteriza-tion of a high-elbow, fluoroscopic electronic portal imagingdevice for portal dosimetry. Phys Med Biol 2000;45:197–216.

[50] Partridge M, Evans PM, Symonds-Tayler JR. Optical scatteringin camera-based electronic portal imaging. Phys Med Biol1999;44:2381–96.

[51] Partridge M, Symonds-Tayler JR, Evans PM. IMRT verificationwith a camera-based electronic portal imaging system. PhysMed Biol 2000;45:N183–96.

[52] Dirkx ML, Heijmen BJ. Testing of the stability of intensitymodulated beams generated with dynamic multileaf collima-tion, applied to the MM50 racetrack microtron. Med Phys2000;27:2701–7.

[53] Dirkx ML, Kroonwijk M, de Boer JC, Heijmen BJ. Dailydosimetric quality control of the MM50 racetrack microtronusing an electronic portal imaging device. Radiother Oncol1995;37:55–60.

[54] Antonuk LE, El-Mohri Y, Huang W, et al. Initial performanceevaluation of an indirect-detection, active matrix flat-panelimager (AMFPI) prototype for megavoltage imaging. Int JRadiat Oncol Biol Phys 1998;42:437–54.

[55] Antonuk LE, Yorkston J, Huang W, et al. A real-time, flat-panel, amorphous silicon, digital X-ray imager. Radiographics1995;15:993–1000.

[56] McDermott LN, Louwe RJ, Sonke JJ, van Herk MB, MijnheerBJ. Dose–response and ghosting effects of an amorphoussilicon electronic portal imaging device. Med Phys2004;31:285–95.

[57] Winkler P, Georg D. An intercomparison of 11 amorphoussilicon EPIDs of the same type: implications for portaldosimetry. Phys Med Biol 2006;51:4189–200.

[58] Parent L, Seco J, Evans PM, Fielding A, Dance DR. Monte Carlomodelling of a-Si EPID response: the effect of spectralvariations with field size and position. Med Phys2006;33:4527–40.

[59] Winkler P, Hefner A, Georg D. Dose–response characteristicsof an amorphous silicon EPID. Med Phys 2005;32:3095–105.

[60] Parent L, Fielding AL, Dance DR, Seco J, Evans PM.Amorphous silicon EPID calibration for dosimetric applica-tions: comparison of a method based on Monte Carloprediction of response with existing techniques. Phys MedBiol 2007;52:3351–68.

[61] Chen J, Chuang CF, Morin O, Aubin M, Pouliot J. Calibration ofan amorphous-silicon flat panel portal imager for exit-beamdosimetry. Med Phys 2006;33:584–94.

[62] Nijsten SM, van Elmpt WJ, Jacobs M, et al. A globalcalibration model for a-Si EPIDs used for transit dosimetry.Med Phys 2007;34:3872–84.

[63] Greer PB. Correction of pixel sensitivity variation and off-axisresponse for amorphous silicon EPID dosimetry. Med Phys2005;32:3558–68.

[64] Van Esch A, Depuydt T, Huyskens DP. The use of an aSi-basedEPID for routine absolute dosimetric pre-treatment verifica-tion of dynamic IMRT fields. Radiother Oncol2004;71:223–34.

[65] Greer PB, Popescu CC. Dosimetric properties of an amorphoussilicon electronic portal imaging device for verification ofdynamic intensity modulated radiation therapy. Med Phys2003;30:1618–27.

[66] Kirkby C, Sloboda R. Consequences of the spectral responseof an a-Si EPID and implications for dosimetric calibration.Med Phys 2005;32:2649–58.

[67] McCurdy BM, Luchka K, Pistorius S. Dosimetric investigationand portal dose image prediction using an amorphous siliconelectronic portal imaging device. Med Phys 2001;28:911–24.

[68] Siebers JV, Kim JO, Ko L, Keall PJ, Mohan R. Monte Carlocomputation of dosimetric amorphous silicon electronicportal images. Med Phys 2004;31:2135–46.

