Use of Receiver Operating Curve Analysis and Machine Learning With an Independent Dose Calculation System Reduces the Number of Physical Dose Measurements Required for Patient-Specific Quality Assurance.

PURPOSE: Our purpose was to assess the use of machine learning methods and Mobius 3D (M3D) dose calculation software to reduce the number of physical ion chamber (IC) dose measurements required for patient-specific quality assurance during corona virus disease 2019. METHODS AND MATERIALS: In this study, 1464 inversely planned treatments using Pinnacle or Raystation treatment planning software (TPS) were delivered using Elekta Versa HD and Varian Truebeam and Truebeam STx linear accelerators between June 2018 and November 2019. For each plan, an independent dose calculation was performed using M3D, and an absolute dose measurement was taken using a Pinpoint IC inside the Mobius phantom. The point dose differences between the TPS and M3D calculation and between TPS and IC measurements were calculated. Agreement between the TPS and IC was used to define the ground truth plan failure. To reduce the on-site personnel during the pandemic, 2 methods of receiver operating characteristic analysis (n = 1464) and machine learning (n = 603) were used to identify patient plans that would require physical dose measurements. RESULTS: In the receiver operating characteristic analysis, a predelivery M3D difference threshold of 3% identified plans that failed an IC measurement at a 4% threshold with 100% sensitivity and 76.3% specificity. This indicates that fewer than 25% of plans required a physical dose measurement. A threshold of 1% on a machine learning model was able to identify plans that failed an IC measurement at a 3% threshold with 100% sensitivity and 54.3% specificity, leading to fewer than 50% of plans that required a physical dose measurement. CONCLUSIONS: It is possible to identify plans that are more likely to fail IC patient-specific quality assurance measurements before delivery. This possibly allows for a reduction of physical measurements taken, freeing up significant clinical resources and reducing the required amount of on-site personnel while maintaining patient safety.

A mathematical framework for virtual IMRT QA using machine learning.

PURPOSE: It is common practice to perform patient-specific pretreatment verifications to the clinical delivery of IMRT. This process can be time-consuming and not altogether instructive due to the myriad sources that may produce a failing result. The purpose of this study was to develop an algorithm capable of predicting IMRT QA passing rates a priori. METHODS: From all treatment, 498 IMRT plans sites were planned in eclipse version 11 and delivered using a dynamic sliding window technique on Clinac iX or TrueBeam Linacs. 3%/3 mm local dose/distance-to-agreement (DTA) was recorded using a commercial 2D diode array. Each plan was characterized by 78 metrics that describe different aspects of their complexity that could lead to disagreements between the calculated and measured dose. A Poisson regression with Lasso regularization was trained to learn the relation between the plan characteristics and each passing rate. RESULTS: Passing rates 3%/3 mm local dose/DTA can be predicted with an error smaller than 3% for all plans analyzed. The most important metrics to describe the passing rates were determined to be the MU factor (MU per Gy), small aperture score, irregularity factor, and fraction of the plan delivered at the corners of a 40 × 40 cm field. The higher the value of these metrics, the worse the passing rates. CONCLUSIONS: The Virtual QA process predicts IMRT passing rates with a high likelihood, allows the detection of failures due to setup errors, and it is sensitive enough to detect small differences between matched Linacs.

Dosimetric characterization of an image-guided stereotactic small animal irradiator.

Small animal irradiation provides an important tool used by preclinical studies to assess and optimize new treatment strategies such as stereotactic ablative radiotherapy. Characterization of radiation beams that are clinically and geometrically scaled for the small animal model is uniquely challenging for orthovoltage energies and minute field sizes. The irradiator employs a commercial x-ray device (XRAD 320, Precision x-ray, Inc.) with a custom collimation system to produce 1-10 mm diameter beams and a 50 mm reference beam. Absolute calibrations were performed using the AAPM TG-61 methodology. Beam's half-value layer (HVL) and timer error were measured with an ionization chamber. Percent depth dose (PDD), output factors (OFs) and off-axis ratios were measured using radiochromic film, a diode and a pinpoint ionization chamber at 19.76 and 24.76 cm source-to-surface distance (SSD). PDD measurements were also compared with Monte Carlo (MC) simulations. In-air and in-water absolute calibrations for the reference 50 mm diameter collimator at 19.76 cm SSD were measured as 20.96 and 20.79 Gy min(-1), respectively, agreeing within 0.8%. The HVL at 250 kVp and 15 mAs was measured to be 0.45 mm Cu. The reference field PDD MC simulation results agree with measured data within 3.5%. PDD data demonstrate typical increased penetration with increasing field size and SSD. For collimators larger than 5 mm in diameter, OFs measured using film, an ion chamber and a diode were within 3% agreement.

