Volume 33, Issue 6, June 2006
Index of content:
- Therapy Scientific Session: Valencia A
- IMRT Optimization and Delivery
33(2006); http://dx.doi.org/10.1118/1.2241568View Description Hide Description
Purpose:IMRT optimization involves several treatment parameters producing a complex, unstable and computationally challenging problem during its search of an optimal plan in a reasonable time. Most parameters have been studied in IMRT optimization except the collimator angle, which is investigated in this study. Method and Materials: Five head‐and‐neck and five prostate cases are selected. The head ‐and‐neck and prostate PTVs range between 79.6–441 cm3 and 86.2–250 cm3 respectively, whereas the OAR volumes vary from 11.0–46.3 cm3 and 41.1–312 cm3 respectively. The patients are treated with five or seven fields equally distributed, 0‐ degree collimator angle, using the 1 cm leaf MLC from a Siemens Primus accelerator based on the plan generated using the Nucletron Oncentra treatment planning system. While dose‐volume constraints are kept the same as in the patient's initial treatment plan, collimator angles are varied systematically (0–90 degrees) and a new treatment plan is optimized for each collimator angle. The number of beamlets, monitor units (MU) and DVHs for each collimator angle are compared. Results: The variation of the total number of beamlets with collimator angle follows the shape of a parabola and peaks at 45° collimator angle for all patients. However, the MUs appear to be relatively independent of the collimator angle. The PTV dose coverage statistics for each patient are relatively independent of the collimator angle. Similar observations are noted for all the OARs, except for the small structures for which differences could be observed in the DVHs between the different collimator angles. Conclusions:Collimator angle does not play a significant role in IMRT optimization, as long as the PTV overage is adequate. This provides an additional freedom to choose from 0–90 degree of the collimator angle for long fields without compromising the coverage with limited MLC range or treatment time.
33(2006); http://dx.doi.org/10.1118/1.2241571View Description Hide Description
Purpose: The essence of inverse planning is how to rank objectively the competitive treatment plans. The purpose of this work is to show the equivalence of recently introduced EUD‐based and the conventional dose‐based objective function and to setup a unified framework in the dose domain for IMRT inverse planning. Methods: The dose‐based objective function such as the quadratic function treats each voxel within a structure equally, whereas the EUD‐based objective function aims to take into account of dose‐volume effect when ranking the candidate plans. For the conventional approach to accomplish what an EUD method does, we partially “break” the implicit constraint that a structure is a mathematically uniform entity with all voxels having the same identity and assign each voxel a voxel‐specific ranking (or penalty scheme). Algorithmically, this is accomplished by assigning a voxel‐specific importance or a voxel specific prescription dose. The seemingly insolvable task of determining a large number of voxel‐specific importance/prescription is dealt by heuristically relating them to the actual local doses. We show that this strategy is a more general ranking scheme, which can not only model the volumetric behavior but also the higher order factors beyond the dose‐volume effect. EUD and other dose‐volume formalisms represent special cases of the general framework. Results: A unified inverse planning framework is established for inverse planning. By assigning higher importance to the voxels within the target volume but close to the margin with a critical structure, while simultaneously assigning higher importance to the voxels within the critical structure close to the target volume, the dose to the target volume is dramatically improved in comparison with the conventional approaches. Conclusions: The EUD‐based approach can be reformulated into the realm of conventional dose‐based formalism without using biological parameters. The formalism allows us to design IMRT plans that would otherwise be unattainable.
33(2006); http://dx.doi.org/10.1118/1.2241572View Description Hide Description
Purpose: In IMRT planning, leaf‐sequencing algorithms are used to translate optimized intensity maps into deliverable aperture shapes. We have developed a novel approach to leaf sequencing called continuous intensity map optimization (CIMO). The CIMO algorithm can significantly reduce the number of segments required for IMRT delivery. A unique feature of the CIMO algorithm is that it operates directly on continuous intensity maps. Consequently, the intensity maps do not need to be divided into discrete levels. Method and Materials: CIMO uses a simulated annealing algorithm to optimize the aperture shapes and weights with a goal of minimizing the discrepancies between the optimized and sequenced fluence maps. We have benchmarked the performance of the CIMO sequencer against the algorithms in the Pinnacle3 and Eclipse treatment planning systems. Results: When the CIMO sequencer was applied to 10 IMRT plans from the Pinnacle3 planning system, the average number of segments was reduced from 133 to 62 (a 54% reduction). The average number of monitor units was reduced from 646 to 555 (a 14% reduction). The CIMO algorithm also provided a 36% reduction in the average root mean square errors between the optimal and sequenced fluence maps. Additionally, the CIMO sequenced plans provided more uniform PTV coverage with comparable sparing of critical structures. The average standard deviation of the PTV dose distribution decreased from 9.6 to 7.4 cGy. When the CIMO algorithm was applied to 5 IMRT patients from the Eclipse planning system, the average number of segments was reduced by 46% and the average number of monitor units was cut by 31%. Equivalent plan quality was observed. Conclusion: As compared with both commercial systems, the CIMO algorithm resulted in equivalent or improved plan quality while providing a significant reduction in the required number of segments and monitor units.