[69] Grein EE, Lee R, Luchka K. An investigation of a newamorphous silicon electronic portal imaging device for transitdosimetry. Med Phys 2002;29:2262–8.

[70] Kirkby C, Sloboda R. Comprehensive Monte Carlo calculationof the point spread function for a commercial a-Si EPID. MedPhys 2005;32:1115–27.

[71] Louwe RJ, McDermott LN, Sonke JJ, et al. The long-termstability of amorphous silicon flat panel imaging devices fordosimetry purposes. Med Phys 2004;31:2989–95.

[72] El-Mohri Y, Antonuk LE, Yorkston J, et al. Relative dosimetryusing active matrix flat-panel imager (AMFPI) technology.Med Phys 1999;26:1530–41.

[73] Partridge M, Hesse B-M, Muller L. A performance comparisonof direct- and indirect-detection flat-panel imagers. NuclInstrum Methods Phys Res A 2002;484:351–63.

[74] Schach von Wittenau AE, Logan CM, Aufderheide 3rd MB,Slone DM. Blurring artifacts in megavoltage radiography witha flat-panel imaging system: comparison of Monte Carlosimulations with measurements. Med Phys 2002;29:2559–70.

[75] Greer PB. Off-axis dose response characteristics of anamorphous silicon electronic portal imaging device. Med Phys2007;34:3815–24.

[76] Wendling M, Louwe RJ, McDermott LN, Sonke JJ, van Herk M,Mijnheer BJ. Accurate two-dimensional IMRT verificationusing a back-projection EPID dosimetry method. Med Phys2006;33:259–73.

[77] Talamonti C, Casati M, Bucciolini M. Pretreatment verifica-tion of IMRT absolute dose distributions using a commercial a-Si EPID. Med Phys 2006;33:4367–78.


[78] McDermott LN, Nijsten SM, Sonke JJ, Partridge M, van HerkM, Mijnheer BJ. Comparison of ghosting effects for threecommercial a-Si EPIDs. Med Phys 2006;33:2448–51.

[79] Siewerdsen JH, Jaffray DA. A ghost story: spatio-temporalresponse characteristics of an indirect-detection flat-panelimager. Med Phys 1999;26:1624–41.

[80] Budgell GJ, Zhang Q, Trouncer RJ, Mackay RI. Improving IMRTquality control efficiency using an amorphous silicon elec-tronic portal imager. Med Phys 2005;32:3267–78.

[81] Boellaard R, van Herk M, Mijnheer BJ. A convolution model toconvert transmission dose images to exit dose distributions.Med Phys 1997;24:189–99.

[82] Chang J, Mageras GS, Chui CS, Ling CC, Lutz W. Relativeprofile and dose verification of intensity-modulated radiationtherapy. Int J Rad Oncol Biol Phys 2000;47:231–40.

[83] Chin PW, Spezi E, Lewis DG. Monte Carlo simulation of portaldosimetry on a rectilinear voxel geometry: a variable gantryangle solution. Phys Med Biol 2003;48:N231–238.

[84] Baker SJ, Budgell GJ, MacKay RI. Use of an amorphous siliconelectronic portal imaging device for multileaf collimatorquality control and calibration. Phys Med Biol2005;50:1377–92.

[85] Curtin-Savard A, Podgorsak EB. An electronic portal imagingdevice as a physics tool. Med Dosim 1997;22:101–5.

[86] Vieira SC, Bolt RA, Dirkx ML, Visser AG, Heijmen BJ. Fast,daily linac verification for segmented IMRT using electronicportal imaging. Radiother Oncol 2006;80:86–92.

[87] Budgell GJ, Zhang R, Mackay RI. Daily monitoring of linearaccelerator beam parameters using an amorphous siliconEPID. Phys Med Biol 2007;52:1721–33.

[88] Chang J, Obcemea CH, Sillanpaa J, Mechalakos J, Burman C.Use of EPID for leaf position accuracy QA of dynamic multi-leaf collimator (DMLC) treatment. Med Phys 2004;31:2091–6.

