Review and recommendations on deformable image registration uncertainties for radiotherapy applications.

Deformable image registration (DIR) is a versatile tool used in many applications in radiotherapy (RT). DIR algorithms have been implemented in many commercial treatment planning systems providing accessible and easy-to-use solutions. However, the geometric uncertainty of DIR can be large and difficult to quantify, resulting in barriers to clinical practice. Currently, there is no agreement in the RT community on how to quantify these uncertainties and determine thresholds that distinguish a good DIR result from a poor one. This review summarises the current literature on sources of DIR uncertainties and their impact on RT applications. Recommendations are provided on how to handle these uncertainties for patient-specific use, commissioning, and research. Recommendations are also provided for developers and vendors to help users to understand DIR uncertainties and make the application of DIR in RT safer and more reliable.

Head and neck cancer patient images for determining auto-segmentation accuracy in T2-weighted magnetic resonance imaging through expert manual segmentations.

PURPOSE: The use of magnetic resonance imaging (MRI) in radiotherapy treatment planning has rapidly increased due to its ability to evaluate patient's anatomy without the use of ionizing radiation and due to its high soft tissue contrast. For these reasons, MRI has become the modality of choice for longitudinal and adaptive treatment studies. Automatic segmentation could offer many benefits for these studies. In this work, we describe a T2-weighted MRI dataset of head and neck cancer patients that can be used to evaluate the accuracy of head and neck normal tissue auto-segmentation systems through comparisons to available expert manual segmentations. ACQUISITION AND VALIDATION METHODS: T2-weighted MRI images were acquired for 55 head and neck cancer patients. These scans were collected after radiotherapy computed tomography (CT) simulation scans using a thermoplastic mask to replicate patient treatment position. All scans were acquired on a single 1.5 T Siemens MAGNETOM Aera MRI with two large four-channel flex phased-array coils. The scans covered the region encompassing the nasopharynx region cranially and supraclavicular lymph node region caudally, when possible, in the superior-inferior direction. Manual contours were created for the left/right submandibular gland, left/right parotids, left/right lymph node level II, and left/right lymph node level III. These contours underwent quality assurance to ensure adherence to predefined guidelines, and were corrected if edits were necessary. DATA FORMAT AND USAGE NOTES: The T2-weighted images and RTSTRUCT files are available in DICOM format. The regions of interest are named based on AAPM's Task Group 263 nomenclature recommendations (Glnd_Submand_L, Glnd_Submand_R, LN_Neck_II_L, Parotid_L, Parotid_R, LN_Neck_II_R, LN_Neck_III_L, LN_Neck_III_R). This dataset is available on The Cancer Imaging Archive (TCIA) by the National Cancer Institute under the collection "AAPM RT-MAC Grand Challenge 2019" (https://doi.org/10.7937/tcia.2019.bcfjqfqb). POTENTIAL APPLICATIONS: This dataset provides head and neck patient MRI scans to evaluate auto-segmentation systems on T2-weighted images. Additional anatomies could be provided at a later time to enhance the existing library of contours.

CT images with expert manual contours of thoracic cancer for benchmarking auto-segmentation accuracy.

