A two-step algorithm for predicting portal dose images in arbitrary detectors

Recently, portal imaging systems have been successfully demonstrated in dosimetric treatment verification applications, where measured and predicted images are quantitatively compared. To advance this approach to dosimetric verification, a two-step model which predicts dose deposition in arbitrary portal image detectors is presented. The algorithm requires patient CT data, source-detector distance, and knowledge of the incident beam fluence. The first step predicts the fluence entering a portal imaging detector located behind the patient. Primary fluence is obtained through ray-tracing techniques, while scatter fluence prediction requires a library of Monte Carlo-generated scatter fluence kernels. These kernels allow prediction of basic radiation transport parameters characterizing the scattered photons, including fluence and mean energy. The second step of the algorithm involves a superposition of Monte Carlo-generated pencil beam kernels, describing dose deposition in a specific detector, with the predicted incident fluence. This process is performed separately for primary and scatter fluence, and yields a predicted dose image. A small but noticeable improvement in prediction is obtained by explicitly modeling the off-axis energy spectrum softening due to the flattening filter. The algorithm is tested on a slab phantom and a simple lung phantom (6 MV). Furthermore, an anthropomorphic phantom is utilized for a simulated lung treatment (6 MV), and simulated pelvis treatment (23 MV). Data were collected over a range of air gaps (10-80 cm). Detectors incorporating both low and high atomic number buildup are used to measure portal image profiles. Agreement between predicted and measured portal dose is better than 3% in areas of low dose gradient (<30%/cm) for all phantoms, air gaps, beam energies, and detector configurations tested here. It is concluded that this portal dose prediction algorithm is fast, accurate, allows separation of primary and scatter dose, and can model arbitrary detectors.     

基于射野影像的剂量重建方法之可行性初探

利用治疗时得到的影像信息重建病人体内所接受的真实剂量是剂量引导下放射治疗（DGRT）技术临床应用的重要内容之一。治疗剂最重建方法之一是：通过匹对模拟的射野影像和治疗时获得的射野影像来反推射野注量分布，然后计算出治疗时病人体内实际接受的剂量：由此重建出的射野注量是否精确受诸多因素的影响．因素之一是是否能真实地模拟射野影像。本文用傅立叶卷积积分法模拟射野影像，采用不同积分半径得剑不同程度散射近似下的模拟射野影像。通过用不同程度散射近似下的模拟射野影像来反推射野注量分布．考察了散射因素对反推射野注量分布的影响。发现：即便射野影像在没有噪音和分辨率足够高的理想情况下，要精确地反推射野注量，必须在模拟射野影像中充分考虑散射因素。     

Investigation of basic imaging properties in digital radiography. 12. Effect of matrix configuration on spatial resolution

We investigated the effects of three matrix configurations (a conventional 512 X 512 matrix, a double-sampling matrix, and a high-resolution 1024 X 1024 matrix) on the resolution properties of an image intensifier (II)-TV digital radiographic imaging system. Expressions for the optical transfer function were derived theoretically for each of the three matrix configurations, and the corresponding digital modulation transfer functions (MTF's) were calculated using the measured MTF's of the presampling analog components and the parameters of the digital system. In addition, digital images of a star pattern phantom and a hand phantom were obtained with the II-TV system for each matrix type. From our results, we conclude that, in order to increase the resolution of an II-TV digital imaging system, it is necessary not only to increase the matrix size, but also to improve the resolution capabilities of analog components such as the II and/or TV camera. In addition, if the system cannot be upgraded to a 1024 X 1024 matrix configuration, a double-sampling technique with an 512 X 512 matrix may be employed.     

