BJR
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF)
Services
Right arrow Similar articles in this journal
Right arrow Similar articles in PubMed
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Langmack, K A
Right arrow Search for Related Content
PubMed
Right arrow PubMed Citation
Right arrow Articles by Langmack, K A
British Journal of Radiology 74 (2001),789-804 © 2001 The British Institute of Radiology

Review article

Portal imaging

K A Langmack, DPhil

Medical Physics Department, Lincoln County Hospital, Greetwell Road, Lincoln, UK


   Abstract
Top
 Abstract
Introduction
 Theoretical aspects of portal...
 Imaging media
 Image registration
 Portal imaging dosimetry
 Future requirements
 References
 
Portal imaging is the acquisition of images with a radiotherapy beam. Imaging theory suggests that the quality of portal images could be much higher if the efficiency of the imaging media in detecting radiation could be improved. Introduction of new media (films and electronic portal imaging devices) has confirmed this by markedly increasing the quality of portal images. Images from these devices can then be used to verify a patient's treatment. Geometric verification requires the portal image to be registered with a reference image. Dosimetric verification requires the portal imager to be calibrated for dose. This review gives a brief overview of the current areas of interest in portal imaging: imaging theory; imaging media, film and electronic portal imaging devices; image registration; and dosimetry using these devices.


   Introduction
Top
 Abstract
 Introduction
Theoretical aspects of portal...
 Imaging media
 Image registration
 Portal imaging dosimetry
 Future requirements
 References
 
Portal imaging is the use of a therapeutic X-ray beam to form an image of the area being irradiated. As these beams are of megavoltage energies, photon interactions with the patient occur mainly by Compton scattering. Compton scattering is almost independent of atomic number, so there is inherently less subject contrast in portal images compared with diagnostic X-rays [1]. Some efforts have been made to use diagnostic X-rays to image the same area as the therapy beam to provide high quality images. This has been done either by adding an X-ray tube onto the linear accelerator (linac) gantry [2–4] or by modifying the linac to produce diagnostic quality X-rays [5–8]. These methods have not been widely adopted for portal imaging. Instead, the major thrust of portal imaging research and development has been to improve the technology employed to take portal images with megavoltage X-rays. This effort has paid off, resulting both in enhanced image quality and in increased ease of image acquisition [1, 9, 10].

The historical and current main use of portal images has been the study of set-up errors in patient treatment [11]. This has resulted in improved treatment accuracy [12] and in quantification of the margins required to account for the uncertainties in treatment delivery [13]. Margin quantification and reduction is an increasingly important issue for all radiotherapy departments with the increasing acceptance that conformal therapy improves patient treatment [14]. A current and developing area of research with portal imaging devices is their use for determining patient dose information. This means that portal imaging is becoming able to give both geometrical and dosimetric information and thus to provide powerful verification tools for advanced techniques such as conformal therapy and intensity modulated radiotherapy.

So, portal imaging has evolved from interpreting indistinct images on films to a major area of radiotherapy physics. It covers several distinct topics: imaging theory; imaging media; film and electronic portal imaging devices; image processing and registration; and dosimetry using these devices. This review is a brief overview of these areas that currently make up the subject of portal imaging.


   Theoretical aspects of portal imaging performance
Top
 Abstract
 Introduction
 Theoretical aspects of portal...
Imaging media
 Image registration
 Portal imaging dosimetry
 Future requirements
 References
 
Boyer et al [1], extending the work of Motz and Danos [15], derived the theoretical characteristics of portal imaging performance. Briefly, Boyer et al assert that the exit radiation beam contains much more information than is extracted by conventional portal imaging systems. The detector quantum efficiency (DQE) is a measure of the efficiency of an imaging system in transferring the input information to the output image. The DQE is reduced by inefficient detection of incident radiation, low spatial resolution of the system and system noise. It has been shown that system noise due to inefficiencies in the various detectors dominates many portal imaging systems [16–18]. Image quality can be improved by increasing the efficiency of the detector system [19, 20]. Once system noise has been reduced, the information content of the image is determined by the signal-to-noise ratio (SNR) in the image, which is controlled by the size of the structure of interest, the subject contrast and the number of X-ray quanta used to form the image [1]. The subject contrast is 10–20 times lower in megavoltage portal imaging compared with diagnostic X-ray imaging. This is offset by the fact that many more photons are used to form the megavoltage image, so the SNR does not decrease much between diagnostic and therapy energy beams [1, 15]. Hence, a portal imaging system with a high DQE should be able to give near diagnostic quality images. Other factors affecting image quality include magnification and scattered radiation. It has been shown that the optimal image magnification for portal images is between about 1 and 2,depending on the imaging medium [21, 22]. Scattered radiation decreases image quality, but anti-scatter grids as employed in diagnostic radiology are ineffective, as the energy of the scattered radiation is too high [7, 22, 23].


   Imaging media
Top
 Abstract
 Introduction
 Theoretical aspects of portal...
 Imaging media
Image registration
 Portal imaging dosimetry
 Future requirements
 References
 
Film
X-ray film is the traditional medium for portal imaging. A way of characterizing such film is to plot its characteristic curve [24]. This is a graph of optical density against log10(exposure), which is sigmoidal in shape and is used to determine three characteristics of the film (gamma, latitude and speed). Strictly, gamma ({gamma}) is the maximum slope of the characteristic curve, but practically it is determined by measuring the slope of the linear part of the curve. It is related to image contrast; the larger the {gamma} the greater the contrast. A film system with a high {gamma} is required for portal imaging. Latitude defines the useful dynamic range of the film and is the extent of the straight portion of the characteristic curve on the log10(exposure) axis. The wider the latitude, the less critical it is to achieve optimal exposure conditions. These exposure conditions depend on the distance between the radiation source and the film, the thickness of the patient and the field size. This necessitates the production of technique charts that give the required exposure as a function of these parameters [25–27]. Film speed for portal films is defined as the dose required to achieve a reasonable optical density. For portal films this is often taken to be around 1.6 [25, 26, 28, 29]. Speed is important as it determines the use of a particular imaging system. If a film is fast (i.e. requires only a few monitor units to produce) then it can be used as a localization film and is only exposed for a small part of the treatment. This also allows a double exposure technique, where a second exposure is given to the same film with the field size increased to allow better anatomical localization of small fields. Slow films can be used for portal verification, where the film is irradiated for the entire treatment [30].

Historically, portal films were industrial direct exposure films [30]. Introduction of metal screens into the film cassette [31–33] meant that some medical X-ray films could be used. The composition of the metal screens has been shown to affect the image quality [33–35], as has the film used inthese cassettes [28, 36–38]. The CEA systems (CEA Medical Imaging Products, UK) determined by Roberts [28] to have the highest {gamma} (4.4) had stainless-steel screens. Another way to improve the image quality is to increase the film {gamma}. The {gamma} increases with the number of hits a film's grain requires to become developable [39]. Light photons carry less energy than the electrons produced by irradiating a metal screen with megavoltage photons. So a system where light photons, for example from a fluorescent screen, are used to expose the film, rather than the electrons generated in a metal screen, should have more contrast. This is indeed the case [10, 26, 30, 40–43]. The Kodak EC system [10] (Figure 1Go) uses a 1 mm thick copper front screen to produce electrons that then interact with a gadolinium oxysulphide intensifying screen to produce light, which exposes the film. For localization port films there is the EC-L system (film speed 1.5 cGy), and for verification films there is the EC-V system (film speed 35 cGy). The {gamma} for the EC-L system is around 6 [10, 26], compared with more traditional systems with {gamma} values ranging from 1.8 to 4.4 [28], giving portal films of much higher quality [26, 41–43]. The reduced granularity of the Kodak EC film [10] is also likely to increase image quality [16].



