Out‐of‐field doses from radiotherapy using photon beams: A comparative study for a pediatric renal treatment

Abstract Purpose First, this experimental study aims at comparing out‐of‐field doses delivered by three radiotherapy techniques (3DCRT, VMAT (two different accelerators), and tomotherapy) for a pediatric renal treatment. Secondly, the accuracy of treatment planning systems (TPS) for out‐of‐field calculation is evaluated. Methods EBT3 films were positioned in pediatric phantoms (5 and 10 yr old). They were irradiated according to four plans: 3DCRT (Clinac 2100CS, Varian), VMAT (Clinac 2100CS and Halcyon, Varian), and tomotherapy for a same target volume. 3D dose determination was performed with an in‐house Matlab tool using linear interpolation of film measurements. 1D and 3D comparisons were made between techniques. Finally, measurements were compared to the Eclipse (Varian) and Tomotherapy (Accuray) TPS calculations. Results Advanced radiotherapy techniques (VMATs and tomotherapy) deliver higher out‐of‐field doses compared to 3DCRT due to increased beam‐on time triggered by intensity modulation. Differences increase with distance to target and reach a factor of 3 between VMAT and 3DCRT. Besides, tomotherapy delivers lower doses than VMAT: although tomotherapy beam‐on time is higher than in VMAT, the additional shielding of the Hi‐Art system reduces out‐of‐field doses. The latest generation Halcyon system proves to deliver lower peripheral doses than conventional accelerators. Regarding TPS calculation, tomotherapy proves to be suitable for out‐of‐field dose determination up to 30 cm from field edge whereas Eclipse (AAA and AXB) largely underestimates those doses. Conclusion This study shows that the high dose conformation allowed by advanced radiotherapy is done at the cost of higher peripheral doses. In the context of treatment‐related risk estimation, the consequence of this increase might be significative. Modern systems require adapted head shielding and a particular attention has to be taken regarding on‐board imaging dose. Finally, TPS advanced dose calculation algorithms do not certify dose accuracy beyond field edges, and thus, those doses are not suitable for risk assessment.


| INTRODUCTION
Recently, early diagnosis and improvements in treatment techniques and therapeutic strategies have led to an increasing success of cancer treatments. 1 Consequently, patients' life expectancy following cancer is increasing and more patients will survive long after treatments. Among the different techniques involved, radiation therapy is nowadays used in more than 50% of cancer treatments 2,3 and its efficacy has been largely acknowledged. The most advanced techniques enable conformal dose distribution to the tumor volume reducing adjacent organs doses. This precision in dose delivery is carried out thanks to multiple beam incidences, beams' intensity modulation, and a precise patient positioning using on-board imaging systems. However, modern radiotherapy inevitably increases the volume of normal tissue exposed to ionizing radiation from treatment beams themselves, from out-of-field radiation, 4 and also from daily imaging. 5 The exposure of normal tissues may lead to adverse effects following treatments. Although deterministic effects only appear at high doses, stochastic effects such as second cancers or cardiac diseases can be related to low doses exposure and can occur years after treatment. As life expectancy after cancer is increasing and modern techniques are now widely used, late treatment-related side effects are becoming an important concern. Besides, this issue is even more important for young patients because of their high organs radiosensitivity and their long-term survival following primary cancer. 6 The risks related to medical exposure are determined from epidemiological studies which aim at correlating doses to observed side effects.
Thus, a precise knowledge of the dose distribution delivered to patients and to healthy organs is needed to enhance the prediction of adverse effects risks and their reduction. Modern treatment planning systems (TPS) enable a precise determination of the doses delivered within the treatment beams but for locations outside the treatment field edge, some tend to largely underestimate the doses 7 . In addition, the accuracy of out-of-field dose determination of some recent TPS algorithms is not well known. Therefore, they cannot be used to estimate related adverse effects risks. Furthermore, there is few experimental data regarding out-of-field doses acquired in realistic conditions (anthropomorphic phantoms), in particular for recent radiotherapy techniques and for pediatric patients.
The Institut Curie (Paris, France) has a long experience in pediatric oncology. In this institute, photon radiotherapy is particularly involved in the treatment of pediatric abdomino-pelvic cancers and among them, our work focuses on renal tumors which are frequently represented. Before the generalization of intensity modulation, the standard radiotherapy modality used to treat those cancers was conventional three-dimensional conformal treatment (3DCRT) consisting in two anteroposterior beams. Now, they are typically treated using volumetric arc therapy (VMAT) using conventional linear accelerators and sometimes using tomotherapy units.
In this context, the first aim of this experimental study is to compare the peripheral doses delivered by two advanced techniques (VMAT and tomotherapy) and by conventional 3DCRT for pediatric abdomino-pelvic cancers treatment. In this study, in addition to the Clinac 2100, the latest generation accelerator Halcyon version 2.0 (Varian Medical System), newly installed at the Institut Curie was also included to deliver VMAT plans as its use may be considered in the future for pediatric patients. This accelerator is a single-beam 6 MV-FFF system equipped with dual-layer multi-leaf collimator (MLC) and fixed primary and secondary collimators. 8,9 Clinically relevant treatment plans according to commonly used planning protocols were prepared for two anthropomorphic pediatric phantoms (5 and 10 yr old). Peripheral dose comparison was performed thanks to EBT3 film measurements placed in these phantoms in order to overcome the possible inaccuracy of TPS algorithms outside the treatment field. 7 The measured out-of-field doses were reconstructed in 3D by linear interpolation using original in-house Matlab scripts. 10 The second aim of the present study is to evaluate the accuracy of two TPS for out-of-field dose calculation. For that purpose, TPS calculated doses obtained for the two phantoms and the different techniques are compared to the experimental doses obtained from the film measurements. The TPS Eclipse™ (Varian MS) and tomotherapy (Accuray Inc., Sunnyvale, CA, USA) were studied.