Optimization of accelerator target and detector for portal imaging using Monte Carlo simulation and experiment

Megavoltage portal images suffer from poor quality compared to those produced with kilovoltage x-rays. Several authors have shown that the image quality can be improved by modifying the linear accelerator to generate more low-energy photons. This work addresses the problem of using Monte Carlo simulation and experiment to optimize the beam and detector combination to maximize image quality for a given patient thickness. A simple model of the whole imaging chain was developed for investigation of the effect of the target parameters on the quality of the image. The optimum targets (6 mm thick aluminium and 1.6 mm copper) were installed in an Elekta SL25 accelerator. The first beam will be referred to as A16 and the second as Cu1.6. A tissue-equivalent contrast phantom was imaged with the 6 MV standard photon beam and the experimental beams with standard radiotherapy and mammography film/screen systems. The arrangement with a thin Al target/mammography system improved the contrast from 1.4 cm bone in 5 cm water to 19% compared with 2% for the standard arrangement of a thick, high-Z target/radiotherapy verification system. The linac/phantom/detector system was simulated with the BEAM/EGS4 Monte Carlo code. Contrast calculated from the predicted images was in good agreement with the experiment (to within 2.5%). The use of MC techniques to predict images accurately, taking into account the whole imaging system, is a powerful new method for portal imaging system design optimization.     

X-ray imaging optimization using virtual phantoms and computerized observer modelling

This study develops and demonstrates a realistic x-ray imaging simulator with computerized observers to maximize lesion detectability and minimize patient exposure. A software package, ViPRIS, incorporating two computational patient phantoms, has been developed for simulating x-ray radiographic images. A tomographic phantom, VIP-Man, constructed from Visible Human anatomical colour images is used to simulate the scattered portion using the ESGnrc Monte Carlo code. The primary portion of an x-ray image is simulated using the projection ray-tracing method through the Visible Human CT data set. To produce a realistic image, the software simulates quantum noise, blurring effects, lesions, detector absorption efficiency and other imaging artefacts. The primary and scattered portions of an x-ray chest image are combined to form a final image for computerized observer studies and image quality analysis. Absorbed doses in organs and tissues of the segmented VIP-Man phantom were also obtained from the Monte Carlo simulations. Approximately 25,000 simulated images and 2,500,000 data files were analysed using computerized observers. Hotelling and Laguerre-Gauss Hotelling observers are used to perform various lesion detection tasks. Several model observer tasks were used including SKE/BKE, MAFC and SKEV. The energy levels and fluence at the minimum dose required to detect a small lesion were determined with respect to lesion size, location and system parameters.     

Optimal steel thickness combined with computed radiography for portal imaging of nasopharyngeal cancer patients

The poor image quality of conventional metal screen-film portal imaging system has long been of concern, and various methods have been investigated in an attempt to enhance the quality of portal images. Computed radiography (CR) used in combination with a steel plate displays image enhancement. The optimal thickness of the steel plate had been studied by measuring the modulation transfer function (MTF) characteristics. Portal images of nasopharyngeal carcinoma patients were taken by both a conventional metal screen-film system and this optimal steel and CR plate combination system. Compared with a conventional metal screen-film system, the CR-metal screen system achieves a much higher image contrast. The measured modulation transfer function (MTF) of the CR combination is greater than conventional film-screen portal imaging systems and also results in superior image performance, as demonstrated by receiver operator characteristic (ROC) analysis. This optimal combination steel CR plate portal imaging system is capable of producing high contrast portal images conveniently.     

A Monte Carlo based three-dimensional dose reconstruction method derived from portal dose images

The verification of intensity-modulated radiation therapy (IMRT) is necessary for adequate quality control of the treatment. Pretreatment verification may trace the possible differences between the planned dose and the actual dose delivered to the patient. To estimate the impact of differences between planned and delivered photon beams, a three-dimensional (3-D) dose verification method has been developed that reconstructs the dose inside a phantom. The pretreatment procedure is based on portal dose images measured with an electronic portal imaging device (EPID) of the separate beams, without the phantom in the beam and a 3-D dose calculation engine based on the Monte Carlo calculation. Measured gray scale portal images are converted into portal dose images. From these images the lateral scattered dose in the EPID is subtracted and the image is converted into energy fluence. Subsequently, a phase-space distribution is sampled from the energy fluence and a 3-D dose calculation in a phantom is started based on a Monte Carlo dose engine. The reconstruction model is compared to film and ionization chamber measurements for various field sizes. The reconstruction algorithm is also tested for an IMRT plan using 10 MV photons delivered to a phantom and measured using films at several depths in the phantom. Depth dose curves for both 6 and 10 MV photons are reconstructed with a maximum error generally smaller than 1% at depths larger than the buildup region, and smaller than 2% for the off-axis profiles, excluding the penumbra region. The absolute dose values are reconstructed to within 1.5% for square field sizes ranging from 5 to 20 cm width. For the IMRT plan, the dose was reconstructed and compared to the dose distribution with film using the gamma evaluation, with a 3% and 3 mm criterion. 99% of the pixels inside the irradiated field had a gamma value smaller than one. The absolute dose at the isocenter agreed to within 1% with the dose measured with an ionization chamber. It can be concluded that our new dose reconstruction algorithm is able to reconstruct the 3-D dose distribution in phantoms with a high accuracy. This result is obtained by combining portal dose images measured prior to treatment with an accurate dose calculation engine.     

