S666 I. J. Radiation Oncology d Biology d Physics Volume 69, Number 3, Supplement, 2007

Author Disclosure: S. Trichter, International Specialty Products, C. Other Research Support; M. Zaider, None; D. Nori, None;A. Sabbas, None; F. Kulidzhanov, None; D. Lewis, International Specialty Products, A. Employment; C.G. Soares, None.

2836 Design of Multi-Purpose Phantom and Automated Software Analysis Tool for Quality Assurance of

Onboard kV/MV Imaging System

W. Mao, L. Xing

Stanford University, Stanford, CA

Purpose/Objective(s): Onboard imaging system consists of multiple components and successful image guidance of the therapeu-tic process depends critically on the proper function of each individual part and the system as a whole. This work takes a systematicapproach and designs a multi-purpose imaging phantom with automated software analysis tool for accurate and efficient exami-nations of various mechanical and radiation properties of the integrated kV/MV devices, including (1) the positional accuracyof kV and MV X-ray sources at any gantry angle; (2) the positional and orientational accuracy of the kV and MV imagers atany gantry angle; and (3) the coincidence of kV and MV beam isocenters.

Materials/Methods: A phantom with multiple embedded metallic fiducials was designed to simultaneously derive eight funda-mental parameters (three parameters for the X-ray source position, three parameters for the position of the imager center, andtwo parameters for the orientation of the imager) characterizing the above three types of tests. An in-house simulation softwarewas used to optimize the shape of the phantom, the number and locations of the fiducials to maximize the detection sensitivityof the system parameters. As a result, a cubic phantom (18 � 18 � 18 cm) with 13 ball bearings (BBs) with a diameter of 4.76mm stood out as an optimal design. To facilitate the data analysis, a software analysis tool was developed to extract the valuesof the system parameters by comparing the measured and predicted BB locations at a given gantry angle. The reliability and ac-curacy of the QA system was tested using a Varian Trilogy system by intentionally introducing a number of errors in system vari-ables. Additionally, a large number of kV and MV projections were collected with gantry rotating over 360� and the eightparameters for each imaging modality were obtained as functions of gantry angle. MV and kV isocenters were then calculated.All results were compared with the system specifications.

Results: A multi-purpose QA phantom and automated data analysis software have been developed for routine QA of the newlyemerged onboard IGRT devices. An accuracy better than 1 mm was found in detecting any intentionally introduced error in thepositions of the kV/MV X-ray sources or the imagers or the kV/MV isocenter mis-alignment. The minimum detectable angularuncertainty was found to be 0.1�. Furthermore, the system was able to reveal any combinational error of the angular and spatialvariables at any gantry angle. The QA system was used to monitor the performance of the Trilogy machine on a weekly basisover a period of 6 months and it was found that the system specifications of various parameters were met in most cases. However,significant errors in the system variables may occur. For example, the offsets of the MV imager center up to 4.5 mm were detectedafter a machine service.

Conclusions: QA of onboard imaging devices should go beyond a simplified kV/MV beam isocenter coincidence check. The QAtool developed here provides us with capability of quantitatively examining all relevant system variables accurately and efficiently.

Author Disclosure: W. Mao, None; L. Xing, None.

2837 Intrahepatic Tumor and Vessel Identification in Intravenous Contrast Enhanced Liver kV Cone Beam CT

R. V. Tse, D. J. Moseley, J. Siewerdsen, C. Eccles, I. Yeung, J. J. Kim, L. A. Dawson

Princess Margaret Hospital, Toronto, ON, Canada

Purpose/Objective(s): Without intravenous (IV) contrast, most liver tumors are not visible on kV cone beam CT (CBCT). Visu-alization of hepatic vessels and tumor could improve accuracy of image guided liver cancer stereotactic body radiation therapy

Proceedings of the 49th Annual ASTRO Meeting S667

(SBRT). The aim of this study was to measure the improvement in vessel and tumor enhancement on contrast CBCT compared tonon-contrast CBCT.

Materials/Methods: On an IRB approved study, IV contrast was given on up to 2 occasions per patient concomitantly with CBCTacquired in the treatment position immediately following SBRT (volume 2 mL/kg, range 120–170, injection rate 3–5 mL/sec).CBCT imaging (120 kVp, 100 mAs, 200–360 degree gantry rotation, 163–200 frames) was initiated at 0–30 seconds followinginjection and acquired over 60–120 seconds. Exhale breath hold (BH) gated CBCT when feasible; otherwise images were obtainedduring free breathing (FB). A non-contrast CBCT was acquired immediately prior to the contrast CBCT. Assessment of change invisualization of tumor and hepatic vasculature between contrast and non-contrast CBCT was made by measuring changes in CTnumber in regions of interest (ROI) within visible vessels or tumor and adjacent liver. The changes in mean CT number in tumorand vessel ROIs were compared to that in ROI in adjacent liver, for non-contrast and IV CBCTs. The changes in contrast to noiseratio (CNR) were also measured.

