Correction of megavoltage cone-beam CT images for dose calculation in the head and neck region

Megavoltage cone-beam computed tomography (MVCBCT) imaging systems are now available for image-guided radiation therapy delivery and verification. In order to use the three-dimensional anatomical information for dose calculation, the MVCBCT image must provide accurate electron density. This work proposes a new method that has been developed to correct for the cupping and missing data artifacts seen on MVCBCT images of the head and neck region. It uses a conventional kilovoltage CT (kVCT) image as a reference for electron density and rigid registration with a MVCBCT image to obtain correction factors. Dose calculations performed on MVCBCT images corrected with the proposed method agree with calculations done on kVCT images within +/- 1% on phantoms. With patients images the agreement is within +/- 13% above the shoulders and +/- 5% below the shoulder line. This level of dose calculation accuracy allows the use of MVCBCT images for dose verification purposes.     

Patient dose considerations for routine megavoltage cone-beam CT imaging

Megavoltage cone-beam CT (MVCBCT), the recent addition to the family of in-room CT imaging systems for image-guided radiation therapy (IGRT), uses a conventional treatment unit equipped with a flat panel detector to obtain a three-dimensional representation of the patient in treatment position. MVCBCT has been used for more than two years in our clinic for anatomy verification and to improve patient alignment prior to dose delivery. The objective of this research is to evaluate the image acquisition dose delivered to patients for MVCBCT and to develop a simple method to reduce the additional dose resulting from routine MVCBCT imaging. Conventional CT scans of phantoms and patients were imported into a commercial treatment planning system (TPS: Phillips, Pinnacle) and an arc treatment mimicking the MVCBCT acquisition process was generated to compute the delivered acquisition dose. To validate the dose obtained from the TPS, a simple water-equivalent cylindrical phantom with spaces for MOSFETs and an ion chamber was used to measure the MVCBCT image acquisition dose. Absolute dose distributions were obtained by simulating MVCBCTs of 9 and 5 monitor units (MU) on pelvis and head and neck patients, respectively. A compensation factor was introduced to generate composite plans of treatment and MVCBCT imaging dose. The article provides a simple equation to compute the compensation factor. The developed imaging compensation method was tested on routinely used clinical plans for prostate and head and neck patients. The quantitative comparison between the calculated dose by the TPS and measurement points on the cylindrical phantom were all within 3%. The dose percentage difference for the ion chamber placed in the center of the phantom was only 0.2%. For a typical MVCBCT, the dose delivered to patients forms a small anterior-posterior gradient ranging from 0.6 to 1.2 cGy per MVCBCT MU. MVCBCT acquisitions in the pelvis and head and neck areas deliver slightly more dose than current portal imaging but render soft tissue information for positioning. Overall, the additional dose from daily 9 MU MVCBCTs of prostate patients is small compared to the treatment dose (<4%). Dose-volume histograms of compensated plans for pelvis and head and neck patients imaged daily with MVCBCT showed no additional dose to the target and small increases at low doses. The results indicate that the dose delivered for MVCBCT imaging can be precisely calculated in the TPS and therefore included in the treatment plan. This allows simple plan compensations, such as slightly reducing the treatment dose, to minimize the total dose received by critical structures from daily positioning with MVCBCT. The proposed compensation factor reduces the number of MU per treatment beam per fraction. Both the number of fractions and the beam arrangement are kept unchanged. Reducing the imaging volume in the cranio-caudal direction can further reduce the dose delivered for MVCBCT. This is a useful feature to eliminate the imaging dose to the eyes or to focus on a specific region of interest for alignment.     

The effect of temporal HU variations on the uncertainty of dose recalculations performed on MVCT images

Over the course of radiation therapy, a patient's anatomy may change substantially. The relatively recent addition of frequent in-room imaging to assist with patient localization has provided a database of images that may be used to recalculate dose distributions for adaptive radiotherapy purposes. The TomoTherapy Hi-Art II unit (Accuray Inc., Sunnyvale, CA, USA) uses a helical scanning geometry and a megavoltage (MV) beam to acquire volumetric patient images. This study evaluated the uncertainty of dose calculations performed on megavoltage CT (MVCT) images as a function of temporal Hounsfield Unit (HU) variations observed in the imaging system over three years on two machines. A baseline error between dose calculations performed on kVCT and MVCT images was established using a series of phantoms. This baseline error ranged from -1.4% to 0.6%. Materials of differing densities were imaged and MVCT numbers were measured periodically. The MVCT number of solid water varied from 5 to 103 HU and consistently increased prior to target replacement. Finally, the dosimetric uncertainty of the temporal HU variation was assessed using MVCT images of typical head and neck, lung and prostate cancer patients. Worst-case MVCT recalculation errors could approach 5%, 7% and 10% for the head and neck, lung and prostate images, respectively. However, if a tolerance of ±30 HU were maintained for the MVCT number of solid water, dosimetric errors were limited to ±2.5%, ±3% and ±4%, respectively.     

