Technical Note: A treatment plan comparison between dynamic collimation and a fixed aperture during spot scanning proton therapy for brain treatment

Abstract

Purpose:

To quantitatively assess the advantages of energy-layer specific dynamic collimation system (DCS) versus a per-field fixed aperture for spot scanning proton therapy (SSPT).

Methods:

Five brain cancer patients previously planned and treated with SSPT were replanned using an in-house treatment planning system capable of modeling collimated and uncollimated proton beamlets. The uncollimated plans, which served as a baseline for comparison, reproduced the target coverage and organ-at-risk sparing of the clinically delivered plans. The collimator opening for the fixed aperture-based plans was determined from the combined cross sections of the target in the beam’s eye view over all energy layers which included an additional margin equivalent to the maximum beamlet displacement for the respective energy of that energy layer. The DCS-based plans were created by selecting appropriate collimator positions for each row of beam spots during a Raster-style scanning pattern which were optimized to maximize the dose contributions to the target and limited the dose delivered to adjacent normal tissue.

Results:

The reduction of mean dose to normal tissue adjacent to the target, as defined by a 10 mm ring surrounding the target, averaged 13.65% (range: 11.8%–16.9%) and 5.18% (2.9%–7.1%) for the DCS and fixed aperture plans, respectively. The conformity index, as defined by the ratio of the volume of the 50% isodose line to the target volume, yielded an average improvement of 21.35% (19.4%–22.6%) and 8.38% (4.7%–12.0%) for the DCS and fixed aperture plans, respectively.

Conclusions:

The ability of the DCS to provide collimation to each energy layer yielded better conformity in comparison to fixed aperture plans.

1. INTRODUCTION

Spot scanning proton therapy (SSPT) reduces integral normal tissue dose relative to intensity modulated x-ray therapy techniques. However, tumor dose conformity for x-ray therapy may be better than proton therapy depending on spot size,^1–6 especially for shallow treatments since the lateral spot size increases with decreasing energy.⁷ However, for small brain lesions, this may not necessarily be the case. The impact of spot size using spot scanning proton therapy for small brain lesions has been investigated and previously determined that the spot sizes required to improve upon photon therapy are significantly smaller than anything currently available —even with collimation.¹ The larger the spot size, the greater the inability to laterally spare normal tissue surrounding the tumor. Widesott et al. found that a spot size of 4 mm is necessary in SSPT to surpass the normal tissue sparing obtained through intensity-modulated photon therapy for head and neck tumors.⁵ Such spot sizes are difficult to achieve with available commercial systems, as evidenced by the measurements performed at the University of Pennsylvania which revealed that the IBA universal nozzle and dedicated nozzle (Louvain-la-Neuve, Belgium) had a lateral in-air spot sizes of 7.5 and 5.2 mm, respectively, at 100 MeV.⁸ As discussed by Newhauser et al., collimation may be necessary to fully exploit the advantages of SSPT.⁹

In order to increase the effectiveness of SSPT, a dynamic collimation system (DCS)-based on two orthogonal pairs of traveling trimmer blades was proposed by Hyer et al.¹⁰ Gelover et al. generated a dosimetric DCS model,¹¹ and the benefits of using the DCS in comparison to uncollimated treatment plans were reported.¹² As implementing the DCS clinically would require significant hardware, software, and quality assurance process development, it is important to also compare its performance against widely available fixed apertures. In this work, the DCS is compared to a fixed aperture for brain tumor SSPT treated with an IBA Universal Nozzle. In particular, this work focuses on shallow tumor sites because they have shown more benefit from collimation due to their larger spot sizes at low energies and less patient scattering. The work of Moignier et al. has shown that the simulated off-axis dose profiles from collimated plans have a significantly sharper lateral penumbra while retaining equivalent target uniformity as compared to the uncollimated plans.¹² Since the DCS is capable of providing sharper penumbras around organ-at-risk (OAR) for each delivered energy layer, it was the inspiration of this study to quantify the advantages of energy-layer specific collimation offered by the DCS in comparison to a per-field fixed aperture for five investigated patient treatment plans.

2. MATERIALS AND METHODS

2.A. Treatment planning with collimation

Our in-house treatment planning system (TPS), called RDX, accounts for the dosimetric properties of proton pencil beam collimation (Fig. 1), optimizes collimator edge positions, and generates conventional (uncollimated) SSPT plans. RDX has been previously used for the calculation of intensity modulated proton therapy (IMPT) plans,^13,14 and the capability of accounting for the effects of collimation and generating asymmetric proton beamlets has been enabled through a modification of Hong’s pencil beam model.^11,15

FIG. 1.

An illustration of the dosimetric distribution for a pencil beam intercepted by two orthogonal trimmers compared to that of an uncollimated (UNCOL) pencil beam. The trimmed spot shifted from its original center, becomes asymmetric, and experiences a reduction in influence.

