C Houser

The use of mixed-integer programming for inverse treatment planning with pre-defined field segments

Complex intensity patterns generated by traditional beamlet-based inverse treatment plans are often very difficult to deliver. In the approach presented in this work the intensity maps are controlled by pre-defining field segments to be used for dose optimization. A set of simple rules was used to define a pool of allowable delivery segments and the mixed-integer programming (MIP) method was used to optimize segment weights. The optimization problem was formulated by combining real variables describing segment weights with a set of binary variables, used to enumerate voxels in targets and critical structures. The MIP method was compared to the previously used Cimmino projection algorithm. The field segmentation approach was compared to an inverse planning system with a traditional beamlet-based beam intensity optimization. In four complex cases of oropharyngeal cancer the segmental inverse planning produced treatment plans, which competed with traditional beamlet-based IMRT plans. The mixed-integer programming provided mechanism for imposition of dose–volume constraints and allowed for identification of the optimal solution for feasible problems. Additional advantages of the segmental technique presented here are: simplified dosimetry, quality assurance and treatment delivery.

A deterministic iterative least-squares algorithm for beam weight optimization in conformal radiotherapy

Currently, inverse treatment planning in conformal radiotherapy is, in part, a trial-and-error process due to the interplay of many competing criteria for obtaining a clinically acceptable dose distribution. A new method is developed for beam weight optimization that incorporates clinically relevant nonlinear and linear constraints. The process is driven by a nonlinear, quasi-quadratic objective function and the solution space is defined by a set of linear constraints. At each step of iteration, the optimization problem is linearized by a self-consistent approximation that is local to the existing dose distribution. The dose distribution is then improved by solving a series of constrained least-squares problems using an established method until all prescribed constraints are satisfied. This differs from the current approaches in that it does not rely on the search for the global minimum of a specific objective function. Essentially, our proposed objective function can be construed as a functional that comprises a class of dose-based quadratic objective functions. Empirical adjustment for appropriate model parameters in the construction of objective function is minimized, since these parameters are in effect adaptively adjusted during optimization. The method is robust in solving difficult clinical cases using either aperture or pencil beam based planning techniques for intensity-modulated radiation therapy.

Reference dosimetry in clinical high‐energy electron beams: Comparison of the AAPM TG‐51 and AAPM TG‐21 dosimetry protocols

A comparison of the determination of absorbed dose to water in reference conditions with high-energy electron beams (Enominal of 6, 8, 10, 12, 15, and 18 MeV) following the recommendations given in the AAPM TG-51 and in the original TG-21 dosimetry protocols has been made. Six different ionization chamber types have been used, two Farmer-type cylindrical (PTW 30001, PMMA wall; NE 2571, graphite wall) and four plane parallel (PTW Markus, and Scanditronix-Wellhöfer NACP, PPC-05 and Roos PPC-40). Depending upon the cylindrical chamber type used and the beam energy, the doses at dmax determined with TG-51 were higher than with TG-21 by about 1%-3%. Approximately 1% of this difference is due to the differences in the data given in the two protocols; another 1.1%-1.2% difference is due to the change of standards, from air-kerma to absorbed dose to water. For plane-parallel chambers, absorbed doses were determined by using two chamber calibration methods: (i) direct use of the ADCL calibration factors N(60Co)D,w and Nx for each chamber type in the appropriate equations for dose determination recommended by each protocol, and (ii) cross-calibration techniques in a high-energy electron beam, as recommended by TG-21, TG-39, and TG-51. Depending upon the plane-parallel chamber type used and the beam energy, the doses at dmax determined with TG-51 were higher than with TG-21 by about 0.7%-2.9% for the direct calibration procedures and by 0.8%-3.2% for the cross-calibration techniques. Measured values of photon-electron conversion kecal, for the NACP and Markus chambers were found to be 0.3% higher and 1.7% lower than the corresponding values given in TG-51. For the PPC-05 and PPC-40 (Roos) chamber types, the values of kecal were measured to be 0.889 and 0.893, respectively. The uncertainty for the entire calibration chain, starting from the calibration of the ionization chamber in the standards laboratory to the determination of absorbed dose to water in the user beam, has been analyzed for the two formalisms. For cylindrical chambers, the observed differences between the two protocols are within the estimated combined uncertainty of the ratios of absorbed doses for 6 and 8 MeV; however, at higher energies (10< or =E< or =18 MeV), the differences are larger than the estimated combined uncertainties by about 1%. For plane-parallel chambers, the observed differences are within the estimated combined uncertainties for the direct calibration technique; for the cross-calibration technique the differences are within the uncertainty estimates at low energies whereas they are comparable to the uncertainty estimates at higher energies. A detailed analysis of the reasons for the discrepancies is made which includes comparing the formalisms, correction factors, and quantities in the two protocols, as well as the influence of the implementation of the different standards for chamber calibration.