TU‐D‐ValA‐04: Adaptive Diffusion Smoothing: A Novel Method to Control IMRT Field Complexity Based On the Diffusion Equation33(2006); http://dx.doi.org/10.1118/1.2241573View Description Hide Description
Purpose: To introduce and evaluate adaptive diffusion smoothing (ADS), a novel procedure designed to preferentially reduce IMRT beam complexity based on any case related parameter inside the optimization.Method and Materials: The diffusion equation was used to develop a procedure in which IMRT beams are smoothed using coefficients defined for each beamlet. The coefficients can be a function of any parameter and dictate the degree of smoothing allowed for each beamlet. The ADS procedure was incorporated into our optimization system as a weighted objective function penalty and several possible ADS coefficient definitions were investigated. Coefficients were designed to promote 1) uniform smoothing everywhere, 2) smoothing based on beamlet intensities, and 3) smoothing based on beamlet gradients with respect to the plan objectives. The method has been validated on a phantom and studied in clinical sites including prostate. Results: The addition of the ADS penalty in the objective function, for all three coefficient types, produced plans with reduced modulation and minimal dosimetric impact in the phantom. Each ADS coefficient definition had merit, but gradient‐based coefficients showed the most potential for reducing beam modulation without affecting dosimetric plan quality. For example, in a prostate plan, this method reduced MU 40% while preserving full target coverage and increasing mean normal tissue doses by less than 2.2 Gy. Conclusion: A unique method based on the diffusion equation and used within the objective function has been developed to control IMRT beam complexity. This method, called adaptive diffusion smoothing, has been applied to phantom and clinical cases, and was able to reduce modulation significantly while preserving dosimetric plan quality. Adaptive diffusion smoothing is a promising tool for ensuring that only the necessary amount of beam modulation is used, promoting more efficient and accurate IMRT planning, QA, and delivery.
Supported in part by NIH Grant P01‐CA‐59827.
TU‐D‐ValA‐05: Self‐Correction of DMLC Delivery Errors Based On Feedback From On Flight Intensity Calculations33(2006); http://dx.doi.org/10.1118/1.2241574View Description Hide Description
Purpose: To deliver DMLC intensity modulated therapy accurately in the presence of errors related to unaccounted target motion, improperly executed motions of leaves and imprecise control of beam intensity rate during delivery.Method and Materials: The intensity delivered to each target point is calculated on‐the‐fly under the assumption of unidirectional leaf motion in the target frame of reference. The delivered intensity is then compared to the intended intensity. If a discrepancy arises, the leading leaf's trajectory is left unchanged and the following leaf's trajectory is modified to correct the error. Using the derived formulas, the following leaf is sped up or slowed down so that it efficiently corrects the error without unnecessarily compromising the delivered intensity to other points of the target. The formulas only consider the discrepancy between the delivered and intended intensity. They are not directly dependent on the actual and intended leaf positions as a function of time. This means that the leaves will not, in general, return to their originally planned trajectories. After correction is complete however, the correct intensity will be delivered to the rest of the target. Results: Two representative examples of self‐correction are shown. The first uses a static target, and shows how an intensity error due to an unintended leaf velocity can be corrected. The second shows a similar error, but instead uses a rigid moving target. Conclusions: An algorithm is developed to modify the following leaf trajectory when a discrepancy between the delivered and intended intensity arises. The results show that it is feasible to correct some errors in DMLC delivery without interrupting delivery. The tolerance for errors can therefore be set much higher than current control algorithms. This should lead to considerably decreased number of beam interruptions at delivery.