[89] Liu G, van Doorn T, Bezak E. Assessment of flatness andsymmetry of megavoltage X-ray beam with an electronicportal imaging device (EPID). Australas Phys Eng Sci Med2002;25:58–66.

[90] Metcalfe P, Tangboonduangjit P, White P. Intensity-modu-lated radiation therapy: overlapping co-axial modulatedfields. Phys Med Biol 2004;49:3629–37.

[91] Williams MJ, Metcalfe P. Verification of a rounded leaf-endMLC model used in a radiotherapy treatment planning system.Phys Med Biol 2006;51:N65–78.

[92] Ploeger LS, Smitsmans MH, Gilhuijs KG, van Herk M. Accuratemeasurement of the dynamic response of a scanning elec-tronic portal imaging device. Med Phys 2001;28:310–6.

[93] Prisciandaro JI, Herman MG, Kruse JJ. Utilizing an electronicportal imaging device to monitor light and radiation fieldcongruence. J Appl Clin Med Phys 2003;4:315–20.

[94] Xing L, Li JG. Computer verification of fluence map forintensity modulated radiation therapy. Med Phys2000;27:2084–92.

[95] Yang Y, Xing L. Quantitative measurement of MLC leafdisplacements using an electronic portal image device. PhysMed Biol 2004;49:1521–33.

[96] Samant SS, Zheng W, Parra NA, et al. Verification ofmultileaf collimator leaf positions using an electronic portalimaging device. Med Phys 2002;29:2900–12.

[97] Parent L, Seco J, Evans PM, Dance DR, Fielding A. Evaluationof two methods of predicting MLC leaf positions using EPIDmeasurements. Med Phys 2006;33:3174–82.

[98] Vieira SC, Dirkx ML, Pasma KL, Heijmen BJ. Fast and accurateleaf verification for dynamic multileaf collimation using anelectronic portal imaging device. Med Phys 2002;29:2034–40.

[99] Partridge M, Evans PM, van Herk M, Ploeger LS, Budgell GJ,James HV. Leaf position verification during dynamic beam

delivery: a comparison of three applications using electronicportal imaging. Med Phys 2000;27:1601–9.

[100] Ploeger LS, Smitsmans MH, Gilhuijs KG, van Herk M. A methodfor geometrical verification of dynamic intensity modulatedradiotherapy using a scanning electronic portal imagingdevice. Med Phys 2002;29:1071–9.

[101] Sonke JJ, Ploeger LS, Brand B, Smitsmans MH, van Herk M.Leaf trajectory verification during dynamic intensity modu-lated radiotherapy using an amorphous silicon flat panelimager. Med Phys 2004;31:389–95.

[102] James HV, Atherton S, Budgell GJ, Kirby MC, Williams PC.Verification of dynamic multileaf collimation using an elec-tronic portal imaging device. Phys Med Biol 2000;45:495–509.

[103] Li W, Siebers JV, Moore JA. Using fluence separation toaccount for energy spectra dependence in computing dosi-metric a-Si EPID images for IMRT fields. Med Phys2006;33:4468–80.

[104] Chytyk K, McCurdy BM. Investigation of tilted dose kernels forportal dose prediction in a-Si electronic portal imagers. MedPhys 2006;33:3333–9.

[105] Yeboah C, Pistorius S. Monte Carlo studies of the exit photonspectra and dose to a metal/phosphor portal imaging screen.Med Phys 2000;27:330–9.

[106] Chang J, Ling CC. Using the frame averaging of aS500 EPID forIMRT verification. J Appl Clin Med Phys 2003;4:287–99.

[107] Kapatoes JM, Olivera GH, Ruchala KJ, Mackie TR. On theverification of the incident energy fluence in tomotherapyIMRT. Phys Med Biol 2001;46:2953–65.

[108] Zeidan OA, Li JG, Ranade M, Stell AM, Dempsey JF. Verifi-cation of step-and-shoot IMRT delivery using a fast video-based electronic portal imaging device. Med Phys2004;31:463–76.

