S594 I. J. Radiation Oncology d Biology d Physics Volume 72, Number 1, Supplement, 2008

information. The CTV1 was mirrored on the opposite side of the patient to identify equivalent normal tissue to calculate the normalSUV. A new CTV1 was then created based on the mean of the normal SUV plus two standard deviations.

Results: Mirror CTV1s were created for 10 patients, and any overlap was eliminated. Nine of the cases had less than 15% overlapbetween the CTV1 and mirror CTV1s. One patient had �29% overlap and a mirror CTV1 mean SUV of 2.66, indicating that thedisease was too bilateral to use this method effectively. Another patient had high PET activity in the tonsils which extended into themirror region and artificially increased the normal mean SUV. Of the remaining eight patients, the CTV1 SUV population meanwas 3.44 ± 1.59, and the normal SUV population mean was 1.74 ± 0.62. The new SUV cutoff values ranged from 2.36 to 3.64, butthe new CTV1s created using only the patient specific SUV cutoffs were smaller than the treated CTV1s for all cases by an averageof 61.9 ± 16.5%. New CTV1s were largely contained within original CTV1s, however an average of 6.3 ± 9.4% of the new CTV1volume was identified outside the original CTV1. The large standard deviation is due to one patient with an SUV threshold of 2.36and 27.9% of the new CTV1 volume outside the original CTV1. This may be because new regions were identified using a thresholdlower than 2.5.

Conclusions: The range of SUV thresholds obtained by using a value two standard deviations above the equivalent normal tissuemean and the subsequent volume change in CTV1 indicates that patient specific information is important in tumor delineation.However, this method is limited to cases where there is primarily unilateral disease.

Author Disclosure: S.M. McGuire, None; K.C. Bylund, None; J.E. Bayouth, None.

2948 RapidArc - Commissioning and QA Procedures

C. Ling1,2, P. Zhang1, Y. Archambault2, J. Bocanek2, G. Tang3, T. LoSasso1

1Memorial Sloan-Kettering Cancer Center, New York, NY, 2Varian Medical Systems, Palo Alto, CA, 3University of Maryland,Baltimore, MD

Purpose/Objective(s): To develop a program of commissioning and quality assurance for the Varian RapidArc radiation deliverysystem. RapidArc performs IMRT using one gantry revolution to deliver a 2 Gy fraction in 2 minutes or less. The key advancedtechnical features are simultaneous dynamic MLC and dose-rate variation during variable-speed gantry rotation.

Materials/Methods: We designed a step-by-step program to evaluate the performance of a RapidArc-enabled Clinac. (1) Toassess the dynamic MLC accuracy, we performed RapidArc plans of picket-fence patterns, for comparison with the same patternat fixed gantry. (2) Using a custom-designed RapidArc plan we irradiated different parts of a film with different dose-rates, toascertain the accuracy of variable dose-rate. (3) We studied the combined use of dynamic MLC and variable dose-rate to achievea designed dose pattern. All tests were performed with radiographic films mounted at isocenter on a fixture that rotated with thegantry (Sun Nuclear Isocenter Mounting Fixture - IMF).

Results: (1) The mechanical stability of the IMF, as measured using a slit light field and a slit X-ray field at different gantry angles,was better than�0.5 mm. Comparison of picket-fence patterns, acquired at stationary gantry angle and during RapidArc delivery,showed shifts of 1 mm or less. ‘‘Intentional errors’’ inserted into the picket fence pattern showed that the sensitivity of this test fordetecting 0.5 mm errors in MLC position. (2) During RapidArc delivery, different parts of a film were exposed to the same dosewith variable dose rates (equivalent to 111, 222, 332, 443, 554, 665 and 776 MU/min or 0.33, 0.66, 0.99, 1.33, 1.67, 2 and 2.33MU/deg) by combining variable dose-rate (up to 600 MU/min) and variable gantry speed (up to 5.54 deg/sec). The observed meanvariation in the delivered dose was 1.3% (range, 0.1- 2.1%). (3) Different parts of a film were exposed to the same dose usingDMLC sliding window technique, combining different leaf speeds (0.5, 1, 2 and 3 cm/sec) with different dose rates (150 - 600MU/min). The mean deviation from the intended dose was 0.7% (range, 0.3 - 1.3%). (4) Log files of machine performance (Dy-nalog files) during RapidArc delivery, comparing intended and achieved parameter values, indicated mean standard deviations of0.04 MU and 0.26 deg at the control points of the RapidArc plans.