PURPOSE: Automatic segmentation offers many benefits for radiotherapy treatment planning; however, the lack of publicly available benchmark datasets limits the clinical use of automatic segmentation. In this work, we present a well-curated computed tomography (CT) dataset of high-quality manually drawn contours from patients with thoracic cancer that can be used to evaluate the accuracy of thoracic normal tissue auto-segmentation systems. ACQUISITION AND VALIDATION METHODS: Computed tomography scans of 60 patients undergoing treatment simulation for thoracic radiotherapy were acquired from three institutions: MD Anderson Cancer Center, Memorial Sloan Kettering Cancer Center, and the MAASTRO clinic. Each institution provided CT scans from 20 patients, including mean intensity projection four-dimensional CT (4D CT), exhale phase (4D CT), or free-breathing CT scans depending on their clinical practice. All CT scans covered the entire thoracic region with a 50-cm field of view and slice spacing of 1, 2.5, or 3 mm. Manual contours of left/right lungs, esophagus, heart, and spinal cord were retrieved from the clinical treatment plans. These contours were checked for quality and edited if necessary to ensure adherence to RTOG 1106 contouring guidelines. DATA FORMAT AND USAGE NOTES: The CT images and RTSTRUCT files are available in DICOM format. The regions of interest were named according to the nomenclature recommended by American Association of Physicists in Medicine Task Group 263 as Lung_L, Lung_R, Esophagus, Heart, and SpinalCord. This dataset is available on The Cancer Imaging Archive (funded by the National Cancer Institute) under Lung CT Segmentation Challenge 2017 (http://doi.org/10.7937/K9/TCIA.2017.3r3fvz08). POTENTIAL APPLICATIONS: This dataset provides CT scans with well-delineated manually drawn contours from patients with thoracic cancer that can be used to evaluate auto-segmentation systems. Additional anatomies could be supplied in the future to enhance the existing library of contours.

Automatic diaphragm segmentation for real-time lung tumor tracking on cone-beam CT projections: a convolutional neural network approach.

PURPOSE: To automatically segment the diaphragm on individual lung cone-beam CT projection images, to enable real-time tracking of lung tumors using kilovoltage imaging. METHODS: The deep neural network Mask R-CNN was trained on 3500 raw cone-beam CT projection images from 10 lung cancer patients, with the diaphragm manually segmented on each image used as a ground truth label. Ground-truth breathing traces were extracted from each patient for both diaphragm hemispheres, and apex positions were compared against the predicted output of the neural network. Ten-fold cross-validation was used to evaluate the segmentation accuracy. RESULTS: The mean diaphragm apex prediction error was 4.4 mm. The mean percentage of projection images for which a successful prediction could me made was 87.3%. Prediction accuracy at some lateral gantry angles was worse due to overlap between diaphragm hemispheres, and the increased amount of fatty tissue. CONCLUSIONS: The neural network was able to track the diaphragm apex position successfully. This allows accurate assessment of the breathing phase, which can be used to estimate the position of the lung tumor in real time.

Efficient Descriptor-Based Segmentation of Parotid Glands With Nonlocal Means.

OBJECTIVE: We introduce descriptor-based segmentation that extends existing patch-based methods by combining intensities, features, and location information. Since it is unclear which image features are best suited for patch selection, we perform a broad empirical study on a multitude of different features. METHODS: We extend nonlocal means segmentation by including image features and location information. We search larger windows with an efficient nearest neighbor search based on kd-trees. We compare a large number of image features. RESULTS: The best results were obtained for entropy image features, which have not yet been used for patch-based segmentation. We further show that searching larger image regions with an approximate nearest neighbor search and location information yields a significant improvement over the bounded nearest neighbor search traditionally employed in patch-based segmentation methods. CONCLUSION: Features and location information significantly increase the segmentation accuracy. The best features highlight boundaries in the image. SIGNIFICANCE: Our detailed analysis of several aspects of nonlocal means-based segmentation yields new insights about patch and neighborhood sizes together with the inclusion of location information. The presented approach advances the state-of-the-art in the segmentation of parotid glands for radiation therapy planning.

Fractionated Lung IMPT Treatments: Sensitivity to Setup Uncertainties and Motion Effects Based on Single-Field Homogeneity.

