Proceedings of the 50th Annual ASTRO Meeting S625

pulmonary disease. The median follow-up of these 49 patients was 31 (10 to 72) months. The 3-year local control, disease-free, and3-year overall survival rates in stages 1A and 1B were 93% and 96% (p = 0.86), 76% and 77% (p = 0.83), , and 90% and 63% (p =0.09), respectively. No acute toxicity was observed. Grade 2 and grade 3 radiation pneumonitis were experienced by one and twopatients, respectively; one suffered fatal bacterial pneumonia.

Conclusions: SBRT at 50 Gy per 5 fractions to the periphery of the PTV calculated by a superposition algorithm is feasible. Highlocal control rates were achieved both in T2 and T1 tumors.

Author Disclosure: A. Takeda, None; N. Sanuki, None; E. Kunieda, None; T. Ohashi, None; Y. Oku, None; T. Takeda, None; N.Shigematsu, None; A. Kubo, None.

3015 Dose Escalation Feasible Due to Gating in Lung Cancer Patients

G. Luxton, J. Antony, B. W. Loo, D. Carlson, P. G. Maxim, L. Xing

Stanford University, Stanford, CA

Purpose/Objective(s): To evaluate the impact of respiratory gating on treatment planning generated on moving and deformingtumors, by analyzing escalated dose achievable using 3D conformal radiation therapy (3D-CRT) on gated lung tumors.

Materials/Methods: Twelve lung patients were considered for this study, and their planning targets were contoured for gated andnon-gated treatments. 3D-CRT plans were generated to obtain a systematic analysis keeping comparable toxicity of normal tissueas the dose constraint. The prescription dose was defined as the dose that covered 95% of the planning target volume. Targets con-toured for both hypothetical treatments had 3D-CRT plans created so that dose prescribed was increased to the upper limit con-sistent with staying below dose constraints of the organs at risk. Linear-quadratic tumor control probability (TCP) model wasused to determine the impact of dose escalation on local tumor control. Normal tissue complication probability (NTCP) was cal-culated using Lyman-Kutcher-Burman model.

Results: Three of twelve patients exhibited significant increases in prescription doses for gated versus non-gated plans. ITV vol-umes for these showed tumor motion .1cm between the respiratory phases of end-inspiration (0%) and end-expiration (50%), inboth the superior and inferior tumor margins. For three of the 12 patients, no dose change was found between gated and non-gatedplans. One patient showed a dose decrease in the gated case due to nonlinearity of tumor motion with respect to the breathing phase,i.e., tumor in 30% phase was closer to an OAR than the tumor at both 0% and 50% phase. Here gated plan, which includes the 30%phase required de-escalation (decrease in dose), even though the gated volume was smaller than non-gated volume, because doseconstraint was reached with a smaller dose at the 30% phase, where the OAR was most proximal to target. When there was morethan one lung tumor for a patient, and one was in the pulmonary region and one in the mediastinum, dose escalation was not pos-sible because the OAR reached the toxicity constraint unaffected by gating. For prescription dose \80 Gy, the gated TCPs weresignificantly greater than those for non-gated TCPs.

Conclusions: In this sample 12 lung patients planning gated treatment allowed reduction in tumor margin and dose escalation up to12%, while staying within clinically acceptable toxicity criteria. Deformation of the tumor assumed importance when the targetdeformed in such a way that proximity of tumor to an OAR was altered, and in one case, choosing only the 0% and 50% phaseswas not appropriate. Tumors in the middle and lower portion of left lung and mediastinal tumors could be treated to higher dosewith gating. NTCP increased with mean lung dose and TCP increased with dose value up to 80 Gy, above which modeled TCP wasunity.

Author Disclosure: G. Luxton, None; J. Antony, None; B.W. Loo, None; D. Carlson, None; P.G. Maxim, None; L. Xing, None.

3016 Impact of Individualized Target Volumes on Stereotactic Body Radiotherapy Treatment Planning (SBRT):

Dosimetric Analysis

C. J. Hampton, W. T. Kearns, J. J. Urbanic, K. P. McMullen, A. W. Blackstock, W. H. Hinson

Wake Forest University School of Medicine, Winston-Salem, NC

Purpose/Objective(s): This work assesses the impact of individualized target volumes derived from helical and 4D-CT on treat-ment plans for lung SBRT patients. The dosimetric impact of individualized target volumes on target and normal tissues is con-trasted with patient population-based target volumes recommended by the RTOG 0236 phase II clinical trial.