2.A | Treatment planning
The heterogeneous ATOM ® dummies (CIRS, Norfolk, VA) representing children aged 5 and 10 yr old were used in this study. They were scanned with a 3 mm slice thickness using an Aquilion LB (Toshiba) CT scanner at the Institut Curie (Paris, France) according to a clinical protocol. Two pediatric patients, morphologically similar to the two phantoms and previously treated at the Institut Curie for renal tumors, were then selected. Those patients were extracted from a pediatric patient cohort treated for renal tumors gathered by a radiotherapist. Different parameters were analyzed in that cohort: patient age, patient size, patient weight, PTV volume, and PTV localization.
The two selected patients corresponded to the median case for each age group. Using the Eclipse™ TPS (Varian Medical System), the patients' CT images and their outlined structures were registered on the phantoms' CT images using deformable registration. In order to obtain realistic planning treatment volumes (PTV) and organs shapes in the phantoms, the registered structures were copied on phantoms CT images and manually processed to avoid any overlap. The PTV including the clinical target volume (CTV) with a 5 mm margin and the vertebrae in its immediate proximity were of 726 cc and of 382 cc for the 5-year-old and 10-year-old phantoms, respectively.
For the 10-year-old phantom, the PTV is central and located slightly to the left of the lumbar rachis. For the 5-year-old phantom, the target extends in front of the left kidney (Fig. 1). The structures obtained from the patients' files are the liver, the kidneys, the spinal cord near the CTV, the digestive system, and the vertebrae. As for clinical practice, only a small portion of the patient body is scanned, additional organs were manually outlined within the phantoms such as eyes, thyroid, lungs, heart, bladder, and rectum (Fig. 1). The RVR COLNOT ET AL.
| 95 (remaining volume at risk), defined by the difference between the volume of the body contour and that of the CTV and the outlined organs, was also studied in this work.
For each phantom, four treatment plans were developed and optimized in accordance with clinical constraints and the institute pediatric experience. Table 1 lists the parameters of the four treatment plans. The VMAT and 3DCRT plans were optimized using the AAA algorithm (Eclipse™, Varian) and then recalculated using Acuros ® (except for the Halcyon plan) in dose to water by keeping the same MU per beam. Tomotherapy plans were calculated with the dedicated Tomotherapy TPS (Accuray). The dose calculation grid includes the entire body of phantoms. To study the two jaw modes available on the tomotherapy system (static and dynamic), the static mode was used for the 5-year-old plan whereas the dynamic delivery was used for the 10-year-old plan. The jaw width was set to 2.5 cm for the two phantoms. All the plans were generated with 6 MV beams, the tomotherapy and Halcyon units use FFF beams. 21 Gy (14 fractions) were prescribed to the PTV and the dose was normalized so that the mean PTV dose matches the prescribed dose. In the following, the VMAT plan performed with the Clinac 2100CS accelerator is referred as VMAT plan whereas the one performed with the Halcyon system is referred as Halcyon plan.