Electron beam treatment verification using measured and Monte Carlo predicted portal images

Electron beam treatments may benefit from techniques to verify patient positioning and dose delivery. This is particularly so for complex techniques such as mixed photon and electron beam radiotherapy and electron beam modulated therapy. This study demonstrates that it is possible to use the bremsstrahlung photons in an electron beam from a dual scattering foil linear accelerator to obtain portal images of electron beam treatments. The possibility of using Monte Carlo (MC) simulations to predict the electron beam treatment portal images was explored. The MC code EGSnrc was used to model a Varian CL21EX linear accelerator (linac) and to characterize the bremsstrahlung photon production in the linac head. It was found that the main sources of photons in the electron beam are the scattering foils, the applicator and the beam-shaping cut-out. Images were acquired using the Varian CL21EX linac and the Varian aS500 electronic portal imager (EPI); four electron energies (6, 9, 12, 16 MeV), and different applicator and cut-out sizes were used. It was possible to acquire images with as little as 10.7 MU per image. The contrast, the contrast-to-noise ratio (CNR), the signal-to-noise ratio (SNR), the resolution and an estimate of the modulated transfer function (MTF) of the electron beam portal images were computed using a quality assurance (QA) phantom and were found to be comparable to those of a 6 MV photon beam. Images were also acquired using a Rando anthropomorphic phantom. MC simulations were used to model the aS500 EPID and to obtain predicted portal images of the QA and Rando phantom. The contrast in simulated and measured portal images agrees within +/-5% for both the QA and the Rando phantom. The measured and simulated images allow for a verification of the phantom positioning by making sure that the structure edges are well aligned. This study suggests that the Varian aS500 portal imager can be used to obtain patient portal images of electron beams in the scattering foil linacs.     

IMRT verification by three-dimensional dose reconstruction from portal beam measurements

A method of reconstructing three-dimensional, in vivo dose distributions delivered by intensity-modulated radiotherapy (IMRT) is presented. A proof-of-principle experiment is described where an inverse-planned IMRT treatment is delivered to an anthropomorphic phantom. The exact position of the phantom at the time of treatment is measured by acquiring megavoltage CT data with the treatment beam and a research prototype, flat-panel, electronic portal imaging device. Immediately following CT imaging, the planned IMRT beams are delivered using the multiple-static field technique. The delivered fluence is sampled using the same detector as for the CT data. The signal measured by the portal imaging device is converted to primary fluence using an iterative phantom-scatter estimation technique. This primary fluence is back-projected through the previously acquired megavoltage CT model of the phantom, with inverse attenuation correction, to yield an input fluence map. The input fluence maps are used to calculate a "reconstructed" dose distribution using the same convolution/superposition algorithm as for the original planning dose calculation. Both relative and absolute dose reconstructions are shown. For the relative measurements, individual beam weights are taken from measurements but the total dose is normalized at the reference point. The absolute dose reconstructions do not use any dosimetric information from the original plan. Planned and reconstructed dose distributions are compared, with the reconstructed relative dose distribution also being compared to film measurements.     