TU‐D‐ValA‐06: Incorporation of Modulation Range Constraints Into a Gradient Search Algorithm for IMRT Optimization33(2006); http://dx.doi.org/10.1118/1.2241575View Description Hide Description
Purpose: Hardware‐sensitive optimization for IMRT results in more conformal dose distributions; however, most planning systems neglect hardware constraints, most likely due to the complexity of incorporating them into gradient search algorithms. An important limitation of delivery systems is the modulation range, which we demonstrate can be incorporated as a hard constraint into gradient search algorithms. Method and Materials: We used a commercial treatment planning system (TPS) for calculation of dose‐deposition coefficients, which were exported to optimization software that we developed. For a beam j comprised of beamlets i, we reformulated the optimization problem such that the fluence of beamlet i, φ i , is the product of the weight of beam j and the transmission of beamlet i, φ i = wjti . A gradient search algorithm determined the optimal values of wj and ti subject to the constraints wj ⩾ 0 and Tminimum ⩽ ti ⩽ 1. The resulting fluence distributions were exported back to the TPS for MLC leaf sequencing and dose calculation. We applied the method to a phantom simulating head and neck re‐irradiation, a situation that requires very low fluence values to protect the spinal cord. Using the same cost function parameters, we varied the minimum transmission constraint. After optimization, leaf sequences for dynamic MLC delivery were determined and the dose calculated. All plans were normalized such that 95% of the target received the prescription dose. Results: Target mean and maximum doses were essentially identical for all transmission constraint values. The maximum spinal cord dose was 33% of the prescription dose for no transmission constraint. The minimum value, 30.8%, occurred for an 8% transmission constraint. Conclusion: Modulation constraint is a good approximation of hardware transmission limits and could be readily incorporated into existing TPSs. Furthermore, initial results suggest that transmission constraint beyond the intrinsic hardware limitation may be beneficial.
TU‐D‐ValA‐07: Prioritized Prescription‐Goal Treatment Planning for IMRT: The Effect of Constraint Moderation33(2006); http://dx.doi.org/10.1118/1.2241576View Description Hide Description
Purpose: Determining the ‘best’ optimization parameters in IMRT planning is typically a time‐consuming trial‐and‐error process with no unambiguous termination point. Recently we and others proposed using a goal‐programming approach which better captures the prioritization of clinical goals without introducing ambiguous user‐input parameters. We consider here the effect of adding ‘slip’ to this method, which allows for slight degradations in metric performance compared to maximum achievable. Method and Materials: In the first phase of the optimization process, only the highest‐order goals are considered (target coverage and dose‐limiting normal structures). In subsequent phases (levels), the achievements of the previous step are turned into hard constraints and lower‐order goals are optimized subject to these constraints. Slip factors were introduced to allow for slight violations of the constraints. Linear as well as quadratic goal terms were evaluated for performance as well as dosimetric ‘steerability.’ The resulting constraints can also be expressed as linear or quadratic equations.Results: Focusing on head and neck cases, we present several examples of treatment plans using prioritized optimization. These are compared to conventional IMRT plans in terms of dosimetric properties and optimization speed. The main advantages of the new optimization method are (1) its ability to generate plans that meet the clinical goals/prescriptions without tuning any weighting factors or dose‐volume constraints, and (2) the ability to conveniently include more terms which represent elements such as beam weight smoothness. Lower level goals can be optimized to the achievable limit without compromising higher order goals. Modest slip factors improved overall performance. Conclusion: The prioritized prescription‐goal planning method including slip factors allows for a more intuitive and human‐time‐efficient way of dealing with conflicting goals compared to the conventional trial‐and‐error method of varying weighting factors.
This research was supported by a NIH grant R01 CA90445 and a grant from TomoTherapy, Inc.
TU‐D‐ValA‐08: Geometric Considerations for Optimizing Beam Directions for IMAT Treatment of Lung Cancer with Mediastinum Nodal Irradiation33(2006); http://dx.doi.org/10.1118/1.2241577View Description Hide Description
Purpose: Based on geometric considerations, we optimized beam directions and arc ranges for intensity‐modulated arc therapy (IMAT) to maximize the sparing of lung and spinal cord while treating the mediastinum nodes and gross disease. Methods and Materials: A phantom with multiple planning target volumes (PTV) and lung with varying geometric shapes was created on Pinnacle 7.6c (Philips Medical Systems). A spinal cord was placed posteriorly. Mean lung and maximum cord dose were obtained for the various geometries for AP/PA and IMAT fields that were designed to spare the cord. We combined AP/PA and IMAT fields in order to achieve an optimal plan for each geometry, and seek correlations between patient geometry and their optimal relative weights. This combination of AP/PA and the IMAT fields were used to treat a patient in 2004 diagnosed with non‐small cell lungcancer with concurrent chemotherapy.Results: For a circular PTV in the phantom, the mean lungdose varied for AP/PA between 50% and 29% of the mean PTV dose, depending on the lung geometry, and for IMAT, between 50% and 46%. The maximum cord dose was 107% for AP/PA versus 33% for IMAT. Optimized composite plan of AP/PA and IMAT were created for each phantom geometry, balancing PTV dose,lung and cord tolerance. For the patient plan, the prescription dose was 60Gy, and the composite AP/PA‐IMAT plan gave a maximum cord dose of 45Gy with mean lungdose =16.1Gy and V20Gy =23%. Patient completed treatment and had no acute or late lung or spinal cord toxicities. Conclusions: IMAT was found to be useful in sparing the cord but not the lung, while AP/PA maximizes lung sparing but not the cord. A combination of AP/PA and IMAT provides an optimal class solution for treatment of lungcancer where mediastinum nodal irradiation is indicated.