Treatment uncertainties in radiotherapy are either systematic or random. This study evaluates the sensitivity of fractionated intensity-modulated proton therapy (IMPT) lung treatments to systematic and random setup uncertainties. Treatments in which single-field homogeneity was restricted to within ±20% (IMPT20%) were compared to full IMPT (IMPTfull) for 10 patients with lung cancer. Four-dimensional Monte Carlo calculations were performed using patient computed tomography geometries with ±5 mm systematic or random setup uncertainties applied over a 35 × 2.5Gy(RBE) treatment course. Fifty fractionated courses were simulated for each patient using both IMPT delivery methods with random setup uncertainties applied each fraction and for 3 energy-dependent spot sizes (big spots, σ≈18-9 mm; intermediate spots, σ≈11-5 mm; and small spots, σ≈4-2 mm). These results were compared to Monte Carlo recalculations of the original treatment plan assuming zero setup uncertainty. Results are presented as the difference in equivalent uniform dose (ΔEUD), V95 (ΔV95), and target dose homogeneity (ΔD1-D99). Over the whole patient cohort, the ΔEUD was 2.0 ± 0.5 (big spots), 1.9 ± 0.7 (intermediate spots), and 1.3 ± 0.4 (small spots) times more sensitive to ±5 mm systematic setup uncertainties in IMPTfull compared to IMPT20% IMPTfull is 1.9 ± 0.9 (big spots), 2.1 ± 1.1 (intermediate spots), and 1.5 ± 0.6 (small spots) times more sensitive to random setup uncertainties than IMPT20% over a fractionated treatment course. The ΔV95 is at least 1.4 times more sensitive to systematic and random setup uncertainties for IMPTfull for all spot sizes considered. The ΔD1-D99 values coincided within uncertainty limits for both IMPT delivery methods for the 3 spot sizes considered, with higher mean values always observed for IMPTfull The paired t-test indicated that variations observed between IMPTfull and IMPT20% were significantly different for the majority of scenarios. Significantly larger variations were observed in ΔEUD and ΔV95 in IMPTfull lung treatments in addition to higher mean ΔD1-D99 The steep intra-target dose gradients in IMPTfull make it more susceptible to systematic and random setup uncertainties.

Contour-Driven Atlas-Based Segmentation.

We propose new methods for automatic segmentation of images based on an atlas of manually labeled scans and contours in the image. First, we introduce a Bayesian framework for creating initial label maps from manually annotated training images. Within this framework, we model various registration- and patch-based segmentation techniques by changing the deformation field prior. Second, we perform contour-driven regression on the created label maps to refine the segmentation. Image contours and image parcellations give rise to non-stationary kernel functions that model the relationship between image locations. Setting the kernel to the covariance function in a Gaussian process establishes a distribution over label maps supported by image structures. Maximum a posteriori estimation of the distribution over label maps conditioned on the outcome of the atlas-based segmentation yields the refined segmentation. We evaluate the segmentation in two clinical applications: the segmentation of parotid glands in head and neck CT scans and the segmentation of the left atrium in cardiac MR angiography images.

Motion mitigation for lung cancer patients treated with active scanning proton therapy.

PURPOSE: Motion interplay can affect the tumor dose in scanned proton beam therapy. This study assesses the ability of rescanning and gating to mitigate interplay effects during lung treatments. METHODS: The treatments of five lung cancer patients [48 Gy(RBE)/4fx] with varying tumor size (21.1-82.3 cm(3)) and motion amplitude (2.9-30.6 mm) were simulated employing 4D Monte Carlo. The authors investigated two spot sizes (σ ∼ 12 and ∼ 3 mm), three rescanning techniques (layered, volumetric, breath-sampled volumetric) and respiratory gating with a 30% duty cycle. RESULTS: For 4/5 patients, layered rescanning 6/2 times (for the small/large spot size) maintains equivalent uniform dose within the target >98% for a single fraction. Breath sampling the timing of rescanning is ∼ 2 times more effective than the same number of continuous rescans. Volumetric rescanning is sensitive to synchronization effects, which was observed in 3/5 patients, though not for layered rescanning. For the large spot size, rescanning compared favorably with gating in terms of time requirements, i.e., 2x-rescanning is on average a factor ∼ 2.6 faster than gating for this scenario. For the small spot size however, 6x-rescanning takes on average 65% longer compared to gating. Rescanning has no effect on normal lung V20 and mean lung dose (MLD), though it reduces the maximum lung dose by on average 6.9 ± 2.4/16.7 ± 12.2 Gy(RBE) for the large and small spot sizes, respectively. Gating leads to a similar reduction in maximum dose and additionally reduces V20 and MLD. Breath-sampled rescanning is most successful in reducing the maximum dose to the normal lung. CONCLUSIONS: Both rescanning (2-6 times, depending on the beam size) as well as gating was able to mitigate interplay effects in the target for 4/5 patients studied. Layered rescanning is superior to volumetric rescanning, as the latter suffers from synchronization effects in 3/5 patients studied. Gating minimizes the irradiated volume of normal lung more efficiently, while breath-sampled rescanning is superior in reducing maximum doses to organs at risk.