For the DCS and fixed aperture-based plans, the proton pencil beams were modeled in RDX by placing the exit window of the DCS or aperture and the range shifter 5 and 12 cm from the patient surface, respectively. For uncollimated plans, the range shifter was placed 5 cm from the surface of the patient. Thus, all plans were calculated with a 5 cm air gap between the last element on the beam line and the patient surface.

2.B. Patient selection and planning methods

Five datasets from patients with brain tumors previously treated with SSPT at University of Pennsylvania (UPenn) Proton Therapy Center, including the simulation CT scan, structure outlines, and computed dose distribution, were obtained through an Institutional Review Board approved research protocol and agreement. The clinically delivered plans were computed with the Eclipse treatment planning system (Varian Medical Systems) for the IBA Universal Nozzle (Louvain-la-Neuve, Belgium). A summary of the patients is presented in Table I.

TABLE I.

Plan characteristics with tumor diagnosis, planning target volume (PTV), dose prescription, and beam orientations.

Plan	Diagnosis	PTV volume (cm3)	Dose prescription (Gy)	Beam orientations
P1	Chordoma	42	50	Apex right lateral
P2a	Germinoma	P2-A: 423	P2-A: 23.4	Right and left laterals
		P2-B: 81.9	P2-B: 21.6
P3	Ependymoma	150	54	Apex left superior oblique
P4	Craniopharyngioma	91	54	Right lateral right superior oblique
P5	Low grade glioma	92	54	Left lateral left superior oblique

^aP2 was composed of two sub-plans, P2-A and P2-B, for the initial target and the non-simultaneous boost, respectively.

The clinically delivered plans were reproduced with RDX without collimation in order to obtain a baseline plan to evaluate the dosimetric improvement achieved by collimation. The lateral and longitudinal spot spacings were chosen at 4 and 5 mm, respectively, for all plans. The uncollimated plans closely matched the clinically delivered plans on the basis of the dose-volume histograms for both the target and the OARs as shown in Fig. 2.

FIG. 2.

Dose-volume histogram showing the concordance in PTV coverage on the basis of target and OARs coverage between the clinically delivered plan at UPenn (left) and the simulated uncollimated plan reproduced in RDX (right).

Once the optimization objectives were established for the uncollimated plans, they were used as a starting point for both the DCS and fixed aperture-based plans. For the DCS-based plans, an algorithm was developed to select trimmer positions that maximized the dose inside of the target to the dose outside of the target for each row of beam spots delivered at a particular energy. The pencil beam scanning target volume (PBSTV) used by UPenn treatments, which included an expansion to account for range uncertainty, were used for both trimmed and uncollimated plans. A Raster-style delivery pattern was assumed for the delivery of DCS-based plans. Ideally, spot-specific collimation would yield the optimal collimation possible but at the cost of a high time penalty during treatment. To reduce the time penalty associated while using the DCS, a concession between timely delivery and sufficient collimation was devised. During treatment with the DCS, the trimmers were selected as to collimate all beam spots within a row during Raster-style scanning. The trimmer positions were determined from optimizing the dose inside the target region and minimizing the dose outside the target for all beam spots within the row. This optimization was then repeated for all rows in each energy layer for a particular treatment field. For the fixed aperture-based plans, the aperture shape was determined from an expansion of the cross sections of the original target volume at each energy layer by an additional energy-specific margin in the beam’s eye view. The energy-specific margin was determined from the maximum displacement of a beamlet’s center position resulting from collimation at a particular energy calculated from the asymmetric beamlet model developed by Gelover et al.¹¹ Together, with the limitation of the lateral spot spacing, an optimal margin was selected using the method outlined by Wang et al.¹⁶ Specifically, for the fixed aperture collimated plans, each newly expanded cross section was projected onto a single plane, and this two-dimensional projection was then used as the fixed aperture for the entire treatment field. Small changes of the optimization constraints were allowed in order to ensure target coverage equivalence during planning between the collimated and uncollimated plans. Each plan was then optimized using SFUD.

2.C. Evaluation tools

The dose distribution comparison was performed with the RayStation TPS (Raysearch, Stockholm, Sweden). The cumulative DVH curves of the target and OARs were used for initial evaluation. The conformity index (CI 50%), defined as the ratio between the patient volume receiving at least 50% of the prescribed dose and the volume of the target, as well as the mean dose to the 10 mm ring of normal tissue surrounding the target, was calculated in order to quantify the dose conformity improvement afforded by the addition of collimation. For OARs, the mean doses (D_mean) or the maximum doses expressed as D_1%, i.e., the dose corresponding to 1% of the structure volume on the cumulative DVH, are reported depending on the structure radiosensitivity.

3. RESULTS

While the target coverage was identical between the collimated and uncollimated plans (Figs. 3 and 4), a noticeable improvement of the dose conformity was observed for both collimated plans among the five patient plans (Fig. 5 and Table II).

FIG. 3.

