E.L. Chaney

Large deformation three-dimensional image registration in image-guided radiation therapy

In this paper, we present and validate a framework, based on deformable image registration, for automatic processing of serial three-dimensional CT images used in image-guided radiation therapy. A major assumption in deformable image registration has been that, if two images are being registered, every point of one image corresponds appropriately to some point in the other. For intra-treatment images of the prostate, however, this assumption is violated by the variable presence of bowel gas. The framework presented here explicitly extends previous deformable image registration algorithms to accommodate such regions in the image for which no correspondence exists. We show how to use our registration technique as a tool for organ segmentation, and present a statistical analysis of this segmentation method, validating it by comparison with multiple human raters. We also show how the deformable registration technique can be used to determine the dosimetric effect of a given plan in the presence of non-rigid tissue motion. In addition to dose accumulation, we describe a method for estimating the biological effects of tissue motion using a linear-quadratic model. This work is described in the context of a prostate treatment protocol, but it is of general applicability.

A Monte Carlo study of accelerator head scatter

The production of off-focus x rays in the head of a 6 MV linac has been investigated using the EGS4 Monte Carlo code. The purpose of the study was to identify the sources of off-focus radiation and the relative contribution for each source. Even though a particular energy and linac were modeled, the broad conclusions are expected to be general since the effects of head scatter are similar for most conventional head designs, regardless of manufacturer, energy, and model. The head components that were modeled include the exit window of the accelerating structure, target, beam stopper, flattening filter, monitor chamber, primary and secondary collimators, and air. Monoenergetic 6 MeV electrons were followed through the exit window, target, and beam stopper until all energy was expended. Primary- and higher-order x rays produced throughout the head were followed until they were either absorbed or passed through a plane at the isocenter. Sites of off-focus radiation were found to be diffusely distributed throughout the head, with the most intense sources being the primary collimator, flattening filter, and beam stopper. Data analysis shows that the collimator effect is determined primarily by the volume of the extended head-scatter source that is exposed to the point of measurement through the collimating system. The results of this study provide a rationale for developing extended source models to calculate the collimator factor for fields defined by arbitrary collimation. An additional advantage is an improvement in the agreement between measured and calculated isodose distributions.

A finite-size pencil beam model for photon dose calculations in three dimensions

A three-dimensional dose computation model employing a finite-size, diverging, pencil beam has been developed and is demonstrated for Cobalt-60 gamma rays. The square cross-section pencil beam is simulated in a semi-infinite water phantom by convolving the pencil beam photon fluence with the Monte Carlo point dose kernel for Cobalt-60. This finite-size pencil beam is calculated one time and becomes a new data base with which to build larger beams by two-dimensional superposition. The pencil beam fluence profile, angle correction for beam divergence, the Mayneord inverse square correction, radial and angular sampling rates, error propagation, and computation time have been investigated and are reported. Radial and angular sampling rates have a great effect on accuracy and their appropriate selection is important. Percent depth doses calculated by finite-size pencil beam superposition are within 1% of values calculated by full convolution and the agreement with values from the literature is within 6%. The latter disagreement is shown to be due to a low-energy photon component which is not modeled in other calculations. Computation time measurements show the pencil beam method to be faster than full convolution and one implementation of the differential-scatter-air-ratio (dSAR) method.

A comparison of postoperative techniques for carcinomas of the larynx and hypopharynx using 3-D dose distributions