Deformable registration of the planning image (kVCT) and the daily images (MVCT) for adaptive radiation therapy

The incorporation of daily images into the radiotherapy process leads to adaptive radiation therapy (ART), in which the treatment is evaluated periodically and the plan is adaptively modified for the remaining course of radiotherapy. Deformable registration between the planning image and the daily images is a key component of ART. In this paper, we report our researches on deformable registration between the planning kVCT and the daily MVCT image sets. The method is based on a fast intensity-based free-form deformable registration technique. Considering the noise and contrast resolution differences between the kVCT and the MVCT, an 'edge-preserving smoothing' is applied to the MVCT image prior to the deformable registration process. We retrospectively studied daily MVCT images from commercial TomoTherapy machines from different clinical centers. The data set includes five head-neck cases, one pelvis case, two lung cases and one prostate case. Each case has one kVCT image and 20-40 MVCT images. We registered the MVCT images with their corresponding kVCT image. The similarity measures and visual inspections of contour matches by physicians validated this technique. The applications of deformable registration in ART, including 'deformable dose accumulation', 'automatic re-contouring' and 'tumour growth/regression evaluation' throughout the course of radiotherapy are also studied.     

A megavoltage scatter correction technique for cone-beam CT images acquired during VMAT delivery

Kilovoltage cone-beam CT (kV CBCT) can be acquired during the delivery of volumetric modulated arc therapy (VMAT), in order to obtain an image of the patient during treatment. However, the quality of such CBCTs is degraded by megavoltage (MV) scatter from the treatment beam onto the imaging panel. The objective of this paper is to introduce a novel MV scatter correction method for simultaneous CBCT during VMAT, and to investigate its effectiveness when compared to other techniques. The correction requires the acquisition of a separate set of images taken during VMAT delivery, while the kV beam is off. These images--which contain only the MV scatter contribution on the imaging panel--are then used to correct the corresponding kV/MV projections. To test this method, CBCTs were taken of an image quality phantom during VMAT delivery and measurements of contrast to noise ratio were made. Additionally, the correction was applied to the datasets of three VMAT prostate patients, who also received simultaneous CBCTs. The clinical image quality was assessed using a validated scoring system, comparing standard CBCTs to the uncorrected simultaneous CBCTs and a variety of correction methods. Results show that the correction is able to recover some of the low and high-contrast signal to noise ratio lost due to MV scatter. From the patient study, the corrected CBCT scored significantly higher than the uncorrected images in terms of the ability to identify the boundary between the prostate and surrounding soft tissue. In summary, a simple MV scatter correction method has been developed and, using both phantom and patient data, is shown to improve the image quality of simultaneous CBCTs taken during VMAT delivery.     

Investigation of Kodak extended dose range (EDR) film for megavoltage photon beam dosimetry

We have investigated the dependence of the measured optical density on the incident beam energy, field size and depth for a new type of film, Kodak extended dose range (Kodak EDR). Film measurements have been conducted over a range of field sizes (3 x 3 cm2 to 25 x 25 cm2) and depths (d(max) to 15 cm), for 6 MV and 15 MV photons within a solid water phantom, and the variation in sensitometric response (net optical density versus dose) has been reported. Kodak EDR film is found to have a linear response with dose, from 0 to 350 cGy, which is much higher than that typically seen for Kodak XV film (0-50 cGy). The variation in sensitometric response for Kodak EDR film as a function of field size and depth is observed to be similar to that of Kodak XV film; the optical density varied in the order of 2-3% for field sizes of 3 x 3 cm2 and 10 x 10 cm2 at depths of d(max), 5 cm and 15 cm in the phantom. Measurements for a 25 x 25 cm2 field size showed consistently higher optical densities at depths of d(max), 5 cm and 15 cm, relative to a 10 x 10 cm2 field size at 5 cm depth, with 4-5% differences noted at a depth of 15 cm. Fractional depth dose and profiles conducted with Kodak EDR film showed good agreement (2%/2 mm) with ion chamber measurements for all field sizes except for the 25 x 25 cm2 at depths greater than 15 cm, where differences in the order of 3-5% were observed. In addition, Kodak EDR film measurements were found to be consistent with those of Kodak XV film for all fractional depth doses and profiles. The results of this study indicate that Kodak EDR film may be a useful tool for relative dosimetry at higher dose ranges.     