Treatment plans of three example patients. Row 1 (green): chordoma, row 2 (blue): craniopharyngioma, and row 3 and row 4 (red): low-grade glioma. For the first three rows, the respective plans in the columns are uncollimated plan (a), fixed aperture plan (b), and DCS plan (c). The fourth row shows the difference between the uncollimated and DCS plans (a), uncollimated and fixed aperture plans (b), and fixed aperture and DCS plans (c) for the patient data in Row 3. The color scheme, detailed by the color bars, illustrates the relative percentage of the prescribed dose that is delivered to a region.

FIG. 4.

DVH of the target (PTV), a 10 mm normal tissue (NT) ring, and relevant OARs for the following patient cases shown in Fig. 2: Chordoma (top), craniophyaryngioma (middle), and low-grade glioma (bottom). The solid, dotted, and dashed lines correspond to the uncollimated, fixed aperture, and DCS plans, respectively.

FIG. 5.

Mean doses (D_mean) and maximum doses (D_1%) relative to the uncollimated plans. HBrain = healthy brain; L = left; R = right; TL = temporal lobe; ON = optic nerve; Cochl. = cochlea; Hipp. = hippocampus.

TABLE II.

Conformity index for the 50% of the prescription dose (CI 50%) and mean dose (D_mean) to the 10-mm ring surrounding the target for both the uncollimated, fixed aperture and DCS plans. The percentage given in the parentheses indicates the percentage improvement compared to the uncollimated plans.

	CI 50%			Dmean (Gy)
	UNCOL	APERTURE	DCS	UNCOL	APERTURE	DCS
P1	4.18	3.95 (5.5%)	3.26 (22.0%)	39.82	38.64 (2.9%)	35.12 (11.8%)
P2-A	2.58	2.46 (4.7%)	2.00 (22.6%)	19.20	18.64 (2.9%)	16.74 (12.8%)
P2-B	2.87	2.68 (6.6%)	2.24 (22.0%)	16.94	15.78 (6.8%)	14.58 (13.9%)
P3	2.58	2.32 (10.1%)	2.00 (22.6%)	47.97	45.58 (5.0%)	40.99 (14.6%)
P4	3.34	2.96 (11.4%)	2.69 (19.4%)	45.04	42.24 (6.2%)	39.70 (11.9%)
P5	3.16	2.78 (12.0%)	2.54 (19.6%)	44.72	41.53 (7.1%)	37.15 (16.9%)

It can be seen in Fig. 4 that the overall sparing among OAR’s and lateral tissue is improved by use of the DCS for these five patient cases. For most OARs, the plans generated using the DCS reduced the dose delivered to the same OAR using a fixed aperture. There were some OARs in which the DCS only provided slight sparing compared to a fixed aperture; for instance, for the optic chiasm and left cochlea of P2-B, the optic chiasm and pituitary gland of P4, and the left cochlea of P5. This was primarily due to the location and proximity of the organ with respect to the beam. Additionally, in the case of the optic chiasm and pituitary gland of P4 and the left cochlea of P5, both collimated plans did not improve the uncollimated plan because the OAR location was at the target boundary or inside of the target.

4. DISCUSSION

Collimation reduces the dose to normal tissue adjacent to the target while maintaining target coverage. The benefits of collimation were observed with both approaches (fixed aperture and DCS) among the five patient plans investigated in this study. However, in these cases, the dose distributions generated with the DCS outperformed those from the fixed aperture plans. The reduction of mean dose to normal tissue adjacent to the target, as defined by a 10 mm ring, was on average 13.65% and 5.18% for the DCS and fixed aperture plans, respectively (Table II). The 50% conformity index yielded an average improvement of 21.35% and 8.38% for the DCS and fixed aperture plans, respectively (Table II). Improved effectiveness of the DCS compared to a fixed aperture can be clearly observed for one example patient in Fig. 3 (third and fourth rows) where the minimum amount of dose delivered to the normal tissue outside of the target occurs when the DCS is used. The DCS eliminates the need for beam-specific manufacturing and storage space in exchange for a delivery time penalty of less than 1 min (between 16 and 32 s depending on the target size for dynamic delivery). In theory, the DCS should have a superior performance in comparison to the fixed apertures since the DCS is acting as if different apertures were positioned on the beamline for each energy layer during a SSPT delivery. It is expected that dose distributions might be further improved with the DCS by investigating alternative spot placement schemes and trimmer positioning algorithms.

5. CONCLUSION

Compared to the uncollimated spot scanning proton treatments of the five patient cases investigated in this study, improved dose conformity can be achieved when using collimation. Furthermore, the use of the DCS offered a superior improvement to that of a fixed aperture. This is possible because the DCS has the ability to provide collimation to each energy layer. It is expected that the DCS may replace the need for fixed apertures used during treatment since it has a small footprint, simple cost-effective design, and may be integrated with current or future proton therapy equipment.

CONFLICT OF INTEREST DISCLOSURE

This study was funded by Ion Beam Applications (IBA) (Louvain-la-Neuve, Belgium). None of the authors have any relevant conflicts of interest to disclose.