View larger version (41K):
[in this window]
[in a new window]
  		Figure 1. Cross-sectional representation of the Kodak portal imaging device.

 
Although these new portal film systems produce high quality images, it has been shown that simple visual inspection of images is unable to identify placement errors of up to 5 mm reliably [44]. Hence, for high precision radiotherapy, computerized tools for measuring set-up errors are required. A digital image is required to use these tools. Digitizing high quality portal images is straightforward, but it would be more convenient to start with a digital image. Electronic portal images are digital and have been shown to be a good substitute for portal films [45]. The rest of this section will briefly review some electronic portal imaging devices. More in-depth reviews of these devices are provided in the literature [1, 9].

Electronic portal imaging devices
Electronic portal imaging devices (EPIDs; Figure 2Go) have many potential advantages over traditional X-ray film for portal imaging. The images obtained are immediately available and so can be used interactively to adjust patient or field position during radiotherapy. The images are digital, which aids image processing, contrast enhancement and image matching. Moreover, digital archiving saves space and allows for rapid recall of images over a network. The disadvantages are that they have been bulky or unwieldy devices that are of limited practicality with disappointing image quality. This is changing with the introduction of more modern technology, such as amorphous silicon-based devices, but image quality still remains an issue.



View larger version (103K):
[in this window]
[in a new window]
  		Figure 2. Examples of commercially available electronic portal imaging devices: (a) Electa iView system; (b) Varian PortalVision system; (c)Siemens BeamView system; and (d) Eliav PortPro, a portable system.

 
Commissioning and quality assurance of EPIDs
There is limited literature on the commissioning and quality assurance (QA) of EPIDs compared with their design characteristics [46–50]. The latest UK advice [46] states that EPID quality control isstill developing so it is difficult to give a comprehensive set of tests. The report gives the minimal performance of an EPID on a contrast detail phantom as "a 1% contrast object of 5 mm diameter should be detectable with a dose of 10 cGy at 6 MV". It also gives monthly and annual QA tests. The monthly tests include: contrast and spatial resolution and sensitivity; magnification; position of the central axis of the machine; and field uniformity. Annual tests are: image distortion; reproducibility of position; and contrast sensitivity. Other authors [47, 48] have been somewhat more expansive. These papers split commissioning and quality control into five areas: physical operation and safety; image acquisition, resolution and sensitivity calibration; image storage, analysis and handling; reference image acquisition; and clinical operations. The details and frequencies of the tests required are device-specific, but this approach gives a good guide to areas to be addressed in any commissioning and QA programme.

When designing a QA programme for EPIDs the consequence of mechanical movement of the imager with gantry angle on image quality should be borne in mind [51]. Most EPID systems require a background and a flood-field image during calibration to correct for any variations in detector efficiency, beam flatness, and generally to remove fixed pattern noise [1]. For this to work there must be exact registration between the calibration data set and the image data. Movement of the imaging system with gantry angle can result in misregistration of the two data sets, degrading imaging performance at angles other than the calibration gantry angle [51].

Quantitation of image quality can be difficult owing to its multifaceted nature. It is determined practically using two phantom types: contrast detail phantoms that examine the visibility of an object as a function of its contrast and size [47, 52–54]; and spatial resolution phantoms [55, 56]. Both types of phantom have been shown to be useful in showing changes in image quality with time and in determining the required calibration frequency to keep EPID performance optimal [47, 55].

Camera-based systems
Camera-based EPIDs have formed the basis of several commercially available systems and have been well reviewed (Table 1Go) [1, 9]. They consist of an X-ray-to-light converter, a mirror and a TV camera (Figure 3Go). Variations in linac output during irradiations have been shown to decrease the image quality with this type of imager [57]. The X-ray-to-light converter is usually a metal plate with an attached phosphor [1, 58]. Various attempts have been made to increase the efficiency of the light output of these screens. Increasing phosphor thickness up to 350–400 mg cm-2 has been shown to be effective [58, 59], but only 10–20% of the light generated in the phosphor escapes [60]. Replacement of this plate with a glass scintillator does not increase the efficiency of radiation-to-light conversion [61], but using a customized caesium iodide array may do so [62]. The X-ray-to-light converter is not the weak link in this type of EPID. In fact, recent Monte Carlo simulations indicate that the DQE(f) is nearly X-ray quantum absorption limited for metal plate/phosphor systems up to spatial frequencies of 0.4 cycles mm-1 [18].


View this table:
[in this window]
[in a new window]
  		Table 1. Commercially available electronic portal imaging devices

 


View larger version (8K):
[in this window]
[in a new window]
  		Figure 3. Cross-sectional representation of a camera- based electronic portal imaging device. The X-ray-to-light converter is usually a metal plate with a phosphor screen, although scintillators have been used.

 
The DQE of these systems is limited by electronic noise in the camera system [59, 63–65] and poor optical coupling between the light emitter and the camera system (only 0.01% of the emitted photons reach the camera) [9, 58, 59]. Various attempts have been made to ameliorate these problems by improving the optical systems of these devices. Lower f-number lenses, larger sensor arrays and hence smaller magnifications between the input image and the camera detector have all been shown to increase the optical coupling between the converter and the camera [62, 63, 66]. In addition, increasing the gain of thecamera system improves the DQE of these systems, with DQE(0) reaching about 1% [20, 64, 65]. Another problem with these systems is light scatter in the optical path producing glare. This may be reduced by incorporation of a louvre grid [67]. Versions of these devices have also been designed with optical fibres replacing the mirror. This may allow more flexible systems to be constructed, but does not increase the optical coupling above an efficient lens system [68].

Liquid ion chamber arrays
This type of EPID also forms the basis of a commercial system and has been reviewed elsewhere (Table 1Go) [1, 9]. The principles of construction of the system are quite straightforward [69, 70]. It consists of two boards, each with 256 lines of electrodes, orientated perpendicularly to each other forming a 256 x 256 element matrix ionization chamber. The distance between electrodes is 1.27 mm, giving a detector size of 32.5 cm x 32.5 cm (Figure 4Go). One set of electrodes can be connected to a 300 V supply, the other set is attached to a bank of 256 electrometers. A 1 mm thick steel plate on the front surface of the chamber provides build-up. The gap between the electrodes is filled with iso-octane or trimethylpentane, liquids that integrate the ion current over about 0.5 s owing to the inherent low mobility ion transport in such liquids [1]. An image takes around 5 s to acquire, so only 10% of available photons are used to form an image [1]. Image acquisition requires each of the 65 536 ion chambers to be interrogated. This is done 256 at a time by connecting one of the high voltage electrodes to the 300 V supply and reading each of the 256 electrometers for about 20 ms. The high voltage is then switched to the next line and the process repeated until all 256 lines have been used. A variety of acquisition modes are available that alter the way switching is performed [1, 71]. This requires the EPID to be synchronized with the pulse repetition frequency of the linac, and that this be stable otherwise image artefacts (lines or bands) will be observed [57]. This can limit its use in verification of dynamic techniques, such as intensity modulated radiotherapy where the dose rate can be varied during the treatment.