An x-ray image guidance system for small animal stereotactic irradiation.

An x-ray image-guided small animal stereotactic irradiator was developed and characterized to enable tumor visualization and accurate target localization for small field, high dose irradiation. The system utilizes a custom collimation system, a motorized positioning system (x, y, θ), a digital imaging panel and operating software, and is integrated with a commercial x-ray unit. The essential characteristics of the irradiator include small radiation fields (1-10 mm), high dose rate (>10 Gy min(-1)) and submillimeter target localization. The software enables computer-controlled image acquisition, stage motion and target localization providing simple and precise automated target localization. The imaging panel was characterized in terms of signal-to-noise ratio (SNR), contrast-to-noise ratio (CNR) and spatial resolution. Overall localization accuracy and precision were assessed. SNR, CNR and spatial resolution are 24 dB, 21 dB and 2.8 lp mm(-1), respectively, and localization accuracy is approximately 65 µm with 6 µm precision. With the aid of image guidance, system performance was subsequently used to evaluate radiation response in a rat orthotopic lung tumor effectively sparing normal tissues and in a mouse normal lung. The capabilities of 3D treatment and cone-beam computed tomography are presented for 3D localization and delivery as a work in progress.

Stereotactic radiotherapy for the treatment of pituitary adenomas.

PURPOSE: The aim of this study was to retrospectively review local control and morbidity following stereotactic radiotherapy (SRT) for pituitary adenoma. METHODS: Between 1997 and 2004, 39 patients with pituitary adenomas received SRT. Median age was 56 years (range: 13 to 90 years). Thirty-three patients underwent incomplete transsphenoidal surgery prior to SRT and six had unresectable tumors. The largest tumor dimension varied from 1.7 to 6 cm (median: 3 cm). Tumor volume varied from 1.2 to 56 mL (median 10.5 mL). Thirty-five tumors were < or = 1 mm from the optic chiasm/nerve. Thirty-three tumors were non-functional. SRT was delivered by a dedicated linear accelerator (Novalis, Heimstetten, Germany). Beam collimation was achieved by a fixed circular collimator (five patients) or a micro-multileaf collimator (34 patients). Total dose varied from 4500 to 5040 cGy (median: 4860 cGy) and was prescribed at the 90 % isodose line. RESULTS: After a median follow-up of 32 months (range: 12 to 94 months), the local control rate was 100 %. Tumor size was stable in 26 patients and decreased in 13 patients. Hormone normalization did not occur following SRT. New endocrine deficiency occurred in six patients. No patient developed cranial nerve injury or second malignancy following treatment. CONCLUSIONS: SRT achieves a high rate of local control and a low rate of treatment-induced morbidity. SRT is applicable to pituitary adenomas in close proximity to the optic apparatus and tumors in excess of three centimeters in the greatest dimension. Further follow-up is necessary to establish the long-term outcome following SRT for pituitary adenomas.

Linear accelerator stereotactic radiosurgery for the treatment of gelastic seizures due to hypothalamic hamartoma.

PURPOSE: There are reports of successful gamma-knife stereotactic radiosurgery (SRS) for the treatment of gelastic seizures associated with a hypothalamic hamartoma. The authors reviewed the results of linear accelerator (LINAC) radiosurgery for patients with medically refractory gelastic seizures due to a sessile hypothalamic hamartoma. METHODS: Three patients with gelastic seizures received SRS between 2003 and 2004. All patients had associated partial complex and/or generalized seizures. One patient demonstrated aggressive behavior. Sessile hamartomas varying in diameter from 6 to 14 mm were identified by MRI. SRS was delivered to a single isocenter by a dedicated LINAC equipped with either a circular beam collimator or a micromultileaf collimator. Patients received 1500 to 1800 cGy prescribed at the 90 to 95 % isodose line. Seizure outcome was scored according to Engel's classification. RESULTS: Two patients became free of gelastic and partial complex/generalized seizures seven and nine months after radiosurgery. These patients remain free of seizures at 17 and 15 months, respectively, after treatment (Engle Class IA). One patient experienced a decline in gelastic seizure frequency nine months after treatment (Engle Class II) without significant reduction in aggressive behavior. Follow-up MRI demonstrated no change in the size or signal characteristics of any tumor. No patient developed post-treatment cranial neuropathy or hypothalamic-pituitary suppression. CONCLUSIONS: LINAC SRS represents a safe and effective therapeutic alternative for patients with medically refractory gelastic seizures due to unresectable hypothalamic hamartomas. Radiosurgery is associated with a latency of several months from treatment to reduction in seizure frequency. Further follow-up is required to establish the duration of seizure control following radiosurgery.

