Comparison of treatment planning approaches for spatially fractionated irradiation of deep tumors

Abstract Purpose The purpose of this work was to compare the dosimetry and delivery times of 3D‐conformal (3DCRT)‐, volumetric modulated arc therapy (VMAT)‐, and tomotherapy‐based approaches for spatially fractionated radiation therapy for deep tumor targets. Methods Two virtual GRID phantoms were created consisting of 7 “target” cylinders (1‐cm diameter) aligned longitudinally along the tumor in a honey‐comb pattern, mimicking a conventional GRID block, with 2‐cm center‐to‐center spacing (GRID2 cm) and 3‐cm center‐to‐center spacing (GRID3 cm), all contained within a larger cylinder (8 and 10 cm in diameter for the GRID2 cm and GRID3 cm, respectively). In a single patient, a GRID3 cm structure was created within the gross tumor volume (GTV). Tomotherapy, VMAT (6 MV + 6 MV‐flattening‐filter‐free) and multi‐leaf collimator segment 3DCRT (6 MV) plans were created using commercially available software. Two tomotherapy plans were created with field widths (TOMO2.5 cm) 2.5 cm and (TOMO5 cm) 5 cm. Prescriptions for all plans were set to deliver a mean dose of 15 Gy to the GRID targets in one fraction. The mean dose to the GRID target and the heterogeneity of the dose distribution (peak‐to‐valley and peak‐to‐edge dose ratios) inside the GRID target were obtained. The volume of normal tissue receiving 7.5 Gy was determined. Results The peak‐to‐valley ratios for GRID2 cm/GRID3 cm/Patient were 2.1/2.3/2.8, 1.7/1.5/2.8, 1.7/1.9/2.4, and 1.8/2.0/2.8 for the 3DCRT, VMAT, TOMO5 cm, and TOMO2.5 cm plans, respectively. The peak‐to‐edge ratios for GRID2 cm/GRID3 cm/Patient were 2.8/3.2/5.4, 2.1/1.8/5.4, 2.0/2.2/3.9, 2.1/2.7/5.2 and for the 3DCRT, VMAT, TOMO5 cm, and TOMO2.5 cm plans, respectively. The volume of normal tissue receiving 7.5 Gy was lowest in the TOMO2.5 cm plan (GRID2 cm/GRID3 cm/Patient = 54 cm3/19 cm3/10 cm3). The VMAT plans had the lowest delivery times (GRID2 cm/GRID3 cm/Patient = 17 min/8 min/9 min). Conclusion Our results present, for the first time, preliminary evidence comparing IMRT‐GRID approaches which result in high‐dose “islands” within a target, mimicking what is achieved with a conventional GRID block but without high‐dose “tail” regions outside of the target. These approaches differ modestly in their ability to achieve high peak‐to‐edge ratios and also differ in delivery times.

regions outside of the target. These approaches differ modestly in their ability to achieve high peak-to-edge ratios and also differ in delivery times.