Materials/Methods: Six patients were simulated using the Elekta Stereotactic Body Frame (Elekta, Inc., Norcross, GA), three ofwhich met the criteria for the use of abdominal compression to force shallow breathing by the patient. 4D-CT, was used for allpatients to create patient-specific internal target volumes (ITV) either as a complement to or in the absence of the compressiondevice. Our SBRT simulation/treatment planning protocol introduces a hybrid internal target volume (ITVInd) defined by the com-bination of 4D-CT-derived maximum intensity projections (MIP), with gross tumor volumes (GTV1 and GTV2) contoured from 2sequential helical CT acquisitions interspaced by patient repositioning. Non-gated PET imaging was also included at our physi-cians’ discretion. Isotropic margins of 5mm were added to ITVInd as a ‘‘safety margin" creating PTVInd. 3D treatment planswere created with D95%=Rx dose for PTVInd, and minimal dose to normal tissues, especially the lung (V20\10 Gy). For compar-ison, treatment plans were also created for a second PTV (PTV0236) based on the expansion of GTV2 with population-based mar-gins of 5 mm (10 mm S/I) as recommended by RTOG 0236.

Results: Separate treatment plans providing the desired coverage for PTV0236 and PTVInd were successfully created with an in-crease in heterogeneity (average increase = 8%) accompanying the selection of the Rx isodose needed to cover PTVInd for all cases.When the plan used to cover PTV0236 was overlaid on top of PTVInd, target coverage was on average reduced by 8.3%. 100%coverage was maintained for only one patient whiles the worst case indicated underdosing of PTVInd by 14%. The portion ofPTVInd not overlapped by PTV0236 was on average covered by a Dmean=96% of the Rx isodose. For all patients, V20 differedby no greater than 2% for the two plans even when no abdominal compression is available and PTVInd is largest. All treatmentplans were able to achieve V20 less than 10%. When individualized margins are used along with abdominal compression, themean lung dose is decreased by an average of 11 cc with one patient achieving a reduction of 27cc.

S626 I. J. Radiation Oncology d Biology d Physics Volume 72, Number 1, Supplement, 2008

Conclusions: Underdosing of the PTVInd may occur for some patients. Individualized planning target volume margins taking intoaccount data quantifying setup uncertainties and physiological tumor motion can be used for stereotactic body radiotherapy withoutincreasing irradiated lung volumes beyond accepted values.

Author Disclosure: C.J. Hampton, None; W.T. Kearns, None; J.J. Urbanic, None; K.P. McMullen, None; A.W. Blackstock, None;W.H. Hinson, None.

3017 Hyperpolarized Helium MRI and the Importance of Ventilation Tumor Volumes (VTVs) before and after

RT for Non-small Cell Lung Cancer

A. M. Allen1, Y. Sun2, H. B. Caglar1, P. Zygmanski1, J. H. Killoran1, M. Albert1

1Dana-Farber Cancer Institute/Brigham and Women’s Hospital, Boston, MA, 2Brigham and Women’s Hospital, Boston, MA

Purpose/Objective(s): Hyperpolarized 3Helium MRI (HeMRI) is a noble gas MRI which produces 3-D ventilation images of thelungs. We sought to use HeMRI before and after RT for Non-Small Cell Lung Cancer (NSCLC) and compare Ventilation TumorVolumes (VTVs) to tumor volumes obtained from CT and PET images. We also compared changes in VTVs to PFT changes afterRT.

Materials/Methods: Five patients with stage III NSCLC receiving RT (median dose 60 Gy) were enrolled on a prospective IRBapproved functional imaging protocol. Each patient underwent CT, FDG-PET and HeMRI at the time of simulation and then im-mediately (within 10 days) following their therapy. 2/5 patients underwent an additional set of scans 6 weeks post-therapy. Thethree scans (CT, FDG-PET and HeMRI) were co-registered in the ECLIPSE treatment planning system. The CT and PETGTVs were drawn as well as the ventilation tumor volume (VTV) on HeMRI. Patients underwent pulmonary function tests ateach time point.