[109] Warkentin B, Steciw S, Rathee S, Fallone BG. Dosimetric IMRTverification with a flat-panel EPID. Med Phys2003;30:3143–55.

[110] Cozzi L, Fogliata A, Nicolini G. Pre-treatment verification ofintensity modulated photon beams with films and electronicportal imaging–two years of clinical experience. Z Med Phys2004;14:239–50.

[111] Moran JM, Roberts DA, Nurushev TS, Antonuk LE, El-Mohri Y,Fraass BA. An active matrix flat panel dosimeter (AMFPD) forin-phantom dosimetric measurements. Med Phys2005;32:466–72.

[112] Nicolini G, Fogliata A, Vanetti E, Clivio A, Cozzi L. GLAaS: anabsolute dose calibration algorithm for an amorphous siliconportal imager. Applications to IMRT verifications. Med Phys2006;33:2839–51.

[113] Hesse BM, Nill S, Tucking T, Oelfke U. 3D in vivo dosimetry bymeans of entrance dosimetry for dose guided radiotherapy.Radiother Oncol 2006;81:S46.

[114] Poppe B, Thieke C, Beyer D, et al. DAVID–a translucentmulti-wire transmission ionization chamber for in vivo veri-fication of IMRT and conformal irradiation techniques. PhysMed Biol 2006;51:1237–48.

[115] Ansbacher W. Three-dimensional portal image-based dosereconstruction in a virtual phantom for rapid evaluation ofIMRT plans. Med Phys 2006;33:3369–82.

[116] van Elmpt WJ, Nijsten SM, Schiffeleers RF, et al. A MonteCarlo based three-dimensional dose reconstruction methodderived from portal dose images. Med Phys 2006;33:2426–34.

[117] van Elmpt WJ, Nijsten SM, Dekker AL, Mijnheer BJ, Lambin P.Treatment verification in the presence of inhomogeneitiesusing EPID-based three-dimensional dose reconstruction. MedPhys 2007;34:2816–26.

[118] van Elmpt W, Nijsten S, Mijnheer B, Dekker A, Lambin P. Thenext step in patient-specific QA: 3D dose verification ofconformal and intensity-modulated RT based on EPID dosimetry


and Monte Carlo dose calculations. Radiother Oncol2008;86:86–92.

[119] Renner WD, Norton K, Holmes T. A method for deconvolutionof integrated electronic portal images to obtain incidentfluence for dose reconstruction. J Appl Clin Med Phys2005;6:22–39.

[120] Steciw S, Warkentin B, Rathee S, Fallone BG. Three-dimen-sional IMRT verification with a flat-panel EPID. Med Phys2005;32:600–12.

[121] van Zijtveld M, Dirkx ML, de Boer HC, Heijmen BJ. 3D dosereconstruction for clinical evaluation of IMRT pretreatmentverification with an EPID. Radiother Oncol 2007;82:201–7.

[122] van Zijtveld M, Dirkx M, Heijmen B. Correction of conebeamCT values using a planning CT for derivation of the ‘‘dose ofthe day’’. Radiother Oncol 2007;85:195–200.

[123] McNutt TR, Mackie TR, Reckwerdt P, Paliwal BR. Modelingdose distributions from portal dose images using the convo-lution/superposition method. Med Phys 1996;23:1381–92.

[124] McNutt TR, Mackie TR, Paliwal BR. Analysis and convergenceof the iterative convolution/superposition dose reconstruc-tion technique for multiple treatment beams and tomother-apy. Med Phys 1997;24:1465–76.

[125] Leunens G, Van Dam J, Dutreix A, van der Schueren E. Qualityassurance in radiotherapy by in vivo dosimetry. 1. Entrancedose measurements, a reliable procedure. Radiother Oncol1990;17:141–51.