View larger version (22K):
[in this window]
[in a new window]
  		Figure 4. Cross-sectional representation of a liquid ion chamber array.

 
Unlike camera- and amorphous silicon-based systems, there has been no modelling of the DQE of this system. The SNR of this system has been modelled and determined experimentally, with the quantum efficiency found to be about 1%[72]. The maximum spatial resolution is 2.3 mm x 2.9 mm, increasing to 2.3 mm x 4.5 mm depending on acquisition mode, and the noise levels vary from 0.13% to 0.28 % [1]. The characteristic curve of the system has also been measured in a variety of situations. These results again show that the response varies substantially with acquisition mode, and that detector contrast increases at low dose rates [73].

Amorphous silicon
Active matrix flat panel imagers (AMFPIs) based on hydrogenated amorphous silicon (a-Si:H) photodiodes and thin film transistors (TFTs) are a major area of current research [74–76]. Portal imaging devices based on this technology are starting to become commercially available (Table 1Go). There are several possible designs [75], but the one with the most development uses a front metal sheet, usually 1 mm copper, with a gadolinium oxysulphide phosphor to convert X-rays to light. Other studies of metal plate/phosphor thickness indicate that 1 mm tungsten or 1.5 mm steel bonded to 1 mm thick phosphor seems to produce a good compromise in imaging efficiency [18]. This means that the initial X-ray detection step is the same as camera-based devices [18, 59]. The light is detected using an array of a-Si:H photodiodes controlled by a-Si:H TFT (Figure 5Go). The photodiodes are electronically read and form the pixels of the image. The advantage of this device over other EPIDs is that it has the potential to form quantum noise-limited images [76, 77]. In a contrast detail study, a prototype AMFPI performed as well as Kodak EC-L film [76].



View larger version (41K):
[in this window]
[in a new window]
  		Figure 5. Cross-sectional representation of an amorphous silicon electronic portal imaging device. The illustration shows only one pixel.

 
There have been several studies examining the DQE of these devices [18, 59, 76, 77]. Comparing the quantum accounting diagrams of AMFPIs and camera-based systems, the real gain for AMFPIs is in the efficiency of light detection [59]. Consideration of the thickness of the phosphor screen shows that thicker screens produce more photons, but they have worse spatial resolution. This means that there is no real gain with screens thicker than about 400 mg cm-2 [59].

For one prototype device [77], consisting of a 1.5 mm thick copper plate, a 134 mg cm-2 gadolinium oxysulphide phosphor screen and Schottky type diodes, has a system noise level at only 1% of the total noise, so the device is quantum noise-limited. Further analysis shows that this device should behave as if 1% of the incident X-rays were detected by the EPID, which is similar to the liquid ionization chamber array discussed above. The DQE of this device is 70–80% greater than that for a camera-based EPID. A problem with this device is that there are image artefacts due to pulsing of the linac that result in much lower visual image quality than might be expected. In a comparison with Kodak EC-L film, this device shows similar anatomy butmore artefacts, indicating the need for frame averaging.

Studies using another prototype device, comprising a copper screen with a 133 mg cm-2 thick coat of gadolinium oxysulphide but with n-type-semi-conductor–insulator–p-type-semi-conductor diodes [76, 78], show that this device is quantum noise-limited down to doses of 0.3 cGy. This system has a DQE(0) equivalent to camera-based systems (0.86%), and increasing the phosphor thickness to 400 mg cm-2 improves the DQE to 1.6%.

Kubo et al [79] report their experience with a prototype commercial amorphous silicon EPID based on the device described above [76]. This device required a fixed pulse repetition frequency from the linac, so it could not image during dynamically wedged fields. Also, the system saturated at an unattenuated dose of 3 monitor units (MU), which meant that with the acquisition speeds available, ranging from 1.31–6.20 s for the conditions investigated, dose rates of 400 MU min-1 and above could not be used for imaging. There are time penalties of 3–4 s for data transmission and image display processing, which slows down image display.

Other portal imaging devices
Storage phosphors. Storage phosphors and digital fluoroscopy have been investigated for portal imaging [80–86]. Digital storage phosphor radiography is a radiographic technique where a phosphor plate replaces film as the radiation detector. Laser scanning of the irradiated phosphor induces luminescence, detection of which is used to form a digital image [81]. The quality of these images is at least as good as those obtained with older style portal films [83, 85] and may have advantages over some current EPID systems in the verification of dynamic treatments [86]. However, like film, they still require developing before an image can be obtained.

Amorphous selenium. When amorphous selenium is irradiated it conducts electricity and an electrostatic image can be formed. This property has been used to form radiographic images [87] and is under active investigation for portal imaging [75, 88–91]. It is possible to combine a metal plate/amorphous selenium detector with aflat panel amorphous silicon TFT to form a direct-detecting imaging device [75]. This is like the amorphous silicon devices described above, with the phosphor photodiode layers replaced with amorphous selenium. It has been shown that the DQE of a metal+amorphous selenium converter is similar to that of a metal plate +phosphor converter of the same mass thickness [91]. Therefore, the relative advantages that this type of EPID may have over other a-Si:H devices depends on the comparative size of the quantum sinks owing to the coupling of the subassemblies and on any gain in the ease of device fabrication.

Scanning detectors. A variety of scanned linear arrays of radiation detectors have also been used to form portal images. These include scintillation crystals coupled to photodiodes [56, 92, 93] and silicon diodes [94].


   Image registration
Top
 Abstract
 Introduction
 Theoretical aspects of portal...
 Imaging media
 Image registration
Portal imaging dosimetry
 Future requirements
 References
 
The primary role of portal imaging has been to ensure the patient is in the correct position during treatment [9], with regular portal imaging decreasing the size and frequency of field placement errors [12, 95]. Computerized measurement techniques are better than manual approaches at detecting field displacements [44, 96]. A quantitative comparison of images is aided by image registration [9, 97]. This process requires a reference image, which shows the patient in the correct position, and the treatment portal image(s) to be compared. The reference image may be a simulator X-ray, another portal image, a digitally reconstructed radiograph or a digitally reconstructed portal image [97]. To make image registration a routine process in clinical practice one requires an integrated system that combines the functions of preparation of the reference image, portal image field edge detection, field edge matching, anatomy matching and presentation of results [98].

Quantifying patient displacement by image registration is a two-stage process [97]. First, the portal image field edge is found and aligned with the reference image field boundary. This determines field size and shape errors and establishes a common frame of reference for the two images. The second stage is to quantify any anatomical translations or rotations.

There have been several methods employed to detect field edges [97, 99–103]. The simplest approach is thresholding (Figure 6Go). If a histogram of frequency against grey level is plotted for a typical image, there are generally two peaks. The lower peak (grey levels 150 to 600 in Figure 6bGo) represents the area outside the field, the higher peak (levels 1000 to 1400 in Figure 6bGo) is the irradiated field. If the image has enough dynamic range, then the irradiated field can be segmented from the background by simply setting a global threshold level (800 in the example) [97]. Another approach is to apply an edge enhancement filter and to join the highest responding pixels [100]. These approaches may be combined, using a global threshold to approximately find the edge then applying a directional Sobel filter to find the point of maximal grey level gradient to refine the edge position [99].