Photon scatter in portal images: physical characteristics of pencil beam kernels generated using the EGS Monte Carlo code

Pencil beam kernels describing scattered photon fluence behind homogeneous water slabs at various air gap distances were generated using the EGS Monte Carlo code. Photon scatter fluence was scored in separate bins based on the particle's history: singly scattered, multiply scattered, and bremsstrahlung and positron annihilation photons. Simultaneously, the mean energy and mean angle with respect to the incident photon pencil beam were tallied. Kernels were generated for incident photon pencil beams exhibiting monoenergetic spectra of 2.0 and 10.0 MeV, and polyenergetic spectra representative of 6 and 24 MV beams. Reciprocity was used to generate scatter fractions on the central axis for various field sizes, phantom thicknesses, and air gaps. The scatter kernels were further characterized by full width at half-maximum estimates. Modulation transfer functions were calculated, providing theoretical estimates of the limit of performance of portal imaging systems due to the intrinsic scattering of photon radiation through the patient.     

Quantitative photoacoustic imaging: correcting for heterogeneous light fluence distributions using diffuse optical tomography

The specificity of molecular and functional photoacoustic (PA) images depends on the accuracy of the photoacoustic absorption spectroscopy. The PA signal is proportional to the product of the optical absorption coefficient and local light fluence; quantitative PA measurements of the optical absorption coefficient therefore require an accurate estimation of optical fluence. Light-modeling aided by diffuse optical tomography (DOT) can be used to map the required fluence and to reduce errors in traditional PA spectroscopic analysis. As a proof-of-concept, we designed a tissue-mimicking phantom to demonstrate how fluence-related artifacts in PA images can lead to misrepresentations of tissue properties. To correct for these inaccuracies, the internal fluence in the tissue phantom was estimated by using DOT to reconstruct spatial distributions of the absorption and reduced scattering coefficients of multiple targets within the phantom. The derived fluence map, which only consisted of low spatial frequency components, was used to correct PA images of the phantom. Once calibrated to a known absorber, this method reduced errors in estimated absorption coefficients from 33% to 6%. These results experimentally demonstrate that combining DOT with PA imaging can significantly reduce fluence-related errors in PA images, while producing quantitatively accurate, high-resolution images of the optical absorption coefficient.     

Patient-specific dosimetry of conventional and intensity modulated radiation therapy using a novel full Monte Carlo phase space reconstruction method from electronic portal images

Electronic portal imagers have promising dosimetric applications in external beam radiation therapy. In this study a patient dose computation algorithm based on Monte Carlo (MC) simulations and on portal images is developed and validated. The patient exit fluence from primary photons is obtained from the portal image after correction for scattered radiation. The scattered radiation at the portal imager and the spectral energy distribution of the primary photons are estimated from MC simulations at the treatment planning stage. The patient exit fluence and the spectral energy distribution of the primary photons are then used to ray-trace the photons from the portal image towards the source through the CT geometry of the patient. Photon weights which reflect the probability of a photon being transmitted are computed during this step. A dedicated MC code is used to transport back these photons from the source through the patient CT geometry to obtain patient dose. Only Compton interactions are considered. This code also produces a reconstructed portal image which is used as a verification tool to ensure that the dose reconstruction is reliable. The dose reconstruction algorithm is compared against MC dose calculation (MCDC) predictions and against measurements in phantom. The reconstructed absolute absorbed doses and the MCDC predictions in homogeneous and heterogeneous phantoms agree within 3% for simple open fields. Comparison with film-measured relative dose distributions for IMRT fields yields agreement within 3 mm, 5%. This novel dose reconstruction algorithm allows for daily patient-specific dosimetry and verification of patient movement.     

Photon scatter in portal images: accuracy of a fluence based pencil beam superposition algorithm

The accuracy of a pencil beam algorithm to predict scattered photon fluence into portal imaging systems was studied. A data base of pencil beam kernels describing scattered photon fluence behind homogeneous water slabs (1-50 cm thick) at various air gap distances (0-100 cm) was generated using the EGS Monte Carlo code. Scatter kernels were partitioned according to particle history: singly-scattered, multiply-scattered, and bremsstrahlung and positron annihilation photons. Mean energy and mean angle with respect to the incident photon pencil beam were also scored. This data allows fluence, mean energy, and mean angular data for each history type to be predicted using the pencil beam algorithm. Pencil beam algorithm predictions for 6 and 24 MV incident photon beams were compared against full Monte Carlo simulations for several inhomogeneous phantoms, including approximations to a lateral neck, and a mediastinum treatment. The accuracy of predicted scattered photon fluence, mean energy, and mean angle was investigated as a function of air gap, field size, photon history, incident beam resolution, and phantom geometry. Maximum errors in mean energies were 0.65 and 0.25 MeV for the higher and lower energy spectra, respectively, and 15 degrees for mean angles. The ability of the pencil beam algorithm to predict scatter fluence decreases with decreasing air gap, with the largest error for each phantom occurring at the exit surface. The maximum predictive error was found to be 6.9% with respect to the total fluence on the central axis. By maintaining even a small air gap (approximately 10 cm), the error in predicted scatter fluence may be kept under 3% for the phantoms and beam energies studied here. It is concluded that this pencil beam algorithm is sufficiently accurate (using International Commission on Radiation Units and Measurements Report No. 24 guidelines for absorbed dose) over the majority of clinically relevant air gaps, for further investigation in a portal dose prediction algorithm.     