2.B | Radiochromic film dosimetry and phantom irradiation
For dose measurements on phantoms, EBT3 radiochromic films (Ashland) were used. EBT3 films were chosen for out-of-field dose measurements as their response proves to have little dependence with energy. [12][13][14][15][16] Moreover, they allow 2D dose measurements and they can easily be housed in anthropomorphic phantoms. Measurements were performed according to a rigorous protocol (cutting, calibration, readout) developed in our laboratory. 11 In particular, it is based on a pixel-to-pixel background subtraction method with the use of the red channel only in order to overcome the limits of the multichannel correction method at low doses. 15 Measurements uncertainty was assessed as described in. 11 This protocol leads to dose measurements with a standard deviation of 2.9% (1 sigma) in the 0.5-4.0 Gy dose range (reaching 4.5% (1 sigma) for doses below 0.5 Gy).
Besides, this protocol was confronted to a Farmer ion chamber for off-axis dose measurements in a previous work 17 and demonstrates good agreement with maximum discrepancies of 20% up to 6 cGy.
The calibration was made at the Institut Curie with a Clinac 2100CS linear accelerator (Varian MS); calibration films were irradiated the same day as the phantoms. The films were calibrated from approximately 1.3 cGy (2 MU) to 24 Gy using 18 dose points (2 films per dose) between tissue-equivalent slabs at 10 cm in depth with a 10 × 10 cm 2 field (SSD = 100 cm). Besides, two overlapping fit curves (from 0.013 Gy to 5 Gy and from 3 Gy to 24 Gy) have been used to make sure that low doses are perfectly represented by the final calibration curve. For the phantoms' measurements, 31 films and 25 films were cut to fit between the slices of the 10-year-old and 5-year-old phantoms, respectively (Fig. 2 left). Those measurements were performed for each radiotherapy technique.
The phantoms were filled with the EBT3 films and irradiations were performed according to the prepared treatment plans at the Institut Curie (Fig. 2 right). For each plan, the 14 fractions were delivered successively without any imaging between them. Thus, no doses from on-board imaging were delivered to the films. The irradiations lasted one hour and a half, one hour, and less than half an hour for the tomotherapy, VMAT with Clinac 2100CS, and the 3DCRT and Halcyon plans, respectively.

2.C | Data analysis
Data and films analyses were carried out using in-house Matlab scripts (Matlab R2013b and Image Processing Toolbox, The Math-Works, Inc.). 10 The different steps of the analysis are described as follows: 1. Conversion of films optical density into absorbed dose to water,   (σ = 85.6%), 93.3% (σ = 148%), and 16.4% (σ = 108%) for tomotherapy, VMAT, and Halcyon, respectively. Moreover, the largest discrepancies reach 350%, 592%, and 424% at 30 cm from field edge for tomotherapy, VMAT, and Halcyon, respectively. T A B L E 1 Treatment plans prepared in this study for the two phantoms and for the four radiotherapy techniques.  Table 3.

Radiotherapy
Given the results presented in Tables 2 and 3    3D doses delivered outside the treatment field border Table 4 reports the mean doses reconstructed in 3D expressed in percent of the prescribed dose for organs located outside beams in the 10-year-old phantom. They range from 0.02% of the prescribed dose (pituitary and eye) to 1.70% of the prescribed dose (heart).

1-D comparison
The lowest dose levels are obtained for the head and neck organs (4 mGy for pituitary).

3.B | Performance of TPS algorithms
In this part, only the results obtained with the 5-year-old phantom are presented as results are identical between the two phantoms.  Regarding the AAA algorithm, the agreement between calculation and measurements proves to be different depending on radiotherapy technique (Fig. 5).

Organs doses
Mean dose differences obtained in 2D at films levels and in 3D using the reconstruction tool for organs located outside the beams are given in Table 6. Thus, the outlined organs which are not crossed by films are consequently not listed in this  is observed for the TPS. Table 7 summarizes dose discrepancies between TPS and measurements for the different organs delineated in the phantom with in 2D

Organs doses
at the film levels and in 3D outside the beams. Finally, Figure 8 represents DVHs obtained with the TPS and the 3D reconstruction tool.
T A B L E 5 Relative difference (%) between 3D interpolated doses for the 10-year-old phantom.