A Monte Carlo tutorial and the application for radiotherapy treatment planning.

Monte Carlo-based treatment planning algorithms are advancing rapidly and will certainly be implemented as part of conventional treatment planning systems in the near future. This paper was designed as a basic tutorial for using the Monte Carlo method as applied to radiotherapy treatment planning. The tutorial addresses the basic transport differences between photon and electron transport as well as the sampling distributions. The implementation of a virtual linac source model and the conversion from the Monte Carlo source modeling reference plane into the treatment reference plane is discussed. The implementation of a thresholding algorithm for converting CT electron density to patient specific materials is also presented. A 6-field prostate boost treatment is used to compare a conventional treatment planning algorithm (pencil beam model) with a Monte Carlo simulation algorithm. The agreement between the 2 calculation methods is good based upon the qualitative comparison of the isodose distribution and the dose-volume histograms for the prostate and the rectum. The effects of statistical uncertainty on the Monte Carlo calculation are also presented.

Comparative behaviour of the dynamically penalized likelihood algorithm in inverse radiation therapy planning.

This paper presents a description of tests carried out to compare the behaviour of five algorithms in inverse radiation therapy planning: (1) The Dynamically Penalized Likelihood (DPL), an algorithm based on statistical estimation theory; (2) an accelerated version of the same algorithm: (3) a new fast adaptive simulated annealing (ASA) algorithm; (4) a conjugate gradient method; and (5) a Newton gradient method. A three-dimensional mathematical phantom and two clinical cases have been studied in detail. The phantom consisted of a U-shaped tumour with a partially enclosed 'spinal cord'. The clinical examples were a cavernous sinus meningioma and a prostate case. The algorithms have been tested in carefully selected and controlled conditions so as to ensure fairness in the assessment of results. It has been found that all five methods can yield relatively similar optimizations, except when a very demanding optimization is carried out. For the easier cases. the differences are principally in robustness, ease of use and optimization speed. In the more demanding case, there are significant differences in the resulting dose distributions. The accelerated DPL emerges as possibly the algorithm of choice for clinical practice. An appendix describes the differences in behaviour between the new ASA method and the one based on a patent by the Nomos Corporation.

Dynamic arc radiosurgery field shaping: a comparison with static field conformal and noncoplanar circular arcs.

PURPOSE: Recent advances in field-shaping technology and linac multileaf collimator (MLC) integration have resulted in new approaches to performing stereotactic radiosurgery. We present a modeling study comparing the absolute dose distributions from three radiosurgery delivery techniques: a conventional approach utilizing noncoplanar circular arcs, a static field conformal approach, and a dynamic arc field-shaping approach. In the latter, the MLC leaves more in a continuous fashion, conforming to the beam's-eye-view projection of the target at every increment along the path of an arc. METHODS AND MATERIALS: For the analysis, we devised a simulated target consisting of three overlapping spheres. This was chosen because it offered a straightforward planning approach for all three techniques, primarily the multiple isocenter approach. In addition, three representative cases were selected from the prior radiosurgery experience. These range in increasing size, from 0.50 to 9.79 cm(3), and in complexity, requiring from 3 isocenters to 16 in the case of circular arcs. In each situation, the goals were twofold: (1) to cover the entire volume with as high an appropriate isodose level (90% in the case of the conformal and dynamic arc techniques, 50% in the case of circular collimators) while (2) minimizing the dose to normal brain and where applicable, any adjacent radiation-sensitive structures. Because of the latter requirement, a single isocenter circular arc approach was ruled out for the analysis. RESULTS: In the case of large or irregularly shaped lesions, the circular arc technique requires multiple isocenters, producing a high level of dose heterogeneity within the target volume. Both the static field and dynamic arc conformal techniques, as with all single isocenter approaches, produce a highly homogeneous dose throughout the target region. For a given large dose, peripheral dose is decreased as additional beams or arc degrees are added with either of the conformal approaches. Dose--volume histogram analysis evaluating the peripheral dose shows that, in many cases, dose to surrounding structures can be reduced through the use of a conformal static or dynamic arc approach over the conventional multiple isocenter, circular arc techniques. CONCLUSIONS: Dynamic arc shaping is an efficient and effective method for accurately delivering a homogeneous target dose while simultaneously minimizing peripheral dose in radiosurgery applications.