| INTRODUCTION
Spatially fractionated radiotherapy was initially used in the era of low energy x-rays to allow for safe delivery of radiation to internal tumors while allowing for skin and superficial tissue sparing. 1,2 The radiation beam, often delivered through a single field, was spatially fractionated into small beamlets by sieve-like blocking to form a grid pattern (GRID therapy). Tissues, such as skin, in the blocked portion of the treatment field were thought to promote healing/repair of normal tissues irradiated to high dose in the beamlet paths. In the era of skin-sparing megavoltage photon irradiation, GRID therapy has continued to play a role in radiation oncology, mostly in the treatment of bulky tumors. 1,2 The GRID treatment has typically been delivered in one high-dose (15-20 Gy) fraction, often followed by conventionally fractionated treatment courses which target the entire tumor. The radiation field is partitioned by commercially available blocks or by MLC leaf patterns which reproduce the effect of these blocks. 3 The treatments are often delivered in a single field.
Many studies have shown excellent tumor response results with this approach and there is a great deal of interest in the radiobiology of GRID treatments. [1][2][3][4][5] Upfront treatment with GRID may influence oxygenation in tumors as well as induce bystander effects. 4,5 Despite the successes with conventional GRID therapy, it has dosimetric limitations in the treatment of very deep-seated tumors.
F I G . 1. (a) Target arrangement for two virtual GRID phantoms consisting of seven cylinders (1-cm diameter) aligned longitudinally along the GTV in a honey-comb pattern, mimicking a conventional GRID block with 2-cm center-to-center spacing (GRID 2 cm ) and 3-cm center-to-center spacing (GRID 3 cm ), all contained within a larger cylinder. The larger cylinder is 8 and 10 cm in diameter for the GRID 2 cm and GRID 3 cm arrangements, respectively. Inline and crossline profiles are identified by the arrows. (b) Schematic of quantities evaluated for plan assessment. GTV is defined by the purple outline, valley (which is the GTV minus the GRID target) is defined by the light purple, and the GRID target is defined by the purple filled circles within the GTV. The ring is the 2-mm ring around the GTV. The peak-to-edge dose ratio (PEDR) is defined as the ratio of the mean dose to the GRID target to the mean dose of the valley. The peak-to-valley dose ratio (PVDR) is defined as the ratio of the mean dose to the GRID target to the mean dose to the 2-mm edge. The solid black line is the volume of the normal tissue.
Due to attenuation characteristics, the maximum dose from a single photon beam is delivered to a shallow depth, with decreasing dose as the deep tumor is reached. Some investigators have tried to mitigate this dosimetric problem with the use of parallel opposed fields. 6 Another solution, which maintains the unique geometry of high-dose "islands" within tumors inherent with GRID and takes advantage of high-energy x-ray attenuation features, is to paint three-dimensional target structures throughout the tumor which mimic conventional two-dimensional GRID blocks and then use conformal or intensity-modulated planning with the goal to deliver high doses to these areas (instead of the entire tumor). This approach has previously been studied using tomotherapy. 7 In this report, we compare a tomotherapy-based approach with two other approaches: volumetric modulated arc therapy (VMAT) and a simple 3D conformal (3DCRT)-based planning technique using cylindrical target structures.

2.A | GRID structures
The virtual GRID structures were generated by DICOMan which is an open source software (University of Arkansas, Little Rock, Arkansas). 8 DICOMan allows for a GRID target to be created within the target volume. The diameter of the cylinders and the center-to-center distance between the cylinders can be configured to the patient's anatomy. To mimic an ideal geometry, two virtual GRID phantoms were created consisting of seven cylinders (1-cm diameter) aligned longitudinally within a larger cylinder (the "GTV") in a honey-comb pattern, mimicking a conventional GRID block, with 2 cm (GRID 2 cm ) and 3 cm (GRID 3 cm ) center-to-center spacing [ Fig. 1(a)]. The larger cylinders (GTV) were 8 and 10 cm in diameter for the GRID 2 cm and GRID 3 cm arrangements, respectively. The "normal tissue" outside of the GTV was defined as a 5-cm ring around the GTV. In the dosimetric analysis, we defined the "peak" dose as the mean dose of the GRID target. We defined the "edge" as a 2-mm ring just outside of the GTV, and the edge dose as the mean dose to this structure.

2.B | Phantom treatment planning
All treatment plans were generated using the Pinnacle v. 9.10 treatment planning system (TPS) (Philips Amsterdam DE) or the tomotherapy TPS (Accuray, Sunnyvale, California). All 3DCRT, VMAT, and GRID-block plans were planned to be delivered by an Elekta linac equipped with Agility MLC (Versa HD, Elekta Inc., Stockholm, Sweden). Three treatment techniques were used for comparison: 3DCRT, VMAT, and tomotherapy. For the GRID 2 cm arrangement, the 3DCRT plans used six beams with gantry angles ranging from 30 to 330°(at 60°increments) intended to align beam paths with groups of individual cylinders. For each gantry angle, three equally weighted segments were created where the field shape corresponded to the outline of a single cylinder. For the GRID 3 cm arrangement, the 3DCRT plan used the same gantry angles as the GRID 2 cm plan, with six additional beams at gantry angles 0°, 60°, 120°, 180°, 240°, and 300°. For these additional gantry angles, a single segment with the field shape corresponding to the central cylinder, with lower weight, was used.
These collimator angles were selected to minimize the creation of "dose islands" between grid targets at varying gantry angles. The VMAT GRID 3 cm arrangement used two full arcs with the collimator rotated at 90°(for the 180°-182°arc) and 270°(for the 182°-180 arc). For tomotherapy, two plans were created with field widths of 5.01 cm (TOMO 5 cm ) and 2.5 cm (TOMO 2.5 cm ) with the same pitch of 0.43. Table 1 shows a list of target and planning structure objectives used for the VMAT and tomotherapy plans. We used 6 MV energy for all plans.
T A B L E 1 List of target and planning structures for VMAT and Tomotherapy plans.