Quantification of proton dose calculation accuracy in the lung.

PURPOSE: To quantify the accuracy of a clinical proton treatment planning system (TPS) as well as Monte Carlo (MC)-based dose calculation through measurements and to assess the clinical impact in a cohort of patients with tumors located in the lung. METHODS AND MATERIALS: A lung phantom and ion chamber array were used to measure the dose to a plane through a tumor embedded in the lung, and to determine the distal fall-off of the proton beam. Results were compared with TPS and MC calculations. Dose distributions in 19 patients (54 fields total) were simulated using MC and compared to the TPS algorithm. RESULTS: MC increased dose calculation accuracy in lung tissue compared with the TPS and reproduced dose measurements in the target to within ±2%. The average difference between measured and predicted dose in a plane through the center of the target was 5.6% for the TPS and 1.6% for MC. MC recalculations in patients showed a mean dose to the clinical target volume on average 3.4% lower than the TPS, exceeding 5% for small fields. For large tumors, MC also predicted consistently higher V5 and V10 to the normal lung, because of a wider lateral penumbra, which was also observed experimentally. Critical structures located distal to the target could show large deviations, although this effect was highly patient specific. Range measurements showed that MC can reduce range uncertainty by a factor of ~2: the average (maximum) difference to the measured range was 3.9 mm (7.5 mm) for MC and 7 mm (17 mm) for the TPS in lung tissue. CONCLUSION: Integration of Monte Carlo dose calculation techniques into the clinic would improve treatment quality in proton therapy for lung cancer by avoiding systematic overestimation of target dose and underestimation of dose to normal lung. In addition, the ability to confidently reduce range margins would benefit all patients by potentially lowering toxicity.

Motion interplay as a function of patient parameters and spot size in spot scanning proton therapy for lung cancer.

PURPOSE: To quantify the impact of respiratory motion on the treatment of lung tumors with spot scanning proton therapy. METHODS AND MATERIALS: Four-dimensional Monte Carlo simulations were used to assess the interplay effect, which results from relative motion of the tumor and the proton beam, on the dose distribution in the patient. Ten patients with varying tumor sizes (2.6-82.3 cc) and motion amplitudes (3-30 mm) were included in the study. We investigated the impact of the spot size, which varies between proton facilities, and studied single fractions and conventionally fractionated treatments. The following metrics were used in the analysis: minimum/maximum/mean dose, target dose homogeneity, and 2-year local control rate (2y-LC). RESULTS: Respiratory motion reduces the target dose homogeneity, with the largest effects observed for the highest motion amplitudes. Smaller spot sizes (σ ≈ 3 mm) are inherently more sensitive to motion, decreasing target dose homogeneity on average by a factor 2.8 compared with a larger spot size (σ ≈ 13 mm). Using a smaller spot size to treat a tumor with 30-mm motion amplitude reduces the minimum dose to 44.7% of the prescribed dose, decreasing modeled 2y-LC from 87.0% to 2.7%, assuming a single fraction. Conventional fractionation partly mitigates this reduction, yielding a 2y-LC of 71.6%. For the large spot size, conventional fractionation increases target dose homogeneity and prevents a deterioration of 2y-LC for all patients. No correlation with tumor volume is observed. The effect on the normal lung dose distribution is minimal: observed changes in mean lung dose and lung V20 are <0.6 Gy(RBE) and <1.7%, respectively. CONCLUSIONS: For the patients in this study, 2y-LC could be preserved in the presence of interplay using a large spot size and conventional fractionation. For treatments using smaller spot sizes and/or in the delivery of single fractions, interplay effects can lead to significant deterioration of the dose distribution and lower 2y-LC.