VMAT/ 3DCRT
Tomotherapy/ 3DCRT Halcyon/ 3DCRT accelerator. These lower dose levels are also due to lower beam-on time and reduced stray radiation induced by FFF beams in comparison with the FF beams of conventional accelerators. This larger exposure of healthy tissues induced by intensity-modulated radiotherapy due to MU and beam-on time increase was also shown by Wang and Xu. 18 They also demonstrate the more beams are used in IMRT, the more peripheral doses are delivered to patients. Moreover, Stathakis et al. 19 report that the whole-body dose is increased by a factor 3 with IMRT in comparison to 3DCRT for a prostate treatment. They also highlight that this factor depends on beam energy, field size, and collimator rotation which affects leakage through this piece.
Although tomotherapy needs long treatment times, the recent Hi-Art system includes an additional shielding composed of lead disks and of a 20 cm thick tungsten system. These components enable to reduce peripheral doses delivered far from the target in comparison with conventional accelerators. As former tomotherapy systems did not include this shielding, they used to deliver higher peripheral doses. 20 This shielding is the reason why tomotherapy better spares tissues located far from PTV in our work. Ramsey  than with dynamic jaws. 26,27 This result is illustrated by Fig. 9 showing calculated profiles on the 5-year-old phantom using fixed and dynamic jaws with the Tomotherapy TPS. In our study, this TPS proves to be reliable for the out-of-field dose calculation: it allows to confirm the increase in penumbra triggered by static jaws and the slight increase of peripheral dose. The dose difference between the two profiles decreases with distance to the target reaching 7% at 30 cm from field edge.
As a consequence, our study reports an increase in dose to the pelvis: this result raises concern given the gonads radio-sensitivity that may lead to an increased side effects risk. Moreover, a larger dose to the heart has been measured in tomotherapy in comparison to VMAT and 3DCRT for this phantom. This increase in dose can be involved in a higher cardiac risk following treatment. 28  In fact, the works of Mackie et al. 41 and Papanikolaou et al. 42 , on which this convolution/superposition algorithm is based, report point kernels calculated up to 30 cm from the voxel center in the lateral direction and up to 85 cm in depth. The lateral distance is in agreement with our results and those obtained by Schneider et al. 40

| CONCLUSION
In this work, out-of-field doses delivered by four radiotherapy techniques have been evaluated by means of film measurements in pediatric anthropomorphic phantoms. Modern techniques enable higher dose conformation compared to 3DCRT; however, this improvement is done at the cost of higher peripheral doses. This study points out a factor of 3 on dose between modern treatments and 3DCRT for organs located far from PTV. This larger exposure raises major concern as it might increase the risks of developing adverse effects following radiotherapy especially for pediatric patients surviving long after the treatments. Among advanced radiotherapy techniques, the latest generation Varian Halcyon system seems a promising treatment option as delivering lower dose levels than conventional accelerator and incorporating kV-CBCT imager. This technique was also the most efficient in terms of treatment delivery time. To our knowledge, our study is the first to report healthy tissue doses delivered with the new Varian Halcyon system. It enables to situate this new treatment option in relation to the other older techniques. The original methods developed and applied to renal pediatric treatments in this work can be used to study other radiotherapy techniques or tumor localization. The conclusions obtained in this work cannot be easily extended to other localizations in particular regarding the dose distribution close to the PTV as it depends on PTV size and its relative distance to normal organs. However, the results obtained away from PTV are more general as the doses delivered are highly dependent on accelerator head design and treatment efficacy and less on morphology. These general findings are important to provide clinical data regarding modern pediatric radiotherapy treatments using latest generation accelerators.
In light of the results presented in this study, it would be interesting to complete the dosimetric comparison between techniques by adding daily imaging dose determination as they largely contribute to increase the exposure of healthy tissues to radiation.
Finally, TPS performances were evaluated in terms of normal tissue dose calculations. Unlike Eclipse™(AAA and Acuros ® ), the Tomotherapy TPS enables a precise dose calculation up to 30 cm from field edge. This study is useful in providing clinical information on the uncertainties of healthy organ doses calculated by two modern treatment planning system for which very few data are available in literature.