Results: The initial Gross Tumor Volumes (GTV)on CT ranged from 261 to 26 ccs. FDG-PET volumes between 15 cc and 196cc. The HeMRI volumes ranged from 573 cc to 81 cc. After RT tumor response was seen in all patients. The CT volumes de-creased to 100 to 28 cc, PET from 53-4 cc, and HeMRI from 198 to 32 cc. Unlike CT and PET GTVs, the changes in VTV weremore varied. In Patients #1 and #5 (right apex lesion and left lingular lesions) The change in VTV correlated precisely with thechange in both CT and PET GTVs and PFTs pre and post therapy showed no changes. Patient #2 had a LUL GTV initiallyoccupying 33% of the left lung on CT. However, because it was obstructing bronchus the VTV represented 95% of the leftlung on HeMRI. After RT, the GTVct decreased by 59%. The GTVpet by 71% and VTV by 64%. However, since the lungwas no longer obstructed the percentage of the lung occupied by the GTVct was 13% (20% change) compared to VTV mea-suring 24% of the left lung (a 70% change). The correlated with a 10% increase in FEV1 after RT. Patient #3 had a 50% re-duction in GTVct and 60% reduction in GTVpet, but had no change in either VTV or PFTs. Patient #4 underwent a 70%reduction in GTVct and VTV and 80% reduction in GTVpet. She had a 10% improvement in FEV1 but a 20% decrease inDLCO. The trend continued with a third HeMRI scan which showed doubling in VTV at 6 weeks despite a complete disappear-ance of the GTVs. The DLCO decreased an additional 30% and the patient was diagnosed with grade III radiation pneumonitis 1week later. When isodose clouds were overlaid and compared to VTV volumes, a the change in VTVs post- RT was seen inregions of lung receiving ./= 45 Gy.

Conclusions: HeMRI may have many implications: including changing functional lung volumes for organ tolerances, guiding ad-vanced planning and correlating with early findings of radiation pneumonitis.

Author Disclosure: A.M. Allen, None; Y. Sun, None; H.B. Caglar, None; P. Zygmanski, None; J.H. Killoran, None; M. Albert,None.

3018 Assessment of the Clinical Accuracy of Real-time Respiratory Motion Tracking by Analysis of Log Files

B. Heijmen, J. Nuyttens, J. Poll, J. Prevost, P. Levendag, M. Hoogeman

Erasmus MC, Rotterdam, Netherlands

Purpose/Objective(s): Phantom experiments have proven that the accuracy of Synchrony real-time respiratory motion trackingwith robotic radiosurgery (CyberKnife, Accuray Inc., CA) is 0.7 ± 0.3 mm. It is unknown to what extent this accuracy can beachieved in real patient cases. Irregular breathing, varying phase-relationships between internal and external markers, and rapidbase-line shifts might decline the accuracy. The aim of this study was to quantify in real patient treatments the correlation modelerror, describing deviations between real tumor positions and positions derived from the applied correlation model, and the pre-diction error, related to inaccuracies in accounting for the total time delay in the system.

Materials/Methods: Data in log files of 47 lung cancer patients (167 treatment fractions), previously treated with respiratorymotion tracking, were analyzed. The logs were filtered to remove parts for which the data and the delivery of a clinicallyused beam did not match in time. For each treatment fraction, correlation model errors were quantified by its mean and standarddeviation. The correlation model error was compared with the geometrical error if no respiratory tracking would have been used.The prediction error was calculated by subtracting the predicted position from the actual measured position after 192.5 ms, whichis the time lag of our current system. For each treatment fraction separately the prediction error was quantified by the mean andstandard deviation of the error. In addition, prediction error was recalculated for a time lag of 115 ms with the new hybrid pre-diction algorithm (1).

Results: The correlation model errors were small, with overall mean values for the involved 167 fractions below 0.2 ± 0.3 mm (1SD) for all directions. Standard deviations describing intra-fraction variations around the fraction mean error were 0.9 mm (range,0.2 - 1.9 mm) for SI, 0.7 mm (0.1 - 1.9 mm) for LR, and 0.9 mm (0.2 - 2.5 mm) for AP. Without the use of motion tracking thesevariations would have been 2.9 mm (0.2 - 8.1 mm), 1.4 mm (0.2 - 5.5 mm), and 1.7 mm (0.2 - 4.4 mm). The SD of the predictionerror was highly correlated with motion amplitude (R = 0.9). The error increased to a maximum of 2.9 mm in the SI direction formotion amplitude of 22 mm. This error was reduced to 1.4 mm by using the hybrid prediction algorithm.