TU‐D‐ValA‐09: Optimized Removal of the Tongue‐And‐Groove Underdose Via Constrained Partial Synchronization and Variable Depth Recursion33(2006); http://dx.doi.org/10.1118/1.2241578View Description Hide Description
Purpose:IMRT leaf sequencers that remove tongue‐and‐groove underdosages (TGU) significantly increase the number of segments by using full synchronization (FS). This work aims to minimize the increase in the number of segments by using Constrained Partial Synchronization (CPS) and Variable Depth Recursion (VDR). Method and Materials: Adjacent leaves simultaneously expose cells in an intensity map to deliver Vs, determined by cell values V0 and V1 (V0 <= V1). The tongue‐and‐groove ratio, TGR, is (V1‐Vs)/(V0‐Vs). In FS, TGR = infinity, while for CPS it is constrained to be greater than the minimum required to remove TGUs. The TGU for various TGRs were measured with film to determine this minimum TGR. The extraction and sweep leaf sequencing processes used in VDR optimization were modified to use CPS. The algorithm was tested on 1400 random maps (3 to 16 levels) and 42 clinical maps. Results: For a Siemens MLC, the TGUs are 18%, 4%, and 0, for TGRs of 1, 1.5, and 2, respectively. For the clinical cases, the average increase in the number of segments relative to VDR with no synchronization was 7% for TGR=1.5, and 32% for VDR with FS. A fully synchronized sweeping window produces 170% more segments. For the random maps, an unsynchronized sweeping window produced 1.5 to 2.5 times as many segments as VDR with CPS. Similar results can be obtained for interdigitating MLCs. This has implications for direct aperture optimization (DAO) algorithms that use the sweeping window as a starting point (Pinnacle), for which TGU has been observed. The concept of CPS can be applied to DAO algorithms, by choosing appropriate levels for each segment. Conclusion: CPS combined with VDR removes the tongue‐and‐groove underdose while minimizing the number of segments.
- Teletherapy Delivery III and Dose Calculation
TH‐D‐ValA‐01: Preliminary Investigation of Multi‐Pass Respiratory Gated Helical Tomotherapy (MRG‐HT)33(2006); http://dx.doi.org/10.1118/1.2241923View Description Hide Description
Purpose: To investigate the feasibility of multi‐pass respiratory gated helical tomotherapy (MRG‐HT) technique by measuring the dosimetric degradation caused by target motion, and the resulting dosimetric improvements made by the MRG‐HT technique Method and Materials: MRG‐HT was simulated on the HT unit (Tomotherapy Inc.) by delivering a sequence of HT plans modified by blocking beam projections remaining outside the gating window (50 % duty cycle, end‐expiration gating). After each plan delivery, the couch was reset and the succeeding plan was commenced at a different respiratory phase to allow the previously “blocked” beam projections to fall within the gating window. EDR2 film was placed inside a moving phantom (Modus Medical Devices Inc.), which was driven sinusoidally in the longitudinal direction. For each of the two scenarios with different experimental parameters, the resulting dose profiles along the axis of gantry rotation were obtained for a stationary target treated with the original plan, a moving target treated with the original plan, and a moving target treated with the MRG‐HT technique. Results: In the first scenario, target motion led to underdosing 7.5 +/− 0.5 mm of the target and overdosing additional 2.0 +/− 0.5 mm of normal tissues. With the gating technique, these values were reduced to 4.5 +/− 0.5 mm and 1.0 +/− 0.5 mm, respectively. Further improvement could be achieved using a smaller gating window. In the second scenario, selected parameters caused significant dose modulations of 8.5 +/− 1.0 % inside the target, which was effectively eliminated by the MRG‐HT technique. Conclusion: MRG‐HT seems to be a very promising technique for gated HT. Its dosimetric improvement was demonstrated in the form of reducing target underdosing and dose modulations. In future, more sophisticated experiments will be made to simulate lungcancer treatment, including the effect of non‐sinusoidal motion and lung inhomogeneity.
