Computational Environment For Radiotherapy Research Research Articles

ROE (Radiotherapy Outcomes Estimator): An open-source tool for optimizing radiotherapy prescriptions

Background and objectivesRadiotherapy prescriptions currently derive from population-wide guidelines established through large clinical trials. We provide an open-source software tool for patient-specific prescription determination using personalized dose-response curves. MethodsWe developed ROE, a plugin to the Computational Environment for Radiotherapy Research to visualize predicted tumor control and normal tissue complication simultaneously, as a function of prescription dose. ROE can be used natively with MATLAB and is additionally made accessible in GNU Octave and Python, eliminating the need for commercial licenses. It provides a curated library of published and validated predictive models and incorporates clinical restrictions on normal tissue outcomes. ROE additionally provides batch-mode tools to evaluate and select among different fractionation schemes and analyze radiotherapy outcomes across patient cohorts. ConclusionROE is an open-source, GPL-copyrighted tool for interactive exploration of the dose-response relationship to aid in radiotherapy planning. We demonstrate its potential clinical relevance in (1) improving patient awareness by quantifying the risks and benefits of a given treatment protocol (2) assessing the potential for dose escalation across patient cohorts and (3) estimating accrual rates of new protocols.

Quantitative Relaxometry for Target Localization and Response Assessment in Ultra-Hypofractionated MR-Guided Radiotherapy to the Prostate and DIL

<h3>Purpose/Objective(s)</h3> MR-guided systems, such as hybrid MR-Linacs, provide a novel platform for the acquisition and development of image-based biomarkers that can provide crucial information during daily online plan adaptation. Relaxometry has been shown to be sensitive to changes in tissue composition or function during radiotherapy. We investigate the added value of longitudinal relaxometry for target localization and response assessment. <h3>Materials/Methods</h3> The study included 42 unfavorable intermediate risk prostate patients undergoing ultra-hypofractionated MR-guided adaptive therapy to the prostate and seminal vesicles CTV with an integrated boost to the dominant intraprostatic lesion (DIL). The prescription was 800/900 cGy x 5 fractions, every other day. The longitudinal (T1) and transverse (T2) relaxation times were measured in the MR-Linac. T1 maps were acquired using the variable flip-angle method: TR/TE=20/2.3 ms, FA={4°, 22°}, FOV=320 × 320 × 200 mm, voxel = 0.7 × 0.7 × 3 mm³, scan time = 1.5 min. T2 maps were acquired using the multi-echo spin-echo method: TR/TE/ΔTE = 4000/22/11 ms, 8 echoes, FOV = 320 × 320 × 100 mm, voxel = 0.5 × 0.5 × 3 mm³, scan time = 5.5 min. The bias and reproducibility of T1/T2 mapping was assessed over the course of eight weeks using a NIST-traceable phantom. Longitudinal phantom measurements were also used to establish baseline values for T1/T2. The analysis was implemented in the Computational Environment for Radiotherapy Research. Daily T1 and T2 maps of prostate patients were rigidly registered to the pre-beam online-planning T2w images and T1/T2 values were analyzed within the CTV/DIL contours. <h3>Results</h3> The median bias with respect to nominal values for T1, T2 was -1.1%, 3.4%. When compared to baseline, the median bias for T1, T2 was 0.6%, 0.1% and the coefficient of variation was 2.1%, 0.9%, respectively. Maximum translation and rotation corrections within the T1/T2 maps were 1.9±0.3mm and 0.9°±0.2°. Across all patients and fractions, T1_CTV=1766±150ms, T1_DIL=1559±245ms, T2_CTV=127±19ms, T2_DIL=111±19ms. T1 and T2 were significantly longer in the CTV. Variations over fraction number are shown in Table 1. Linear regression showed there is a significant increase of mean T1 in the DIL (p=0.05) and CTV (p=0.05). The increase in T1 from fraction 1 to fraction 5 was 12.3% and 6.3% in DIL and CTV, respectively. The trend for T2 was not significant for either DIL or CTV. <h3>Conclusion</h3> T1 and T2 quantitative relaxometry is a surrogate measure of unbound water content and motility and shows the potential to serve as a predictor for changes in the local tumor microenvironment during MR-guided adaptive radiotherapy.

