2.1 Nozzle and Monte Carlo model

A spot-scanning beam line with an IBA universal nozzle was commissioned for RayStation 6. The commissioned model supports a proton energy range of 100-226 MeV, which corresponds with WETs of between 7.71 cm and 31.9 cm [19]. The virtual source-to-axis distances in the horizontal and vertical directions were 230 cm and 193 cm, respectively. The simplified nozzle model and the beam spot size (1δ at the isocenter) as a function of energy are shown in Fig. 1a and Fig. 1b, respectively. The snout position can be adjusted from 8.5 cm to 57.5 cm relative to the isocenter. A polymethylmethacrylate RS with a thickness of 6.5 cm (and density of 1.19 g/cm³, corresponding to a 7.5 cm WET) was available, which can shift the range of a 100 MeV proton beam close to the skin. A brass PSBC (with a density of 8.4 g/cm³ and a thickness of 6.5 cm) was mounted on the source side and an RS was mounted on the patient side (the location sequence of the PSBC and RS was unchangeable because it was fixed by IBA). The inner edge shape of the PSBC was automatically confirmed by RayStation 6 according to the target’s shape.

Fig. 1Full-screen View

a. The nozzle model. (RS: range-shifter, PSBC: patient-specific block collimator) b. The relationship between spot size and energy.

The MC-based dose engine [20], which is a fast MC algorithm in RayStation 6, was developed independently by RaySearch. In the MC code, primary protons and secondary ions (protons, deuterons, and alphas) can be accurately simulated and transported. Neutral reaction products (neutrons and gammas) are not transported but the induced absorbed energies are counted by given fractions. The proton source is modelled as it arrives at the nozzle entrance. The RS and PSBC can be modelled in the beam line such that proton transportation in the RS and PBSC can be correctly simulated [16]. The high accuracy of the MC dose engine in RayStation 6 has been studied and verified in clinical circumstances [18,21].

2.2 Regular targets

A computed tomography dataset, which was generated with an empty phantom image set during commissioning, was imported into RayStation 6. A 40 × 40 × 40 cm region of interest (ROI) was constructed and filled with water. The dose grid resolution used for the dose calculation was set to 2 × 2 × 2 mm. For superficial targets with an RS, two uniform targets with regular shapes and uniform densities were defined: a cubic target of 6 × 6 cm in the coronal plane and extending to 5 cm in depth, and a spherical target with a diameter of 6 cm. The geometric centers of both targets were placed at a depth of 6 cm from the water phantom surface. For comparison with cases without an RS, the same targets (cube and sphere) were located at a depth of 15 cm. The gantry angle of the proton beam was selected to be 0° and the air gap between the RS and the water phantom was 5 cm throughout all plans.

In accordance with the settings above, eight plans were created, as shown in Table 1. For the plans using the PSBC, the external expansion distance of the PSBC inner edge was 1 cm. All plans were optimized and evaluated using an intensity-modulated spot-scanning proton beam under the MC-based dose engine. The spot spacing, weight, and number of spots at each energy layer were determined by the optimization algorithm of RayStation 6. The optimized beam energies and the numbers of layers shown in Table 1 were kept constant for targets of the same shape and at the same depth, regardless of whether the PSBC was placed. The mean dose was 200 cGy within the target volume.

2.3 Clinical cases

Three patient cases including brain, abdomen, and prostate tumors were selected to verify the dose distributions with and without the PSBC. For all patient plans, the dose grid was set to 2 × 2 × 2 mm and two beams with well-chosen gantry angles for each plan were employed. The primary parameters are shown in Table 2. Six plans including plan 1 and plan 2 for brain tumors, plan 3 and plan 4 for abdomen tumors, and plan 5 and plan 6 for prostate tumors were optimized using the MC dose engine. The dose–volume histograms (DVHs) were acquired and the mean doses to the ROIs were calculated for evaluation purposes.

3.1 Homogeneous targets

Fig. 2 shows the cross-section views of the two-dimensional (2D) dose distributions at the isocenter for the cubic and spherical targets with homogeneous densities. The maximum dose level was about 200 cGy, which was spread over the whole target volume. The edge shadows were relatively larger for the uncollimated beams, as illustrated in Fig. 2a1, Fig. 2b1, Fig. 2c1, and Fig. 2d1. When the PSBC was used, the dose levels of the edge shadows were obviously lower than those without the PSBC and the widths were reduced, as shown in Fig. 2a2, Fig. 2b2, Fig. 2c2, and Fig. 2d2. The dose differences are shown in Fig. 2a3, Fig. 2b3, Fig. 2c3, and Fig. 2d3. A reduction in the maximum dose at the outer edge of target area reached 40 cGy, which is significant (~20%) relative to the overall prescription dose (200 cGy). The dose differences at the field edges decreased with increasing depth.