33(2006); http://dx.doi.org/10.1118/1.2241924View Description Hide Description
Purpose: In recent years there has been an explosion of research work concerning the topic of charged ion acceleration using high‐power lasers. The maximum particle energy and the shape of the distribution function are the two main parameters influencing the potential utilization of the new technology in radiation therapy. The energy spectrum depends on the quality of the laser beam as well as the geometric shape of the interaction target. Recent experiments have shown that it is possible to generate quasi‐monoenergetic protons using a double target system. However, influence of the laser prepulse has not yet been fully quantified. The purpose of this study is to find how the laser prepulse changes the property of the target and under which conditions can one expect the highest proton energy. Method and Materials: A radiation‐magnetodynamics code with inline atomic kinetic modeling was used to simulate interaction of the laser prepulse with initially cold double‐layer target. The simulated target system consists of an aluminum substrate on which a thin layer of hydrogen is located. Different initial substrate thicknesses and laser contrast ratios were simulated to find an optimal interaction conditions. Results of hydrodynamic simulations were used as initial conditions for the simulation of the main pulse interaction with the altered target. Results: It is shown that interaction of the prepulse with the target leads to its partial ionization with subsequent formation of a shock wave. Propagating shock wave destroys initial configuration of matter in a substrate, greatly expanding it in laser's propagation direction. Results of simulations suggest that maximal proton energy is achieved using lasers with smallest contrasts, incident on targets of thicknesses 1–2 μm. Conclusions: Presence of laser prepulse impedes acceleration efficiency of protons. Minimizing the contrast ratio or shortening the duration of the prepulse should result in higher maximal proton energies.
TH‐D‐ValA‐03: An Improved Lung Model, Incorporating Realistic Random Anatomical Features, for Monte Carlo‐Based Dosimetry33(2006); http://dx.doi.org/10.1118/1.2241925View Description Hide Description
Purpose: To investigate the effects of incorporating a more physically‐realistic lung model, preserving random anatomical features of the lung, on MC‐based dose distributions. Methods: A random lung model was built based on morphological data. The model homogenizes the lung parenchyma with structures of “chunk” sizes less than 0.05 cm, and models all larger chunks (branches of the bronchial and vessel trees, up to ∼1.5 cm) as randomly‐positioned 2‐D cylinders. The MC code PENELOPE was employed to calculate dose distributions in a water phantom containing a lung region, modeled by either a homogenized lung (as used in conventional planning) or the random lung model. Dose calculations used 6 and 18 MV photon beams with four different field sizes.Results: Depth dose curves in the random lung model illustrate significant perturbations when the structure size is comparable to the field size. For the 1×1 cm field size, large differences (up to 34% of Dmax) exist in the largest structures due to the loss of CPE with small field size. For large field sizes (10×10 cm or higher), little difference is observed between the random and the homogeneous models. The additional attenuation of the large structures also results in a region of dose reduction behind the lung.Conclusion: A new random lung model reveals significant dose perturbations from the homogeneous model, and shows that the homogeneous model breaks down when the field size is comparable to the structure size. This work is of importance in IMRT planning, where beamlets are used, or in the treatment of small tumors, where small field sizes are used in the planning. This work suggests that in such cases, a more precise description of the lung geometry, e.g. a high resolution CT‐based pixel‐by‐pixel density map, may be necessary for accurate dosimetry.
