Proceedings of the 51st Annual ASTRO Meeting S643

day online plan adaptation. Dosimetric changes observed in this limited sample were rather small and generally correlated to tumorshrinkage and weight loss.

Author Disclosure: P.J. Gagnon, None; J. Tanyi, None; L. Ku, None; T. McDonald, None; J.M. Holland, None; C.R. Thomas,None; M. Fuss, Varian Medical Systems, B. Research Grant; Varian Medical Systems, Philips, BrainLab, F. Consultant/AdvisoryBoard.

3027 Fiducial Marker (FM) Matching versus Bone Matching in Image Guided Intensity Modulated Radiation

Therapy (IG-IMRT) of the Prostate Bed following Radical Prostatectomy

M. A. Wood1, N. J. Aherne1,2, J. P. Herden1, J. L. Dean1, J. D. Hill1, A. Last1, C. Everitt1, C. D. Fuller3, A. Roy4,T. P. Shakespeare1,2

1North Coast Cancer Institute, New South Wales, Australia, 2RCS Faculty of Medicine, University of New South Wales, CoffsHarbour, Australia, 3University of Texas Health Sciences at San Antonio Department of Radiation Oncology, San Antonio, TX,4University of Texas at San Antonio Department of Management Science and Statistics, San Antonio, TX

Purpose/Objective(s): The aim of this study is to determine the extent of interfraction motion of the prostatic bed (PB) with respectto bony anatomy during post prostatectomy IMRT and whether bone matching reflects PB position. While the use of soft tissuematching has been shown to be superior to matching to pelvic anatomy in those receiving definitive prostate IMRT, little is knownabout it’s application in post prostatectomy IMRT.

Materials/Methods: Twenty two patients received 726 fractions (66 Gy/33#) of salvage IMRT to the PB. All patients had trans-rectal insertion of three FM into the PB at the vesico-urethral anastomosis and bladder neck with daily online matching to FM.During treatment, patients were setup to skin tattoos, FM position was then visualized on electronic portal images (EPI) (iViewGT,Elekta) and moved to match FM position at planning. After IRB approval, each EPI was retrospectively aligned to the simulationdigitally reconstructed radiographs based on bony anatomy (EPI-BM) and on PB fiducials (EPI-FM). The difference between EPI-BM and EPI-FM alignments (PB Motion) was calculated for each EPI image pair in the X (Rt/Lt),Y (Sup/Inf) and Z (Ant/Post)axes. Statistical analysis was performed using the Linear Mixed Effects model for intermethod comparison with bias control, aswell as an overall correlation coefficient to assess agreement between methods.

Results: There were 2037 EPIs with 4,074 data points available for comparison in the X, Y and Z directions. Intramethod agree-ment was better for EPI-BM (variance X, Y, Z is 0.06, 0.03, 0.06, standard error range 0.0018 - 0.0034) than for EPI - FM (varianceX, Y, Z is 0.07, 0.06, 0.10, standard error range 0.0034 - 0.0058), indicating soft tissue motion is more variable than bone. Inter-method agreement between subjects is low (EPI-BM X, Y, Z is 0.01, 0.02, 0.03 and EPI-FM is X,Y,Z 0.01, 0.01, 0.01), indicatingpoor agreement. The overall correlation coefficient between the two methods was poor at 0.73, 0.30, 0.54 in the X, Y and Z axesrespectively. There is poor correlation in all planes between EPI-BM and EPI-FM measurements. The mean (+/- SD) prostate bedmotion in centimeters in the X, Y, Z axes is 0.05 +/- 0.19, 0.04 +/- 0.31, 0.02 +/- 0.30. The calculated 95% limits of agreement incentimeters between methods were 0.51, 0.41, and 0.6 in the X, Y and Z axes, respectively. Consequently, it can be expected that 95% of the time EPI-FM and EPI-BM will only agree within 0.6 cm.

Conclusions: In patients receiving postoperative IMRT, PB motion is independent of bony anatomy, and there is poor agreementbetween EPI-FM and EPI-BM. PB motion is greatest in the SI and AP directions and when using bone matching, may result ingeographic misses, particularly with IMRT.

Author Disclosure: M.A. Wood, None; N.J. Aherne, None; J.P. Herden, None; J.L. Dean, None; J.D. Hill, None; A. Last, None; C.Everitt, None; C.D. Fuller, None; A. Roy, None; T.P. Shakespeare, None.

3028 Model-based Segmentation for Soft Tissue Contouring in the Pelvis on Fan-beam (FBCT) and Cone-beam

Computed Tomographic (CBCT) Imaging: A Contouring Variability Study

E. Weiss1,2, J. Wu1, F. Sleeman1, J. Bryant1, P. Mitra1, M. Myers1, T. Ivanova1, M. Murphy1, J. Williamson1

1VCU, Richmond, VA, 2Goettingen University, Goettingen, Germany

Purpose/Objective(s): Model-based segmentation (MBS) of computed tomography images is a promising new tool to reduce thecontouring work load associated with image-guided and 4D radiotherapy. We compared manual segmentation (MS) to MBS toassess contouring accuracy and clinical usefulness both in regular pelvic CTs and CBCTs.

Materials/Methods: In 4 patients prostate, bladder and rectum were delineated independently by 5 observers both on CBCTs andFBCTs. Contouring was performed twice using a) conventional MS and b) MBS combined with manual contour editing. Contour-ing variability was assessed as a function of anatomical location using a surface-based method of analysis, and by comparing dif-ferences in volumes, volume coincidence ratios (common/encompassing volume) and contouring time.

Results: Mean ratios of MS/MBS volumes showed good agreement for both imaging modalities with MBS contours being slightlylarger. Similarly, volume coincidence ratios were very similar for MS and MBS with Prostate: FBCT 0.58 vs. 0.58 - CBCT 0.45 vs.0.48; Bladder: FBCT 0.72 vs. 0.74 - CBCT 0.61 vs. 0.63; Rectum: FBCT 0.59 vs. 0.59 - CBCT 0.51 vs. 0.51 for MS vs. MBSrespectively. Ratios of MS/MBS contouring times varied between 0.89 and 1.16 for FBCT and 1.01 and 1.15 for CBCT contouring.Surface-based evaluation of standard deviations around the average mesh per structure showed mean SDs [cm] for prostate of 0.15and 0.16 for FBCT and 0.2 and 0.21 for CBCT, for bladder 0.07 and 0.08 for FBCT and 0.14 and 0.16 for CBCT, for rectum 0.13and 0.12 for FBCT and 0.15 and 0.18 for CBCT for MS and MBS respectively. As a function of location, the largest contouringuncertainties for prostate were mostly posterior with MS and inferior with MBS, for bladder mostly posterior and inferior for MSand MBS, and for rectum consistently in the anterior lower third for MS and MBS.

Conclusions: In this study, MS and MBS produced very similar contours on FBCT and CBCT with no reduction in contouringvariability resulting from use of the MBS contour library. Due to the necessity of additional editing of contours there was no rel-evant reduction in contouring time using MBS.

Author Disclosure: E. Weiss, None; J. Wu, None; F. Sleeman, None; J. Bryant, None; P. Mitra, None; M. Myers, None; T. Ivanova,None; M. Murphy, None; J. Williamson, None.