2153 A Method to Account For Dose Fractionation By Using a Modified Equivalent Uniform Dose Algorithm

C. Park, N. Lee, Y. Kim, J.M. Quivey, T. Phillips, L.J. Verhey, P. Xia

Radiation Oncology, University of California, San Francisco, CA

Purpose/Objective: IMRT results in a more inhomogeneous dose within the tumor target when compared to conventionaldelivery. The concept of equivalent uniform dose has been proposed as a way to concisely report the inhomogeneous dosedistribution based on radiobiologic response of the tumor. However, dose fractionation is not accounted for in a newly proposedgeneralized EUD (gEUD). In this study, we propose a modified EUD (mEUD) to account for dose fractionation without losingthe simplicity of the gEUD.

Materials/Methods: The gEUD, which is fitted phenomenologically from clinical data, is given as (�vi Dia)1/a, where vi and Di are

fractional volume and the dose at that volume, respectively, and a is the tissue specific parameter determined empirically fromcomparing treatment plans and clinical outcomes. This gEUD can be thought of as dose-domain proxy for tumor control probability(TCP) and normal tissue complication probability (NTCP). gEUD has many advantages over TCP/NTCP such as ease of calculation,robustness with imprecise parameters, and utilization of clinical experience. The mEUD is simply the gEUD with addition of thebiologically effective dose (BED) and a correction factor. mEUD is given as (1/c)(�vi BEDi

a)1/a. BED is the dose given in infinitelysmall fractions that will yield the same biologic response as the dose in question and is given as BED�nd[1�d/(�/�)], where n isthe number of fraction, d is the dose per fraction, and �and � are linear and quadratic constants of the survival fraction. C is thecorrection factor which is given as 1�(D0/n0)/(�/�). D0and n0 are the typical prescription dose and fraction number for the particularinstitution. mEUD is interpreted as the uniform dose given in fractions typical of that particular institution that will yield the sameTCP/NTCP as the inhomogeneous IMRT dose distribution in question. This modification also contributes to the more precisedetermination of the a parameter by making it account only for the functional subunit architecture. The gEUD and the mEUD werecalculated for 45 nasopharyngeal cancer patients treated with IMRT. The calculations were carried out when 70 Gy is delivered tothe GTV or PTV70 simultaneously with 59.4 Gy is delivered to the high-risk subclinical disease or PTV59.4. In addition, the parotidglands and the spinal cord were also calculated.

Results: The calculated average gEUDs were 72.3 Gy (� 2.6 Gy) for the PTV70, 54.4 Gy (� 7.2 Gy) for the PTV59.4, 26.7Gy (� 4.5 Gy) for the parotids, and 34.1 Gy (� 6.9 Gy) for the spinal cord. mEUD was calculated incorporating 33 fractionsin the formula. The calculated average mEUDs were 71.9 Gy (� 4.0) for PTV70, 50.2 Gy (� 8.2) for PTV59.4, 27.6 Gy ( �5.3) for the parotid glands, and 35.1 Gy (� 8.4) for the spinal cord.

Conclusions: The concept of EUD was a good starting point to incorporate radiobiologic response as a means to report andanalyze inhomogeneous IMRT treatment dose distribution. Recent generalization of EUD using generalized means made itpossible for the EUD concept to be applied both to tumor and normal tissue. While the gEDU has numerous advantages suchas being very intuitive and convenient to manipulate for such purposes as computer optimization, one of the criticisms has beenthat it failed to account for fractionation. Such shortcoming has been a result of having only one parameter, a, for each tissue,ignoring the role of any other clinical variables other than the total dose. mEUD combining gEUD with BED preserves alladvantages of gEUD while reflecting the fractionation effect and accounting for linear- and quadratic- survival characteristics.

2154 The Utility of Modulated Electron Beams in Intensity Modulated Radiotherapy, Evaluated UsingAutomatic Selection of Beam Energies and Orientations

S.K. Das,1 M. Bell,2 L.B. Marks,1 J.G. Rosenman2

1Radiation Oncology, Duke University Medical Center, Durham, NC, 2Radiation Oncology, University of North Carolina,Chapel Hill, NC

Purpose/Objective: To develop an algorithm for selection of beam orientations and energies for intensity modulatedradiotherapy (IMRT) of mixed photon and electron beams. To apply this algorithm to study the utility of modulated electronbeams in the context of IMRT planning.

Materials/Methods: An algorithm for the selection of beam orientation, type (photon/electron), and energy was developed. The goalof the algorithm is to select, for a user-specified number of beams, the optimal intensity modulated beam arrangement (beamorientations, and photon/electron energy for each orientation). The optimization objective is to deliver a prescribed dose to target,while minimizing normal structure doses corresponding to critical volume limits. The algorithm consists of two stages. The first stageconsiders a large set of predetermined feasible orientations. For a user-specified number of beams, multiple candidate arrangementsare selected using a metric that is a function of target and normal structure dose coverages. In the second stage, these candidatearrangements are sequentially intensity modulated using a fast heuristic procedure (validated for accuracy against a rigorous, but muchslower, procedure that combined gradient-descent and simulated annealing). The candidate arrangement with the least normalstructure doses at critical volume limits is chosen as optimal for that particular number of beams. The algorithm was applied to breast(50 Gy prescribed) and head-and-neck (70 Gy prescribed) cases. For both cases, dose-volume limits were imposed on normalstructures - heart and lung for the breast case, and optic nerves, optic chiasm, optic globes, and cord for the head-and-neck case. Forthe breast case, the algorithm was employed to select optimal arrangements with 2, 4, and 6 beams (axial) and, for the head-and-neckcase, optimal arrangements with 3, 4, 5 and 7 beams (axial & non-axial).