Results: Eight pairs of contrast and non-contrast CBCT scans were obtained in 5 patients (5 exhale BH, 3 FB). Tumors and intra-hepatic vasculature were not visible on the non-contrast CBCT scans. In 5 scans (3 patients, 2 with BH), hepatic vasculature wasvisible on contrast CBCT as regions of relative increased contrast. In 5 scans (4 patients, 2 with BH), hepatic tumor was visible oncontrast CBCT as regions of hypodensity relative to the liver. A significant difference in vessel contrast was observed betweennon-contrast and contrast CBCTs. The mean absolute difference in CT number between vessel and liver for contrast CBCTand non-contrast CBCT was 32.2 and 1.6 respectively (p \ 0.05), with no significant difference in CNR (1.03, 0.16; p = 0.25).Differences in tumor contrast were also observed (Figure 1, CNR 2.8) but were not significant for all patients. Figure 1 illustratesa contrast CBCT (a) showing a 1 cc red ROI on tumour and a 1 cc green ROI on adjacent liver. (b) The tumour shows a decrease inlinear attenuation coefficient compared to liver.

Conclusions: The addition of IV contrast to CBCT enhances the visibility of hepatic vessels and tumor, particularly in breath holdscans. Optimization of IV contrast timing and methods to reduce artifacts are being investigated to improve tumor visualization.

Author Disclosure: R.V. Tse, None; D.J. Moseley, None; J. Siewerdsen, None; C. Eccles, None; I. Yeung, CIHR grant, B. Re-search Grant; J.J. Kim, None; L.A. Dawson, NCIC grant, B. Research Grant; Elekta clinical research funding, C. Other ResearchSupport.

2838 Evaluation of Head and Neck Patient Setup and Immobilization Errors

B. Robison, R. Seibert, C. R. Ramsey

Thompson Cancer Center, Knoxville, TN

Purpose/Objective(s): Image-Guided Radiation Therapy allows setup errors to be corrected prior to treatment delivery. For Head& Neck (H&N) cases, positioning errors can originate from changes in the patient’s setup, immobilization, or anatomy. With CT-based IGRT, it is also possible to quantify the source of these positional errors. The primary objective of this study was to identifythe cause of positioning errors that resulted in unacceptable random and systematic errors.

Materials/Methods: A total of 30 H&N patients were imaged with Megavoltage CT (MVCT) prior to treatment for a total of 992imaging sessions. 29 patients were immobilized using custom Uni-frame aquaplastic masks. In addition, a Precise Bite positioningmouthpiece was used in 5 of the 30 patients. The positional offsets from the external alignment marks were recorded for each pa-tient and each fraction based on MVCT image registration with the treatment planning CT images. For each patient, the investi-gators evaluated the images for 1.) Patient position on the headrest, 2.) Position inside the aquaplastic mask, 3.) Changes in themask shape, 4.) Weight loss, and 5.) Radiation response.

Results: All 30 patients had a decrease in the cross-sectional volume during the course of treatment due to a combination of weightloss and/or tissue response to radiation. Patients 1–6 had random setup errors that were outside the 2s errors for the population asa whole. 20% were incorrectly positioned on the headrest, 20% were incorrectly positioned inside the aquaplastic mask, 13% hadchanges in the mask shape, 17% had weight loss, and 60% had a measurable radiation response. Patient 1 was treated without animmobilization mask, and thus represents the worse case scenario. Patient 2 had a GTV that was initially 8-cm that completelyresponded during the course of treatment. Patients 3 to 6 all had problems with the initial setup and construction of the immobi-lization mask. Patient 4 had the mask fabricated without removing a necktie, which is against institutional policy (Table).

Conclusions: All H&N patients immobilized with aquaplastic masks will have residual motion inside the mask due to weight loss,incorrect positioning, and/or response to treatment. Errors during the fabrication of the mask and selection of the incorrect headrestaccounted for 75% of the setup errors greater than 4-mm. Identification of these errors during CT simulation would greatly improvesetup reproducibility. In addition, the amount of motion inside the mask can be greatly improved by using a simple bite block(1s \ 1.6-mm), even in the presence of tumor regression.