B Choi

Determination of patient-specific internal gross tumor volumes for lung cancer using four-dimensional computed tomography

BackgroundTo determine the optimal approach to delineating patient-specific internal gross target volumes (IGTV) from four-dimensional (4-D) computed tomography (CT) image data sets used in the planning of radiation treatment for lung cancers.MethodsWe analyzed 4D-CT image data sets of 27 consecutive patients with non-small-cell lung cancer (stage I: 17, stage III: 10). The IGTV, defined to be the envelope of respiratory motion of the gross tumor volume in each 4D-CT data set was delineated manually using four techniques: (1) combining the gross tumor volume (GTV) contours from ten respiratory phases (IGTVAllPhases); (2) combining the GTV contours from two extreme respiratory phases (0% and 50%) (IGTV2Phases); (3) defining the GTV contour using the maximum intensity projection (MIP) (IGTVMIP); and (4) defining the GTV contour using the MIP with modification based on visual verification of contours in individual respiratory phase (IGTVMIP-Modified). Using the IGTVAllPhases as the optimum IGTV, we compared volumes, matching indices, and extent of target missing using the IGTVs based on the other three approaches.ResultsThe IGTVMIP and IGTV2Phases were significantly smaller than the IGTVAllPhases (p < 0.006 for stage I and p < 0.002 for stage III). However, the values of the IGTVMIP-Modified were close to those determined from IGTVAllPhases (p = 0.08). IGTVMIP-Modified also matched the best with IGTVAllPhases.ConclusionIGTVMIP and IGTV2Phases underestimate IGTVs. IGTVMIP-Modified is recommended to improve IGTV delineation in lung cancer.

Comparison of Rigid and Adaptive Methods of Propagating Gross Tumor Volume Through Respiratory Phases of Four-Dimensional Computed Tomography Image Data Set

PURPOSE To compare three different methods of propagating the gross tumor volume (GTV) through the respiratory phases that constitute a four-dimensional computed tomography image data set. METHODS AND MATERIALS Four-dimensional computed tomography data sets of 20 patients who had undergone definitive hypofractionated radiotherapy to the lung were acquired. The GTV regions of interest (ROIs) were manually delineated on each phase of the four-dimensional computed tomography data set. The ROI from the end-expiration phase was propagated to the remaining nine phases of respiration using the following three techniques: (1) rigid-image registration using in-house software, (2) rigid image registration using research software from a commercial radiotherapy planning system vendor, and (3) rigid-image registration followed by deformable adaptation originally intended for organ-at-risk delineation using the same software. The internal GTVs generated from the various propagation methods were compared with the manual internal GTV using the normalized Dice similarity coefficient (DSC) index. RESULTS The normalized DSC index of 1.01 +/- 0.06 (SD) for rigid propagation using the in-house software program was identical to the normalized DSC index of 1.01 +/- 0.06 for rigid propagation achieved with the vendors research software. Adaptive propagation yielded poorer results, with a normalized DSC index of 0.89 +/- 0.10 (paired t test, p <0.001). CONCLUSION Propagation of the GTV ROIs through the respiratory phases using rigid- body registration is an acceptable method within a 1-mm margin of uncertainty. The adaptive organ-at-risk propagation method was not applicable to propagating GTV ROIs, resulting in an unacceptable reduction of the volume and distortion of the ROIs.

A technique for reducing patient setup uncertainties by aligning and verifying daily positioning of a moving tumor using implanted fiducials.