[126] Essers M, Lanson JH, Leunens G, Schnabel T, Mijnheer BJ. Theaccuracy of CT-based inhomogeneity corrections and in vivodosimetry for the treatment of lung cancer. Radiother Oncol1995;37:199–208.

[127] Pasma KL, Kroonwijk M, Quint S, Visser AG, Heijmen BJ.Transit dosimetry with an electronic portal imaging device(EPID) for 115 prostate cancer patients. Int J Rad Oncol BiolPhys 1999;45:1297–303.

[128] Nijsten SM, Mijnheer BJ, Dekker AL, Lambin P, Minken AW.Routine individualised patient dosimetry using electronicportal imaging devices. Radiother Oncol 2007;83:65–75.

[129] D’Andrea M, Laccarino G, Carpino S, Strigari L, Benassi M.Primary photon fluence extraction from portal imagesacquired with an amorphous silicon flat panel detector:experimental determination of a scatter filter. J Exp ClinCancer Res 2007;26:125–32.

[130] Piermattei A, Fidanzio A, Stimato G, et al. In vivo dosimetryby an aSi-based EPID. Med Phys 2006;33:4414–22.

[131] Piermattei A, Fidanzio A, Azario L, et al. Application of apractical method for the isocenter point in vivo dosimetry bya transit signal. Phys Med Biol 2007;52:5101–17.

[132] Pasma KL, Heijmen BJ, Kroonwijk M, Visser AG. Portal doseimage (PDI) prediction for dosimetric treatment verificationin radiotherapy. I. An algorithm for open beams. Med Phys1998;25:830–40.

[133] He X, Van Esch A, Reymen R, Huyskens D. Evaluation of anelectronic portal imaging device for transit dosimetry. ActaOncol 1999;38:591–6.

[134] Dirkx ML, de Boer JC, Heijmen BJ. Improvement of radio-therapy treatment delivery accuracy using an electronicportal imaging device. Radiat Prot Dosimetry 2006;121:70–9.

[135] van Elmpt WJ, Nijsten SM, Mijnheer BJ, Minken AW. Exper-imental verification of a portal dose prediction model. MedPhys 2005;32:2805–18.

[136] McCurdy BM, Pistorius S. Photon scatter in portal images:accuracy of a fluence based pencil beam superpositionalgorithm. Med Phys 2000;27:913–22.

[137] McCurdy BM, Pistorius S. Photon scatter in portal images:physical characteristics of pencil beam kernels generatedusing the EGS Monte Carlo code. Med Phys 2000;27:312–20.

[138] McCurdy BM, Pistorius S. A two-step algorithm for predictingportal dose images in arbitrary detectors. Med Phys2000;27:2109–16.

[139] Dahlgren CV, Ahnesjo A, Montelius A, Rikner G. Portal doseimage verification: formalism and application of the col-lapsed cone superposition method. Phys Med Biol2002;47:4371–87.

[140] Dahlgren CV, Eilertsen K, Jorgensen TD, Ahnesjo A. Portaldose image verification: the collapsed cone superpositionmethod applied with different electronic portal imagingdevices. Phys Med Biol 2006;51:335–49.

[141] Pasma KL, Vieira SC, Heijmen BJ. Portal dose image predic-tion for dosimetric treatment verification in radiotherapy. II.An algorithm for wedged beams. Med Phys 2002;29:925–31.

[142] Kroonwijk M, Pasma KL, Quint S, Koper PC, Visser AG,Heijmen BJ. In vivo dosimetry for prostate cancer patientsusing an electronic portal imaging device (EPID); demonstra-tion of internal organ motion. Radiother Oncol1998;49:125–32.

[143] Vieira SC, Dirkx ML, Heijmen BJ, de Boer HC. SIFT: a methodto verify the IMRT fluence delivered during patient treatmentusing an electronic portal imaging device. Int J Rad OncolBiol Phys 2004;60:981–93.

[144] Fielding AL, Evans PM, Clark CH. Verification of patientposition and delivery of IMRT by electronic portal imaging.Radiother Oncol 2004;73:339–47.