Results: For increasing numbers of beams selected, the number of electron beams increased for the breast case (0, 1, 3 beams for the2-, 4-, 6-beam plans, respectively). However, not more than one electron beam was selected for the 3-, 5-, 7-beam head-and-neckplans. For both cases, increasing the number of selected beams led to decreasing normal structure volumes for doses � 15 Gy (seeFigure). However, target homogeneity decreased with increasing number of selected beams for the breast case, and increased withincreasing number of selected beams for the head-and-neck case. We surmise that the nature of the contribution of electron beamsis different in the two cases. For the breast case, multiple electron beams likely play the role of decreasing lung dosage. They do soby providing dose to the extreme medial and lateral target edges which drape around lung, thereby reducing penetration into lung.For the head-and-neck case, the single, approximately anteriorly oriented, electron beam likely serves the role of increasing targethomogeneity by providing dose to shallow anterior target areas.

Conclusions: Modulated electrons and photons offer potential dosimetric improvements over modulated photons alone. Atechnique to select beam orientation, type (photon/electron), and energy for IMRT is presented. Results of applying thistechnique indicate that modulated electron beams play important roles, complementary to modulated photons, that differ basedon treatment site.

S433Proceedings of the 45th Annual ASTRO Meeting

2155 Using IMRT to Repair Unacceptable Dose Distributions of Prostate Implants

X. Li, J.Z. Wang, P.P. Amin, M. Earl, D. Shepard

Radiation Oncology, University of Maryland, Baltimore, MD

Purpose/Objective: Unacceptable dose distributions (e.g., cold spots) in prostate permanent implants are frequently seen, dueto many anatomical and/or technical reasons including pubic arch interference, prostate volume/shape change, seed migration,image artifacts, and seed placement errors. While the brachytherapy is typically not repeated, we propose to use image-guidedintensity-modulated radiotherapy (IMRT) to repair these unacceptable dose distributions. To explore its feasibility, a dosimetricand biologic treatment planning study is carried out.

Materials/Methods: The equivalent uniform dose (EUD) is applied to both brachytherapy and IMRT and is used to determinethe required dose for IMRT to compensate an underdosing from an implant. For prostate tumor, the EUD is estimated basedon the LQ model with the most recent parameters derived from clinical data (��0.15Gy-1, �/��3.1Gy). The EUD for normaltissue is computed based on the Lyman model. For an unsatisfactory implant, the EUD value in each sub-region (voxel),presently termed as voxel equivalent dose (VED), is calculated based on the dose distribution from post-implant CT. Theobtained 3D VED distribution is used to form an objective function that aims to add a non-uniform dose distribution tocompensate the underdosing regions within the prostate and, in the mean time, to spare the normal structures. This objectivefunction is then used in the IMRT planning based on the same CT data. Either fixed-gantry or rotational IMRT is planned inorder to deliver the required non-uniform dose distribution for different patient situations. The fixed gantry IMRT is plannedusing a commercial system (Corvus), and the rotational IMRT, termed as IMAT, is planned either by forward planning usinga commercial system (Precise) or by inverse planning using an in-house developed algorithm (direct aperture optimization). Asoftware tool is developed to (1) calculate the EUD value and the 3D VED distribution from the physical dose distribution foreither brachytherapy or IMRT, (2) to display the combined VED distribution in 3D, and (3) to evaluate plans in terms of theEUD and the dose volume histogram (DVH) based on VED. To demonstrate the feasibility, the entire planning process iscarried out on selected patient cases.

Results: The required doses of IMRT for various unacceptable implant dose levels are tabulated in the table. The values in thetable are physical doses (Gy) to be delivered in 2Gy daily fraction and they are calculated for two combined plans with EUDvalues of 70 and 80 Gy. According to the table, for example, an underdosing by 45Gy from an 125I implant with a intendedprescription of 145Gy (EUD � 70Gy) can be compensated by an IMRT dose of 26Gy. Our results show that the required 3DIMRT dose distributions are highly non-uniform for patient cases studied. Dosimetric planning performed to achieve these dosedistributions indicates that different unacceptable implant can be repaired by a preferable form of IMRT with satisfactory targetcoverage and normal structure sparing. For example, if underdosing is mainly in apex-anterior region (e.g., due to pubic archinterference), fixed-gantry IMRT is preferred because of its low integral dose and simplicity. If underdosing is mainly inanterior-lateral peripheral region (e.g., due to ultrasound image artifacts from large prostate), IMAT is advantageous becauseof its superior sparing of urethra. Because of its highly non-uniform dose distributions and/or small treatment volumes, IMRTneeds to be guided by daily imaging (e.g., ultrasound).

Conclusions: It is dosimetrically and biologically feasible for using image-guided IMRT to repair an unacceptable prostateimplant. Necessary tools to address the biological and dosimetric issues for the clinical implementation of this approach aredeveloped.

S434 I. J. Radiation Oncology ● Biology ● Physics Volume 57, Number 2, Supplement, 2003