This study aimed to validate and implement a methodology in which fiducials implanted in the periphery of lung tumors can be used to reduce uncertainties in tumor location. Alignment software that matches marker positions on two‐dimensional (2D) kilovoltage portal images to positions on three‐dimensional (3D) computed tomography data sets was validated using static and moving phantoms. This software also was used to reduce uncertainties in tumor location in a patient with fiducials implanted in the periphery of a lung tumor. Alignment of fiducial locations in orthogonal projection images with corresponding fiducial locations in 3D data sets can position both static and moving phantoms with an accuracy of 1 mm. In a patient, alignment based on fiducial locations reduced systematic errors in the left–right direction by 3 mm and random errors by 2 mm, and random errors in the superior–inferior direction by 3 mm as measured by anterior–posterior cine images. Software that matches fiducial markers on 2D and 3D images is effective for aligning both static and moving fiducials before treatment and can be implemented to reduce patient setup uncertainties. PACS number: 81.40.Wx

SU‐FF‐T‐328: Open Source, ImageJ Based, Web Accessible Tool for Treatment Plan Evaluation

Purpose: To develop a cross platform, open source, web accessible, Java based treatment plan visualization and evaluation software using the ImageJ toolkit. Method and Materials: The application can be launched from any web browser with Java Web Start. The treatment plan and structures may be imported in PinnacleTM or in DICOM format. For Pinnacle format, the patients and plans may be selected from the clinical database. This system supports the standard viewing of isodose lines, regions of interest and dose volume histograms generated in real time. To assist in plan evaluation, the software can automatically compare the DVHs to planning guidelines to ensure both target coverage and normal tissue tolerances. The constraints can be volumetric or absolute. Conclusion: We have written an open source software tool that allows one to review a treatment plan remotely from any web browser. To facilitate research, dose volumes can be imported in binary format and dose volumes and DVHs can be exported as binary or ASCII. Future plans include a fast dose calculation engine for both photon and proton treatments and better support of batch processing through macros. Conflict of Interest: SRA with Philips Medical Systems.

SU‐FF‐J‐31: Comparison of Image Registration Tools Used for Daily KV Imaging‐Based Patient Set‐Up

Purpose: To assess daily set‐up of thoracic patients, based on kV images and kV cone‐beam CT(CBCT), and to compare different tools for analysis of kV images.Method and Materials: Nine thoracic patients who had daily kV imaging as a part of their routine treatment were evaluated. A total of 205 images (18–26 images per patient) were available for analysis. Patients were set up daily to skin marks, and orthogonal kV images were acquired. Shifts used for patient treatment were determined by side‐by‐side visual inspection of images in a commercial R&V system. Weekly CBCTimages were obtained, and patient alignment was done using an in‐house 3D‐3D rigid registration tool. All images were retrospectively analyzed by a single observer using in‐house 2D‐3D image registration tools. In this software, split windows and shifting of the reference image were used to determine the shifts. Results: The standard deviation of the shifts actually made, for all 9 patients, were 3.8, 4.5, and 4.6 mm (L/R, S/I and A/P respectively). The standard deviation of the differences between actual shifts and those obtained retrospectively by manual 2D‐3D matching were 2.7, 3.0, and 2.7 mm (L/R, S/I and A/P respectively). The standard deviation of the differences between manual kV shifts and CBCT were 2.3, 2.5, and 1.5 mm (L/R, S/I and A/P respectively). The largest systematic difference observed was between manual 2D‐3D and CBCT shifts (1.0 mm in the A/P direction). Conclusion: This study demonstrates that PTV margins are non‐zero and need to be determined even when using daily kV imaging. The differences between different manual methods of evaluating daily patient images are smaller than the actual shifts made, but still substantial. Reduction of this variation, through use of better tools, training, or improvements in image quality, may improve patient set‐up and allow for further margin reduction.

