Abstract
Background/Aim: In spot-scanning proton therapy, intra-fractional anatomical changes by organ movement can lead to deterioration in dose distribution due to beam range variation. To explore a more robust treatment planning method, this study evaluated the dosimetric characteristics and robustness of two proton therapy planning methods for liver cancer. Patients and Methods: Two- or three-field treatment plans were created for 11 patients with hepatocellular carcinoma or metastatic liver cancer using a single-field uniform dose (SFUD) technique. The plans were optimized using either beam-specific planning target volume (BSPTV) or worst-case optimization (WCO). The target coverage for the gross tumor volume (GTV), planning target volume (PTV), and organs at risk (OAR) parameters related to toxicity were calculated from the perturbed dose distributions, considering setup and range uncertainties. Statistical analyses of the BSPTV and WCO plans were performed using the Wilcoxon signed-rank sum test (p<0.05). The calculation times for a single optimization process were also recorded and compared. Results: The robustness of the WCO plans in the worst-case scenario was significantly higher than that of the BSPTV plan in terms of GTV target coverage, prevention of maximum dose increase to the gastrointestinal tract, and the dose received by normal liver regions. However, there were no significant differences in PTV, and the calculation time required to create the WCO plan was considerably longer. Conclusion: In SFUD proton therapy for liver cancer, the WCO plans required a longer optimization time but exhibited superior robustness in GTV coverage and sparing of OARs.
- Beam-specific planning target volume
- liver cancer
- single field uniform dose
- spot-scanning proton therapy
- worst-case optimization
Primary and secondary cancers of the liver are the third most common causes of cancer-associated deaths worldwide (1, 2). Radiotherapy has become standard treatment for patients with medically inoperable liver cancer. In this treatment, it is important to minimize low-dose exposure to normal liver volume, in addition to the mean dose, to limit radiation-induced liver toxicity (3-6). With recent developments in radiotherapy techniques, there is increasing evidence that radiation is safe and effective, with a high local control rate for liver cancer (7, 8). Compared to conventional X-ray radiotherapy, a proton beam has a finite range, which is determined by its energy, and deposits the majority of its dose over a short distance at the Bragg peak, thus improving tumor control while simultaneously reducing toxicity (4, 9-16). However, the proton beam is also known to be susceptible to various uncertainties, such as patient setup error, intra- and inter-fractional anatomical changes, and respiratory motion during treatment (4, 11, 17, 18). All these factors can significantly affect the quality of proton therapy dose distribution by varying the beam range, thus leading to inadequate target coverage and/or excessive doses to organs at risk (OARs).
The movement of the liver with respiratory motion leads to deterioration in dose distribution by the intra-fractional anatomical changes in the patient’s body. Thus, a method for controlling motion is essential for liver proton therapy. Methods, such as four-dimensional (4D) treatment planning, abdominal pressure, respiratory gating, and rescanning are frequently used for this purpose (19). Furthermore, in the dose calculation process, two types of optimization methods are currently utilized—beam-specific planning target volume (BSPTV) (20-22) and worst-case optimization (WCO) (23, 24)—to mitigate treatment delivery uncertainties. The BSPTV method was originally designed for proton therapy treatment planning using the single-field uniform dose (SFUD) technique, which ensures that a uniform dose is delivered to the target volume from each beam field (20, 25). This method provides a significant advantage in maintaining higher treatment quality over the traditional method through use of a single planning target volume (PTV) margin for all beams because the magnitude of each margin can be individualized for each field. The WCO method directly uses the gross tumor volume (GTV) or the clinical target volume (CTV) as the target, which no longer requires an expanded PTV margin. This method optimizes the worst-case dose distribution by considering the setup and range uncertainties, with the aim of maintaining target coverage while minimizing the deviation of the OAR doses.
The effectiveness of the BSPTV and WCO methods has been elucidated in terms of plan robustness against the aforementioned uncertainties. However, to our knowledge, no comparative study between these two methods has been conducted to date. Therefore, in this study, we aimed to evaluate the dosimetric characteristics and robustness of two types of proton therapy planning methods. Our findings will provide clinical insights for treatment decision-making in patients with liver cancer undergoing proton therapy.