If a head and neck cancer originates low in the neck with a primary site below the shoulders, a technical challenge to the radiation oncologist exists in that the entire neck needs treatment while avoiding overlap of multiple fields on the spinal cord. No standard solution to this problem exists. We have developed a 3-D treatment planning tool that can be used to develop and compare 3-D treatment plans and dose distributions. Using this tool, we have studied the following techniques for the postoperative treatment of carcinomas of the larynx and hypopharynx, tumors that often embody the problems discussed above: (a) the mini-mantle technique used at the Massachussetts General Hospital, (b) a 3-field technique used at the University of Florida at Gainesville (UF 3-field), (c) a 3-field technique used at our institution and at many others (standard 3-field), and (d) the kicked out lateral technique used at our institution and at others. The 3-D dose distributions from these plans are compared. With 100% delivered just anterior to the vertebral body at mid-neck, the mini-mantle technique results in large 120% hot spots laterally and anteriorly in the neck. Near the mastoid tips, however, the dose falls to 100%. The upper neck nodes may be underdosed since this is 20% cooler than the lateral-anterior neck dose (where a large 120% hot spot exists). The spinal cord is adequately blocked. The two 3-field techniques result in small hot spots at the junction of the lateral and anterior fields. Because different methods are used to prevent overlap at the spinal cord, these hot spots occur anteriorly in the standard 3-field technique and laterally in the UF 3-field technique. The spinal cord block results in untreated neck tissue which can be supplemented with electrons in the standard 3-field technique, but is left untreated in the UF 3-field technique. Both techniques result in a generous length of spinal cord which does not receive full dose. The kicked out lateral technique treats the entire neck and reconstructed pharynx without matching fields at midneck. The upper mid mediastinum is underdosed 10-20% despite being within the posterior inferior portion of the beam. This could be minimized by using a tissue compensator. Unless there is significant subglottic extension or significant risk of disease in the upper mediastinum, we favor treating these malignancies with the kicked out lateral technique, which avoids the problem of junctioning lateral and anterior fields and provides a fairly homogeneous dose distribution.

Clustering on image boundary regions for deformable model segmentation

We present a novel approach, clustering on local image profiles, for statistically characterizing image intensity in object boundary regions. In deformable model segmentation, a driving consideration is the geometry to image match, the degree to which the target image conforms to some template within the object boundary regions. The template should account for variation over a training set and yet be specific enough to drive an optimization to a desirable result. Using clustering, a template can be built that is optimal over the training data in the metric used, such as normalized correlation. We present a method that first determines local cross-boundary image profile types in the space of training data and then builds a template of optimal types. Also presented are the results of a study using this approach on the human kidney in the context of medial representation deformable model segmentation. The results show an improvement in the automatic segmentations using the cluster template, over a previously built template.

SU-FF-T-362: PLanUNC as An Open-Source Radiotherapy Planning System for Research and Education

Purpose: PLanUNC is a radiotherapy planning software package that has been under development and clinical use at the University of North Carolina for approximately 20 years. Under a joint grant from the NCRR and NCI (R01 RR 018615), PLanUNC has been documented, commented, and prepared for distribution as a freely available open‐source treatment planning tool for use as an adaptable and common platform for radiotherapy research. Method and Materials: The software and source code have been made available to qualifying users through a web portal located at http://planunc.radonc.unc.edu. Licenses for PLanUNC are available without fee to institutions, departments, and other facilities engaged in research and education involving radiation therapy.Results: Recent research milestones demonstrating the extensibility and increasing utility of PLanUNC include tools for 4D planning, interfaces with ITK segmentation and registration tools, daily correction of patient positioning, and interfaces with a variety of Monte Carlo dose engines. PLanUNC is currently supported for Linux and Windows operating systems, but has been successfully compiled on Alpha, Macintosh, Solaris, and other platforms. Conclusion: Licensed users will have access to PLanUNC source code, user and development documentation, annual training workshops, and limited support from UNC and the PLanUNC research community. PLanUNC is distributed as source code, making it customizable and extensible to meet the needs of a diverse range of research applications.

SU‐FF‐I‐58: A Software Toolkit for Multi‐Image Registration and Segmentation in IGRT and ART

Purpose: Extracting information from images for image‐guided and adaptive radiation therapy usually involves segmentation and/or registration with a reference image. The purpose of this research was to develop and make freely available a software toolkit for research use that combines several registration and segmentation methods and is applicable to problems involving multiple sequential images.Method and Materials: Two automatic approaches have been integrated with standard interactive methods. One automatic approach solves PDEs describing visco‐elastic flow to register a set of target images with a reference image. Segmentations can then be transferred from the reference image to target images via corresponding deformation fields. The second method segments images via Bayesian posterior optimization of a 3D deformable shape model called an m‐rep. User efficiency has been improved by the ability to queue, view, work with, and compare multiple images and segmentations during a single session. The software was developed in C++ and runs on Windows and Linux. Results: The GUI, including a window with panes for standard orthogonal 2D views linked to a second 3D window, preserves a common look and feel for all three approaches. A set of treatment images of the same patient initially are rigidly registered to a reference image. The images are then segmented by the methods of choice. Unsatisfactory automatic results are edited with conventional methods. Registered images and segmentations are output in DICOM RT format. Conclusion: The toolkit combines multiple approaches for image registration and segmentation in a system specifically designed to support research on IGRT and ART. A cost‐free license to download and use the software is available at http://planunc.radonc.unc.edu. Research supported by NCI/NCRR R01 RR 01861501.