Impact of transrectal ultrasound- and computed tomography-based seed localization on postimplant dosimetry in prostate brachytherapy

PURPOSE To study the impact of seed localization, as performed by different observers using linked (125)I seeds, on postimplant dosimetry in prostate brachytherapy and, to compare transrectal ultrasound (TRUS)-based with CT-based approach for the dosimetric outcomes. METHODS AND MATERIALS Nineteen permanent prostate implants were conducted using linked (125)I seeds. Postimplant TRUS and CT images were acquired and prostate glands were, after implantation, delineated on all images by a single oncologist, who had performed all 19 seeding procedures. Six observers independently localized the seeds on both TRUS and CT images, from which the principle dosimetric parameters V(100) (volume of prostate that received the prescribed dose), V(150) (volume of prostate that received 150% of the prescribed dose), and D(90) (minimal dose delivered to 90% of the prostate) were directly calculated for each patient. A single-factor analysis of variance was first applied to determine interobserver variability in seed localization. A nonparametric comparison of the approach using TRUS and CT was then carried out by the Wilcoxon paired-sample test. RESULTS Analysis from the analysis of variance for TRUS showed that the null hypothesis for equal means, could not be rejected for all six observers based on a significance level alpha=0.05. TRUS-based and CT-based approaches were then cross compared by the Wilcoxon paired-sample test, which suggested that the null hypothesis was insignificant for V(100) and D(90), but was significant for V(150). CONCLUSIONS Both TRUS- and CT-imaging modalities provided indistinguishable postimplant dosimetry results as far as V(100) and D(90) were concerned. There was comparable observer independence between TRUS- and CT-based seed localization for linked-seed implant procedures. With other advantages that TRUS-imaging modality had over CT in the evaluation of postimplant dosimetry, TRUS would be a preferred choice in conjunction with linked seeds for intraoperative procedures in prostate brachytherapy.

SU‐FF‐T‐382: Special Dosimetric/Measurement Considerations in Commissioning a Novel Integrated MiniMLC Linear Accelerator

Purpose: To present an overview of commissioning a novel integrated miniMLC linear accelerator system, including the complete steps in final clinical implementation. Special precautions due to the collimator design are identified. Reference datasets for electrons are included. Method and Materials: Elekta BeamModulator has 40 pairs of 4mm leaves. There are no movable backup diaphragms and the field is defined only by the open leaves. Smaller field sizes and more precise leaf positioning (<1mm) are specified. The film scanning system was verified to within 0.2mm accuracy. Microchambers were used for in‐water scanning. Brass and graphite miniphantoms were constructed for head‐scatter factor measurements of fields from 0.8cm×0.8cm to 16cm×16cm. A complete set of scan and point measurement data was collected for photons and electrons. Beams were modeled in XIO and PrecisePlan and an independent calculation program (RADCALC). Before actual clinical implementation, periodic QA baselines were established, and site specific IMRT plans and QA measurements were performed on phantoms. Results and Conclusion: An extensive and comprehensive program was employed in commissioning. Beam data collection and calibration were internally verified by at least two independent measurements and checked against standard datasets. Treatment planning system modeling followed the guidelines of TG53. When compromises had to be made, the best fits were chosen for situations mimicking IMRT segments (4cm×4cm and 4.8cm×4.8cm). QA measurements of 3D conformal plans and IMRT plans achieved the following agreement statistics: 3mm DTA, 3% difference, produced pass rate of 97.8% average (2.6%STD). Dose point measurements with chambers agreed to plan values within 3.6%. After comparisons between 3D dose, independent monitor unit(MU) and manual calculations; Radcalc and XIO independent MU calculation programs were deemed unusable (the discrepancy reaching 5.4%), due to incorrect modeling of head‐scatter factors for this collimator. Additionally, measured electron cone factors varied up to 13% from standard Elekta linacs.

Output variation from an intensity modulating dynamic collimator

Intensity modulated radiation therapy (IMRT) offers a method of delivering radiation dose conforming to the shape of targets while minimizing the dose to the surrounding tissue and nearby critical organs. One popular device is the NOMOS MIMiC Collimator coupled to the CORVUS treatment planning system. The MIMiC collimator, mounted on a linac head, opens and closes one or more of its 40 small leaves as determined by the planner while the linac delivers radiation and the gantry rotates. This dynamic IMRT allows the intensity to be modulated yielding a highly conformal dose distribution. However, the dose output becomes a function of the detailed manner in which the leaves open and close, since the opening and closing are not instantaneous. We investigate the effect of switch rates and delay in the open/close events on the output profiles. The output is enhanced as the switch rate increases. The enhancement factor at any point of measurement is dependent on its distance from the central plane. We interpret these variations in terms of a simple model, which includes the effect of leaf travel time during the process of opening and closing. We also include the time delay in establishing the specified pressure in the pneumonic device, which controls the opening and closing of the leaves. The information presented here offers a means for incorporating these output changes into the planning system. This may avoid the current situation where many patient plans need to be renormalized based on the actual measurement taken during the delivery of the specified intensity pattern to a phantom.