Study of Analysis the Sensitivity of the Computational Environment for Radiological Research Field Size Based on Two Dimensional Dose Distribution for Water Phantom Cases

Purpose This study analysed the sensitivity of the field size from variations in the target volume dimensions, depth, and position. The variations in the target volume analysis were used to determine the width of the field size. Thus, the quality control of the radiation beam can be obtained. Materials and Methods The computed tomography (CT) image of the IBA Dose 1 type of water phantom consists of 350 slices. Variations in the dimension of the target volume were modelled in 10×10×10 cm3, 10×12×10 cm3 , 10.2×10×10.2 cm3, and 15×15×15 cm3. Beam parameters use one beam of irradiation on the central axis 0°, 6 MV energy, 100 cm source-skin distance (SSD), beamlet delta x, and y set to 0.1 cm. Dose distribution in the form of the XZ isodose curve and dose profile was used to observe the field size. Results In this study, the isodose curve was successfully displayed in the XZ isodose curve. The field size’s sensitivity has been successfully reviewed from variations of the target volume, depth, and position. The target X and Z direction analysis is used in determining the width and length of the field size. Conclusion The analysis related to the field size sensitivity study was obtained from a relatively valid calculation. The field size was evaluated with variations in depth of 1.5 cm, 5 cm, 10 cm, and variations in positions of 10 cm, 12 cm, 14 cm, 18 cm, and 20 cm. This study will be used as a reference to validate the distribution of computational environment for radiotherapy research (CERR) dose in the future. Thus, the accuracy of the dose calculation can be obtained.

A method for generating intensity-modulated radiation therapy fields for small animal irradiators utilizing 3D-printed compensator molds.

The purpose of this study was to investigate the feasibility of using fused deposition modeling (FDM) three-dimensional (3D) printer to generate radiation compensators for high-resolution (~1mm) intensity-modulated radiation therapy (IMRT) for small animal radiation treatment. We propose a novel method incorporating 3D-printed compensator molds filled with NaI powder. The inverse planning module of the computational environment for radiotherapy research (CERR) software was adapted to simulate the XRAD-225Cx irradiator, both geometry and kV beam quality (the latter using a phase space file provided for XRAD-225Cx). A nine-field IMRT treatment was created for a scaled-down version of the imaging and radiation oncology core (IROC) Head and Neck IMRT credentialing test, recreated on a 2.2-cm-diameter cylindrical phantom. Optimized fluence maps comprising nine fields and a total of 2564 beamlets were calculated at resolution of 1.25×1.25mm2 . A hollow compensator mold was created (using in-house software and algorithm) for each field using 3D printing with polylactic acid (PLA) filaments. The molds were then packed with sodium iodide powder (NaI, measured density ρNaI =2.062g/cm3 ). The mounted compensator mold thickness was limited to 13.8mm due to clearance issues with couch collision. At treatment delivery, each compensator was manually mounted to a customized block tray attached to the reference 40×40mm2 collimator. Compensator reproducibility among three repeated 3D-printed molds was measured with Radiochromic EBT2 film. The two-dimensional (2D) dose distributions of the nine fields were compared to calculated 2D doses from CERR using gamma comparisons with distance-to-agreement criteria of 0.5-0.25mm and dose difference criteria of 3-5%. Good reproducibility of 3D-printed compensator manufacture was observed with mean error of ±0.024Gy and relative dose error of ±4.2% within the modulated part of the beam. Within the limit of 13.8mm compensator height, a maximum radiation blocking efficiency of 91.5% was achieved. Per field, about 45.5g of NaI powder was used. Gamma analysis on each of the nine delivered IMRT fields using radiochromic films resulted in eight of nine treatment fields with >90% pass rate with 5%/0.5mm tolerances. However, low gamma passing rate of 49-66% (3%/0.25mm to 5%/0.5mm) was noted in one field, attributed to fabrication errors resulting in over-filling the mold. The nine-field treatment plan was delivered in under 30min with no mechanical or collisional issues. We show the feasibility of high spatial resolution IMRT treatment on a small animal irradiator utilizing 3D-printed compensator shells packed with NaI powder. Using the PLA mold with NaI powder was attractive due to the ease of 3D printing a PLA mold at high geometric resolution and the well-balanced attenuation properties of NaI powders that prevented the mold from becoming too bulky. IMRT fields with 1.25-mm resolution are capable with significant fluence modulation with relative dose accuracy of ±4.2%.