SU‐FF‐J‐01: A Comparision of 4DCT with Breath‐Hold CT for Determination of Tumor Motion with Respiration

Purpose: Internal target volumes (ITVs) have been determined using both breath‐hold CT scans (BHCTs) and four‐dimensional CT (4DCT) to assess the extent of tumor motion during normal respiration. The purpose of this work is to compare the differences in tumor excursion when measured with BHCT and 4DCT. Method and Materials: All 4DCT and BHCT datasets in this study were acquired as a part of the radiotherapy simulation process using a commercial 4DCT system (Discovery ST, GE Healthcare, Waukesha, WI). Respiratory tracking was accomplished using a commercial system (RPM, Varian Medical Systems, Palo Alto, CA). A visual prompt from this system was displayed to patients to assist them in holding their breath at the correct level during BHCTs. The locations of the tumors with respect to a reference dataset (4DCT end‐expiration) was determined using a rigid‐body cross‐correlation algorithm that found the location on each dataset that best matched the region of the physician‐determined gross tumor volume (GTV) on the reference dataset. The patient did not move between the 4DCT and BHCT scans, thus differences in tumor location were due to tumor motion rather than bulk patient motion. Results: For 20 patients, the average difference in displacement of the GTV between BHCT and 4DCT scans was 5 mm at end‐inspiration and 3 mm at end‐expiration with maximum differences of 12 mm and 10 mm respectively. GTV motion on BHCTs was always greater than or equal to the motion on the 4DCT. The direction of tumor motion was also found to be different between 4DCT and BHCT images with the average difference in the vector angles being 14°. Conclusion: The results of this work suggests that patients being treated during normal breathing should be simulated during normal breathing (4DCT) and those to be treated using a breath‐hold technique should be simulated using BHCT.

SU‐E‐T‐523: Comparison of Fiducial and Bony‐Anatomy Based Alignment for Prostate Localization in Proton Therapy

Purpose: To quantify and compare differences in localization when aligning proton therapy patients using carbon fiducials versus bony‐anatomy and to evaluate the dosimetric consequences of these differences. Methods: 250 pairs of AP and lateral daily kV images were obtained for 16 prostate cancer patients treated at our institution. Prior to treatment 2–3 carbon fiducials were implanted in each patient. Patients were treated to 76 Co‐Gy‐ Equivalent in 38 fractions, and immobilized with a knee‐foot cradle. A water‐filled endorectal balloon was used to suppress prostate motion. Before each fraction, therapists aligned patients on fiducials and acquired a set of post‐shift images to confirm alignment. Residual errors for fiducial alignment were collected from post‐shift images using a point alignment tool; for bony‐anatomy alignment, a 2D‐3D method was used. The dosimetric implications were analyzed using verification plans offset by the maximum difference between couch shifts determined using the two methods. Results: The average systematic component of residual shifts was less than 0.1 cm for fiducials and bony‐anatomy in all directions. The standard deviation of systematic components was less than 0.1 cm for fiducials and was 0.23, 0.26, and 0.08 cm for bony‐anatomy in the AP, SI, and RL directions respectively. The random component of residual shifts was less than 0.1 cm for fiducials and was 0.16, 0.17, and 0.05 cm for bonyanatomy. Incorporating the maximum difference in shifts into verification plans, CTV coverage was found to be minimally compromised with no less than 96% of the CTV receiving full dose in any plan and V70 of the rectum/bladder varied by a maximum of 13 percentage points in either direction Conclusions: Systematic error for each method was small, while random error was larger in the AP and SI directions for bony‐anatomy than fiducials. Verification plans revealed minimal CTV coverage degradation for the maximum shift differences.

SU‐E‐T‐850: Selection of Initial DVH Constraints for IMRT Planning Based on Target Dose Gradient Characteristics