View larger version (14K):
[in this window]
[in a new window]
  		Figure 6. (a) The principle of thresholding. The background pixels have grey level values between 250 and 600, whilst the irradiated field pixel values fall between 1000 and 1400. If a threshold of 800 is set, then the irradiated field will be outlined. (b) Histogram of frequency against grey level for a typical image.

 
Next, any deviation between the size and shape of the reference and portal field edge can be detected and reported [104, 105]. The two fields are then aligned. Common algorithms for this include moment normalization [97, 106, 107] or chamfer matching [9, 103, 108]. A problem with some implementations of moment normalization is that the images are required to be square and to have the same pixel size. If they do not, then errors in magnification and alignment can occur. Saying that, these methods can be fast and accurate, with alignment accuracies of 0.5 mm in shift, 0.3° in rotation and 0.004 in magnification achievable in 2 s on a Pentium II [107].

The final stage of the process is to match anatomical structures in both images to quantify displacements. The simplest way to achieve this is to identify structures in common (landmarks) between the two images and to translate/rotate the portal image onto the reference image [9, 109, 110]. This can be made less dependent on identifying exactly corresponding points by drawing open curves to identify the structures (e.g. pelvic rim) and then registering these curves [111]. Semi-automation of the process by automatically extracting bony ridges from the portal images and then using chamfer matching to register the extracted features to corresponding features drawn on the reference image has proved successful [112–115], with reported accuracies of less than 2 mm or 1° [113, 115]. This technique becomes less robust as the amount of information in the images decreases, with the failure rate increasing from 1% in large anteroposterior pelvic fields to 15% in small anteroposterior pelvic fields [116]. Other algorithms include core-based image analysis[117], cross-correlation [97, 118–120] and wavelet analysis [103]. Some of these algorithms work without any operator delineation of structures [97, 103, 119]. Once the deviations of the portal image from the reference image have been quantified, a decision needs to be made as to whether an adjustment in the patient's position is required. These decision rules may need to be site specific [114], but systems can be trained to make these decisions automatically [121].

Most algorithms assume that the two images can be superimposed on each other by simple two-dimensional (2D) rigid body transformations (scaling, translation and rotation). This is not always the case, as these images are 2D projections of a 3D structure, and out-of-plane rotations as small as 2° have a detrimental effect on some image registration algorithms. Progress has been made in using 3D analysis of the data [122–125]. In fact, 3D analysis proves more robust than conventional 2D analysis, with only 6% of 3D matched results requiring manual adjustment compared with 44% of 2D matched images [125].


   Portal imaging dosimetry
Top
 Abstract
 Introduction
 Theoretical aspects of portal...
 Imaging media
 Image registration
 Portal imaging dosimetry
Future requirements
 References
 
As well as containing information regarding patient anatomy, portal images provide information on patient dosimetry. Hence, portal imaging can potentially be used for both dosimetric and geometric verification of radiotherapy treatments. There are two main approaches to dosimetric verification. The first approach is to calibrate the imager in terms of patient exit dose and to compare this with the exit dose map from the treatment planning system. The second approach is to use a model to calculate the portal dose image and to compare that with the measured portal dose image. Both approaches require use of the planning CT data to provide a calculated dose map; the difference is the point at which the comparison takes place. In either case, if the patient's anatomy changes between the planning CT and the treatment, a difference will be found between the calculated and measured dose images.

Portal image dose prediction has received much attention. An early approach applied the delta volume method [126]. This worked well with 60Co radiation and little or no air gap between the patient and the detector. It does not work well with larger air gaps (greater than 10 cm) and so methods involving the superposition of Monte Carlo generated point-interaction dose kernels (convolution/superposition method) were developed [127]. Further work has increased the applicability to larger air gaps and higher energies [128, 129] and arbitrary detectors [130]. The scatter signal in these dose maps has also been modelled using methods ranging from full Monte Carlo simulations and pencil beam models to analytic functions [23, 131–135]. This allows the separation of the primary and scattered fluences and hence calculation of the actual dose delivered to the patient. Calculation methods include iterative modification of the CT data [136] and back-projection techniques [137–139].

There have been specific devices designed with portal dosimetry in mind [80, 93], but the general approach has been to use a portal imaging device for both purposes. The calibration of an image (grey level to dose) has been performed in several different ways [73, 140–143]. Each portal imaging medium has been studied to calibrate for dosimetry. These studies will now be summarized.

Film
Portal imaging film, such as Kodak X-Omat V, is calibrated for dose by plotting a sensometric curve for the radiation beam quality being measured. To do this, a film is exposed with a series of known doses. After developing, the background-corrected optical density is measured for each dose. The sensitometric curve is a plot of this optical density against dose. There have been several attempts to calibrate portal film to measure exit dose directly [144–148]. Exit dose is defined as the dose at a depth dmax (where dmax is the build-up depth of the radiation) inside the exit surface of the patient. For portal films taken at short distances from the surface of the patient, reasonably accurate exit doses (within 2%) can be determined from film densities transformed to doses using a sensitometric curve and applying theinverse-square law [146, 147]. This approach decreases in accuracy as the air gap between the patient and the film is increased. This is due to the changing proportion of scattered radiation in thebeam [147, 148] and also affects EPIDs. The best accuracy in relative exit dose profiles may beobtained by applying an inverse-square law correction for images taken close to the patient, while no correction should be applied for films taken at distances of 20–30 cm from the patient [148]. It has also been shown that film portal image dosimetry systematically underestimates doses behind high density inhomogeneities and overestimates doses behind low density inhomogeneities [147, 148].

Camera-based systems
Camera-based EPIDs have been configured for dosimetry [149–151]. These have been shown to have a good linear dose response, and exit doses have been measured to within 3% of those obtained with silicon diodes [150, 152]. Heijmen and colleagues have performed much of the work on the dosimetric aspects of these EPIDs and this will be briefly summarized here [128, 153–161]. These devices are very stable, with a day-to-day variation in response of 0.4% (1 SD) [154]. The measured grey scale values are linearly proportional to transmitted portal doses with a proportionality constant that is independent of the thickness of a flat, water-equivalent absorber in the beam, but that does significantly depend on field size [154]. This field size dependence of response is caused by light scatter in the optical system (glare) and can be removed by deconvolving the signal with a point spread function describing the light scatter in the system [154, 156]. This deconvolution method has also been used to remove any spatial non-uniformities andany non-linearities in response [156]. This approach allows calibration of camera EPIDs for open, wedged and intensity modulated 25 MV photon beams to within 1% (1 SD) of ionization chamber measurements [156]. A method to calculate the portal dose image with an accuracy of 1–2% has been developed using the patient's CT scans [128]. This has been used to performed clinical in vivo dosimetry on prostate cancer patients [155, 158]. Both studies found good agreement (within 2%) on the beam central axis, but found off-axis discrepancies of up to 15% owing to variations in the position and amount of bowel gas between the CT scan and the treatment. In fact, theoretical studies show that it is possible to detect some organ motions with portal imaging dosimetry [161]. Other applications of camera EPID dosimetry developed by this group include the verification of compensator thickness [159], some linac quantity control measurements [153] and the pre-treatment verification of dynamic intensity modulated radiotherapy fields [157].