Monte-Carlo Modelling For Evaluation Of Two Different Calculation Algorithms

Modeling of linac head (VARIAN Trilogy) for 6 MeV photon beam was performed using BEAMnrc code package (BEAMnrc 2017). The DOSXYZnrc code was used to determine the percentage depth dose (PDD profiles) and beam profiles for different symmetric square field sizes, i.e., 5cm X 5cm; and 40cm X 40cm. The DICOM images of Alderson Radiation Therapy (ART) RANDO Phantom was used. Four field 3D-CRT treatment plans were generated using AAA, PBC, and Monte-Carlo (MC). It was found that nominal energy of 5.7 MeV with FWHM of 1.2 mm provides best matching of modelled and working linac. All three 3D-CRT plans calculated with AAA, PBC and MC on a Pelvic Rando phantom, were compared using CERR (Computational Environment for Radiotherapy Research) and MATLAB 2013b. It was found that AAA and PBC have comparable results, although in case of tissue interfaces and inhomogeneous media AAA provides better accuracy in comparison to PBC. It can also be observed that AAA and PBC underestimate doses in comparison to MC in the soft muscle tissue which includes OARs such as bladder, bowel bag, and PTV TOTAL. It was found that both AAA and PBC fails to account for tissue air interface accurately and shows variation of 30-40% whereas for surface dose variation of +/-10% was observed. In homogeneous media (muscle tissue) AAA and PBC underestimate doses in comparison to MC. These commercially available algorithms overestimates and underestimates dose values as compared to MC based dose calculation for low and high dose regions specially.

OA051] Toward a new treatment planning system accounting for in-vivo proton range verification in proton therapy

Purpose Proton therapy is widely used in state-of-the-art radiotherapy. However, precision of proton therapy could be affected by proton range uncertainties due to e.g. anatomy change, patient setting uncertainties. Prompt gamma (PG) imaging is investigated to monitor proton range in vivo by detecting PG emitted by nuclear de-excitation processes in the beam path. Despite ideal correlation between the penetration depth of PG signal and proton range in most cases, few literature studies reported that this correlation could deteriorate because of tissue heterogeneities. Hence, in this study, we investigate a new treatment planning approach accounting for PG imaging. Methods A research computational platform, combining Monte Carlo (MC, Geant4) pre-calculated pencil beams (subPB) with the Matlab-based TPS engine CERR (A computational Environment for Radiotherapy Research), was introduced for treatment planning. To optimize current treatment plans considering PG imaging, first, a MC treatment plan is created using a particle extension of CERR. Thereby, the PG emission and dose distribution for each subPB is obtained. Secondly, the PG fall-off position is evaluated and compared to the proton range (80% distal dose fall-off) for all subPBs. SubPBs with reliable PG-dose correlation are visualized in terms of beam-eye-view location and individual dose delivery. Few of these “good-correlation” subPBs are boosted in the new treatment plan via manual or automated selection to enable good detectability. Then, a re-optimized plan is created and recalculated on the same patient CT using Geant4 for evaluation. Results The new treatment planning approach was applied to 3 head & neck tumor patients. The dose distribution of the original and re-optimized treatment plan was found to be of comparable dosimetric quality. The positions and PG-dose correlations of highest-intensity subPBs in the original plan were random, while in the optimized plan the highest-intensity subPBs are with good PG-dose correlation and uniformly distributed across the entire field for a reliable and thorough monitoring. Conclusions A new treatment planning approach has been proposed, which could improve the treatment planning process in proton therapy by integrating the PG-based in vivo monitoring of the beam range for a safer and more controllable proton treatment. EU-MSCA GA n. 675265 (OMA).

Isobio software: biological dose distribution and biological dose volume histogram from physical dose conversion using linear-quadratic-linear model.

PurposeTo develop an in-house software program that is able to calculate and generate the biological dose distribution and biological dose volume histogram by physical dose conversion using the linear-quadratic-linear (LQL) model.Material and methodsThe Isobio software was developed using MATLAB version 2014b to calculate and generate the biological dose distribution and biological dose volume histograms. The physical dose from each voxel in treatment planning was extracted through Computational Environment for Radiotherapy Research (CERR), and the accuracy was verified by the differentiation between the dose volume histogram from CERR and the treatment planning system. An equivalent dose in 2 Gy fraction (EQD2) was calculated using biological effective dose (BED) based on the LQL model. The software calculation and the manual calculation were compared for EQD2 verification with pair t-test statistical analysis using IBM SPSS Statistics version 22 (64-bit).ResultsTwo and three-dimensional biological dose distribution and biological dose volume histogram were displayed correctly by the Isobio software. Different physical doses were found between CERR and treatment planning system (TPS) in Oncentra, with 3.33% in high-risk clinical target volume (HR-CTV) determined by D90%, 0.56% in the bladder, 1.74% in the rectum when determined by D2cc, and less than 1% in Pinnacle. The difference in the EQD2 between the software calculation and the manual calculation was not significantly different with 0.00% at p-values 0.820, 0.095, and 0.593 for external beam radiation therapy (EBRT) and 0.240, 0.320, and 0.849 for brachytherapy (BT) in HR-CTV, bladder, and rectum, respectively.ConclusionsThe Isobio software is a feasible tool to generate the biological dose distribution and biological dose volume histogram for treatment plan evaluation in both EBRT and BT.