Infrared patient positioning for stereotactic radiosurgery of extracranial tumors.

We report on a novel, non-invasive patient positioning system for radiosurgery of extracranial tumors. The system consisted of infrared cameras and reflective markers attached to the skin. Because localization accuracy is critical in radiosurgery, we performed a theoretical analysis of the accuracy of the system. A computer simulation program modeled errors in marker position, and was used to predict errors in targeting and study methods for minimizing errors. The use of redundant markers improved the overall accuracy of targeting. Experimental data was collected using a rigid torso phantom and correlated with theoretical results. The accuracy of the infrared system was compared with existing systems.

Radiosurgery and stereotactic radiation therapy of skull base meningiomas: proposal of a grading system.

OBJECTIVE: The development of a grading system to guide treatment selection, and predict treatment difficulty and outcome of skull base meningiomas infiltrating the cavernous sinus which are managed by stereotactic radiosurgery (SRS) and stereotactic radiotherapy (SRT), based on an 8-year experience with stereotactic radiation of skull base meningiomas. METHODS: T1 gadoliniun-enhanced magnetic resonance imaging (MRI) of 40 patients with skull base meningiomas, with or without prior surgery, who underwent radiosurgery or stereotactic radiation therapy from 1991 to 1998 at the UCLA Medical Center were reviewed, and the result of treatment was related to the tumor grade. Grade was based on tumor infiltration of the cavernous sinus and extension into adjacent structures. Treatment was performed with a linac-based system. The dose prescribed to the periphery of the tumor for SRS patients (n = 34) ranged from 12 to 22 Gy, and the maximum dose delivered to the tumor ranged from 24 to 46 Gy. SRT (n = 6). Treatment was planned using a single isocenter, usually prescribed to the 90% isodose volume, bringing the fractionation scheme to the maximal tolerance of the optic apparatus. The periphery dose ranged from 24 to 46 Gy with a maximum dose of 45 to 51 Gy. Clinical and MRI follow-up was performed every six months for the first 3 years and every year thereafter. RESULTS: Grade I meningiomas were restricted to the cavernous sinus (n = 12). Grade II cavernous sinus meningiomas extended to the clivus and/or the petrous bone, without compression of the brainstem (n = 9). Grade III meningiomas had superior and/or anterior extension with compression of the optic nerve or tract (n = 9). Grade IV tumors compressed the brain stem (n = 8), and Grade V were bilateral lesions (n = 2). Tumor control rates were 90% for Grade I, 86% for Grade II, 86% for Grade III, 42% for Grade IV and no control for tumors Grade V. Complications were not related to tumor grade. CONCLUSION: This grading system correlated with outcome and difficulty in planning radiosurgery. Failure of treatment was more likely to occur in patients with higher Grade tumors.

Radiosurgery performed with the aid of a 3-mm collimator in the subthalamic nucleus and substantia nigra of the vervet monkey.

OBJECT: Radiosurgery for functional neurosurgery performed using a linear accelerator (LINAC) has not been extensively characterized in preclinical studies. In the present study, the properties of a newly designed 3-mm-diameter collimator were evaluated in a dedicated LINAC, which produced lesions in the basal ganglia of vervet monkeys. Lesion formation was determined in vivo in three animals by examining magnetic resonance (MR) images to show the dose-delivery precision of targeting and the geometry and extent of the lesions. Postmortem immunohistochemical studies were conducted to determine the extent of lesion-induced radiobiological effects. METHODS: In three male vervet monkeys, the subthalamic nucleus (STN; one animal) and the pars compacta of the lateral substantia nigra (SN; two animals) were targeted by a Novalis Shaped Beam Surgery System that included a 3-mm collimator and delivered a maximum dose of 150 Gy. Magnetic resonance images obtained 4, 5, and 9 months posttreatment were reviewed, and the animals were killed so that immunohistological characterizations could be made. CONCLUSIONS: The generation of precise radiosurgical lesions by a 3-mm collimator was validated in studies that targeted the basal ganglia of the vervet monkey. The extent of the lesions created in all animals remained restricted in diameter (< 3 mm) throughout the duration of the studies, as assessed by reviewing MR images. Histological studies showed that the lesions were contained within the STN and SN target areas and that there were persistent increases in glial fibrillary acidic protein immunoreactivity. Increases in immunoreactivity for tyrosine hydroxylase, the serotonin transporter, and the GluR1 subunit of the alpha-amino-3-hydroxy-5-methyl-4-isoxazole propionate glutamate receptor in penumbral regions of the lesion were suggestive of compensatory neuronal adaptations. This radiosurgical approach may be of particular interest for the induction of lesions of the STN and SN in studies of experimental parkinsonism, as well as for the development of potential radiosurgical treatments for Parkinson disease.