Conclusions: We have developed a prototypical program of commissioning and quality assurance for the Varian RapidArc radi-ation delivery system. The program tests the critical technical features of simultaneous dynamic MLC and dose-rate variation dur-ing variable-speed gantry rotation. The test results indicate that the Clinac control system accurately implements the designedRapidArc plans, with excellent agreement between the planned and delivered dose patterns.

Author Disclosure: C. Ling, Varian, A. Employment; P. Zhang, None; Y. Archambault, None; J. Bocanek, None; G. Tang, None;T. LoSasso, None.

2949 Dynamic Contrast Enhanced Magnetic Resonance Imaging (DCE-MRI) Assessment of Tumor Physiology

in Head and Neck Cancer: A Comparison of Intra Patient and Inter Patient Variability

O. I. Craciunescu1, D. Yoo1, E. W. Cleland1, D. P. Barboriak1, M. Dolguikh2, N. Muradyan2, M. Kasibhatla1, M. Carroll1,D. M. Brizel1

1Duke University Medical Center, Durham, NC, 2CAD Sciences, Inc., White Plains, NY

Purpose/Objective(s): DCE-MRI quantitatively measures tumor physiology and treatment induced changes including vasculartransfer (PERM, Ktrans), extracellular volume fraction (EVF, Ve), and initial area under the curve (iAUC), calculated from an intra-tumor region of interest (ROI). Optimal ROI delineation is not established. The valid use of DCE-MRI requires that the variationmeasured within a tumor be less than that observed between tumors in different patients. This work evaluated the impact of ROIselection on assessment of intra and inter patient variability in a prospective clinical trial.

Materials/Methods: Patients received hyperfractionated RT + concurrent cisplatin with synchronous Avastin (A) and Tarceva(T). Patients were randomized to receive 2 weeks (lead-in) of A, T, or A+T prior to targeted chemoradiation with A+T. DCE-MRI images acquired on a 1.5T GE Signa Exite scanner at baseline and at the end of lead-in were analyzed. A dynamic 3D spoiledgradient echo sequence was used before and after bolus injection of Gd DTPA (Magnevist): TR = 6.4 ms, field of view (FOV) = 24cm, matrix size: 256 x 256, single scan duration = 10 sec, number of scans = 31-33, 10 mm thickness, flip angle = 60o. Full time

Proceedings of the 50th Annual ASTRO Meeting S595

point pharmacokinetic image analysis was performed with CADvue (CAD Sciences White Plains, NY) to measure PERM, EVF,and iAUC for Gd. Four different sets of ROIs were generated to cover the whole primary tumor; 4 separate and different ROI setswere created for nodes when present: whole tumor (Whole), 3-4 slices including the slice containing the most enhancing voxels(Partial), a single slice containing the most enhancing voxels (SliceMax), and the most enhancing 5% voxels in SliceMax (95Max).The coefficient of variation (CV) was calculated to establish intra patient variability among different ROI sets and to establish interpatient variability for each ROI selection method. The ratio between each intra patient CV and the inter patient CV was calculated(RCV).

Results: Baseline primary tumor RCVs for the different ROIs not including 95Max ranged as follows: PERM 0.21-0.22, EVF0.20-0.26, and iAUC 0.13-0.14. Corresponding nodal RCVs: PERM 0.18-0.19, EVF 0.17-0.19, and iAUC 0.20-0.21. The postlead-in primary RCVs were PERM 0.18-0.23, EVF 0.11-0.11, and iAUC 0.10-0.10. The nodal RCVs were PERM 0.15-0.16,EVF 0.14-0.15, and iAUC 0.08-0.09. In all cases with use of 95Max, intra patient CV’s were much larger and nearly all RCVswere .1.

Conclusions: Distinction between tumors cannot be made using 95Max ROIs . The other 3 strategies are viable and equivalent forusing DCE-MRI to measure head and neck cancer physiology. Single slice (Slice Max) is preferred for its simplicity. Future effortwill correlate these parameters with treatment outcome.

Author Disclosure: O.I. Craciunescu, None; D. Yoo, None; E.W. Cleland, None; D.P. Barboriak, None; M. Dolguikh, CAD Sci-ences, Inc., A. Employment; N. Muradyan, CAD Sciences, Inc., A. Employment; M. Kasibhatla, None; M. Carroll, None; D.M.Brizel, Genentech Clinical Trial Support, C. Other Research Support.

2950 Comparison of Head and Neck IMRT Setup Accuracy and Reproducibility using Different On-board

Imaging Systems

N. Dogan, J. Shumadine, H. Saleh, S. Song

Virginia Commonwealth University, Richmond, VA

Purpose/Objective(s): To assess the setup errors obtained with three on-board image guidance systems for Head-and-Neck(H&N) patients receiving intensity modulated radiotherapy (IMRT).