Consensus Contouring Guidelines for Postoperative Stereotactic Body Radiation Therapy for Metastatic Solid Tumor Malignancies to the Spine

Todevelop consensus contouring guidelines for postoperative stereotactic body radiation therapy (SBRT) for spinal metastases. Ten spine SBRT specialists representing 10 international centers independently contoured the clinical target volume (CTV), planning target volume (PTV), spinal cord, and spinal cord planning organ at risk volume (PRV)for 10 representative clinical scenarios in postoperative spine SBRT for metastatic solid tumor malignancies. Contours were imported into the Computational Environment for Radiotherapy Research. Agreement between physicians was calculated with an expectation minimization algorithm using simultaneous truth and performance level estimation with κ statistics. Target volume definition guidelines were established by finding optimized confidence level consensus contours using histogram agreement analyses. Nine expert radiation oncologists and 1 neurosurgeon completed contours for all 10 cases. The mean sensitivity and specificity were 0.79 (range, 0.71-0.89) and 0.94 (range, 0.90-0.99) for the CTV and 0.79 (range, 0.70-0.95) and 0.92 (range, 0.87-0.99) for the PTV), respectively. Mean κ agreement, which demonstrates the probability that contours agree by chance alone, was 0.58 (range, 0.43-0.70) for CTV and 0.58 (range, 0.37-0.76) for PTV (P<.001 for all cases). Optimized consensus contours were established for all patients with 80% confidence interval. Recommendations for CTV include treatment of the entire preoperative extent of bony and epidural disease, plus immediately adjacent bony anatomic compartments at risk of microscopic disease extension. In particular, a "donut-shaped" CTV was consistently applied in cases of preoperative circumferential epidural extension, regardless of extent of residual epidural extension. Otherwise more conformal anatomic-based CTVs were determined and described. Spinal instrumentation was consistently excluded from the CTV. We provide consensus contouring guidelines for common scenarios in postoperative SBRT for spinal metastases. These consensus guidelines are subject to clinical validation.

The Comparison Study of Quadratic Infinite Beam Program on Optimization Instensity Modulated Radiation Therapy Treatment Planning (IMRTP) between Threshold and Exponential Scatter Method with CERR® In The Case of Lung Cancer

This research compares the quadratic optimization program on Intensity Modulated Radiation Therapy Treatment Planning (IMRTP) with the Computational Environment for Radiotherapy Research (CERR) software. We assumed that the number of beams used for the treatment planner was about 9 and 13 beams. The case used the energy of 6 MV with Source Skin Distance (SSD) of 100 cm from target volume. Dose calculation used Quadratic Infinite beam (QIB) from CERR. CERR was used in the comparison study between Gauss Primary threshold method and Gauss Primary exponential method. In the case of lung cancer, the threshold variation of 0.01, and 0.004 was used. The output of the dose was distributed using an analysis in the form of DVH from CERR. The maximum dose distributions obtained were on the target volume (PTV) Planning Target Volume, (CTV) Clinical Target Volume, (GTV) Gross Tumor Volume, liver, and skin. It was obtained that if the dose calculation method used exponential and the number of beam 9. When the dose calculation method used the threshold and the number of beam 13, the maximum dose distributions obtained were on the target volume PTV, GTV, heart, and skin.

Prospective review of radiotherapy trials through implementation of standardized multicentre workflow and IT infrastructure.

We sought to develop a process that would allow us to perform a prospective review of outlining in trials using expert reviewers based in multiple centres. We implemented a specific information technology infrastructure and workflow that could serve all organizations involved in the radiotherapy quality assurance (RTQA) process. Data were processed and packaged in the computational environment for radiotherapy research (CERR) binary format and securely transmitted to the expert reviewer at the designated remote organization. It was opened and reviewed using the distributed CERR-compiled application, and a standardized report was sent to the respective centre. Centres were expected to correct any unacceptable deviations and resubmit outlining for approval prior to commencing treatment. 75% of reviews were completed and fed back to centres within 3 working days. There were no delays in treatment start date. Our distributed RTQA review approach provides a method of prospective outlining review at multiple centres, without compromising the quality, delaying the start of treatment or the need for significant additional infrastructure resources. Future progress in the area of prospective individual case review will need to be supported by additional resources for clinician time to undertake the reviews. Trial groups around the world have formulated different approaches to address the need for the prospective review of radiotherapy (RT) data with clinical trials, in line with available resources. We report a UK solution that has allowed the workload for outlining review to be distributed across a wider group of volunteer reviewers without the need for any additional infrastructure costs and has already been adopted within the UK RT trials community.