Evaluation of a general practitioner referral service for manual treatment of back and neck pain.

AIM: To describe and evaluate a community-based musculoskeletal service, commissioned by National Health Services North East Essex Primary Care Trust (PCT), in terms of patient-reported outcomes and satisfaction. BACKGROUND: Persistent musculoskeletal conditions, including back and neck pain, are costly in terms of primary and secondary healthcare resources. Most patients are assessed and managed by general practitioners (GPs), with referral when necessary to secondary care services. METHOD: Patients consulting for at least four weeks for back or neck pain were referred by their GP according to patient preference to either a chiropractor or osteopath or physiotherapist working in the independent sector. Patients completed questionnaires at baseline and at discharge from the service. RESULTS: Questionnaire data were obtained from 696 patients, 97% of whom were seen within two weeks. About half (51%) had had their pain for less than three months, and of the remainder 49% for more than 12 months. Patients received on average six treatments. Using the Bournemouth Questionnaire, the Bothersomeness scale and the Global Improvement Scale, approximately two-thirds (64.6%, 67.8% and 69.9%, respectively) reported improvement at discharge, and approximately 65% a significant reduction in medication. Almost all (99.5%) patients were satisfied with the service. Similarly, almost all (97%) patients were discharged from the service with advice on self-management; the remainder were recommended for secondary care referral. CONCLUSION: This service improved patient access and choice resulting in shorter waiting times and effective outcomes. An impact analysis of the first 12 months of the service by the PCT showed a reduction in primary care consultations and in inappropriate referrals to secondary care.

A fast double template convolution isocenter evaluation algorithm with subpixel accuracy.

PURPOSE: To design a fast Winston Lutz (fWL) algorithm for accurate analysis of radiation isocenter from images without edge detection or center of mass calculations. METHODS: An algorithm has been developed to implement the Winston Lutz test for mechanical/ radiation isocenter agreement using an electronic portal imaging device (EPID). The algorithm detects the position of the radiation shadow of a tungsten ball within a stereotactic cone. The fWL algorithm employs a double convolution to independently find the position of the sphere and cone centers. Subpixel estimation is used to achieve high accuracy. Results of the algorithm were compared to (1) a human observer with template guidance and (2) an edge detection/center of mass (edCOM) algorithm. Testing was performed with high resolution (0.05 mm/px, film) and low resolution (0.78 mm/px, EPID) image sets. RESULTS: Sphere and cone center relative positions were calculated with the fWL algorithm for high resolution test images with an accuracy of 0.002 +/- 0.061 mm compared to 0.042 +/- 0.294 mm for the human observer, and 0.003 +/- 0.038 mm for the edCOM algorithm. The fWL algorithm required 0.01 s per image compared to 5 s for the edCOM algorithm and 20 s for the human observer. For lower resolution images the fWL algorithm localized the centers with an accuracy of 0.083 +/- 0.12 mm compared to 0.03 +/- 0.5514 mm for the edCOM algorithm. CONCLUSIONS: A fast (subsecond) subpixel algorithm has been developed that can accurately determine the center locations of the ball and cone in Winston Lutz test images without edge detection or COM calculations.

Dosimetric impact of motion in free-breathing and gated lung radiotherapy: a 4D Monte Carlo study of intrafraction and interfraction effects.