Liquid ion chamber arrays
The construction and read-out characteristics of liquid ion chamber arrays means that they measure dose rate rather than dose [140]. The response is non-linear, with the pixel value being proportional to the square-root of the dose rate at low dose rates [140, 162, 163]. They have been calibrated to obtain radiological thickness maps for compensator design, with the overall error being approximately 4 mm in thickness or 2% in dose [163]. They can also be used for profile verification and absolute isocentric dose measurement for intensity modulated fields using integrated dose rate maps [164].

Boellaard and colleagues [141–143, 165–168] have made major contributions to the use of these EPIDs for dosimetry. They used additional build-up material on the front of the EPID to achieve electron equilibrium and went on to show that the relationship between ionization current and dose rate is described within 1% (1 SD) by an equation of the form , where I is the ionization current, a and b are constants that depend on energy, pulse repetition frequency and image acquisition mode, and D· is the dose rate. For images obtained under a typical clinical situation, the contribution of the square-root and linear term to the EPID signal is 94% and 6%, respectively. The dose response relationship must be determined for each EPID and accelerator setting [142].

Further work has developed a set of algorithms to deconvolve the scatter dose [141, 143] and to calculate the exit dose distribution [166]. This is a five-step process. First, the transmission dose is measured with an air gap of 40 cm or more between the detector and the patient. Then, thecontribution of scattered radiation from the patient on the EPID is subtracted from the transmission dose, yielding the dose distribution from directly transmitted (primary) radiation. This is possible because the contribution of scatter dose is small and the scatter-to-primary ratio is almost constant at these distances. The third step is to apply an inverse-square law correction to give the primary exit dose at the patient. Next, a convolution model uses this primary exit dose to reconstruct the 2D contribution of scatter dose. Finally, the total exit dose is calculated by adding primary and scatter dose. If inhomogeneities are present, then the model uses the radiological path length calculated from the ratio of transmission doses with and without the patient. The exit dose can be determined from portal images with an accuracy of 1.2% (1 SD) compared with ionization chamber measurements for open beams and homogeneous phantoms. In the presence of wedges and for inhomogeneous phantoms, the average relative accuracy slightly deteriorates to 1.7% (1 SD). This is without the use of CT data, but does require knowledge of the number of monitor units set for the treatment because it is dose rate that is measured directly [166].

This model has been extended to the calculation of mid-plane dose [167]. After calculating primary and scatter contribution to exit dose, the mid-plane dose is calculated assuming a homogeneous patient, or at least one with symmetrical inhomogeneities. This method can be used to assess the mid-plane dose for most clinical situations within 2% relative to ionization chamber measurements. In the presence of large asymmetrical inhomogeneities (e.g. lungs), discrepancies of about 8% have been found (for small field sizes) owing to the absence of lateral electron equilibrium. For large field sizes, the agreement between measured and predicted mid-plane dose was within 3%.

Applying these methods in the clinic, measurements were performed to verify the mid-plane dose during radiotherapy of larynx cancer with 4 MV beams, breast and lung cancer with 8 MV beams and prostate cancer with 8 MV and 18 MV beams. Mid-plane doses from portal dose measurements were compared with mid-plane doses calculated by a 3D treatment planning system (TPS). For the larynx treatment the measured 2D mid-plane dose agreed within 2% with TPS calculations over most parts of the field. Portal dose measurements and TPS calculations agreed to within 2.5% for most of the prostate and lung irradiations. For a few of the prostate and lung treatments, larger local differences were found owing to differences between the actual patient anatomy and the planning CT data. This was a result of variable gas in the rectum and anatomical changes in the lung [168].

Amorphous silicon
There has been one preliminary study of the use of a a-Si:H EPID for portal imaging dosimetry [169]. This study used the device in one of two ways: direct mode, with metal screen/phosphor converter removed; indirect mode, with the converter present. In the direct mode the EPID acts like an array of silicon diodes. Calibration was straightforward, with signal levels acquired within the first 25% of pixel charge capacity being linear with dose to better than 99% and dose rate independent. In the indirect mode the detector exhibits large differences compared with ion chamber measurements owing to the overresponse of the phosphor to low energy scattered radiation. To enable such a device to act as both an imager and dosimeter, calibration routines similar to those used with camera-based EPIDs are required. This has not yet been reported [169].


   Future requirements
Top
 Abstract
 Introduction
 Theoretical aspects of portal...
 Imaging media
 Image registration
 Portal imaging dosimetry
 Future requirements
References
 
With the introduction of commercial EPIDs, portal imaging for geometric verification of radiotherapy treatments is becoming increasingly used and accepted in the clinic. Imaging protocols are being developed that identify and reduce set-up errors; it is hoped that these improvements in geometric accuracy will lead to improved patient outcome [12, 170, 171]. Unfortunately, the image registration algorithms in current commercially available systems are at best semi-automatic and require a fair amount of operator input to give accurate results. Consequently, it is quite time consuming to process and analyse the images, which limits the application of imaging protocols that require multiple images unless dedicated imaging staff are available. To overcome this limitation there is a need to develop and implement image registration algorithms in commercial systems that are fully automatic [97, 103, 119]. This may also enable automatic analysis of the large numbers of images that can be produced during modern high precision radiotherapy. An interesting area of development is the combined use of fast imaging amorphous silicon EPIDs and fuzzy logic decision-making algorithms for on-line motion detection [79, 121]. This may be important if intensity modulated radiotherapy is to fulfil itspotential. Also, there is a need to develop commercial imagers capable of imaging intensity modulated radiotherapy fields [157, 164].

Portal imaging dosimetry modules also need to be developed and implemented in an easy to use manner if its application is to move out of the research setting. A possible example of such a system would be for a planning system to export predicted portal dose images [97, 130] with the portal imaging system calibrated to give a measured portal dose image [130, 172, 173]. This could be used both as a dosimetric verification tool for individual treatment fractions and to give actual dose delivered to a patient during the entire course of treatment. The routine implementation of such a scheme is still a long way off.

Task Group 58 of the American Association of Physicists in Medicine (AAPM) has produced a report on the clinical use of electronic portal imaging [174]. This report covers the system characteristics, limitations and software tools currently available commercially. It also addresses the acceptance testing, commissioning and QA required by these systems. It does not include amorphous silicon devices. Those implementing electronic portal imaging clinically will find this report very useful.


   Acknowledgments
 
I wish to thank Electa Oncology Systems Ltd (Crawley, UK), Oncology Systems Ltd (OSL) (Shrewsbury, UK), Siemens Medical Engineering (Bracknell, UK) and Varian Medical Systems (Crawley, UK) for their help in preparing parts of this article.

Received for publication January 4, 2001. Revision received June 5, 2001. Accepted for publication June 21, 2001.