Modulated photon radiotherapy (XMRT): an algorithm for the simultaneous optimization of photon beamlet energy and intensity in external beam radiotherapy (EBRT) planning

This is a proof of principle study on an algorithm for optimizing external beam radiotherapy in terms of both photon beamlet energy and fluence. This simultaneous beamlet energy and fluence optimization is denoted modulated photon radiotherapy (XMRT). XMRT is compared with single-energy intensity modulated radiotherapy (IMRT) for five clinically relevant test geometries to determine whether treating beamlet energy as a decision variable improves the dose distributions. All test geometries were modelled in a cylindrical water phantom. XMRT optimized the fluence for 6 and 18 MV beamlets while IMRT optimized with only 6 MV and only 18 MV. CERR (computational environment for radiotherapy research) was used to calculate the dose deposition matrices and the resulting dose for XMRT and IMRT solutions. Solutions were compared via their dose volume histograms and dose metrics, such as the mean, maximum, and minimum doses for each structure. The homogeneity index (HI) and conformity number (CN) were calculated to assess the quality of the target dose coverage. Complexity of the resulting fluence maps was minimized using the sum of positive gradients technique. The results showed XMRT’s ability to improve healthy-organ dose reduction while yielding comparable coverage of the target relative to IMRT for all geometries. All three energy-optimization approaches yielded similar HI and CNs for all geometries, as well as a similar degree of fluence map complexity. The dose reduction provided by XMRT was demonstrated by the relative decrease in the dose metrics for the majority of the organs at risk (OARs) in all geometries. Largest reductions ranged between 5% to 10% in the mean dose to OARs for two of the geometries when compared with both single-energy IMRT schemes. XMRT has shown potential dosimetric benefits through improved OAR sparing by allowing beam energy to act as a degree of freedom in the EBRT optimization process.

SU‐E‐J‐253: The Radiomics Toolbox in the Computational Environment for Radiological Research (CERR)

Purpose:To present an open source and free platform to facilitate radiomics research — The “Radiomics toolbox” in CERR.Method:There is scarcity of open source tools that support end‐to‐end modeling of image features to predict patient outcomes. The “Radiomics toolbox” strives to fill the need for such a software platform. The platform supports (1) import of various kinds of image modalities like CT, PET, MR, SPECT, US. (2) Contouring tools to delineate structures of interest. (3) Extraction and storage of image based features like 1st order statistics, gray‐scale co‐occurrence and zonesize matrix based texture features and shape features and (4) Statistical Analysis. Statistical analysis of the extracted features is supported with basic functionality that includes univariate correlations, Kaplan‐Meir curves and advanced functionality that includes feature reduction and multivariate modeling. The graphical user interface and the data management are performed with Matlab for the ease of development and readability of code and features for wide audience. Open‐source software developed with other programming languages is integrated to enhance various components of this toolbox. For example: Java‐based DCM4CHE for import of DICOM, R for statistical analysis.Results:The Radiomics toolbox will be distributed as an open source, GNU copyrighted software. The toolbox was prototyped for modeling Oropharyngeal PET dataset at MSKCC. The analysis will be presented in a separate paper.Conclusion:The Radiomics Toolbox provides an extensible platform for extracting and modeling image features. To emphasize new uses of CERR for radiomics and image‐based research, we have changed the name from the “Computational Environment for Radiotherapy Research” to the “Computational Environment for Radiological Research”.

Correlation of Dosimetric and Clinical Factors With the Development of Esophagitis and Radiation Pneumonitis in Patients With Limited-Stage Small-Cell Lung Carcinoma