Patients and Methods
Patient data. This study used treatment-planning computed tomography (CT) images of 11 patients who were previously treated for hepatocellular carcinoma (HCC) or metastatic liver cancer (MLC) using spot-scanning proton therapy between November 2022 and February 2024 at Shonan Kamakura General Hospital. Treatment-planning expiratory phase CT images were acquired using a Somatom Confidence CT scanner (Siemens, Forchheim, Germany) with a reconstruction resolution of 0.977×0.977×2 mm3 and a respiratory gating system (AZ-733V; Anzai Medical, Co., Ltd, Tokyo, Japan).
After acquiring the CT images, experienced radiation oncologists delineated the GTV on the expiratory phase CT images using MIM Maestro version 7.2.10 (MIM Software Inc., Cleveland, OH, USA). Additionally, OARs, such as the liver, stomach, small intestine, and colon, were delineated. The CTV was defined by adding margins of 5 mm within the liver and 2 mm outside the liver to the GTV. The PTV was created by adding a 3 mm internal margin to the CTV and excluding the area that was extended by 7 mm around the gastrointestinal tract. The clinical characteristics of the patients are summarized in Table I.
Patient characteristics: tumor type, prescription dose, target volumes, and locational relationship between the gastrointestinal tract and target.
Ethics approval and consent to participate. Ethical approval and patient consent to participate were obtained prior to the initiation of this study. This study was approved by The Institutional Review Board of the Shonan Kamakura General Hospital (Tokushukai Group Ethics Committee, No. 2561). All methods were performed in accordance with relevant guidelines and regulations. Given its retrospective nature, the review board waived the need for informed consent by offering an opt-out option on the institution’s homepage (Tokushukai Group Ethics Committee, https://www.mirai-iryo.com/service/index.php#s03).
Proton therapy treatment planning. Treatment planning was performed using the VQA Treatment Planning System (TPS; Hitachi Ltd., Tokyo, Japan) commissioned for the PROBEAT-M1 proton beam delivery system (Hitachi Ltd.). The system uses spot scanning using a synchrotron, and the beam energy ranged from 70.2 to 228.7 MeV, corresponding to a penetration depth of 4-32 g/cm2, with a spot size of 7.0-2.5 mm in air at the isocenter. The maximum field size at the isocenter was 30×40 cm2.
For each patient, a two- or three-field coplanar spot-scanning proton treatment plan was created using the SFUD and rescanning techniques. In this study, layered rescanning was applied, in which each isoenergy layer was individually rescanned and the number of rescans was uniformly set to four. The prescribed doses and fractions were 66 Gy [relative biological effectiveness (RBE); RBE value: 1.1] in 10 fractions for peripheral tumors, 72.6 Gy (RBE) in 22 fractions for tumors adjacent to the gastrointestinal tract, and 64 Gy (RBE) in eight fractions for metastatic tumors. The gantry angle was determined based on the patient’s anatomy to minimize the maximum dose to the gastrointestinal tract and the volume of the normal liver that received radiation.
Treatment planning was optimized using the BSPTV or WCO methods. For the BSPTV plans, the beam-specific planning target volume was created as the evaluative PTV (PTVeval), which defined the PTV plus a lateral margin of 5 mm and a beam margin of 3.5% (26). The BSPTV plan was optimized to ensure that the maximum prescribed dose did not exceed 110% of the prescription dose (Dmax) to the stomach and that the small intestine did not exceed 50 Gy (RBE). Furthermore, the Dmax of the colon did not exceed 55 Gy (RBE), and the volume of the normal liver [defined as the volume of liver minus GTV that received less than 2 Gy (RBE)] was maintained as low as possible and normalized to the prescribed dose of 95% for the PTV. The BSPTV plans were used in clinical practice following approval form the oncologist.
For fair comparison purposes, the WCO was performed using the same dose constraints as the BSPTV, in which the worst-case minimax approach, considering a 5-mm setup uncertainty in the left-right, superior-inferior, and anterior-posterior directions (six scenarios in total), along with a 3.5% range uncertainty, was applied (27). In the WCO, the PTV was the target and was normalized to the prescribed dose of 95%.
