Further, the window-sweeps can be made complementary, so as to preserve an approximate uniform dose to the PTV.Because this (artificially-induced) modulation is orthogonal to the apparent motion of the PTV caused by geometric rotation,the optimized final treatment can be expected to inherit from the SW initialization some of its modulation, while remainingdeliverable at every step.

The inherent mathematical simplicity of the treatment representation allows for a priori analytic computation of beam weightsleading to a dose that best fits the prescription. These beam weights are optimized at each internal step of the optimizationalgorithm, the output of which is therefore not only a leaf sequence but also ideal weights to use for that sequence.

The leaf-position representation also suggests the possibility of optimization in the Fourier domain, due to the fact that(neglecting higher-order effects present in the dosimetry model) the dose delivered between a given pair of leaves is a squarewave, and thus admits of a simple Fourier representation based on sinc functions. We thus intend to optimize treatment in theFourier domain, using a Fourier representation of computed dose to adjust leaf positions. The Fourier representation may betruncated at some wave number cutoff for increased efficiency.

Results: Although research is ongoing, preliminary results indicate that a single-arc treatment plan with the desired properties -fidelity to PTV, avoidance of sensitive structures - is feasible and can be obtained within a reasonable amount of computing time.Numerical experiments on phantom test structures produce dose distributions showing adherence to PTV and avoiding OARs.

Conclusions: The technological development of IMRT has led to the ability to deliver radiation treatments whose dosedistributions conform exceptionally well to a given target volume of irregular shape, while allowing sparing of dose to sensitivestructures. However, by representing a treatment in terms of fluence, IMRT suffers from the drawback that a given treatmentis the result of an additional field segmentation step. This step introduces additional error (because of discrete approximationof fluence maps) and treatment time (because the result must be delivered in a number of arcs). A treatment plan with themodulation properties of IMRT, yet which can be delivered as a continuous arc, is clearly desirable. Constraints on leafmovement, and geometric aspects of the PTV which involve symmetry, present barriers to successful development of such atreatment. The SWAT method holds the potential for overcoming these barriers, because it forces modulation - and breaks thesymmetry - in a systematic, and MLC-deliverable, way, by careful choice of initial conditions.

Supported in part by a grant from Varian Medical Systems.

2478 Inverse Planning Optimization of 3DCRT for Extracranial Stereotactic Radioablation

R. M. Cardinale,1,2 L. Mao,1 R. Smith,1 B. Kavanagh,3 B. Chon,1 D. Fein,1 S. Benedict2

1Radiation Oncology, Princeton Medical Center, Princeton, NJ, 2Medical College of Virginia Hospitals, Richmond, VA,3University of Colorado Health Sciences Center, Denver, CO

Purpose/Objective: The goal of standard 3DCRT treatment planning for extracranial stereotactic radioablation (ESR) is tomaximize planning target volume (PTV) dose conformity and limit both high and intermediate isodose volumes outside of thePTV. Most targets are concave in nature and are isolated within organs with parallel architecture making full-fledged IMRTtreatment unnecessary. We have previously shown that proper selection of beam shape and block (MLC) margin on the PTVfor 3DCRT has a very significant effect on normal tissue sparing for ESR (IJROBP 1999;45:515–520). However, iterative,forward-based planning to determine proper blocking is time consuming and user dependent. We sought to determine if beamweighting and block aperture shape and margin for 3DCRT ESR treatment plans could be optimized using inverse planningalgorithms.

Materials/Methods: Representative patient cases of lung and liver tumors were chosen for analysis. Inverse planning wasperformed with multiple noncoplanar beams using a modification of our IMRT planning system with the goal of determiningthe optimal shape and weighting of each beam. The step and shoot mode was used and only a single segment was allowed foreach beam. The following constraints were used; 99% PTV coverage by 20 Gy (prescription isodose), minimization of50–100% isodose volumes of surrounding non-PTV tissue, no PTV dose homogeneity constraint. The resultant beam aperturesand weighting were used to compute a 3D plan. Separately, the best iterative forward-3D plans were calculated and comparedto the optimized inverse 3D plan. DHF (dose homogeneity factor-maximum dose/prescription dose), EUD (equivalent uniformdose), CI (conformity index-100% volume/PTV), and non-PTV normal tissue contained in the 100, 90, 80, and 50% ofprescription isodose volumes were calculated for all plans.

Results: The median PTV size and beam number were 22 cm3 and 12 respectively. Inverse generated 3D plans were highlyunpredictable in terms of beam shape and weighting. Optimized block apertures were highly irregular, and were often placedwell inside and outside the PTV. The mean block aperture margin per case (and per beam) was 0.0 mm with significantvariations ranging from 5 mm inside the PTV to 5 mm outside the PTV (one standard deviation). The inverse plan was alwayssuperior or equal to the most favorable forward 3D plans in terms of DI and V100, 90, 80 and 50 and required considerableless physics planning time. The median DHF and CI for inverse plans were 1.70 and 1.20 respectively.

Conclusions: It is especially important to optimize ESR treatment plans because of the very high fractional doses delivered.The risk of tissue injury is volume dependent because most tumor targets are located within parallel organs. We have shownfor the cases selected that block shape, block margin, and beam weighting can be optimized by a simplified inverse planningprocess that uses just one segment per beam. These plans maximized resultant 3D plans in terms of conformality and minimizednontarget volumes compared to forward planning. As a byproduct of optimal normal tissue sparing potentially desirable highdose gradients within the PTV occur that may increase tumor control probabilities.

2479 Variability of Tumor Control Under Localization Uncertainty in External Beam Radiation Therapy(EBRT) of Prostate Cancer

K. Purushothaman,1 S. Chelikani,1 Z. Chen,2 R. E. Peschel,2 J. P. Knisely,2 R. Nath,2 J. S. Duncan1

1Diagnostic Radiology, Yale University, New Haven, CT, 2Therapeutic Radiology, Yale University, New Haven, CT

Purpose/Objective: The purpose of this study is to investigate the effect of random variations in prostate localization betweenfractions of EBRT on the uncertainty in tumor control probability (TCP). Such TCP uncertainty may become significant asimprovements to TCP are sought via various methods such as margin reduction, dose escalation, and better localization.

S627Proceedings of the 46th Annual ASTRO Meeting