The purpose of this study was to investigate if interfraction and intrafraction motion in free-breathing and gated lung IMRT can lead to systematic dose differences between 3DCT and 4DCT. Dosimetric effects were studied considering the breathing pattern of three patients monitored during the course of their treatment and an in-house developed 4D Monte Carlo framework. Imaging data were taken in free-breathing and in cine mode for both 3D and 4D acquisition. Treatment planning for IMRT delivery was done based on the free-breathing data with the CORVUS (North American Scientific, Chatsworth, CA) planning system. The dose distributions as a function of phase in the breathing cycle were combined using deformable image registration. The study focused on (a) assessing the accuracy of the CORVUS pencil beam algorithm with Monte Carlo dose calculation in the lung, (b) evaluating the dosimetric effect of motion on the individual breathing phases of the respiratory cycle, and (c) assessing intrafraction and interfraction motion effects during free-breathing or gated radiotherapy. The comparison between (a) the planning system and the Monte Carlo system shows that the pencil beam algorithm underestimates the dose in low-density regions, such as lung tissue, and overestimates the dose in high-density regions, such as bone, by 5% or more of the prescribed dose (corresponding to approximately 3-5 Gy for the cases considered). For the patients studied this could have a significant impact on the dose volume histograms for the target structures depending on the margin added to the clinical target volume (CTV) to produce either the planning target (PTV) or internal target volume (ITV). The dose differences between (b) phases in the breathing cycle and the free-breathing case were shown to be negligible for all phases except for the inhale phase, where an underdosage of the tumor by as much as 9.3 Gy relative to the free-breathing was observed. The large difference was due to breathing-induced motion/deformation affecting the soft/lung tissue density and motion of the bone structures (such as the rib cage) in and out of the beam. Intrafraction and interfraction dosimetric differences between (c) free-breathing and gated delivery were found to be small. However, more significant dosimetric differences, of the order of 3%-5%, were observed between the dose calculations based on static CT (3DCT) and the ones based on time-resolved CT (4DCT). These differences are a consequence of the larger contribution of the inhale phase in the 3DCT data than in the 4DCT.

Comparison of target registration errors for multiple image-guided techniques in accelerated partial breast irradiation.

PURPOSE: External beam accelerated partial breast irradiation requires accurate localization of the target volume for each treatment fraction. Using the concept of target registration error (TRE), the performance of several methods of target localization was compared. METHODS AND MATERIALS: Twelve patients who underwent external beam accelerated partial breast irradiation were included in this study. TRE was quantified for four methods of image guidance: standard laser-based setup, kilovoltage imaging of the chest wall, kilovoltage imaging of surgically implanted clips, and three-dimensional surface imaging of the breast. The use of a reference surface created from a free-breathing computed tomography scan and a reference surface directly captured with three-dimensional video imaging were compared. The effects of respiratory motion were also considered, and gating was used for 8 of 12 patients. RESULTS: The median value of the TRE for the laser, chest wall, and clip alignment was 7.1 mm (n=94), 5.4 mm (n=81), and 2.4 mm (n=93), respectively. The median TRE for gated surface imaging based on the first fraction reference surface was 3.2 mm (n=49), and the TRE for gated surface imaging using the computed tomography-based reference surface was 4.9 mm (n=56). The TRE for nongated surface imaging using the first fraction reference surface was 6.2 mm (n=25). CONCLUSIONS: The TRE of surface imaging using a reference surface directly captured with three-dimensional video and the TRE for clip-based setup were within 1 mm. Gated capture is important for surface imaging to reduce the effects of respiratory motion in accelerated partial breast irradiation.

Fluoroscopic tracking of multiple implanted fiducial markers using multiple object tracking.

When treating mobile tumors using techniques such as beam gating or beam tracking, precise localization of tumor position is required, which is often realized by fluoroscopically tracking implanted fiducial markers. Multiple markers placed inside or near a tumor are often preferred to a single marker for the sake of accuracy. In this work, we propose a marker tracking system that can track multiple markers simultaneously, without confusing them, and that is also robust enough to continue tracking even when the markers are moving behind bony anatomy. The integrated radiotherapy imaging system (IRIS), developed at the Massachusetts General Hospital (MGH), was used to take fluoroscopy videos for marker tracking. The tracking system integrates marker detection with a multiple object tracking process, inspired by the multiple hypothesis marker tracking (MHT) process. It also utilizes breathing pattern information to help tracking. Four criteria are used to identify tracking failure, and when tracking failure occurs, the system can immediately inform the user. (In the clinical environment, the system would immediately disable the treatment beam.) In this paper, two liver patients with implanted fiducial markers were studied, and the studies were performed retrospectively to assess the effectiveness of the new tracking system. For both patients, LAT and AP fluoroscopic videos were studied. In order to better test the proposed tracking system, artificial markers were added around the real markers to disturb the tracking of the real markers. The performance of the proposed system was compared to that of a conventional tracking system (one that did not use multiple object tracking). The performance of the new system was also investigated with and without consideration of the breathing pattern information. We found that the conventional tracking system can easily miss tracking markers in the presence of artificial markers, and it cannot detect the tracking failures. On the other hand, our proposed system can track markers well and can also successfully detect tracking failures. Failure rate was calculated on a per-frame-per-marker basis for the proposed tracking system. When the system considered breathing pattern information, it had a 0% failure rate 75% of the time and 0.4% failure rate 25% of the time. However, when the system did not consider breathing patterns, it had a much higher failure rate, in the range of 1.2%-12%. Both examples of the proposed system yielded low e(95) (the maximum marker tracking error at 95% confidence level)-less than 1.5 mm.

