Validation of the dosimetry of total skin irradiation techniques by Monte Carlo simulation

Abstract Purpose To validate the dose measurements for two total skin irradiation techniques with Monte Carlo simulation, providing more information on dose distributions, and guidance on further technique optimization. Methods Two total skin irradiation techniques (stand‐up and lay‐down) with different setup were simulated and validated. The Monte Carlo simulation was primarily performed within the EGSnrc environment. Parameters of jaws, MLCs, and a customized copper (Cu) filter were first tuned to match the profiles and output measured at source‐to‐skin distance (SSD) of 100 cm where the secondary source is defined. The secondary source was rotated to simulate gantry rotation. VirtuaLinac, a cloud‐based Monte Carlo package, was used for Linac head simulation as a secondary validation. The following quantities were compared with measurements: for each field/direction at the treatment SSDs, the percent depth dose (PDD), the profiles at the depth of maximum, and the absolute dosimetric output; the composite dose distribution on cylindrical phantoms of 20 to 40 cm diameters. Results Cu filter broadened the FWHM of the electron beam by 44% and degraded the mean energy by 0.7 MeV. At SSD = 100 cm, MC calculated PDDs agreed with measured data within 2%/2 mm (except for the surface voxel) and lateral profiles agreed within 3%. At the treatment SSD, profiles and output factors of individual field matched within 4%; dmax and R80 of the simulated PDDs also matched with measurement within 2 mm. When all fields were combined on the cylindrical phantom, the dmax shifted toward the surface. For lay‐down technique, the maximum x‐ray contamination at the central axis was (MC: 2.2; Measurement: 2.1)% and reduced to 0.2% at 40 cm off the central axis. Conclusions The Monte Carlo results in general agree well with the measurement, which provides support in our commissioning procedure, as well as the full three‐dimensional dose distribution of the patient phantom.


| INTRODUCTION
Total skin electron irradiation (TSEI or TSI) has been employed for more than 50 yr as one of the most effective treatment techniques of malignant skin diseases such as mycosis fungoides, and cutaneous lymphomas. 1,2 The clinical goal is to deliver a uniform dose to the whole skin, without damaging inner organs. According to the recommendation of AAPM Task Group No.30 (TG-30), the field size of the composite electron beam at the patient treatment plane must be approximately 200 cm in height by 80 cm in width to encompass the largest patient. Within the rectangular field, a vertical uniformity of ±8% and a horizontal uniformity of ±4% over the central 160 by 60 cm area are desired. 1 TG-30 also contains explicit requirements of penetration depth and accompanying megavoltage x-ray background dose: a penetration depth ranges from approximately 5 to 15 mm or more than 50% isodose surface encompasses most lesion; the x-ray contamination exposed to the rest of body should be as low as reasonably achievable, with a level <1% of the electron dose at dose maximum.
Multiple techniques have been used for TSI treatment. The most commonly used technique is called Stanford six dual-field method. [1][2][3] In this method, the patient stands on a platform at an extended source source-to to-skin distance (SSD) and the platform should be optimized based on individual facility room design and linac to maximize the field uniformity. Other techniques also exist that are variant of the Stanford method. 4,5 For example, the platform on which the patient stands can rotate at a constant speed while radiation is on. In addition, some facilities utilize an additional scattering filter that can degrade the energy and broaden the beam, 4 ensuring an appropriate dose homogeneity. Hence, a small treatment room with a shorter SSD can be used to perform TSI treatment.
Although both Stanford technique and rotational technique 4 provide a uniform dose distribution at the treatment plane, it requires that the patient remains standing for the entire treatment duration, which could be a safety issue for patients who are too weak to stand and therefore unable to endure the procedure. To tackle the problem, a TSI lay-down technique was first proposed by Wu et al. 6 and further modified by Deufel and Antolak,7,8 which maintained the advantage of the Stanford technique while allowing the patient on a more comfortable lying position during the treatment. In this technique, AP/PA fields were delivered with the patient's umbilicus positioned directly under the Linac head; LAO/LPO/RAO/RPO fields were delivered at a 60°gantry angle with junction fields. Furthermore, Deufel and Antolak 7 proposed a novel hybrid method, in which a customized filter was designed to broaden the electron beam to compensate for the reduced SSD (target to floor distance is usually <230 cm). As a result of the addition of the filter, a single beam can be used for oblique fields to satisfy the dose uniformity requirement, eliminating the need for field junctioning, hence the setup time was reduced and treatment efficiency can be significantly improved.
Currently, there is no commercial treatment planning system for TSI. The commissioning of the TSI is specific to each linac room and a comprehensive list of dosimetric measurements is required. The treatment planning is largely based on the measurement data from the commissioning process and doses at a few representative points over the skin are calculated and measured for each patient.
Monte Carlo (MC) simulation is a sophisticated method in dose estimation and has been gradually accepted as an alternative dose calculation in the radiotherapy field. 9  the collimators (X and Y jaws) were set to the standard 36 × 36 cm 2 field at isocenter and electron applicator were removed to maximize the file size. MLC was fully retracted. The beams were weighted equally with gantry angles set to 251°and 289°(tilted ±19°from 270°) for optimal dose uniformity. The patient stands on a platform that can rotate at steps of 60°, and the patient's surface is positioned 300 cm away from the source (200 cm from isocenter). To improve the treatment efficiency and reduce patient fatigue, we delivered the total six pairs of electron beams for the first and second fraction, when in vivo dose measurements were performed; then we split a single fraction into two sub-fractions: three pairs of beams at 120°apart with twice the MUs were used on one day and the other three pairs on the other day, and they alternated at each fraction so a complete six pairs of beams are used to achieve prescribed dose after every two fractions.