Structures
Objectives Weights

2.D | Dosimetry analysis
The peak-to-valley dose ratio (PVDR) was defined as the ratio of the mean dose to the GRID target to the mean dose of the valley.
The peak-to-edge dose ratio (PEDR) was defined as the ratio of the mean dose to the GRID target to the mean dose to the 2-mm ring [ Fig. 1(b)]. The volume of the normal tissue receiving 7.5 Gy (V 7.5 Gy ) and 5 Gy (V 5 Gy ) was quantified. To evaluate superficial tissue sparing, we evaluated the dosimetric differences for 0.03 cc (D 0.03 cc ) and 10 cc (D 10 cc ) of skin for the patient data.   (Table 2). Similar to the plans for the GRID 2 cm arrangement, the TOMO 5 cm and TOMO 2.5 cm plans had the longest delivery times.
T A B L E 2 Comparison of delivery time, peak-to-edge dose and peak-to-valley dose ratios for 3D conformal, VMAT, and Tomotherapy plans across the GRID 2 cm and GRID 3 cm cylinder arrangements. 3DCRT, three-dimensional conformal radiotherapy; TOMO 5 cm , tomotherapy plans with 5 cm field width; TOMO 2.5 cm , tomotherapy plans with 2.5 cm field width; D mean , mean dose of structure; D max , maximum dose to structure; PEDR, peak-to-edge dose ratio; PVDR, peak-to-valley dose ratio; Normal Tissue V 7.5 , volume of normal tissue receiving 7.5 Gy; Normal Tissue V 5 , volume of normal tissue receiving 5 Gy. The mean delivery times were consistently lower for the GRID 3 cm arrangement for all plans ( For a representative patient, a single axial slice is shown for the six treatment plans in Fig. 5. Visually, it can be seen that the VMAT and tomotherapy plans resulted in the lowest dose surrounding the PTV. Similar to the phantom plans, the 3DCRT plans had the highest PEDR and PVDR (Table 2 and Table 3). However, when looking at the percentage of normal tissue receiving 7.5 and 5.0 Gy, these values were lowest with the VMAT and tomotherapy plans (Table 3).
Specifically, the TOMO 2.5 cm plan had the lowest D max to the cauda

| DISCUSSION
In this study, we used tomotherapy-, VMAT-, and 3DCRT-based approaches to generate spatially fractionated radiation treatment patterns in deep-seated tumors. We painted cylindrical targets throughout tumor volumes as a template to guide treatment planning. These cylinders were a 3D representation of the 2D circular patterns seen with conventional GRID blocks. In the initial phantom studies, the simple 3DCRT approach for the GRID 2 cm and GRID 3  A key clinical issue moving forward is to determine the appropriate number and spacing of "high-dose islands" targets within tumors.
Although the value of a high PEDR result seems clear, the relevance of high PVDR values is less clear. It should be noted that the optimal PVDR and PEDR have not been extensively explored in previous work, studies have reported valley to peak ratios ranging from 0.0008 to 2.5 7,10,11 and PEDRs ranging from 5 to 20. 11 Our approach in this work was to construct, in three-dimensions, the two-dimensional pattern achieved with a conventional GRID-block based on the historical successes with this approach. However, it should also be noted that in conventional GRID irradiation that, for a given slice perpendicular to the axis of the beam, the dose homogeneity in the tumor increases with depth. Thus, at the 2D level, the PVDR approaches 1 at depth. Finally, we should also acknowledge that the treatment planning time required for a GRID-block treatment is substantially less than approaches proposed in this study.
This may limit our technique to patients who do not need to be treated immediately.
In summary, we demonstrated that all of the studied approaches are capable of delivering high-dose radiation to cylindrical structures within large tumors, yielding spatially fractionated radiation dose distributions over the length of the tumor. Selection of one approach over another may depend on the shape and depth of the GTV in the patient and the type and extent of surrounding critical structures. To evaluate the efficacy of these approaches in patients, clinical trials are required.

ACKNOWLEDG MENTS
The authors would like to thank the Johns Hopkins Radiation Oncology and Molecular Sciences department for their support in this project.

CONFLI CT OF INTEREST
Authors have nothing to declare.