Modeling scintillator-photodiodes as detectors for megavoltage CT

The use of cadmium tungstate (CdWO4) and cesium iodide [CsI(Tl)] scintillation detectors is studied in megavoltage computed tomography (MVCT). A model describing the signal acquired from a scintillation detector has been developed which contains two steps: (1) the calculation of the energy deposited in the crystal due to MeV photons using the EGSnrc Monte Carlo code; and (2) the transport of the optical photons generated in the crystal voxels to photodiodes using the optical Monte Carlo code DETECT2000. The measured detector signals in single CdWO4 and CsI(Tl) scintillation crystals of base 0.275 x 0.8 cm2 and heights 0.4, 1, 1.2, 1.6 and 2 cm were, generally, in good agreement with the signals calculated with the model. A prototype detector array which contains 8 CdWO4 crystals, each 0.275 x 0.8 x 1 cm3, in contact with a 16-element array of photodiodes was built. The measured attenuation of a Cobalt-60 beam as a function of solid water thickness behaves linearly. The frequency dependent modulation transfer function [MTF(f)], noise power spectrum [NPS(f)], and detective quantum efficiency [DQE(f)] were measured for 1.25 MeV photons (in a Cobalt-60 beam). For 6 MV photons, only the MTF(f) was measured from a linear accelerator, where large pulse-to-pulse fluctuations in the output of the linear accelerator did not allow the measurement of the NPS(f). A two-step Monte Carlo simulation was used to model the detector's MTF(f), NPS(f) and DQE(f). The DQE(0) of the detector array was found to be 26% and 19% for 1.25 MeV and 6 MV photons, respectively. For 1.25 MeV photons, the maximum discrepancies between the measured and modeled MTF(f), relative NPS(f) and the DQE(f) were found to be 1.5%, 1.2%, and 1.9%, respectively. For the 6 MV beam, the maximum discrepancy between the modeled and the measured MTF(f) was found to be 2.5%. The modeling is sufficiently accurate for designing appropriate detectors for MVCT.     

Calibration of megavoltage cone-beam CT for radiotherapy dose calculations: correction of cupping artifacts and conversion of CT numbers to electron density

Megavoltage cone-beam CT (MV CBCT) is used for three-dimensional imaging of the patient anatomy on the treatment table prior to or just after radiotherapy treatment. To use MV CBCT images for radiotherapy dose calculation purposes, reliable electron density (ED) distributions are needed. Patient scatter, beam hardening and softening effects result in cupping artifacts in MV CBCT images and distort the CT number to ED conversion. A method based on transmission images is presented to correct for these effects without using prior knowledge of the object's geometry. The scatter distribution originating from the patient is calculated with pencil beam scatter kernels that are fitted based on transmission measurements. The radiological thickness is extracted from the scatter subtracted transmission images and is then converted to the primary transmission used in the cone-beam reconstruction. These corrections are performed in an iterative manner, without using prior knowledge regarding the geometry and composition of the object. The method was tested using various homogeneous and inhomogeneous phantoms with varying shapes and compositions, including a phantom with different electron density inserts, phantoms with large density variations, and an anthropomorphic head phantom. For all phantoms, the cupping artifact was substantially removed from the images and a linear relation between the CT number and electron density was found. After correction the deviations in reconstructed ED from the true values were reduced from up to 0.30 ED units to 0.03 for the majority of the phantoms; the residual difference is equal to the amount of noise in the images. The ED distributions were evaluated in terms of absolute dose calculation accuracy for homogeneous cylinders of different size; errors decreased from 7% to below 1% in the center of the objects for the uncorrected and corrected images, respectively, and maximum differences were reduced from 17% to 2%, respectively. The presented method corrects the MV CBCT images for cupping artifacts and extracts reliable ED information of objects with varying geometries and composition, making these corrected MV CBCT images suitable for accurate dose calculation purposes.     

A megavoltage CT scanner for radiotherapy verification.

We have further developed a system for generating megavoltage CT images immediately prior to the administration of external beam radiotherapy. The detector is based on the scanner of Simpson (Simpson et al 1982)--the major differences being a significant reduction in dose required for image formation, faster image formation and greater convenience of use in the clinical setting. Attention has been paid to the problem of ring artefacts in the images. Specifically, a Fourier-space filter has been applied to the sinogram data. After suitable detector calibration, it has been shown that the device operates close to its theoretical specification of 3 mm spatial resolution and a few percent contrast resolution. Ring artefacts continue to be a major source of image degradation. A number of clinical images have been presented. The next stage of this work is to use the system to make clinical measurements of patient set-up inaccuracies building on our work making such measurements from digital portal images (Evans et al 1992).     