A pencil beam algorithm with a triple Gaussian kernel model (28) and a grid resolution of 2×2×2 mm3 was used for dose calculations. The triple Gaussian kernel model that comprised a primary component (multiple Coulomb scattering) and a halo component (inelastic, nonelastic, and elastic nuclear reactions) enhanced the precision of the dose calculation in the spot-scanning technique. A previous study reported that the model reproduced the Monte Carlo simulation results well in the liver region (28).
Evaluation of plan robustness. After creating each original plan, the robustness of the plans against the setup and range uncertainties was evaluated. Perturbed plans that considered the setup uncertainty scenario were created by isotropically shifting the plan isocenter by 5 mm in six directions (LR, AP, and SI) and recalculations. The perturbed plans that incorporated the range uncertainty scenario were recalculated with the CT density changed by ±3.5%. Consequently, nine scenario plans, including the original plan for each optimization method, were created for each patient.
Evaluation indices and statistical analysis. To quantify plan quality, the target coverage V100% (volume receiving 100% of the prescription dose) was calculated for each GTV and PTV, in both the BSPTV and WCO original and worst-case plans from the perturbed plans. Furthermore, patients were classified into two groups based on whether the GTV was within or greater than 7 mm from the gastrointestinal tract margin, and the target (GTV and PTV) V100% was compared between cohorts.
For the OARs, the following toxicity-relevant parameters were evaluated: Dmax values for the stomach, small intestine (including duodenum and jejunum), and colon, and the volume of normal liver recovered to less than 2 Gy (RBE). Only patient data for each OAR (stomach, small intestine, and colon) within 2 cm of the PTV were used for a more relevant dosimetric evaluation of OAR doses (29). The calculation times for a single optimization for the BSPTV and WCO plans were also recorded. The Wilcoxon signed-rank sum test was used to calculate two-tailed p-values for the differences in the evaluation indices between the BSPTV and WCO plans. Statistical significance was set at p<0.05.
Results
Dosimetric analysis. Figure 1 shows the representative dose distributions of BSPTV and WCO plans for patients, L06 (upper) and L07 (lower). The WCO had a steeper dose gradient in the peripheral area of the target than the BSPTV. Table II summarizes the patient-specific dosimetric parameters of the BSPTV and WCO plans. All plans met the clinical criteria for the target and OARs at our institute.
Representative dose distributions for the BSPTV (a,c) and the WCO (b,d) plans. The left panel shows the dose distributions in the axial section, and the right panel shows the enlarged dose distributions near the boundary area between the target and OARs. BSPTV: Beam-specific planning target volume; WCO: worst-case optimization; GTV: gross tumor volume; PTV: planning target volume; RBE: relative biological effectiveness.
Patient-specific dosimetric parameters for BSPTV and WCO original plans.
Figure 2 shows the boxplots of GTV V100%, PTV V100%, stomach Dmax, small intestine Dmax, colon Dmax, and normal liver volume for the BSPTV and WCO plans in the worst-case scenario (upper) and all scenarios (lower). In the worst-case scenario, the median values (minimum–maximum values) of the target coverage in the BSPTV and WCO plans were 86.7% (81.6%-99.7%) and 98.9% (84.2%-100.0%) for the GTV and 81.8% (70.2%-91.1%) and 87.1% (77.2%-91.0%) for the PTV, respectively. The WCO plans maintained a higher robustness for GTV coverage (p<0.05), whereas no significant difference was observed in the PTV coverage between the two plans. With respect to the OARs, the median values (minimum–maximum values) of the Dmax in BSPTV and WCO plans were 56.7 Gy (RBE) [12.4 Gy (RBE) −71.9 Gy (RBE)] and 52.1 Gy (RBE) [8.2 Gy (RBE) −59.3 Gy (RBE)], 66.1 Gy (RBE) [61.8 Gy (RBE) −73.9 Gy (RBE)] and 60.0 Gy (RBE) [57.6 Gy (RBE) −63.0 Gy (RBE)], 66.6 Gy (RBE) [41.3 Gy (RBE) −71.7 Gy (RBE)] and 63.5 Gy (RBE) [30.7 Gy (RBE) −64.6 Gy (RBE)] for stomach, small intestine, colon, respectively. The WCO plans had a significantly lower gastrointestinal Dmax than the BSPTV plans (p<0.05). The median values (minimum–maximum values) of the normal liver volume were 624.9 cc (416.5 cc-1197.5 cc) and 646.3 cc (434.5 cc-1,224.8 cc) for BSPTV and WCO plans. Compared