33(2006); http://dx.doi.org/10.1118/1.2241926View Description Hide Description
Purpose: To quantitatively evaluate the consistency of cone beam CT(CBCT) data and the deviation from helical CT if used for calculating dose to heterogeneous material. Methods and Materials: A Gammex RMI 467 tissue characterization phantom was used to generate CT numbers for both the Elekta XVI CBCT and the Picker 5000CT. For both devices, the RMI phantom was positioned using the manufacturer's recommendations. Hounsfield numbers (HU) were obtained using a default window and level and a ROI of 1cm diameter over the center of the rod (insert). The numbers were recorded and a HU vs. electron density correction curve was generated. For the cone beam portion, we used the Elekta XVI CBCT with M10 collimator and no filter. CBCT data, conventional CT data and correction curves were transferred to our CMS XIO treatment planning system. To verify the consistency of CBCT numbers, we added a 1cm of bolus to the phantom and repeated the procedure. A single field treatment plan was generated with heterogeneity and non heterogeneity corrections for CBCT and CT datasets, and for nonbolused and bolused geometries. The Gammex phantom tissue equivalent rods were modified to accept TLD for measurement verification. Results: Calculated doses for treatment plans generated without heterogeneity correction for data sets from CT and CBCT agreed within ∼1%. CBCT numbers for the same insert differed with a small change in setup (ie. adding 1cm bolus). When heterogeneity correction was turned on, a difference of 14% was found between CT and CBCT nonbolused plans. TLD measured doses agreed with those from the calculated CT set. Conclusion: Without heterogeneity correction, the dose calculation with CT and CBCT has excellent agreement suggesting a consistency in contour acquisition. One has to be extremely cautious in using CBCT data for heterogeneity corrections since significant error can occur.
33(2006); http://dx.doi.org/10.1118/1.2241927View Description Hide Description
Purpose: To characterize the CT‐density conversion for a cone‐beam CT(CBCT) system, and to evaluate the accuracy of dose calculation using CBCTimages in head & neck (H&N) and abdominal sites. Method and Materials: A small (16 cm) and a large (32 cm) acrylic phantom with inserts of known electron densities were scanned on conventional CT and CBCT scanners. CBCTimages were acquired using full and half fan (small phantom only), with and without bow‐tie filter. CT‐density conversion tables were measured for each combination. Head and pelvis phantoms, and H&N and prostate patients were imaged.IMRT treatment plans were designed based on conventional CT and transferred to CBCT after image registration (deformable registration for H&N patient). The CBCT dose distributions were then recalculated using the measured CT‐density curves. Results: For the small phantom, minimal differences were found between CT and CBCT numbers. Different conditions (half/full fan, with/without bow‐tie) resulted in maximum differences of approximately 100 HU. For the large phantom, differences between the CT numbers and CBCT numbers were greater than 350 HU at both low and high densities. For the H&N patient and the head phantom, dose distributions calculated on CT and CBCT differed by less than 2%. For the pelvis phantom and prostate patient, there were substantial differences in the dose distributions. In the worst case, using CBCTimages and the measured CBCT‐density conversion curve, the mean prostate dose was 14% higher than that calculated using the conventional CT. Due to artifacts in the CBCT, the differences in dose were larger using the CBCT‐density curves than using the default CT‐density curves. Using unit density for CBCT is a good approximation for pelvis dose calculation. Conclusion: Artifacts in CBCTimages and CT‐density conversion may lead to considerable errors in dose calculation using the large pelvis CBCTimages.
TH‐D‐ValA‐06: A Novel, Heterogeneity Inclusive, Pencil‐Beam Based Algorithm to Improve Lung IMRT Using the Corvus Planning System33(2006); http://dx.doi.org/10.1118/1.2241928View Description Hide Description
Purpose: We investigate a new finite‐size pencil‐beam algorithm for calculating absorbed photondose in heterogeneous media of arbitrarily varying density for inverse planning in CORVUS treatment planning system and evaluate its performance modeling heterogeneous systems and in optimization of an IMRTlung plan. Method and Materials: A new FSPB is developed by extending a phenomenological model (Med. Phys 26:1893–1990, 1999) for the central‐axis absorbed dose in therapeuticphoton beams for heterogeneous media. The model's parameters are rescaled based on the density of the medium. A differential equation is introduced to model the interface build‐up processes of CAX primary and scatter dose. Primary dose profiles are calculated using density‐dependent kernel integration, interpolated in the FSPB axis direction and evaluated depending on the density at the point of interest. Scatter dose profiles are computed using scatter integration and evaluated locally. Results: The new heterogeneity inclusive FSPB was implemented in a development version of CORVUS. Original and new FSPB dose calculations were compared with Monte Carlo calculations performed using PEREGRINE. For a heterogeneous semi‐slab phantom and for an IMRTlung plan, the dose distribution generated by the new FSPB agrees well with MC results, while the original one shows substantial discrepancies. IMRT plan optimizations were carried out using both original and new FSPB, and then a final dose calculation was performed using PEREGRINE. The plan calculated using the new FSPB shows better target conformality than the one computed using the original FSPB. Conclusion: The new FSPB possesses greatly improved accuracy as demonstrated in a variety of phantom and patient cases, both for dose calculation and IMRT optimization. FSPB best features were preserved with little extra computational overhead promising accurate and fast inverse planning and real‐time dose sculpting and dose volume histogram manipulation.