Evaluation of uncertainty predictions and dose output for model-based dose calculations for megavoltage photon beams

In many radiotherapy clinics an independent verification of the number of monitor units (MU) used to deliver the prescribed dose to the target volume is performed prior to the treatment start. Traditionally this has been done by using methods mainly based on empirical factors which, at least to some extent, try to separate the influence from input parameters such as field size, depth, distance, etc. The growing complexity of modern treatment techniques does however make this approach increasingly difficult, both in terms of practical application and in terms of the reliability of the results. In the present work the performance of a model-based approach, describing the influence from different input parameters through actual modeling of the physical effects, has been investigated in detail. The investigated model is based on two components related to megavoltage photon beams; one describing the exiting energy fluence per delivered MU, and a second component describing the dose deposition through a pencil kernel algorithm solely based on a measured beam quality index. Together with the output calculations, the basis of a method aiming to predict the inherent calculation uncertainties in individual treatment setups has been developed. This has all emerged from the intention of creating a clinical dose/MU verification tool that requires an absolute minimum of commissioned input data. This evaluation was focused on irregular field shapes and performed through comparison with output factors measured at 5, 10, and 20 cm depth in ten multileaf collimated fields on four different linear accelerators with varying multileaf collimator designs. The measurements were performed both in air and in water and the results of the two components of the model were evaluated separately and combined. When compared with the corresponding measurements the resulting deviations in the calculated output factors were in most cases smaller than 1% and in all cases smaller than 1.7%. The distribution describing the calculation errors in the total dose output has a mean value of -0.04% and a standard deviation of 0.47%. In the dose calculations a previously developed correction of the pencil kernel was applied that managed to contract the error distribution considerably. A detailed analysis of the predicted uncertainties versus the observed deviations suggests that the predictions indeed can be used as a basis for creating action levels and tracking dose calculation errors in homogeneous media.     

A biologically-based algorithm for companding computerized tomography (CT) images

Computerized Tomography (CT) images are High Dynamic Range (HDR) images of the X-ray attenuation coefficients of the body's tissues. The inability to see abnormalities in tissues with marked differences in their X-ray attenuation coefficients, in a single CT window, poses a significant clinical problem in radiology. In order to provide proper contrast, which reveals all the required clinical details within each specifically imaged tissue, a single CT slice must be viewed by a radiologist four times: the first viewing focuses on the lung window; the second viewing focuses on the soft tissues window; the third viewing focuses on the liver window; and the fourth viewing focuses on the bone window. In order to enhance the ability to perform a complete diagnosis, while decreasing diagnostic time, we developed the BACCT (Biologically-based Algorithm for Companding CT images) method. Our algorithm compresses and expands (compands) the HDR CT image into a single, low dynamic range image. Before performing the companding procedure, unique processing is required which involves operations that enhance and stretch the image. The performance of our algorithm has been demonstrated on a large repertoire of CT body images. All the clinically required CT information is exposed in each CT slice in a single image. The algorithm compands the CT images in a fully automatic way. Collaborating radiologists have already tested the results of our algorithmic method, and reported that the images seem to provide all the necessary information. However, clinical tests for statistical reliability are still required.     

Development of a 30-week-pregnant female tomographic model from computed tomography (CT) images for Monte Carlo organ dose calculations

Assessment of radiation dose and risk to a pregnant woman and her fetus is an important task in radiation protection. Although tomographic models for male and female patients of different ages have been developed using medical images, such models for pregnant women had not been developed to date. This paper reports the construction of a partial-body model of a pregnant woman from a set of computed tomography (CT) images. The patient was 30 weeks into pregnancy, and the CT scan covered the portion of the body from above liver to below pubic symphysis in 70 slices. The thickness for each slice is 7 mm, and the image resolution is 512x512 pixels in a 48 cm x 48 cm field; thus, the voxel size is 6.15 mm3. The images were segmented to identify 34 major internal organs and tissues considered sensitive to radiation. Even though the masses are noticeably different from other models, the three-dimensional visualization verified the segmentation and its suitability for Monte Carlo calculations. The model has been implemented into a Monte Carlo code, EGS4-VLSI (very large segmented images), for the calculations of radiation dose to a pregnant woman. The specific absorbed fraction (SAF) results for internal photons were compared with those from a stylized model. Small and large differences were found, and the differences can be explained by mass differences and by the relative geometry differences between the source and the target organs. The research provides the radiation dosimetry community with the first voxelized tomographic model of a pregnant woman, opening the door to future dosimetry studies.     