BackgroundThe purpose of the study was to correlate clinical and dosimetric factors with the development of esophagitis and radiation pneumonitis in patients with limited-stage small-cell lung carcinoma (LS SCLC). Patients and MethodsOne hundred eighteen patients who received curative intent chemoradiotherapy for LS SCLC and had electronically archived radiation treatment plans were included. The medical charts were reviewed for clinical data. The treatment plan was reviewed for critical structure delineation and dose delivered. Treatment planning data were analyzed using Computational Environment for Radiotherapy Research (V3.3). Dosimetric parameters were correlated with the risk of toxicity using Spearman rank correlation. ResultsRadiotherapy dose was 40 Gy in 15 fractions (fx) (n = 80) and 45 Gy in 30 fractions twice per day (n = 38). The 6-month cumulative incidence of Grade ≥ 2 radiation pneumonitis was 6.5% and 7.9% for the 40 Gy/15 fx and 45 Gy/30 fx groups, respectively (P = .40). The 3-month cumulative incidence of Grade 3 esophagitis was 7.5% and 13.2% for the 40 Gy/15 fx and 45 Gy/30 fx groups, respectively (P = .31). Grade ≥ 3 pneumonitis was correlated with volume of lung receiving 20 Gy (V20) and mean lung dose. Grade ≥ 3 esophagitis was correlated with mean esophagus dose and minimum dose to the hottest 45% of the esophagus (D45). ConclusionMean lung dose and V20 were significant predictors of radiation pneumonitis in LS SCLC. Mean esophageal dose and D45 were significant predictors of esophagitis. These 2 treatment schedules have similar toxicity profiles.

TU‐F‐17A‐01: BEST IN PHYSICS (JOINT IMAGING‐THERAPY) – An Automatic Toolkit for Efficient and Robust Analysis of 4D Respiratory Motion

Purpose:To provide an automatic image analysis toolkit to process thoracic 4‐dimensional computed tomography (4DCT) and extract patient‐specific motion information to facilitate investigational or clinical use of 4DCT.Methods:We developed an automatic toolkit in MATLAB to overcome the extra workload from the time dimension in 4DCT. This toolkit employs image/signal processing, computer vision, and machine learning methods to visualize, segment, register, and characterize lung 4DCT automatically or interactively. A fully‐automated 3D lung segmentation algorithm was designed and 4D lung segmentation was achieved in batch mode. Voxel counting was used to calculate volume variations of the torso, lung and its air component, and local volume changes at the diaphragm and chest wall to characterize breathing pattern. Segmented lung volumes in 12 patients are compared with those from a treatment planning system (TPS). Voxel conversion was introduced from CT# to other physical parameters, such as gravity‐induced pressure, to create a secondary 4D image. A demon algorithm was applied in deformable image registration and motion trajectories were extracted automatically. Calculated motion parameters were plotted with various templates. Machine learning algorithms, such as Naive Bayes and random forests, were implemented to study respiratory motion. This toolkit is complementary to and will be integrated with the Computational Environment for Radiotherapy Research (CERR).Results:The automatic 4D image/data processing toolkit provides a platform for analysis of 4D images and datasets. It processes 4D data automatically in batch mode and provides interactive visual verification for manual adjustments. The discrepancy in lung volume calculation between this and the TPS is &lt;±2% and the time saving is by 1–2 orders of magnitude.Conclusion:A framework of 4D toolkit has been developed to analyze thoracic 4DCT automatically or interactively, facilitating both investigational and clinical use. More software tools are under development for automatic analysis of other 4D imaging modalities.This research is in part supported by NIH (U54CA137788/132378) and City Seed FY13(Grant No. 93348‐14 01). AY would like to thank MSKCC summer medical student research program supported by National Cancer Institute and hosted by Department of Medical Physics at MSKCC.

Radiation Therapy Oncology Group (RTOG) Protocol 0915: A Randomized Phase 2 Study Comparing 2 Stereotactic Body Radiation Therapy (SBRT) Schedules for Medically Inoperable Patients With Stage I Peripheral Non-Small Cell Lung Cancer