[145] Spezi E, Lewis DG. Full forward Monte Carlo calculation ofportal dose from MLC collimated treatment beams. Phys MedBiol 2002;47:377–90.

[146] Spies L, Evans PM, Partridge M, Hansen VN, Bortfeld T. Directmeasurement and analytical modeling of scatter in portalimaging. Med Phys 2000;27:462–71.

[147] Spies L, Bortfeld T. Analytical scatter kernels for portalimaging at 6 MV. Med Phys 2001;28:553–9.

[148] Spies L, Partridge M, Groh BA, Bortfeld T. An iterativealgorithm for reconstructing incident beam distributions fromtransmission measurements using electronic portal imaging.Phys Med Biol 2001;46:N203–211.

[149] Swindell W, Evans PM. Scattered radiation in portal images: aMonte Carlo simulation and a simple physical model. MedPhys 1996;23:63–73.

[150] Hansen VN, Swindell W, Evans PM. Extraction of primarysignal from EPIDs using only forward convolution. Med Phys1997;24:1477–84.

[151] Partridge M, Evans PM. The practical implementation of ascatter model for portal imaging at 10 MV. Phys Med Biol1998;43:2685–93.

[152] McNutt TR, Mackie TR, Reckwerdt P, Papanikolaou N, PaliwalBR. Calculation of portal dose using the convolution/super-position method. Med Phys 1996;23:527–35.

[153] Mohammadi M, Bezak E, Reich P. The use of extended doserange film for dosimetric calibration of a scanning liquid-filled ionization chamber electronic portal imaging device. JAppl Clin Med Phys 2007;8:69–84.

[154] Mohammadi M, Bezak E, Reich P. Comparison of two-dimensional transmitted dose maps: evaluation of existingalgorithms. Australas Phys Eng Sci Med 2006;29:179–87.

[155] Kapatoes JM, Olivera GH, Reckwerdt PJ, Fitchard EE,Schloesser EA, Mackie TR. Delivery verification in sequentialand helical tomotherapy. Phys Med Biol 1999;44:1815–41.

[156] Kirby MC, Williams PC. The use of an electronic portalimaging device for exit dosimetry and quality control mea-surements. Int J Radiat Oncol Biol Phys 1995;31:593–603.

[157] Kirby MC, Williams PC. Measurement possibilities using anelectronic portal imaging device. Radiother Oncol1993;29:237–43.


[158] Boellaard R, van Herk M, Uiterwaal H, Mijnheer B. Two-dimensional exit dosimetry using a liquid-filled electronicportal imaging device and a convolution model. RadiotherOncol 1997;44:149–57.

[159] Boellaard R, Essers M, van Herk M, Mijnheer BJ. New methodto obtain the midplane dose using portal in vivo dosimetry.Int J Radiat Oncol Biol Phys 1998;41:465–74.

[160] Boellaard R, van Herk M, Uiterwaal H, Mijnheer B. Firstclinical tests using a liquid-filled electronic portal imagingdevice and a convolution model for the verification of themidplane dose. Radiother Oncol 1998;47:303–12.

[161] McDermott LN, Wendling M, van Asselen B, et al. Clinicalexperience with EPID dosimetry for prostate IMRT pre-treatment dose verification. Med Phys 2006;33:3921–30.

[162] McDermott LN, Wendling M, Sonke JJ, van Herk M, MijnheerBJ. Replacing pretreatment verification with in vivo EPIDdosimetry for prostate IMRT. Int J Radiat Oncol Biol Phys2007;67:1568–77.

[163] Jarry G, Verhaegen F. Patient-specific dosimetry of conven-tional and intensity modulated radiation therapy using anovel full Monte Carlo phase space reconstruction methodfrom electronic portal images. Phys Med Biol2007;52:2277–99.

[164] Partridge M, Ebert M, Hesse BM. IMRT verification by three-dimensional dose reconstruction from portal beam measure-ments. Med Phys 2002;29:1847–58.