Experimental radiosurgery simulations using a theoretical model of cerebral arteriovenous malformations.

BACKGROUND AND PURPOSE: A novel biomathematical arteriovenous malformation (AVM) model based on electric network analysis was used to investigate theoretically the potential role of intranidal hemodynamic perturbations in elevating the risk of rupture after simulated brain AVM radiosurgery. METHODS: The effects of radiation on 28 interconnected plexiform and fistulous AVM nidus vessels were simulated by predefined random or stepwise occlusion. Electric circuit analysis revealed the changes in intranidal flow, pressure, and risk of rupture at intervals of 3 months during a 3-year latency period after simulated partial/complete irradiation of the nidus using doses <25 and >/=25 Gy. An expression for risk of rupture was derived on the basis of the functional distribution of the critical radii of component vessels. The theoretical effects of radiation were also tested on AVM nidus vessels with progressively increasing elastic modulus (E:) and wall thickness during the latency period, simulating their eventual fibrosis. RESULTS: In an AVM with E=5. 0x10(4) dyne/cm(2), 4 (14.3%) of a total 28 sets of AVM radiosurgery simulations revealed theoretical nidus rupture (risk of rupture >/=100%). Three of these were associated with partial nidus coverage and 1 with complete treatment. All ruptures occurred after random occlusion of nidus vessels in AVMs receiving low-dose radiosurgery. Intranidal hemodynamic perturbations were observed in all cases of AVM rupture; the occlusion of a fistulous component resulted in intranidal rerouting of flow and escalation of the intravascular pressure in adjacent plexiform components. Risk of rupture was found to correlate with nidus vessel wall strength: a low E: of 1.9x10(4) dyne/cm(2) resulted in a 92.8% incidence of AVM rupture, whereas a higher E: of 7.0x10(4) dyne/cm(2) resulted in only a 3.6% incidence of AVM rupture. A dramatic reduction in rupture incidence was observed when increasing fibrosis of the nidus was modeled during the latency period. CONCLUSIONS: It was found that the theoretical occurrence of AVM hemorrhage after radiosurgery was low, particularly when radiation-induced fibrosis of nidus vessels was considered. When rupture does occur, it would appear from a theoretical standpoint that the occlusion of intranidal fistulas or larger-caliber plexiform vessels could be a significant culprit in the generation of critical intranidal hemodynamic surges resulting in nidus rupture. The described AVM model should serve as a useful research tool for further theoretical investigations of cerebral AVM radiosurgery and its hemodynamic sequelae.

A virtual source model for Monte Carlo modeling of arbitrary intensity distributions.

A photon virtual source model was developed for simulating arbitrary, external beam, intensity distributions using the Monte Carlo method. The source model consists of a photon fluence grid composed of a matrix of square elements, located 25-cm downstream from the linear accelerator target. Each particle originating from the fluence map is characterized by the seven phase space parameters, position (x, y, z), direction (u, v, w), and energy. The map was reconstructed from fluence and energy spectra acquired by modeling components of the linear accelerator treatment head using the Monte Carlo code MCNP4B. The effect of contaminant electrons is accounted for by the use of a sub-source derived from a phase-space simulation of a 25-MV linac treatment head using the code BEAM. The BEAM sub-source was incorporated into the MCNP4B phase-space model and is sampled using a field-size dependent sampling ratio. A Gaussian blurring kernel is convolved with the photon fluence map to account for the finite focal spot size and scattering effects from structures such as the flattening filter and MLC leaves. Depth dose and profile source calculations for 6-MV and 25-MV photon beams, for 5 x 5 cm2, 10 x 10 cm2, and 15 x 15 cm2 field sizes, are in good agreement with measurement and are well within acceptability criteria suggested by the AAPM Task Group Report No. 53. Irregular field calculations compared with film measurement and with a 3-D pencil beam algorithm show that the source model is capable of accurately simulating arbitrary MLC fields.