Purpose/Objective(s): Multiple studies based on 18F-flourodeoxyglucose positron-emission tomography (FDG-PET) have reported correlations between simple SUV-based measures and outcomes in oropharyngeal cancer (OC). We investigated the use of shape and intensity feature datamining (‘radiomics’) and machine learning techniques to develop predictive models for local failure (LF), regional failure (RF), or distant metastasis (DM) following definitive chemoradiation therapy (CRT). Materials/Methods: From December 2002 to March 2009, all stage III-IV OC patients treated at our institution with CRT with retrievable pretreatment and post-treatment PET scans were identified (N Z 174). Pretreatment PET scans were converted to Computational Environment for Radiotherapy Research (CERR) format and 24 representative shape and intensity features of FDG-avid disease regions were extracted for each scan. Using machine learning feature selection methods, multiparameter logistic regression models were built incorporating clinical factors and imaging features. Rates of LF, RF, and DM were analyzed using logistic regression and Cox proportional-hazards regression with competing-risks analysis. Confidence intervals are 95%. Models were characterized by the Spearman rank correlation coefficient (Rs) and area under the ROC curve (AUC). All model-building methods were tested for significant correlation by leave-one-out cross-validation (LOOCV), and for LF with an independent patient cohort from a collaborating institution. Results: Median follow-up time was 55 months (5-112 months). Five-year estimated LF, RF, and DM rates were 7.4% (CI 3.2-11.6%), 6.1% (CI 2.39.9%), and 19.6% (13.4-25.8%), respectively. Smoking history (69% were long time smokers) was not significant. LF correlated with image intensity skewness and metabolic tumor volume (MTV), model Rs Z 0.277 (p < .001), AUC Z 0.826. RF correlated with shape heterogeneity and MTV, model Rs Z 0.283 (p < .001), AUC Z 0.866. DM correlated with image intensity kurtosis and MTV, model Rs Z 0.334 (p < .001), AUC Z 0.744. All model building methods produced significant correlations on LOOCV, with model Rs Z 0.207, 0.248, and 0.238 (p Z .007, .001, and .002) and AUC Z 0.743, 0.820, and 0.674 for LF, RF, and DM; the model for LF retained significant correlation in the independent patient cohort, Rs Z .186 (p Z .046), AUC 0.625. Conclusions: In the largest study of its kind to date, predictive models constructed using FDG-PET intensity and shape features significantly correlated to LF, RF, and DM in the test cohort and on LOOCV, and with LF with an independent population. Such models could assist in patient selection for dose reduction or escalation, or to identify high-risk patients for additional adjuvant therapy, but further understanding of the biological basis of image features will be necessary. Author Disclosure: M.R. Folkert: None. J.H. Oh: None. J. Setton: None. A.P. Apte: None. W.L. Thorstad: None. H. Schoder: None. N.Y. Lee: None. J.O. Deasy: E. Research Grant; Varian Medical Systems.

Generation of Consensus Contour Atlas for Esophageal Cancer IMRT

• Distributed CT sim datasets from three test cases to an expert GI radiation oncology panel . • Clinical info and diagnostic PET-CT provided • Test cases consisted of a GE junction, distal esophageal, and mid/upper esophageal cancer. • Instructed panelists to generate GTV contours based on this information. • Then, using a reference GTV, panelists generated CTV contours according to CALGB 80803 trial guidelines. • Imported contours into the Computational Environment for Radiotherapy Research (CERR) for quantitative analysis. • Calculated Simultaneous Truth and Performance Level Estimation (STAPLE) to generate consensus contours. • Calculated kappa statistics to characterize level of agreement between panelists.

SU‐E‐J‐87: Improvements to the Computational Environment for Radiotherapy Research

Purpose: To present new tools in CERR to analyze image registration, as well as other software updates and additions. Method: Two new features were added to CERR to evaluate the accuracy of image registration: (1) gamma‐based metric for comparing base and registered scans: All scan quantitative scan‐comparison methods rely on using image intensities. This means, these methods cannot quantify spatial misalignment between base and registered images. The gamma metric (similar to the ones used for dose comparisons) was modified for image‐comparisons. (2) a color‐coded mirror grid, based on mutual information or mean‐square‐error metrics: Color‐coding blocks within the grid help users focus on the well vs. poorly registered areas. In addition to these features, other noteworthy additions to CERR include (1) the ability to use colormaps for scans. This is an improvement over grayscale‐only scan display and is particularly useful for images produced in nuclear medicine. (2) the ability to change scan‐windowing by dragging the mouse along a view, similar to PACS viewers. Interactive windowing is particularly useful while delineating structures. (3) The ability to import Image‐annotations. The DICOM modality PR is used to store graphical and text annotations for images. These are widely used in diagnostic radiology and could be extracted from PACS. CERR can now imports these objects and store them under the planC{indexS.GSPS} element. Results: A 3‐d gamma metric was implemented in CERR for comparing image registration. It was evaluated on prostate cancer registration with criteria of 4mm and 50HU. The fused images were compared against the Gamma metric map. It was observed that the regions that matched up poorly had gamma value above 1. Conclusion: A novel gamma based approach to compare image registration allows for specification of acceptance criteria. Features added to CERR increase its utility in ART, registration, segmentation and collaboration.