to the BSPTV plans, the normal liver volumes in the WCO plans remained significantly larger in the worst-case scenario (p<0.05). For all scenarios, the median values (minimum–maximum values) of the target coverage in the BSPTV and WCO plans were 97.7% (93.7%-99.8%) and 99.3% (97.8%-100.0%) for the GTV, and 91.9% (81.6%-99.7%) and 99.9% (88.4%-100.0%) for the PTV, respectively. The WCO plans maintained a higher robustness for GTV coverage (p<0.05), whereas no significant difference was observed in the PTV coverage between the two plans. With respect to the OARs, the median values (minimum–maximum values) of the Dmax in BSPTV and WCO plans were 43.3 Gy (RBE) [1.6 Gy (RBE) −71.9 Gy (RBE)] and 42.0 Gy (RBE) [1.4 Gy (RBE) −59.3 Gy (RBE)], 49.7 Gy (RBE) [21.5 Gy (RBE) −73.9 Gy (RBE)] and 49.6 Gy (RBE) [32.4 Gy (RBE) −63.0 Gy (RBE)], 45.9 Gy (RBE) [12.2 Gy (RBE) −71.7 Gy (RBE)] and 46.6 Gy (RBE) [9.2 Gy (RBE) −64.6 Gy (RBE)] for stomach, small intestine, colon, respectively. The WCO plans had a significantly lower small intestine and colon Dmax than the BSPTV plans (p<0.05), whereas no significant difference was observed in the stomach Dmax between the two plans. The median values (minimum–maximum values) of the normal liver volume were 674.1cc (416.5 cc-1,243.4 cc) and 692.9 cc (434.5 cc-1,272.2 cc) for BSPTV and WCO plans. Compared with the BSPTV plans, the normal liver volumes in the WCO plans were significantly larger in the worst-case scenario (p<0.05).
Box and whisker plots of the GTV V100%, PTV V100%, stomach Dmax, small intestine Dmax, colon Dmax, and normal liver volume (receiving less than 2 Gy (RBE)) for the BSPTV and WCO plans. The upper panel shows the worst-case plans and the lower panel shows the all perturbed scenario plans. BSPTV: Beam-specific planning target volume; WCO: worst-case optimization; GTV: gross tumor volume; PTV: planning target volume; RBE: relative biological effectiveness.
Table III shows the significant differences that were identified in the GTV and PTV V100% between the classified groups, based on the distance from the GTV to the gastrointestinal tract for both plans in the worst-case scenario. In the case when the GTV was located 7 mm or farther from the gastrointestinal tract, the WCO plans exhibited superior robustness in target coverage compared to the BSPTV plans (p<0.05). However, no significant differences were observed between the two plans in the other cases.
Target coverage and distance from target to gastrointestinal tract for BSPTV and WCO plans in the worst-case scenario.
Calculation time. Table IV shows the calculation times of a single optimization for the BSPTV and WCO plans. The ratio of WCO plan time to BSPTV plan time ranged from 1.9 to 6.7 (median 3.9). Thus, the calculation times for the WCO plans were considerably longer than those for the BSPTV plans.
Calculation time of a single optimization for the BSPTV and WCO plans.
Discussion
Both original SFUD plans with BSPTV or WCO methods for liver cancer met our Institution’s criteria, and the radiation oncologists approved all plans for clinical practice. However, in evaluating the robustness of the two plans against setup and range uncertainties, the results revealed that the WCO method led to a significantly more robust plan quality for GTV coverage than the BSPTV method while improving the sparing of OARs. In the BSPTV method, individual proton beam spots are placed inside the PTVeval along the penumbra and beam margin, and the spot weights are optimized to deliver a uniform dose across the PTV, under the nominal scenario. Thus, the resulting dose distributions are generally not robust to setup and range uncertainties, which change the tissue density along the beam path (30). The WCO method can optimize the spot weight, while simultaneously considering possible perturbations (23, 24). This difference in the optimization process resulted in a greater plan robustness of the WCO for the GTV and OARs. In addition, the WCO method significantly increased the volume of the normal liver in the original plans and reduced the deterioration in the normal liver volume caused by uncertainties in the worst-case plans. WCO plans use the tumor itself as the primary target for irradiation and eliminates the need for a geometrically-expanded PTV. This prevents unnecessary irradiation of extra normal tissues that surround the target and results in higher-dose conformity compared to BSPTV plans, which irradiate the entire PTV.