[165] Chen J, Morin O, Aubin M, Bucci MK, Chuang CF, Pouliot J.Dose-guided radiation therapy with megavoltage cone-beamCT. Br J Radiol 2006;79 Spec No. 1:S87–98.

[166] Louwe RJ, Damen EM, van Herk M, Minken AW, Torzsok O,Mijnheer BJ. Three-dimensional dose reconstruction of breastcancer treatment using portal imaging. Med Phys2003;30:2376–89.

[167] McDermott LN, Wendling M, Nijkamp J, et al. 3D in vivo doseverification of entire hypo-fractionated IMRT treatments usingan EPID and cone-beam CT. Radiother Oncol 2008;86:35–42.

[168] Kapatoes JM, Olivera GH, Ruchala KJ, Smilowitz JB, Reck-werdt PJ, Mackie TR. A feasible method for clinical deliveryverification and dose reconstruction in tomotherapy. MedPhys 2001;28:528–42.

[169] Kapatoes JM, Olivera GH, Balog JP, Keller H, Reckwerdt PJ,Mackie TR. On the accuracy and effectiveness of dose recon-struction for tomotherapy. Phys Med Biol 2001;46:943–66.

[170] Low DA, Harms WB, Mutic S, Purdy JA. A technique for thequantitative evaluation of dose distributions. Med Phys1998;25:656–61.

[171] Low DA, Dempsey JF. Evaluation of the gamma dose distri-bution comparison method. Med Phys 2003;30:2455–64.

[172] Bakai A, Alber M, Nusslin F. A revision of the gamma-evaluation concept for the comparison of dose distributions.Phys Med Biol 2003;48:3543–53.

[173] Jiang SB, Sharp GC, Neicu T, Berbeco RI, Flampouri S,Bortfeld T. On dose distribution comparison. Phys Med Biol2006;51:759–76.

[174] Childress NL, Bloch C, White RA, Salehpour M, Rosen II.Detection of IMRT delivery errors using a quantitative 2Ddosimetric verification system. Med Phys 2005;32:153–62.

[175] Budgell GJ, Perrin BA, Mott JH, Fairfoul J, Mackay RI.Quantitative analysis of patient-specific dosimetric IMRTverification. Phys Med Biol 2005;50:103–19.

[176] Gillis S, De Wagter C, Bohsung J, Perrin B, Williams P,Mijnheer BJ. An inter-centre quality assurance network forIMRT verification: results of the ESTRO QUASIMODO project.Radiother Oncol 2005;76:340–53.

[177] Stock M, Kroupa B, Georg D. Interpretation and evaluation ofthe gamma index and the gamma index angle for theverification of IMRT hybrid plans. Phys Med Biol2005;50:399–411.

[178] Spezi E, Lewis DG. Gamma histograms for radiotherapy planevaluation. Radiother Oncol 2006;79:224–30.

[179] van Zijtveld M, Dirkx ML, de Boer HC, Heijmen BJ. Dosimetricpre-treatment verification of IMRT using an EPID; clinicalexperience. Radiother Oncol 2006;81:168–75.

[180] Wendling M, Zijp LJ, McDermott LN, et al. A fast algorithmfor gamma evaluation in 3D. Med Phys 2007;34:1647–54.

[181] Pasma KL, Dirkx ML, Kroonwijk M, Visser AG, Heijmen BJ.Dosimetric verification of intensity modulated beams pro-duced with dynamic multileaf collimation using an electronicportal imaging device. Med Phys 1999;26:2373–8.

[182] Evans PM, Donovan EM, Partridge M, et al. The delivery ofintensity modulated radiotherapy to the breast using multiplestatic fields. Radiother Oncol 2000;57:79–89.

[183] Donovan E, Bleakley N, Denholm E, et al. Randomised trial ofstandard 2D radiotherapy (RT) versus intensity modulatedradiotherapy (IMRT) in patients prescribed breast radiother-apy. Radiother Oncol 2007;82:254–64.

a literature review of electronic portal imaging for radiotherapy

Documents