Measurement of the source size of a 6- and 18-MV radiotherapy linac.

The quality of portal imaging is strongly affected by the source size of the radiotherapy machine. The effective source size of the dual-energy clinac 1800 (6 and 18 MV) was measured with a 50 microns wide and 120 mm long slit formed by two tungsten-copper alloy blocks. A series of slit images were obtained by translating the slit horizontally. The images were analyzed using a microdensitometer. The measured data was simulated using an analytical model of the source and its size was derived by a best-fit analysis. For both energies the FWHM was found to be 1.5 +/- 0.1 mm.     

Methods for fully automated verification of patient set-up in external beam radiotherapy with polygon shaped fields

Methods have been developed to perform fully automated comparisons of the radiotherapy simulator and portal images in the most common cases where the field edges consist of straight lines. The field defining wires or rods in the simulator image, and the edges of the portal image, are localized by means of a truncated Radon transform. Edges are enhanced with the use of a generalized Laplacian operator. From the detected field outlines, a geometrical mapping function is determined that rotates, scales, and translates one image with respect to the other. A subsequent match of field shapes is executed. If satisfactory agreement is established, the anatomical structures of the simulator and portal images are compared by correlating a new pair of images that only contain the intensity ridges representing bone outlines. The correlation determines the rotation and translation that must be applied to align the anatomical structures. With the simulator image already processed, the remaining automatic processing takes approximately 2 min on a SUN Sparcstation 2. By use of a priori knowledge of the fields, the computation time needed after acquisition of the portal image may be reduced to about 40 s for 512 x 512 images and about 12 s for 256 x 256 images.     

Real-time distortionless high-factor compression scheme

Nowadays, digital subtraction angiography systems must be able to sustain real-time acquisition (30 frames per second) of 512 x 512 x 8 bit images and store several sequences of such images on low cost and general-purpose mass memories. Concretely, that means a 7.8 Mbytes per second rate and about 780 Mbytes disk space to hold a 100-s cardiac examination. To fulfill these requirements at competitive cost, a distortionless compressor/decompressor system can be designed: during acquisition, the real-time compressor transforms the input images into a lower quantity of coded information through a predictive coder and a variable-length Huffman code. The process is fully reversible because during review, the real-time decompressor exactly recovers the acquired images from the stored compressed data. Test results on many raw images demonstrate that real-time compression is feasible and takes place with absolutely no loss of information. The designed system indifferently works on 512 or 1024 formats, and 256 or 1024 gray levels.     

Characterization of the susceptibility artifact around a prostate brachytherapy seed in MRI

Magnetic distortions surrounding a typical brachytherapy seed (IMC6711, OncoSeed) within a clinical magnetic resonance imager were modeled for a number of different seed orientations with respect to the main magnetic field. From these distortion maps, simulated images were produced. The simulated images were then compared to images experimentally acquired using a spin echo technique on a Philips 1.5 T magnetic resonance imaging scanner. The modeled images were found to conform very well to those acquired experimentally, thus allowing one to establish where the seed is positioned within the complex image distortion patterns. The artifact patterns were dependent on the orientation of the seed with the main magnetic field, as well as the direction of the read encode gradient. While all imaging schemes which employ a unidirectional linear read encode trajectory should produce the artifacts modeled in this article, sequences other than spin echo may produce additional artifacts. Gradient echo and steady-state free precession imaging techniques were also performed on the seed for comparison.