Comparison of three IMRT inverse planning techniques that allow partial esophagus sparing in patients receiving thoracic radiation therapy for lung cancer

The purpose of this study is to compare 3 intensity-modulated radiation therapy (IMRT) inverse treatment planning techniques as applied to locally-advanced lung cancer. This study evaluates whether sufficient radiotherapy (RT) dose is given for durable control of tumors while sparing a portion of the esophagus, and whether large number of segments and monitor units are required. We selected 5 cases of locally-advanced lung cancer with large central tumor, abutting the esophagus. To ensure that no more than half of the esophagus circumference at any level received the specified dose limit, it was divided into disk-like sections and dose limits were imposed on each. Two sets of dose objectives were specified for tumor and other critical structures for standard dose RT and for dose escalation RT. Plans were generated using an aperture-based inverse planning (ABIP) technique with the Cimmino algorithm for optimization. Beamlet-based inverse treatment planning was carried out with a commercial simulated annealing package (CORVUS) and with an in-house system that used the Cimmino projection algorithm (CIMM). For 3 of the 5 cases, results met all of the constraints from the 3 techniques for the 2 sets of dose objectives. The CORVUS system without delivery efficiency consideration required the most segments and monitor units. The CIMM system reduced the number while the ABIP techniques showed a further reduction, although for one of the cases, a solution was not readily obtained using the ABIP technique for dose escalation objectives.

SU-GG-T-48: Impact of Seed Localization On Post-Implant Dosimetry in Prostate Brachytherapy

Purpose: To study the impact of seed localization, as performed by different observers using linked I‐125 seeds, on post‐implant dosimetry in prostate brachytherapy and to compare TRUS‐based with CT‐based approach for the dosimetric outcomes. Method and Materials: Permanent prostate implant is conducted using linked I‐125 seeds. Both post‐implant TRUS and CTimages were acquired and the prostate glands were delineated on each of those images by a single oncologist, who performed the seeding procedures for all 19 patients under study. Six observers then independently localized the seeds on both TRUS and CTimages, based on which the principle dosimetric parameters V100, V150 and D90 were directly calculated for each patient. A single‐factor analysis of variance (ANOVA) is first applied to determine inter‐observer variability in seed localization. A non‐parametric comparison of the approach using the two imaging modalities, TRUS and CT, is then carried out by the Wilcoxon paired‐sample test. Results: Analysis from the ANOVA for TRUS (V100: P=0.655; V150: P=0.994; D90: P=0.734) and CT (V100: P=0.901; V150: P=0.999; D90: P=0.99) shows that the null hypothesis for equal means, cannot be rejected for all six observers based on a significance level α=0.05. TRUS‐based and CT‐based approaches are then cross‐compared by the Wilcoxon paired‐sample test, which suggests that the null hypothesis is not rejectable for V100 and D90, but is for V150. Conclusion: Both TRUS and CTimaging modalities provide indistinguishable post‐implant dosimetry results as far as V100 and D90 are concerned. TRUS‐based seed localization has comparable observer independence as CT‐based seed localization for linked‐seed implant procedures. In view of other advantages that TRUSimaging modality has over CT in the evaluation of post‐implant dosimetry,TRUS can therefore be an alternative gold standard to CT and would be a preferred choice together with linked‐seed for intra‐operative procedures ins prostate brachytherapy.

SU-FF-J-12: A Phantom Study to Compare 2D Electronic Portal Imaging with 3D KV Cone-Beam Imaging

Purpose: To compare patient positioning using a 2D Electronic Portal Imaging Device(EPID) with a 3D registration technique that uses a kV Cone Beam (CB) device. Introduction: The Elekta kV CB device represents a significant advancement in on‐line imaging for patient positioning and structure delineation. There exists an obvious interest in studying the robustness of this new technology. The phantom‐based study reported here compares the cone‐beam imaging capability with conventional EPID 2D imaging and registration. Additionally, a technique was developed to generate digitally reconstructed portal images (DRPIs) from the CBCT and compare them to EPIDimages.Materials and Methods: A Rando anthropomorphic phantom had surgical screws placed in the pelvic region. A treatment isocenter was selected in the pelvic region of the CT dataset of the phantom. The phantom was initially positioned using the CB volume imaging capability. Known shifts of the phantom were introduced and the phantom was repositioned using the CB. The CB corrections were compared to the corrections obtained using the internal fiducials and EPIDimaging. Additionally, the EPID‐based positioning was mimicked by simulating the 2D imaging using the 3D CB data set. The digitally reconstructed images(DRR) were generated from the CBCT set and used as DRPIs. They were compared to both the diagnostic CTDRRs and the EPIDimages.Results: and Conclusion: The CB and EPID shifts were well correlated as determined by linear regression analysis. The correlation between the DRPIs and the EPIDs was even higher using the same analysis. However, on the shift‐by‐shift basis the CB was more accurate than the EPID: it gave more accurate shift determination than the EPID for approximately 90% test shifts. The DRPIs had the advantage over the CB 2Dimages because they included the treatment isocenter, field edges and distance scale.