Robust fluoroscopic respiratory gating for lung cancer radiotherapy without implanted fiducial markers.

For gated lung cancer radiotherapy, it is difficult to generate accurate gating signals due to the large uncertainties when using external surrogates and the risk of pneumothorax when using implanted fiducial markers. We have previously investigated and demonstrated the feasibility of generating gating signals using the correlation scores between the reference template image and the fluoroscopic images acquired during the treatment. In this paper, we present an in-depth study, aiming at the improvement of robustness of the algorithm and its validation using multiple sets of patient data. Three different template generating and matching methods have been developed and evaluated: (1) single template method, (2) multiple template method, and (3) template clustering method. Using the fluoroscopic data acquired during patient setup before each fraction of treatment, reference templates are built that represent the tumour position and shape in the gating window, which is assumed to be at the end-of-exhale phase. For the single template method, all the setup images within the gating window are averaged to generate a composite template. For the multiple template method, each setup image in the gating window is considered as a reference template and used to generate an ensemble of correlation scores. All the scores are then combined to generate the gating signal. For the template clustering method, clustering (grouping of similar objects together) is performed to reduce the large number of reference templates into a few representative ones. Each of these methods has been evaluated against the reference gating signal as manually determined by a radiation oncologist. Five patient datasets were used for evaluation. In each case, gated treatments were simulated at both 35% and 50% duty cycles. False positive, negative and total error rates were computed. Experiments show that the single template method is sensitive to noise; the multiple template and clustering methods are more robust to noise due to the smoothing effect of aggregation of correlation scores; and the clustering method results in the best performance in terms of computational efficiency and accuracy.

Determination of the initial beam parameters in Monte Carlo linac simulation.

For Monte Carlo linac simulations and patient dose calculations, it is important to accurately determine the phase space parameters of the initial electron beam incident on the target. These parameters, such as mean energy and radial intensity distribution, have traditionally been determined by matching the calculated dose distributions with the measured dose distributions through a trial and error process. This process is very time consuming and requires a lot of Monte Carlo simulation experience and computational resources. In this paper, we propose an easy, efficient, and accurate method for the determination of the initial beam parameters. We hypothesize that (1) for one type of linacs, the geometry and material of major components of the treatment head are the same; the only difference is the phase space parameters of the initial electron beam incident on the target, and (2) most linacs belong to a limited number of linac types. For each type of linacs, Monte Carlo treatment planning system (MC-TPS) vendors simulate the treatment head and calculate the three-dimensional (3D) dose distribution in water phantom for a grid of initial beam energies and radii. The simulation results (phase space files and dose distribution files) are then stored in a data library. When a MC-TPS user tries to model their linac which belongs to the same type, a standard set of measured dose data is submitted and compared with the calculated dose distributions to determine the optimal combination of initial beam energy and radius. We have applied this method to the 6 MV beam of a Varian 21EX linac. The linac was simulated using EGSNRC/BEAM code and the dose in water phantom was calculated using EGSNRC/DOSXYZ. We have also studied issues related to the proposed method. Several common cost functions were tested for comparing measured and calculated dose distributions, including chi2, mean absolute error, dose difference at the penumbra edge point, slope of the dose difference of the lateral profile, and the newly proposed Kappaalpha factor (defined as the fraction of the voxels with absolute dose difference less than alpha%). It was found that the use of the slope of the lateral profile difference or the difference of the penumbra edge points may lead to inaccurate determination of the initial beam parameters. We also found that in general the cost function value is very sensitive to the simulation statistical uncertainty, and there is a tradeoff between uncertainty and specificity. Due to the existence of statistical uncertainty in simulated dose distributions, it is practically impossible to determine the best energy/radius combination; we have to accept a group of energy/radius combinations. We have also investigated the minimum required data set for accurate determination of the initial beam parameters. We found that the percent depth dose curves along or only a lateral profile at certain depth for a large field size is not sufficient and the minimum data set should include several lateral profiles at various depths as well as the central axis percent depth dose curve for a large field size.