2.A.2 | TSI lay-down technique
Treatment was performed on the same TrueBeam Stx in 6 MeV HDTSe mode. Ten electron beams were used in this technique, based on the relative position between patient and linac. Secondary collimators (X and Y jaws) were set to 30 × 40 cm 2 at isocenter and electron applicator was removed. A customized Cu scattering filter is placed on the interface mount: A 0.25 mm copper disk is positioned between two 1 mm polycarbonate rectangular layers, and corners of polycarbonate layers were trimmed by 5 cm. 8 The setup of the TSI lay-down technique is shown in Fig. 1;, the patient is modeled by a 30 cm diameter cylinder. For anterior-posterior (AP) and posterior-anterior (PA) positions, the patient's umbilicus is positioned below the isocenter at SSD of approximately 195 cm, and the patient is oriented perpendicular to the linac waveguide so the cranial-caudal axis is in the plane of gantry rotation. The beams from AP and PA are identical, each consists of three overlapping sub-beams with gantry angles of 300°, 0°, and 60°, named AP_G300, AP_G0, AP_G60, and similarly PA_G300, PA_G0, and PA_G60, respectively. Each individual beam is carefully weighted to achieve a uniform dose relative profile at 1cm depth (close to d max ).  Table 1. The amount of MU in lay-down technique is approximately twice of those in the stand-up technique, this is partly due to the attenuation effect of the copper filter, and the use of three sub-fields in the AP/PA direction rather than two in the stand-up technique. This may lead to an increased risk of ozone production in the treatment room, which was found to be within acceptable limits during commissioning. However, since the patient just needs to lie on the platform and flipped for one

2.B | Monte Carlo simulation
Several previous studies show that MC-based dose calculations are sensitive to details of the source as well as the geometry of the linac head. 15 Since the electron beam is normally shaped by the jaws and applicators, field size larger than 25 × 25 cm 2 is rarely used in the clinic for standard electron therapy. Only few MC studies have covered the MC simulation of large electron fields. 15,16 In order to investigate the reliability of the simulation, we also used two independent MC codes, EGSnrc and VirtuaLinac, to simulate the Linac head.