Research sponsored by North American Scientific, NOMOS Radiation Oncology Division.
TH‐D‐ValA‐07: Clinical Use of Monte Carlo in Proton Therapy: Pencil‐Beam Algorithm Vs Monte Carlo for Proton Therapy of Skull‐Base and Para‐Spinal Tumors33(2006); http://dx.doi.org/10.1118/1.2241929View Description Hide Description
Introduction: Pencil beam algorithms rely on kernels to model proton range in density‐scaled water equivalent material. Monte Carlo dose calculation methods are more accurate by design. This study addresses the issue of clinical significance of differences between a commercial pencil‐beam algorithm and Monte Carlo dose calculation.
Skull‐base or para‐spinal tumors are challenging for dose calculations due to interfaces between high and low density areas in the irradiated volumes. In addition, for para‐spinal cases, there are often metallic implants that not only distort the CTimage but also affect the accuracy of dose calculations. Methods:Protontreatment planning is done at our institution by using FOCUS/XiO (CMS Inc.). To do full Monte Carlo based dose calculation the treatment head was modeled including a simulation of the modulator wheel rotation as well as aperture and compensator. The patient CT data information was converted into materials with explicit element composition and density. All secondary particles were tracked.
Treatment plans designed with the treatment planning program were recalculated with the Monte Carlo code and compared by analyzing dose distributions and dose volume histograms. Results: In general, we found good agreement. Small differences are in part due to the difference between dose to medium and dose to water. Significant differences were found in and near air cavities as well as in areas affected by high density implants. Conclusion: For some tumors in the head and neck region and near the spine, Monte Carlo based dose calculation shows significant differences compared to pencil‐beam algorithms. Consequently, at our institution, a sub‐set of the patient population receives Monte Carlo dose calculation to assist in treatment planning.
TH‐D‐ValA‐08: Experimental Evaluation of the Accuracy of Contralateral Lung Dose Calculations for IMRT Plans33(2006); http://dx.doi.org/10.1118/1.2241930View Description Hide Description
Purpose: To experimentally determine the accuracy of contralateral lung dose calculated by Eclipse for mesothelioma IMRT plans. Materials and Methods: Two planning techniques were used to create 9 mesothelioma IMRT plans in Eclipse, which were then applied to a Rando torso phantom. Group A consisted of 5 plans with 160° of the contralateral side free from entrance beams while the 4 Group B plans had 80°–90° free. MOSFETs were placed at 10 coplanar points within the contralateral lung of the phantom, and doses measured for each field for each plan. These were compared with doses calculated using the pencil beam algorithm (modified Batho) in Eclipse. Two controls (ipsilateral AP/PA and 40×40cm AP/PA) were evaluated. Results: The 40×40cm AP/PA control plan showed good agreement between Eclipse and experimental results with the difference being 1.04% +/− 3.62% (1 stdev), while the ipsilateral only AP/PA plan showed a disagreement of −33.3% +/− 6.9% (Eclipse was cold compared with measurements) suggesting poor scatter modeling. The overall average difference between measured and calculated daily doses of Group A and B plans was −21.1% +/− 15.7% (−6.98cGy +/− 5.2cGy) and −9.8% +/− 29% (−6.58cGy +/− 19.5cGy), respectively. The differences were dependent on whether the points were in or out of individual fields. The average difference for points in field for Group A and B plans were −5.5% +/−23.4% (−1.169 +/− 4.99cGy) and +3.04% +/− 37.9% (1.57 +/− 19.6cGy), respectively. This increased to −49.2% +/− 10.9% (−5.8 +/− 1.28cGy) and −41% +/− 14.9% (−8.66 +/− 3.12 cGy), respectively, for points outside the field. Conclusions:Eclipse adequately models lung dose in field, but underestimates scatter dose out of field by up to 50%. This should be considered when evaluating contralateral lung DVH's, especially when attempts are made to use IMRT to lower mean lung dose.