Toward Semi-automated Assessment of Target Volume Delineation in Radiotherapy Trials: The SCOPE 1 Pretrial Test Case

To evaluate different conformity indices (CIs) for use in the analysis of outlining consistency within the pretrial quality assurance (Radiotherapy Trials Quality Assurance [RTTQA]) program of a multicenter chemoradiation trial of esophageal cancer and to make recommendations for their use in future trials. The National Cancer Research Institute SCOPE 1 trial is an ongoing Cancer Research UK-funded phase II/III randomized controlled trial of chemoradiation with capecitabine and cisplatin with or without cetuximab for esophageal cancer. The pretrial RTTQA program included a detailed radiotherapy protocol, an educational package, and a single mid-esophageal tumor test case that were sent to each investigator to outline. Investigator gross tumor volumes (GTVs) were received from 50 investigators in 34 UK centers, and CERR (Computational Environment for Radiotherapy Research) was used to perform an assessment of each investigator GTV against a predefined gold-standard GTV using different CIs. A new metric, the local conformity index (l-CI), that can localize areas of maximal discordance was developed. The median Jaccard conformity index (JCI) was 0.69 (interquartile range, 0.62-0.70), with 14 of 50 investigators (28%) achieving a JCI of 0.7 or greater. The median geographical miss index was 0.09 (interquartile range, 0.06-0.16), and the mean discordance index was 0.27 (95% confidence interval, 0.25-0.30). The l-CI was highest in the middle section of the volume, where the tumor was bulky and more easily definable, and identified 4 slices where fewer than 20% of investigators achieved an l-CI of 0.7 or greater. The available CIs analyze different aspects of a gold standard-observer variation, with JCI being the most useful as a single metric. Additional information is provided by the l-CI and can focus the efforts of the RTTQA team in these areas, possibly leading to semi-automated outlining assessment.

SU-E-J-92: CERR: New Tools to Analyze Image Registration Precision.

To present new tools in CERR (The Computational Environment for Radiotherapy Research) to analyze image registration and other software updates/additions. CERR continues to be a key environment (cited more than 129 times to date) for numerous RT-research studies involving outcomes modeling, prototyping algorithms for segmentation, and registration, experiments with phantom dosimetry, IMRT research, etc. Image registration is one of the key technologies required in many research studies. CERR has been interfaced with popular image registration frameworks like Plastimatch and ITK. Once the images have been autoregistered, CERR provides tools to analyze the accuracy of registration using the following innovative approaches (1)Distance Discordance Histograms (DDH), described in detail in a separate paper and (2)'MirrorScope', explained as follows: for any view plane the 2-d image is broken up into a 2d grid of medium-sized squares. Each square contains a right-half, which is the reference image, and a left-half, which is the mirror flipped version of the overlay image. The user can increase or decrease the size of this grid to control the resolution of the analysis. Other updates to CERR include tools to extract image and dosimetric features programmatically and storage in a central database and tools to interface with Statistical analysis software like SPSS and Matlab Statistics toolbox. MirrorScope was compared on various examples, including 'perfect' registration examples and 'artificially translated' registrations. for 'perfect' registration, the patterns obtained within each circles are symmetric, and are easily, visually recognized as aligned. For registrations that are off, the patterns obtained in the circles located in the regions of imperfections show unsymmetrical patterns that are easily recognized. The new updates to CERR further increase its utility for RT-research. Mirrorscope is a visually intuitive method of monitoring the accuracy of image registration that improves on the visual confusion of standard methods.

SU-E-T-95: Investigation of 3D Dosimetry for an Anthropomorphic Spine Phantom.

To evaluate 3D dosimetry for a spinal cord treatment plan delivery using the Radiological Physics Center's (RPC) anthropomorphic spine phantom. The RPC's spine phantom currently uses radiochromic film and thermoluminescent dosimeters (TLD) to evaluate spinal metastases treatments. A second dosimetry insert for the phantom was created to hold a PRESAGE® 3D dosimeter which matched the location of the TLD and film in the original insert. The phantom was CT imaged with each insert and an IMRT treatment plan was developed. The IMRT plan was delivered to the phantom twice; once with each insert. The film and PRESAGE® were scanned on a CCD microdensitometer and optical-CT system, reconstructed to a 2 mm slice width, respectively. The measured dose distributions were compared to the treatment plan calculated dose distribution using RPC in-house developed software or the Computational Environment for Radiotherapy Research (CERR). Film and PRESAGE® dose profiles were taken across several planes and compared for agreement. The distance to agreement (DTA) between the measured data and treatment plan, within the high dose gradient region, was quantified. The PRESAGE® and plan dose profiles agreed to within 2and 1 mm in the AP and SI directions, respectively. The film and plan also agreed to within 2 mm across all profiles. The PRESAGE® 3D dosimeter, based on these preliminary data, shows potential as a dosimeter for the RPC's phantom irradiation studies. Future work will add markers to the PRESAGE® insert to allow for a reproducible registration in CERR and a an optical-CT system, reconstructed to a 2 mm slice width dose calibration protocol will be created. CA 100835.