2.B.1 | Original phase space files
The TrueBeam phase space files of 6 MeV electron beam, which were originally recorded at a plane just above the movable jaw at 26.7 cm away from the target and 73.3 cm from the isocenter using the Geant4 (version 10.0. patch1) code, were provided by the vendor in International Atomic Energy Agency (IAEA) compatible format. 17 To avoid recycling the particles during the end-user simulation, we first concatenated 20 phase space files to generate a source file containing the type (electron, photon, and positron), energy, positions, and direction cosines of more than 7.5 × 10 8 pseudo-particles in the region of interest.

2.B.2 | EGSnrc
The simulation was performed primarily within the EGSnrc environment, 11 which has been accepted as the gold standard in the radio-    Unless explicitly stated, the results were from the EGSnrc code.

2.C.1 | Phase space file analysis
To study the effect of the Cu filter on beam characteristics, we ana-

2.C.6 | Investigation on the effect of patient size
In order to study the dosimetry of TSI technique as a function of different patient sizes, we computed the dose distribution in the cylindrical phantoms with diameters of 20, 25, 30, 35, and 40 cm.
Gantry angle, phantom location on the floor, and MU prescription were the same with the original setup during the commissioning; however, SSDs and relative incident angle to the phantom may change. The quantitative analysis focused on three main parameters that may potentially vary with the phantom size: absolute dose at d max , which affects the actual dose delivery on the patient; B-factor, which represents the cumulative effects of all fields; x-ray contamination, which indicates the potential hazard to the inner organs. with 2%/1 mm criteria was achieved for all points excluding few surface voxels, EGSnrc(EGS) and Virtualinac(VL) agreed within 1%/ 1 mm; for Cu field PDDs, a larger deviation up to 4.5% was observed in the first few voxels of build-up region, but the agreements were still within 2%/1 mm in the fall-off region. Parameters of d max , R 80 , R 50, and contamination at 5 cm depth are listed in Table 2 and the differences between simulation and measurement were all and filtered field;, several representative profiles with collimator setting that would be used for the TSI treatment are shown in Fig. 3(b), and the differences between simulation and measurement were within 3%. At SSD = 100 cm, full width half maximum (FWHM) of the filtered field increased by 44% in average with the same collimator setting, indicting a significant broadening effect of the filter.

3.C | Results in the water phantom at extended
SSDs for TSI  Table 3 shows the output factor at the central axis of the beam from both simulation and measurement, statistical uncertainties were included in the output from the simulation. For AP_G60 and AP_G300 fields, the output was determined by taking the average value of the individual measurement of each field. While both doses at depths of 0 cm (surface) and d max of 1 cm were measured, the surface output was used for final MU determination since the prescription point locates at the surface region.

3.D | Composite dose in cylindrical phantom at extended SSDs for TS
In the cylindrical phantom, simulated PDDs of individual field and composite PDDs are shown in Fig. 6; comparison of d max , R 80 , and surface dose between simulation and measurement are listed in Table 4 and differences are all within 2 mm. We found that d max of

4.A | Error analysis
Although both MC codes were used successfully to obtain an acceptable match to measured electron dose without using the applicator, Virtualinac was found to be slightly more accurate for the large field electron dose calculation; this is likely due to a well-

| CONCLUSIONS
We developed a MC framework to simulate two methods for TSI. TSI stand-up technique is easy to implement and most commonly used in the clinic. TSI lay-down technique is implemented when the patient is unable to endure the stand-up technique. MC simulation was used to examine and validate the dosimetry on water phantoms for the single field and combined fields. The results of our MC calculations were found to be in generally good agreement with the measurements, which provides secondary support in our commissioning procedure. To the best of our knowledge, this is the first time that a TSI lay-down treatment is simulated in MC and compared with the standup technique. In addition to those measurable quantities, the MC simulation can provide further information such as the full dose distribution of the patient phantom, and the ability to investigate and optimize techniques such as different filter design, SSD, and field size variations.

ACKNOWLEDGMENTS
We thank Daren Sawkey, Ph.D., for the support and guidance on the use of VirtuaLinac Monte Carlo package. We thank Rath Gunasingha, Ph.D., for providing workstations in the EGSnrc simulations.

CONF LICT OF I NTEREST
There is no conflict of interest for all authors.