   References
Top
 Abstract
 Introduction
 Theoretical aspects of portal...
 Imaging media
 Image registration
 Portal imaging dosimetry
 Future requirements
 References
 

  1. Boyer AL, Antonuk L, Fenster A, van Herk M, Meertens H, Munro P, et al. A review of electronic portal imaging devices (EPIDs). Med Phys 1992;19:1–16.[Medline]
  2. Sephton R, Hagekyriakou J. A diagnostic-quality electronic portal imaging system. Radiother Oncol 1995;35:240–7.[Medline]
  3. Jaffray DA, Chawla K, Yu C, Wong JW. Dual-beam imaging for online verification of radiotherapy field placement. Int J Radiat Oncol Biol Phys 1995;33:1273–80.[Medline]
  4. Jaffray DA, Drake DG, Moreau M, Martinez AA, Wong JW. A radiographic and tomographic imaging system integrated into a medical linear accelerator for localization of bone and soft-tissue targets. Int J Radiat Oncol Biol Phys 1999;45:773–89.[Medline]
  5. Galbraith DM. Low-energy imaging with high-energy bremsstrahlung beams. Med Phys 1989;16:734–46.[Medline]
  6. Ostapiak OZ, O'Brien PF, Faddegon BA. Megavoltage imaging with low Z targets: implementation and characterization of an investigational system. Med Phys 1998;25:1910–8.[Medline]
  7. Mah DW, Galbraith DM, Rawlinson JA. Low-energy imaging with high-energy bremsstrahlung beams: analysis and scatter reduction. Med Phys 1993;20:653–65.[Medline]
  8. Tsechanski A, Bielajew AF, Faermann S, Krutman Y. A thin target approach for portal imaging in medical accelerators. Phys Med Biol 1998;43:2221–36.[Medline]
  9. Munro P. Portal imaging technology: past, present, and future. Semin Radiat Oncol 1995;5:115–33.[Medline]
  10. Haus AG, Dickerson RE, Huff KE, Monte S, Schlager BA, Atanas M, et al. Evaluation of a cassette–screen–film combination for radiation therapy portal localization imaging with improved contrast. Med Phys 1997;24:1605–8.[Medline]
  11. Booth JT, Zavgorodni SF. Set-up error and organ motion uncertainty: a review. Australas Phys Eng Sci Med 1999;22:29–47.[Medline]
  12. Bel A, Vos PH, Rodrigus PT, Creutzberg CL, Visser AG, Stroom JC, et al. High-precision prostate cancer irradiation by clinical application of an offline patient setup verification procedure, using portal imaging. Int J Radiat Oncol Biol Phys 1996;35:321–32.[Medline]
  13. van Herk M, Remeijer P, Rasch C, Lebesque JV. The probability of correct target dosage: dose–population histograms for deriving treatment margins in radiotherapy. Int J Radiat Oncol Biol Phys 2000;47:1121–35.[Medline]
  14. Tubiana M, Eschwege F. Conformal radiotherapy and intensity-modulated radiotherapy—clinical data. Acta Oncologica 2000;39:555–67.[Medline]
  15. Motz JW, Danos M. Image information content and patient exposure. Med Phys 1978;5:8–22.[Medline]
  16. Munro P, Rawlinson JA, Fenster A. Therapy imaging: a signal-to-noise analysis of metal plate/film detectors. Med Phys 1987;14:975–84.[Medline]
  17. Munro P, Rawlinson JA, Fenster A. Therapy imaging: a signal-to-noise analysis of a fluoroscopic imaging system for radiotherapy localization. Med Phys 1990;17:763–72.[Medline]
  18. Kausch C, Schreiber B, Kreuder F, Schmidt R, Dossel O. Monte Carlo simulations of the imaging performance of metal plate/phosphor screens used in radiotherapy. Med Phys 1999;26:2113–24.[Medline]
  19. Munro P, Bouius DC. X-ray quantum limited portal imaging using amorphous silicon flat-panel arrays. Med Phys 1998;25:689–702.[Medline]
  20. Pang G, Rowlands JA. Electronic portal imaging with an avalanche-multiplication-based video camera. Med Phys 2000;27:676–84.[Medline]
  21. Bissonnette JP, Jaffray DA, Fenster A, Munro P. Optimal radiographic magnification for portal imaging. Med Phys 1994;21:1435–45.[Medline]
  22. Amols HI, Reinstein LE, Lagueux B. A quantitative assessment of portal film contrast as a function of beam energy. Med Phys 1986;13:711–6.[Medline]
  23. Jaffray DA, Battista JJ, Fenster A, Munro P. X-ray scatter in megavoltage transmission radiography: physical characteristics and influence on image quality. Med Phys 1994;21:45–60.[Medline]
  24. Hurter F, Driffield VC. Photo-chemical investigations and a new method of determination of the sensitiveness of photographic plates. J Soc Chem Ind London 1890;9:455–69.
  25. Faermann S. Krutman Y. Generation of portal film charts for 10-MV x rays. Med Phys 1992;19:351–3.[Medline]
  26. Langmack KA, Goss V. Characterization of new portal film systems for radiotherapy verification. Br J Radiol 1999;72:479–84.[Abstract]
  27. Lee PC, Glasgow GP. Technique charts for Kodak's new film–screen systems for portal localization. Med Dosim 1998;23:113–6.[Medline]
  28. Roberts R. Portal imaging with film–cassette combinations: what film should we use? Br J Radiol 1996;69:70–1.[Abstract]
  29. Radiation Therapy Committee Task Group No. 28. Radiotherapy portal imaging quality, AAPM, Report No. 24. Madison, WI: Medical Physics Publishing, 1987.
  30. Haus AG. Film systems for radiotherapy imaging. In: Hale JD, Boyer AL, editors. Imaging in radiation therapy, AAPM Medical Physics monograph No. 24. Madison, WI: Medical Physics Publishing, 1998:179–206.
  31. Hammoudah MM, Henschke K. Supervoltage X-ray beam films. Int J Radiat Oncol Biol Phys 1977;2:571–7.[Medline]
  32. Droege RT, Bjarngard BE. Influence of metal screens on contrast in megavoltage x-ray imaging. Med Phys 1979;6:487–93.[Medline]
  33. Droege RT, Bjarngard BE. Metal screen–film detector MTF at megavoltage x-ray energies. Med Phys 1979;6:515–8.[Medline]
  34. Kihlen B, Cederlund T, Lagergren C, Nordell B, Ruden BI. Improved portal film image quality in radiation therapy with high energy photons. Acta Oncologica 1991;30:369–73.[Medline]
  35. Kirschner H, Busch M, Bornkessel C, Dieckmann K, Duhmke E. Aluminium, steel and tungsten portal film cassettes for the imaging verification of 60Co photon irradiation up to 42 MV in a clinical trial. Strahlenther Onkol 1991;167:480–3.[Medline]
  36. Goretzky EJ, Kind AL. Evaluation of films for use in portal imaging in radiation therapy—a study at the Saskatoon Cancer Centre. Can J Med Radiat Technol 1992;23:117–9.[Medline]
  37. Hamilton CS, Piyaratna N, Ferguson SA, Callaghan TM, Denham JW. Clinical utility of high-resolution portal films. Clin Oncol (R Coll Radiol) 1991;3:90–5.
  38. Li H, Xu Z, Almond PR. Evaluation of several Dupont portal film systems. Med Dosim 1996;21:169–72.[Medline]
  39. Dixon RL, Ekstrand KE. Heuristic model for understanding x-ray film characteristics. Med Phys 1976;3:340–5.[Medline]