TH‐D‐ValA‐09: An Objective Approach to Establishing Tolerances On Photon Beam Modeling Using the Equivalent Uniform Dose33(2006); http://dx.doi.org/10.1118/1.2241931View Description Hide Description
Purpose: To perturb a photon beam model in a controlled manner and to examine the consequences for the Equivalent Uniform Doses (EUD) of the target and organs at risk in external beam radiation therapy of the prostate. Method and Materials: We have developed seven similar but different therapy beam models in the Pinnacle® Treatment Planning System. One model generates beam data close to the golden data provided by Varian® and serves as the reference model for this study. The six other models are modifications of the reference model designed to result in controlled deviations of a particular region of the dose profiles (descending depth dose, build‐up, horns, tail, penumbra and field width). We have analyzed the consequences of planning with these perturbed models on the quality of 4 prostate treatment plans in terms of the EUDs of the PTV, rectum and bladder in comparison with the reference model. Monitor units were kept constant for all plans. Results: to maintain a change in the EUD to the prostate, bladder or rectum of less than 2%, tolerances on the various regions of the dose profiles are as follows: descending depth dose 2%; horns 3%; field width ±1mm. Deviations in the build‐up region and tail of ±10% and ±5% respectively did not change the EUDs of any structure by more than 2%. Conclusion: Currently accepted tolerances on photon beam modeling are broadly internally consistent in so far as they result in similar effects on plan dosimetry, at least for 4 field conformal prostate treatments.
- IMRT Delivery
TH‐E‐ValA‐01: On the Dose Delivered to a Moving Target When Employing Different IMRT Delivery Mechanisms33(2006); http://dx.doi.org/10.1118/1.2241970View Description Hide Description
Purpose: To investigate the influence of target motion on dose distributions generated using unmodulated open fields, solid intensity modulator (SIM), Step and Shoot MLC (SMLC) and dynamic MLC (DMLC). Method and Materials: For two lungcancer cases, four treatment plans were generated using Pinnacle3 7.9t consisting of an open field, SIM, SMLC and DMLC delivery on a Varian Clinac 600C/D equipped with a 120 leaf Millennium MLC. The coordinates (x, y, z, t) of the 4D motion trace for each of the tumors were determined using 4D‐CT from which a 4D motion kernel was generated. For each beam used in the experiment, the beams‐eye view tumor motion due to breathing was simulated using a computerized 2D tabletop apparatus. A MAPcheck diode array was incorporated into the apparatus for dose distribution analysis. Each of the four static treatment plans was delivered to the breathing MAPcheck ten times at various points of the breathing cycle. Results: The variation in diode dose readings within the tumor motion envelope was compared for the open field, solid, segmented, and dynamic IMRT deliveries. The open field provided the most uniform dose to the entire set of tumor mimicking diodes followed by SIM, SMLC, and DMLC IMRT, respectively. On an individual diode by diode basis over ten trials, the open field had the smallest average coefficient of variation of 0.122% followed by SIM (0.98%), SMLC (2.22%) and DMLC (3.88%) IMRT delivery, respectively. Conclusion: For the three IMRT delivery methods (SIM, SMLC, and DMLC), SIM consistently provided a more uniform dose to the tumor over many trials. SMLC performed as well as the solid modulators in many cases or was slightly out performed by SIM. DMLC consistently delivered the least uniform dose to the tumor over many trials.
33(2006); http://dx.doi.org/10.1118/1.2241971View Description Hide Description
Purpose: Intensity‐modulated proton therapy (IMPT) will improve the conformality of protonradiotherapy while preserving target homogeneity and low integral dose characteristics. IMPT is currently delivered using a single scanned pencil beam by placing Bragg peak spots at predetermined points in the patient. Very short pulse lengths and low repetition rates will characterize inexpensive compact proton beam systems of the future such as the dielectric wall accelerator. Their pulse structure is not amenable to scanning a single spot. The use of multiple intensity‐modulated pencil beams delivered in a fan beam is a simple approach to IMPT that can be used to upgrade current proton systems or with future accelerators. Methods and Materials: A fan beam is created with a pair of quadrupolemagnets aligned so that the second magnet amplifies beam divergence initiated by the first magnet. A set of multiple range‐shifters simultaneously adjusts penetration of multiple pencil beams. The actuation for the range‐shifters is done out of the plane of the fan beam. The intensity is modulated for any pencil beam by placing a thickness, greater than the range in the protons, in that portion of the fan beam. This achieves binary (on‐off) modulation as is used in x‐ray tomotherapy. Results: Calculations show that a spot delivery rate increase of 8 times can be achieved with this system which would allow larger target volumes to be practically delivered with IMPT or used with pulsed systems with low repetition rates. The system could be used to deliver spot scanning with multiple pencil beams simultaneously or used to deliverproton tomotherapy. Conclusions: An intensity‐modulated proton therapy system, based on multiple pencil beams issuing from a fan beam geometry, would decrease the delivery time for either continuous or low repetition pulse systems.