CT-based dosimetry calculations for 125I prostate implants.

PURPOSE: To evaluate the Monte Carlo code MCNP4B for low-energy brachytherapy calculations, including the effects of interseed attenuation and patient specific heterogeneities, on the calculated dose distribution from transperineal implantation of 125I. METHODS AND MATERIALS: The Monte Carlo code MCNP4B was used to model and benchmark the absolute dose distribution from two 125I brachytherapy seeds (model 6711 and 6702). Based upon the physical source model, the total photon intensity and differential energy spectrum were evaluated as a function of angle from the transverse bisector of the source. These spectral and intensity data were reformatted to produce probability distributions for sampling from a virtual point source. The virtual source model and a modified version of MCNP4B is then used for simulating arbitrary brachytherapy source configurations within a homogeneous or heterogeneous patient specific computed tomography (CT)-based lattice geometry. RESULTS AND CONCLUSION: Comparison with TG-43 data and the Monte Carlo calculations is excellent with MCNP4B predicting the radial dose function for the 125I 6711 and 6702 sources within 6% for all data points tested. Attenuation effects from neighboring seeds were investigated for pre- and postimplant seed distributions and found to be negligible. Preliminary dosimetry analysis of postimplant seed distributions comparing homogeneous water versus heterogeneous CT simulation geometries indicates an average decrease of approximately 5.6% for the volume of tissue irradiated to a prescription isodose line of 144 Gy.

Dose distributions using kilovoltage x-rays and dose enhancement from iodine contrast agents.

In x-ray phototherapy of brain tumours, the tumour is loaded with iodine and exposed to kilovoltage x-rays. Due to the high photoelectric cross sections of iodine, substantial photoelectric interactions occur. The flux of photoelectrons, characteristic x-rays and Auger electrons produce a localized dose enhancement. A modified computed tomography scanner, CTRx, can be used both for tumour localization and delivery of the dose enhancement therapy. Monte Carlo methods were employed to simulate the treatment of iodinated brain tumours with a CTRx. The calculated results reveal the effect of tumour iodine concentration on dose distribution, the degree of skull bone sparing with the application of multiple arcs, and the homogeneity of tumour dose distribution versus iodine concentration. A comparison with 10 MV stereotactic radiosurgery treatment shows the potential of CTRx treatment relative to conventional treatment modalities.

Fractionated stereotactic radiotherapy: rationale and methods.

Stereotactic radiosurgery (SRS) has become a widely accepted technique for the treatment intracranial neoplasms. Combined with modern imaging modalities, SRS has established its efficacy in a variety of indications. From the outset, however, it was recognized that the delivery of a single large dose of radiation was essentially "bad biology made better by good physics." To achieve the accuracy required to compensate for this biological shortcoming, the application of SRS has required that a neurosurgical head frame of some sort be rigidly attached to the patients head. Historically, this prerequisite has, primarily for practical reasons, precluded the delivery of multiple fractions over multiple days. With recent improvements in immobilization and repeat fixation, the good biology of fractionated delivery has been realized. This technique, which has come to be known as stereotactic radiotherapy (SRT), has significantly expanded the efficacy of the technique through the use of accurate physical targeting coupled with the basic radiobiological principles gleaned from decades of clinical experience.

A CT-based Monte Carlo simulation tool for dosimetry planning and analysis.

The Los Alamos code MCNP4A (Monte Carlo N-Particle version 4A) is currently used to simulate a variety of problems ranging from nuclear reactor analysis to boron neutron capture therapy. A graphical user interface has been developed that automatically sets up the MCNP4A geometry and radiation source requirements for a three-dimensional Monte Carlo simulation using computed tomography data. The major drawback for this dosimetry system is the amount of time to obtain a statistically significant answer. A specialized patch file has been developed that optimizes photon particle transport and dose scoring within the standard MCNP4A lattice geometry. The transport modifications produce a performance increase (number of histories per minute) of approximately 4.7 based upon a 6 MV point source centered within a 30 x 30 x 30 cm3 lattice water phantom and 1 x 1 x 1 mm3 voxels. The dose scoring modifications produce a performance increase of approximately 470 based upon a tally section of greater than 1 x 10(4) lattice elements and a voxel size of 5 mm3. Homogeneous and heterogeneous benchmark calculations produce good agreement with measurements using a standard water phantom and a high- and low-density heterogeneity phantom. The dose distribution from a typical mediastinum treatment planning setup is presented for qualitative analysis and comparison versus a conventional treatment planning system.