Surface dose for megavoltage photon beams outside the treatment field

Measurements made on photon beams from four different radiotherapy machines have demonstrated that skin dose several centimeters outside the boundary of a treatment field may be as much as 20% of the central axis maximum dose. This surface dose has been measured for an AECL Theratron 80, Siemens Mevatron VI, Varian Clinac 20, and CGR Sagittaire for distances up to 12 cm outside the field boundary and for depths up to the depth of maximum central axis dose. This dose has also been measured as a function of field size and of source-to-skin distance. For the lower energy photon beams, this radiation is significantly attenuated in the first 2-3 mm of tissue, while for higher energy beams, a buildup phenomenon with a dmax of 2-3 mm is observed. The magnitude of this radiation is approximately linearly dependent upon field dimension for all energies.     

Correction of scatter in megavoltage cone-beam CT

The role of scatter in a cone-beam computed tomography system using the therapeutic beam of a medical linear accelerator and a commercial electronic portal imaging device (EPID) is investigated. A scatter correction method is presented which is based on a superposition of Monte Carlo generated scatter kernels. The kernels are adapted to both the spectral response of the EPID and the dimensions of the phantom being scanned. The method is part of a calibration procedure which converts the measured transmission data acquired for each projection angle into water-equivalent thicknesses. Tomographic reconstruction of the projections then yields an estimate of the electron density distribution of the phantom. It is found that scatter produces cupping artefacts in the reconstructed tomograms. Furthermore, reconstructed electron densities deviate greatly (by about 30%) from their expected values. The scatter correction method removes the cupping artefacts and decreases the deviations from 30% down to about 8%.     

A preliminary investigation of local tomography for megavoltage CT imaging

We investigate the problem of reconstructing a two-dimensional (2-D) cross-sectional image of a tumor volume from a set of truncated MV projections that are produced by radiation therapy treatment beams. Our proposed approach is conceptually distinct from previously investigated approaches in that it utilizes a noniterative local tomography reconstruction algorithm. A local tomography reconstruction algorithm is implemented and systematically investigated using several sets of simulated and experimental MV projection data. We demonstrate that the conventional (nonlocal) filtered backprojection reconstruction algorithm cannot, in general, accurately reconstruct the edges and boundaries of low-contrast features from truncated MV projection data. We demonstrate that the local tomography algorithm is not adversely affected by projection truncation and can reconstruct accurately the boundaries of low-contrast structures within the region of interest from truncated MV projections.     

Normalized glandular dose (DgN) coefficients for flat-panel CT breast imaging

The development of new digital mammography techniques such as dual-energy imaging, tomosynthesis and CT breast imaging will require investigation of optimal camera design parameters and optimal imaging acquisition parameters. In optimizing these acquisition protocols and imaging systems it is important to have knowledge of the radiation dose to the breast. This study presents a methodology for estimating the normalized glandular dose to the uncompressed breast using the geometry proposed for flat-panel CT breast imaging. The simulation uses the GEANT 3 Monte Carlo code to model x-ray transport and absorption within the breast phantom. The Monte Carlo software was validated for breast dosimetry by comparing results of the normalized glandular dose (DgN) values of the compressed breast to those reported in the literature. The normalized glandular dose was then estimated for a range of breast diameters from 10 cm to 18 cm using an uncompressed breast model with a homogeneous composition of adipose and glandular tissue, and for monoenergetic x-rays from 10 keV to 120 keV. These data were fit providing expressions for the normalized glandular dose. Using these expressions for the DgN coefficients and input variables such as the diameter, height and composition of the breast phantom, the mean glandular dose for any spectra can be estimated. A computer program to provide normalized glandular dose values has been made available online. In addition, figures displaying energy deposition maps are presented to better understand the spatial distribution of dose in CT breast imaging.     

Artificial neural network modelling of megavoltage photon dose distributions

An artificial neural network (NN) has been used to model the two-dimensional dose distributions from a Varian 2100C linac. The network was trained using depth dose data for 6 and 10 MV x-rays, collected during the linac commissioning phase. During training, the number of iterations and hidden nodes was adjusted manually until acceptable agreement between measured and predicted data was obtained. In order to validate the network a subset of the data was set aside and not used for training. This enabled the performance of the network to be investigated in terms of generalization and accuracy, together with its ability to interpolate between different field sizes and positions in the beam. Finally, the network was used to generate data points over a 2D grid so that isodose distributions could be visualized. Good agreement was found between measured data and that produced by the trained neural network.