Materials/Methods: Eleven H&N IMRT patients were selected for this study. Three on-board imaging (OBI) systems, kV Brain-LAB ExacTrac, Varian kV OBI and Cone Beam Computer Tomography (CBCT), were utilized to assess their ability to reduceinter- and intra-fractional setup errors. The patients were initially positioned by laser alignment. First, a set of dual kV ExacTracimages were acquired and the couch shifts were recorded in the lateral (LR), longitudinal (SI) and vertical (AP) directions, but noshift was implemented. Next, the Varian kV OBI images were acquired and the couch shifts were obtained. Daily corrections werebased on the shifts obtained using Varian kV OBI. Finally, at the end of treatment, a second set of ExacTrac images were taken todetermine the intra-fraction errors. Also, a CBCT image was acquired weekly and the shifts were recorded prior to applying anyVarian kV OBI shifts. The inter- and intra-fractional deviations obtained with Varian kV OBI images from the same patients werequantitatively compared to the corresponding setup deviations obtained with ExacTrac and CBCT images.

Results: The inter-fraction systematic and random shifts were -0.06 ± 1.17mm (LR), 0.25 ± 1.181 mm (SI) and -1.03 ± 3.51mm(AP) for kV OBI; -0.39 ± 1.55mm (LR), -0.074 ± 2.21mm (SI) and -1.38 ± 4.08 mm (AP) for kV ExacTrac; and -0.96 ± 1.51mm(LR), 0.22 ± 3.60mm (SI) and -1.46± 5.91mm (AP) for kV CBCT. The mean 3D shift difference between the three systems were\0.5mm, except in the LR direction obtained with CBCT. The mean inter-fraction errors in the AP direction were larger than theLR and SI directions and the results are consistent for all systems. The percentage of shifts in AP direction $3mm and 5mm were36.4% and 18.2% with kV OBI and ExacTrac and 18.2% and 27.3% for kV CBCT respectively. The mean 3D shift distances withboth CBCT and ExacTrac were larger (6.49 ± 3.18mm and 5.63 ± 2.45mm) than for Varian kV OBI (4.94 ± 2.1mm) The intra-fraction systematic and random shifts determined by ExacTrac were 0.56 ± 0.97mm (LR), 0.41 ± 1.75mm (SI) and -1.8 ± 1.7mm(AP) directions. The mean 3D shift was 4.12 ± 1.40mm.

Conclusions: This analysis of setup corrections obtained with three different on-board imaging devices for H&N patients treatedwith IMRT showed some variations and revealed that both residual and intra-fractional setup errors exist even after daily correc-tions are applied. This suggests the importance of quantifying the accuracy and reproducibility for each on-board imaging systemwhich may be useful for assessing H&N treatment planning margins for image guided therapy.

Author Disclosure: N. Dogan, None; J. Shumadine, None; H. Saleh, None; S. Song, None.

2951 Evaluation of the Abilities of kV-kV Radiographic Matching and CBCT Matching to Detect Known Offsets

T. Ogunleye, I. Crocker, E. Elder

Emory University School of Medicine, Atlanta, GA

Purpose/Objective(s): To evaluate the ability of kV-kV orthogonal radiograph matching using an on board imager (OBI) andlinac based cone beam computed tomography (CBCT) to accurately detect sub-centimeter offsets during Image Guided RadioTherapy (IGRT) involving bony anatomy of the head as a reference.

Materials/Methods: A CIRS Radiosurgery Head Phantom containing bone and soft tissue equivalent materials was mounted toa Newport 460P XYZ micrometer stage with sub-millimeter accuracy in all three orthogonal directions. This system in turn wasattached firmly to the linac treatment couch. Orthogonal OBI kV radiographs and the couch lateral, longitudinal and vertical driveswere used to position the phantom to a starting position in agreement with a CT simulator produced digital reconstructed radio-graph. A CBCT was performed to confirm alignment. Once this initial position was found, the linac couch was locked to preventany further motion. Twenty five random offsets in the x, y and z directions were generated using a Microsoft Excel spreadsheet. Theshifts were limited to ±1.0 cm in any one direction from the initial setup position. The offsets were performed with the stage mi-crometers. After the application of each offset, orthogonal radiograph and CBCT alignment procedures were followed and per-formed by a user blinded to the offset values. At no time were the calculated shifts applied to the phantom.