Fig. 2Full-screen View

Two-dimensional dose distributions through the cubic and spherical targets at various depths. The absolute dose (unit: cGy) is represented by colors. a1. The cubic target at a depth of 6 cm with uncollimated fields. a2. The cubic target at a depth of 6 cm with collimated fields. a3. The dose difference between (a1) and (a2). b1. The spherical target at a depth of 15 cm with uncollimated fields. b2. The spherical target at a depth of 15 cm with collimated fields. b3. The dose difference between (b1) and (b2). c1. The cubic target at a depth of 6 cm with uncollimated fields. c2. The cubic target at a depth of 6 cm with collimated fields. c3. The dose difference between (c1) and (c2). d1. The spherical target at a depth of 15 cm with uncollimated fields. d2. The spherical target at a depth of 15 cm with collimated fields. d3. The dose difference between (d1) and (d2).

Fig. 3a shows the lateral penumbral width variation as a function of depth for the cubic targets. The penumbral width increased as the depth increased both with and without the RS and PSBC. As Table 3 shows, if the RS was employed, the penumbral width increased from 4.6 mm to 5.5 mm as the depth increased and from 2.5 mm to 3.6 mm if the PSBC was added; the improvement (reduction of penumbral width) was 46%-35% relative to the non-PSBC case. For targets at a depth of 15 cm, a similar trend was found, as shown by the blue solid line in Fig. 3. The penumbral widths were 3.6 mm to 5.1 mm in the absence of the PSBC, 1.3 mm to 3.1 mm in the presence of the PSBC, and the improvement was from 64% to 39%. The role of the PSBC is evident, and the improvement gradually attenuated with increasing depth in water.

Fig. 3Full-screen View

The relationship between lateral penumbral width and depth in water. a. The cubic targets at depths of 6 cm (red lines) and 15 cm (blue lines). b. The spherical targets at depths of 6 cm (red lines) and 15 cm (blue lines). (Solid lines and dashed lines represent uncollimated and collimated beams, respectively. RS: range-shifter, PSBC: patient-specific block collimator)

The variations in the lateral penumbral widths for the spherical targets at the two depths are shown in Fig. 3b. Because the diameter of the circularly shaped cross-section varied as a function of depth, the width of lateral penumbra first increased (phase 1), then decreased (phase 2), and then increased again (phase 3), which spatially corresponds to the region from the surface of the water phantom to the upper surface of the spherical target, the upper surface to the center of the spherical target, and the center to the lower surface of the spherical target, respectively. As Table 3 lists, the width of the lateral penumbra of the spherical target at a depth of 6 cm was reduced by 36%-30% at phase 1, 30%-32% at phase 2, and 32%-0% at phase 3. For spherical targets at a depth of 15 cm, improvements of 40%-25% at phase 1, 25%-33% at phase 2, and 33%-0% at phase 3 were obtained. Therefore, using a PSBC in IMPT can lead to a decreased lateral penumbra in the field, which could be helpful for protecting the OARs adjacent to the target volume.

3.2 Clinical cases

Fig. 4 shows the 2D dose distributions and the differences between the uncollimated and collimated beams for a brain tumor patient (plan 1 and plan 2). The differences at the fields’ edges were up to 25% of the maximum dose of the planning target volume (PTV). The DVHs are shown in Fig. 4j, in which the dose curves of the PTV and OARs near the target edges are displayed. The dose in the PTV was consistent and the prescription was well-maintained in the PTV regardless of whether the PSBC was used. The doses in the OARs adjacent to the target area, such as in the brain stem, brain stem core, and left cochlea, were clearly reduced, as Fig. 4j illustrates. The mean doses in the left cochlea decreased by as much as 50%, and the other OAR doses were also variously reduced. In general, with the collimated beam, the OARs near the target were better protected than with the uncollimated beam.

Fig. 4Full-screen View

The two-dimensional dose maps of the middle layer in various anatomy planes of patient plan 1 and plan 2. a. The uncollimated fields in the cross plane. b. The collimated fields in the cross plane. c. The dose difference between (a) and (b). d. The uncollimated fields in the coronal plane. e. The collimated fields in the coronal plane. f. The dose difference between (d) and (e). g. The uncollimated fields in the sagittal plane. h. The collimated fields in the sagittal plane. i. The dose difference between (g) and (h). j. The comparative dose–volume histograms of the brain tumor plans. (PTV: plan target volume, Brainstem: brain stem, BrainstemCore: brain stem core, CochleaLt: left cochlea)

Fig. 5a and Fig. 5b show the DVHs of the abdomen tumor and prostate tumor patient plans, respectively. An obvious decrease in unnecessary doses outside the target volumes were found when the PSBC was employed. For the abdomen tumor case, a dummy ROI (block) was established to ensure a low dose to the end of the stomach. When the PSBC was used, the mean doses to the kidney, stomach, and liver were reduced by 35%, 10%, and 9%, respectively. In the prostate tumor patient plans, the maximum reduction of the mean dose was 35% for the bulb of the penis and 20% and 6% for the rectum and bladder, respectively. Above all, there was little change in the target volume but an obvious dose reduction in the ROIs, which was beneficial to the patients.

Fig. 5Full-screen View

a. The dose–volume histograms (DVHs) for the abdomen tumor plans. b. The DVHs for the prostate tumor plans. (The solid and dashed lines represent the uncollimated and collimated beams, respectively. PTV: plan target volume, PenileBulb: penile bulb)