Estimation of the delivered patient dose in lung IMRT treatment based on deformable registration of 4D-CT data and Monte Carlo simulations.

The purpose of this study is to accurately estimate the difference between the planned and the delivered dose due to respiratory motion and free breathing helical CT artefacts for lung IMRT treatments, and to estimate the impact of this difference on clinical outcome. Six patients with representative tumour motion, size and position were selected for this retrospective study. For each patient, we had acquired both a free breathing helical CT and a ten-phase 4D-CT scan. A commercial treatment planning system was used to create four IMRT plans for each patient. The first two plans were based on the GTV as contoured on the free breathing helical CT set, with a GTV to PTV expansion of 1.5 cm and 2.0 cm, respectively. The third plan was based on the ITV, a composite volume formed by the union of the CTV volumes contoured on free breathing helical CT, end-of-inhale (EOI) and end-of-exhale (EOE) 4D-CT. The fourth plan was based on GTV contoured on the EOE 4D-CT. The prescribed dose was 60 Gy for all four plans. Fluence maps and beam setup parameters of the IMRT plans were used by the Monte Carlo dose calculation engine MCSIM for absolute dose calculation on both the free breathing CT and 4D-CT data. CT deformable registration between the breathing phases was performed to estimate the motion trajectory for both the tumour and healthy tissue. Then, a composite dose distribution over the whole breathing cycle was calculated as a final estimate of the delivered dose. EUD values were computed on the basis of the composite dose for all four plans. For the patient with the largest motion effect, the difference in the EUD of CTV between the planed and the delivered doses was 33, 11, 1 and 0 Gy for the first, second, third and fourth plan, respectively. The number of breathing phases required for accurate dose prediction was also investigated. With the advent of 4D-CT, deformable registration and Monte Carlo simulations, it is feasible to perform an accurate calculation of the delivered dose, and compare our delivered dose with doses estimated using prior techniques.

On dose distribution comparison.

In radiotherapy practice, one often needs to compare two dose distributions. Especially with the wide clinical implementation of intensity-modulated radiation therapy, software tools for quantitative dose (or fluence) distribution comparison are required for patient-specific quality assurance. Dose distribution comparison is not a trivial task since it has to be performed in both dose and spatial domains in order to be clinically relevant. Each of the existing comparison methods has its own strengths and weaknesses and there is room for improvement. In this work, we developed a general framework for comparing dose distributions. Using a new concept called maximum allowed dose difference (MADD), the comparison in both dose and spatial domains can be performed entirely in the dose domain. Formulae for calculating MADD values for various comparison methods, such as composite analysis and gamma index, have been derived. For convenience in clinical practice, a new measure called normalized dose difference (NDD) has also been proposed, which is the dose difference at a point scaled by the ratio of MADD to the predetermined dose acceptance tolerance. Unlike the simple dose difference test, NDD works in both low and high dose gradient regions because it considers both dose and spatial acceptance tolerances through MADD. The new method has been applied to a test case and a clinical example. It was found that the new method combines the merits of the existing methods (accurate, simple, clinically intuitive and insensitive to dose grid size) and can easily be implemented into any dose/intensity comparison tool.