SU‐FF‐T‐436: Tools for Integrating Monte Carlo Dose Engines with a Radiotherapy Planning System

Purpose:Monte Carlo simulations represent the gold standard in radiotherapydose calculation. While numerous tools have been developed to facilitate accelerator and patient modeling within a Monte Carlo simulation, there are few commonly available tools for interfacing a Monte Carlodose engine with a fully‐featured treatment planningsoftware package. We report on the development of tools to integrate a Monte Carlodose engine with clinically useful radiotherapy planning software.Method and Materials: The initial release is configured to operate with PLanUNC, a freely available open‐source radiotherapy planning tool. The Monte Carlo integration package consists of several modular scripts and programs that act as a bridge between the treatment planningsoftware and the Monte Carlodose engine. Results: Using PLanUNC as a front end for the Monte Carlo, the user can develop a treatment plan, export beams and patient information to the Monte Carlo, recover the dose distribution, and analyze the results of the calculation in PLanUNC according to isodose, DVH, or EUD, as well as compare the results of the Monte Carlo simulation with results from other calculations. Conclusion: The Monte Carlo interface package facilitates the clinical use of Monte Carlo by allowing a fully‐featured radiotherapy planning suite to be used as a front end, allowing flexible treatment planning and analysis of the Monte Carlo results. The modular nature of the software makes it straightforward to adapt these tools for use with other treatment planningsoftware packages.

Evaluation of lead acrylic as a filter for contaminant electrons in megavoltage photon beams

Electron contamination of megavoltage photonbeams increases the dose at the entrance surface and throughout the buildup region. Measurements on Co‐60, 4‐ and 6‐MV photonbeams show that lead acrylic is an effective filter for removing contaminant electrons while also overcoming a number of disadvantages preventing routine use of previously proposed techniques.

SU-E-J-54: Evaluation of a Toolkit for Automatic Deformable Registration and Segmentation of Treatment Images in Clinical Prostate Cancer IGRT Applications

Purpose: High quality IGRT is only practical when robust tools are available for accurate and fast deformable registration and segmentation of the treatmentimages. We have implemented such a toolkit for the male pelvis using a class of statistically trainable deformable models (SDMs) of anatomical structures, called medial representations (m‐reps). The aim of this study is to evaluate its capabilities and performance in clinical prostate cancerIGRT applications. Methods: An image collection for prostate cancer patients treated with IGRT using a CT‐on‐Rails (CTORs) system was studied retrospectively with this toolkit to determine the actual delivered dose throughout the treatment course. The patient‐specific planning models of the male pelvic organs, including the prostate, bladder, rectum, and femoral heads, were constructed via user‐guided autosegmentation of the planning CT. The planning CT was then automatically registered to the CTORs images via a rigid‐body, soft‐tissue‐based registration method. The planning models were then transformed into initialized treatmentmodels to segment the CTORs images. The essential tissue‐voxel correspondence across treatmentimages was established by this model‐based segmentation process. The dose delivered to the same tissue elements can, therefore, be accumulated for adaptive planning and/or outcome assessment. Results: Preliminary results show that the m‐rep‐based planning image segmentations are clinically acceptable to the physicians and can be readily constructed from the planning CT with nominal user guidance. The automatic image registration with the treatment‐dayimage, occasionally followed by manual refinement, also considered acceptable by clinical staff. Treatmentimage segmentation with intrinsic tissue‐voxel correspondence used substantially less time (5 ± 1.3 minutes) in comparison to manual contouring. Conclusions: Our clinic‐oriented toolkit is effective in segmentation of treatmentimages of the male pelvis. Application for dose accumulation and adaptive IGRT is in progress. This work has been conducted in collaboration with Morphormics, Inc. with grant support from NCI R44 CA119571.