  40. Sephton R, Green M, Fitzpatrick C. A new system for port films. Int J Radiat Oncol Biol Phys 1989;16:251–8.[Medline]
  41. Hermann A, Bratengeier K, Priske A, Flentje M. A new verification film system for routine quality control of radiation fields: Kodak EC-L. Strahlenther Onkol 2000;176:274–7.[Medline]
  42. Ostinelli A, Frigerio M, Gelosa S, Monti AF, Cacciatori M, Cazzaniga F. Contrast improvement in a new radiotherapy imaging system. Strahlenther Onkol 1999;175:182–4.[Medline]
  43. Walker MA, Steinheimer DN, Weir VA, Homco LD, Green RW, Morris EL, et al. A review of portal screen-film technology and five radiologists' evaluations of some existing products. Vet Radiol Ultrasound 1999;40:318–22.[Medline]
  44. Perera T, Moseley J, Munro P. Subjectivity in interpretation of portal films. Int J Radiat Oncol Biol Phys 1999;45:529–34.[Medline]
  45. Thompson V, Bidmead M, Mubata C. Comparison of portal imaging and megavoltage verification films for conformal pelvic irradiation. Br J Radiol 1996;69:1191–5.[Abstract]
  46. Evans PM. Commissioning and QC of imaging devices. In: Mayles WPM, Lake R, McKenzie A, Macaulay EM, Morgan HM, Jordan TJ, et al, editors. Physical aspects of quality control in radiotherapy, IPEM Report 81. York: IPEM, 1999:236–7.
  47. Low DA, Klein EE, Maag D, Umfleet WE, Purdy J. Commissioning and periodic quality assurance of a clinical electronic portal imaging device. Int J Radiat Oncol Biol Phys 1996;34:117–23.[Medline]
  48. Klein EE, Low DA, Maag D, Purdy JA. A quality assurance program for ancillary high technology devices on a dual-energy accelerator. Radiother Oncol 1996;38:51–60.[Medline]
  49. Lescrainier J, Grelot M, Boutinaud C, Bouscayrol H, Champeaux I, Ginestet C. Quality control in electronic portal imaging devices. Commission Imagerie, Societé Française des Physiciens d'Hôpital. Bull Cancer Radiother 1996;83:409–13. [In French.][Medline]
  50. Kirby MC, Atherton S, Carson P, McDonagh C, Sykes JR, Williams PC. The performance of a fluoroscopic electronic portal imaging device modified for portability. Br J Radiol 1999;72:1000–5.[Abstract]
  51. Kirby MC. The consequences of fixed-pattern noise and image movement on electronic portal images. Phys Med Biol 1996;41:2345–56.[Medline]
  52. Harrison RM, Lambert GD. A method and contrast-detail phantom for the quantitative assessment of radiotherapy portal imaging systems. Br J Radiol 1994;67:384–8.[Abstract]
  53. Dong L, Boyer AL. An objective method for evaluating electronic portal imaging devices. Med Phys 1994;21:755–60.[Medline]
  54. Lutz WR, Bjarngard BE. A test object for evaluation of portal films. Int J Radiat Oncol Biol Phys 1985;11:631–4.[Medline]
  55. Rajapakshe R, Luchka K, Shalev S. A quality control test for electronic portal imaging devices. Med Phys 1996;23:1237–44.[Medline]
  56. Morton EJ, Swindell W, Lewis DG, Evans PM. A linear array, scintillation crystal-photodiode detector for megavoltage imaging. Med Phys 1991;18:681–91.[Medline]
  57. Partridge M, Evans PM, Mosleh-Shirazi MA. Linear accelerator output variations and their consequences for megavoltage imaging. Med Phys 1998;25:1443–52.[Medline]
  58. Munro P, Rawlinson JA, Fenster A. A digital fluoroscopic imaging device for radiotherapy localization. Int J Radiat Oncol Biol Phys 1990;18:641–9.[Medline]
  59. Bissonnette JP, Cunningham IA, Munro P. Optimal phosphor thickness for portal imaging. Med Phys 1997;24:803–14.[Medline]
  60. Radcliffe T, Barnea G, Wowk B, Rajapakshe R, Shalev S. Monte Carlo optimization of metal/phosphor screens at megavoltage energies. Med Phys 1993;20:1161–9.[Medline]
  61. Bissonnette JP, Munro P. Evaluation of a high-density scintillating glass for portal imaging. Med Phys 1996;23:401–6.[Medline]
  62. 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 Phys 1998;25:2333–46.[Medline]
  63. Mah DW, Rowlands JA, Rawlinson JA. Measurement of quantum noise in fluoroscopic systems for portal imaging. Med Phys 1996;23:231–8.[Medline]
  64. Althof VG, de Boer JC, Huizenga H, Stroom JC, Visser AG, Swanenburg BN. Physical characteristics of a commercial electronic portal imaging device. Med Phys 1996;23:1845–55.[Medline]
  65. Bissonnette JP, Cunningham IA, Jaffray DA, Fenster A, Munro P. A quantum accounting anddetective quantum efficiency analysis for video-based portal imaging. Med Phys 1997;24:815–26.[Medline]
  66. Swindell W. The lens coupling efficiency in megavoltage imaging. Med Phys 1991;18:1152–3.[Medline]
  67. Partridge M, Evans PM, Symonds-Tayler JR. Optical scattering in camera-based electronic portal imaging. Phys Med Biol 1999;44:2381–96.[Medline]
  68. Wong JW, Binns WR, Cheng AY, Geer LY, Epstein JW, Klarmann J, et al. On-line radiotherapy imaging with an array of fiber-optic image reducers. Int J Radiat Oncol Biol Phys 1990;18:1477–84.[Medline]
  69. Meertens H, van Herk M, Weeda J. A liquid ionisation detector for digital radiography of therapeutic megavoltage photon beams. Phys Med Biol 1985;30:313–21.[Medline]
  70. van Herk M, Meertens H. A matrix ionisation chamber imaging device for on-line patient setup verification during radiotherapy. Radiother Oncol 1988;11:369–78.[Medline]
  71. van Herk M, Bijhold J, Hoogervorst B, Meertens H. Sampling methods for a matrix ionization chamber system. Med Phys 1992;19:409–18.[Medline]
  72. van Herk M. Physical aspects of a liquid-filled ionization chamber with pulsed polarizing voltage. Med Phys 1991;18:692–702.[Medline]
  73. Yin FF, Schell MC, Rubin P. Input/output characteristics of a matrix ion-chamber electronic portal imaging device. Med Phys 1994;21:1447–54.[Medline]
  74. Antonuk LE, Boudry J, Huang W, McShan DL, Morton EJ, Yorkston J, et al. Demonstration of megavoltage and diagnostic x-ray imaging with hydrogenated amorphous silicon arrays. Med Phys 1992;19:1455–66.[Medline]
  75. Antonuk LE, El-Morhi Y, Jee KW. Active matrix flat-panel imagers (AMFPIs) for electronic portal imaging. In: Hacle JD, Boyer AL, editors. Imaging in radiation therapy, AAPM Medical Physics monograph No. 24. Madison, WI: Medical Physics Publishing, 1998:371–92.
  76. Antonuk LE, El Mohri Y, Huang W, Jee KW, Siewerdsen JH, Maolinbay M, et al. Initial performance evaluation of an indirect-detection, active matrix flat-panel imager (AMFPI) prototype for megavoltage imaging. Int J Radiat Oncol Biol Phys 1998;42:437–54.[Medline]
  77. Munro P, Bouius DC. X-ray quantum limited portal imaging using amorphous silicon flat-panel arrays. Med Phys 1998;25:689–702.