Notably, no differences in PTV robustness between the two plans were observed. At our Institution, a planning at-risk volume (PRV) margin of 7 mm was empirically set for the gastrointestinal tract, and the volume that overlapped with the PTV was removed from the PTV, to prevent hotspots and ensure clinical constraints. We presumed that if the targets were closer to the gastrointestinal tract, it would be more difficult to achieve the dose objectives for the target and OARs, due to the trade-off during optimization (31), that is, the effectiveness of the WCO method for plan robustness may be impaired. Therefore, patient data were classified into two groups: 1) GTV 7 mm from the gastrointestinal tract and 2) GTV within 7 mm of the gastrointestinal tract. A significance test between the worst-case BSPTV and WCO plans was then performed for GTV and PTV coverage. The analysis revealed that the WCO method had superior robustness for GTV coverage compared to BSPTV when the GTV was 7 mm from the gastrointestinal tract. However, although the mean PTV V100% values for the BSPTV and WCO plans were 79.8% and 84.7%, respectively, the PTV trend was the same as that in the overall analysis, with no significant difference between the two groups. We then evaluated the dose gradient, which was defined as the dose fall-off from the boundary of the target to the boundary of the nearest OAR divided by the three-dimensional distance, for both original plans. The dose gradient, expressed as a value relative to the prescribed dose, was on average 4.0%/mm and 5.2%/mm in the distal direction with respect to the beam, and 3.3%/mm and 3.9%/mm in the lateral direction with respect to the beam for the BSPTV and WCO plans, respectively. A sharper dose gradient was one of the dosimetric features of the WCO plans, compared with the BSPTV plans. This relationship between the dose gradient and the unique procedure (removing the overlapping PRV from the PTV) may be associated with the robustness of both methods. Further dosimetric investigations with a larger number of diverse patients are essential to elucidate the robustness and generalizability of this PTV coverage plan.
Although the SFUD plans with the WCO method had superior robustness for liver cancer with spot-scanning proton therapy, this approach has one disadvantage in clinical practice. The dose calculation time for creating the WCO plan was on average 3.8 times longer than that for the BSPTV plan and depended on the target volume and number of dose objectives. In cases where an urgent start of treatment is needed, utilizing a plan with the WCO method may not be suitable due to the extension of treatment planning process.
We investigated the dose characteristics of the BSPTV and WCO plans created on a single expiratory phase CT image in this study. However, the liver moves due to respiration, which potentially leads to an interplay effect on the actual dose distribution (3, 32). In the future, 4D dose calculations and evaluations of plan robustness using 4D-CT and breath-hold CT images (33, 34) will be necessary to elucidate the dosimetric effectiveness of the two optimization methods for liver cancer in more detail, which is a limitation of the current study.
Conclusion
This study investigated the dosimetric characteristics, robustness, and optimization calculation time for SFUD proton therapy treatment plans using the BSPTV and WCO methods for liver cancer. Although the WCO plans required longer calculation times, they exhibited higher robustness for GTV coverage and sparing of OAR doses against setup errors and range uncertainties.
Footnotes
Authors’ Contributions
AY and TI contributed substantially to the conception of the study. MY, KM, RS, TY, and TS contributed significantly to the data analysis and interpretation. SS, KT, and WC contributed significantly to the study design and data interpretation. AY and TI drafted the manuscript. All Authors critically reviewed and revised the manuscript and approved the final version for submission.
Conflicts of Interest
The Authors declare that they have no competing interests related to this study or its publication.
- Received August 5, 2024.
- Revision received September 2, 2024.
- Accepted September 12, 2024.
- Copyright © 2024 The Author(s). Published by the International Institute of Anticancer Research.
This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY-NC-ND) 4.0 international license (https://creativecommons.org/licenses/by-nc-nd/4.0).
References
In this issue
Jump to section
Related Articles
Cited By...
- No citing articles found.