Purpose: One major obstacle for IMRT planning is to set up proper DVH constraints for organs at risk (OARs). Inexperienced IMRT planners often set OAR constraints using generic clinical guidelines, which may not always produce the best organ sparing. The goal of this study is to incorporate dose gradient information to set up the most likely achievable OAR dose tolerance. Methods: We investigated the dose falloff characteristics from typical IMRT plans. We found that the fastest dose falloff from a prostate plan can be universally used to describe the best scenario of dose gradient near the PTV. We also found the multi‐field near‐field dose falloff is different from the far‐field dose falloff, which is usually described by the attenuation of a single beam. We used an open‐field un‐modulated dose calculation to simulate the far‐field dose falloff. Also, we calculated the Euclidean distance map to translate the dose fall‐off to regional dose distribution. This dose distribution solely based on the distance from PTV and dose gradient information was reloaded back into the original plan. The DVHs calculated from this dose‐gradient plan represents the best scenario of OAR sparing. Those DVH values for OARs were used as initial estimation of DVH objective function for IMRT planning. Results: We applied our approach to 2 head‐and‐neck patients, 2 lung patients, and 2 prostate patients. We found the 40% dose falloff from the target dose prescription best represents the dose distribution in the near‐field. IMRT plans re‐optimized based on OAR constraints set by the dose‐gradient method produced better OAR sparing than the clinical plans for bladder (in prostate plans), heart (in lung plans), and oral cavity (in head‐and‐neck plans). All other organs produced similar results. Conclusions: We have designed a novel approach to incorporate dose gradient information as dose constraints for IMRT optimization.

SU‐GG‐J‐121: Lung Tumor Setup Using 2D‐3D Registration of Respiratory‐Gated Oblique X‐Rays and the Planning 4D‐CT

Purpose: To investigate the feasibility of using intensity based 2D‐3D registration to directly setup on respiratory gated lungtumors using x‐ray source angles tailored for each patient. Method and Materials: A 2D‐3D registration software was developed in‐house which renders digitally reconstructed radiographs(DRR) in real time with a consumer graphics card. The user was then able to conveniently view DRRs at various source angles to choose the angles where the tumor is less obstructed by surrounding structures during the planning phase. The 2D‐3D software, using mutual information, simultaneously aligned DRRs from the end‐expiration planning 4D‐CT with four end‐expiration projection x‐rays (lateral, Anterior‐Posterior (AP), and 2 oblique) on a sub‐image based on the projection of the tumor. For the analysis, the x‐ray projection data was taken from three lung patients in a separate research protocol for 4D cone‐beam CT. The 2D‐3D results were compared with 3D‐3D alignment of the end‐expiration 4D cone‐beam CT with the end‐expiration planning 4D‐CT on a region around the tumor.Respiration phase information was deduced from a commercial external fiducial system (Real‐time Position Management, Varian Medical Systems). The volumes of the three tumors were 14.0 cc, 45.8 cc, and 17.9 cc. The extents of motion in the Superior‐Inferior (SI) direction were 2 mm, 4 mm, and 11 mm. Results: The mean differences between the 3D‐3D and 2D‐3D registrations were −0.1 mm, 0.1 mm, and 1.6 mm in the lateral, AP, and SI directions respectively. The standard deviations were 2.2 mm, 1.8 mm, and 1.0 mm. Conclusion: We have demonstrated the feasibility of direct lungtumor setup with per‐patient x‐ray source angles. The preferred oblique x‐ray angles were anterior and posterior angles on the same lateral side as the tumor. Future work will investigate the optimal number of x‐rays and the lower limit of the tumors size.

SU‐GG‐T‐396: Automatic Analysis and Presentation of DVH Data Based On Custom Constraints — a Treatment Planning Report Card

Purpose: To automatically compare dose volume histograms (DVHs) against a set of predetermined goals and constraints to facilitate pre‐treatment plan review. Method and Materials: The internal scripting language of the treatment planning system (Pinnacle3, Philips Medical Systems, Milpitas CA) is used to prepare DVH data for evaluation against a set of constraints identified by ROI name. These settings can specify maximum percent volume by dose, maximum dose for a percent volume, or a maximum point dose. The results are displayed in a dialog box and can also be saved or printed for inclusion in the plan documentation. Results: This program functions as planned, and completes its execution quickly. The dosimetrists workflow is affected only marginally. Conclusion: The extra information provided by this analysis has been valuable to the staff responsible for plan review. This work has been supported in part by a Sponsored Research Agreement with Philips Healthcare.