  78. Antonuk LE, El Mohri Y, Siewerdsen JH, Yorkston J, Huang W, Scarpine VE, et al. Empirical investigation of the signal performance of a high-resolution, indirect detection, active matrix flat-panel imager (AMFPI) for fluoroscopic and radiographic operation. Med Phys 1997;24:51–70.[Medline]
  79. Kubo HD, Shapiro EG, Seppi EJ. Potential and role of a prototype amorphous silicon array electronic portal imaging device in breathing synchronized radiotherapy. Med Phys 1999;26:2410–4.[Medline]
  80. Leong J. Use of digital fluoroscopy as an on-line verification device in radiation therapy. Phys Med Biol 1986;31:985–92.[Medline]
  81. Wilenzick RM, Merritt CR, Balter S. Megavoltage portal films using computed radiographic imaging with photostimulable phosphors. Med Phys 1987;14:389–92.[Medline]
  82. Crespi A, Pedroli G, Frigerio F, Pastori R, Stucchi G, Salvini E. Digital radiography with phosphorus memory: verification of the centring of the patient in high-energy radiotherapy. Radiol Med (Torino) 1989;78:452–4.
  83. Gur D, Deutsch M, Fuhrman CR, Clayton PA, Weiser JC, Rosenthal MS, et al. The use of storage phosphors for portal imaging in radiation therapy: therapists' perception of image quality. Med Phys 1989;16:132–6.[Medline]
  84. Barnea G, Navon E, Ginzburg A, Politch J, Roehrig H, Dick CE, et al. Use of storage phosphor imaging plates in portal imaging and high-energy radiography: the intensifying effect of metallic screens on the sensitivity. Med Phys 1991;18:432–8.[Medline]
  85. Scheck RJ, Wendt T, Panzer M. Digital storage phosphor radiography for treatment verification in radiotherapy. Br J Radiol 1993;66:801–6.[Abstract]
  86. Zimmermann JS, Blume J, Wendhausen H, Hebbinghaus D, Kovacs G, Eilf K, et al. The new possibilities of digital luminescence radiography (DLR) and digital imaging data processing for the making of field verification images. Strahlenther Onkol 1995;171:332–9.[Medline]
  87. Yaffe MJ, Rowlands JA. X-ray detectors for digital radiography. Phys Med Biol 1997;42:1–39.[Medline]
  88. Lachaine M, Fallone BG. Monte Carlo detective quantum efficiency and scatter studies of a metal/a-Se portal detector. Med Phys 1998;25:1186–94.[Medline]
  89. Falco T, Wang H, Fallone BG. Preliminary study of a metal/a-Se-based portal detector. Med Phys 1998;25:814–23.[Medline]
  90. Mah D, Rowlands JA, Rawlinson JA. Sensitivity of amorphous selenium to x rays from 40 kVp to 18 MV: measurements and implications for portal imaging. Med Phys 1998;25:444–56.[Medline]
  91. Mah D, Rawlinson JA, Rowlands JA. Detective quantum efficiency of an amorphous selenium detector to megavoltage radiation. Phys Med Biol 1999;44:1369–84.[Medline]
  92. Gildersleve J, Dearnaley D, Evans P, Morton E, Swindell W. Preliminary clinical performance of a scanning detector for rapid portal imaging. Clin Oncol (R Coll Radiol) 1994;6:245–50.
  93. Symonds-Tayler JR, Partridge M, Evans PM. An electronic portal imaging device for transit dosimetry. Phys Med Biol 1997;42:2273–83.[Medline]
  94. Lam KS, Partowmah M, Lam WC. An on-line electronic portal imaging system for external beam radiotherapy. Br J Radiol 1986;59:1007–13.[Abstract]
  95. Michalski JM, Graham MV, Bosch WR, Wong J, Gerber RL, Cheng A, et al. Prospective clinical evaluation of an electronic portal imaging device. Int J Radiat Oncol Biol Phys 1996;34:943–51.[Medline]
  96. Boxwala AA, Chaney EL, Fritsch DS, Raghavan S, Coffey CS, Major SA, et al. Comparison of computer workstation with light box for detecting setup errors from portal images. Int J Radiat Oncol Biol Phys 1999;44:711–6.[Medline]
  97. Dong L. Portal image correlation and analysis. In: Hacle JD, Boyer AL, editors. Imaging in radiation therapy, AAPM Medical Physics monograph No. 24. Madison, WI: Medical Physics Publishing, 1998:415–44.
  98. van Herk M, Bel A, Gilhuijs KG, Vijlbrief RE. A comprehensive system for the analysis of portal images. Radiother Oncol 1993;29:221–9.[Medline]
  99. Bijhold J, Gilhuijs KG, van Herk M, Meertens H. Radiation field edge detection in portal images. Phy Med Biol 1991;36:1705–10.
  100. Leszczynski KW, Shalev S, Cosby NS. The enhancement of radiotherapy verification images by an automated edge detection technique. Med Phys 1992;19:611–21.[Medline]
  101. Wang H, Fallone BG. A robust morphological algorithm for automatic radiation field extraction and correlation of portal images. Med Phys 1994;21:237–44.[Medline]
  102. McGee KP, Schultheiss TE, Martin EE. A heuristic approach to edge detection in on-line portal imaging. Int J Radiat Oncol Biol Phys 1995;32:1185–92.[Medline]
  103. Petrascu O, Bel A, Linthout N, Verellen D, Soete G, Storme G. Automatic on-line electronic portal image analysis with a wavelet-based edge detector. Med Phys 2000;27:321–9.[Medline]
  104. Leszczynski KW, Shalev S, Gluhchev G. Verification of radiotherapy treatments: computerized analysis of the size and shape of radiation fields. Med Phys 1993;20:687–94.[Medline]
  105. Leszczynski K, Loose S. A polygon matching algorithm and its applications to verification of radiation field placement in radiotherapy. Int J Biomed Comput 1995;40:59–67.[Medline]
  106. Bijhold J, Gilhuijs KG, van Herk M. Automatic verification of radiation field shape using digital portal images. Med Phys 1992;19:1007–14.[Medline]
  107. Shu HZ, Ge Y, Luo LM, Wang JW, Yu WX, Bao XD. An orthogonal moment-based method for automatic verification of radiation field shape. Phys Med Biol 2000;45:2897–911.[Medline]
  108. Leszczynski K, Loose S, Dunscombe P. Segmented chamfer matching for the registration of field borders in radiotherapy images. Phys Med Biol 1995;40:83–94.[Medline]
  109. Bijhold J, van Herk M, Vijlbrief R, Lebesque JV. Fast evaluation of patient set-up during radiotherapy by aligning features in portal and simulator images. Phys Med Biol 1991;36:1665–79.[Medline]
  110. Meertens H, Bijhold J, Strackee J. A method for themeasurement of field placement errors in digital portal images. Phys Med Biol 1990;35:299–323.[Medline]
  111. Balter JM, Pelizzari CA, Chen GT. Correlation of projection radiographs in radiation therapy using open curve segments and points. Med Phys 1992;19:329–34.[Medline]
  112. Gilhuijs KG, el Gayed AA, van Herk M, Vijlbrief RE. An algorithm for automatic analysis of portal images: clinical evaluation for prostate treatments. Radiother Oncol 1993;29:261–8.[Medline]
  113. Gilhuijs KG, van Herk M. Automatic on-line inspection of patient setup in radiation therapy using digital portal images. Med Phys 1993;20:667–77.[Medline]
  114. Bissett R, Leszczynski K, Loose S