Script-based automatic volumetric modulated arc therapy planning for hippocampal-sparing whole brain radiotherapy: a dosimetric comparison and efficiency analysis

Introduction

Brain metastases affect approximately 20–40% of cancer patients, with increasing incidence as systemic therapies improve survival (1,2). Whole brain radiotherapy (WBRT) has been a standard treatment for patients with multiple brain metastases, offering therapeutic and prophylactic benefits (3). However, conventional WBRT is associated with neurocognitive toxicity, including memory decline and diminished quality of life (4,5). These adverse effects largely stem from radiation damage to the hippocampus, a critical structure for memory formation containing radiosensitive neural stem cells (6,7).

Hippocampal-sparing WBRT (HS-WBRT) has emerged as an approach to mitigate these neurocognitive deficits while maintaining therapeutic efficacy. The RTOG 0933 trial demonstrated better preservation of memory and quality of life with HS-WBRT compared to historical controls (8), while the NRG CC001 trial confirmed improved neurocognitive function preservation (9).

Although volumetric modulated arc therapy (VMAT) offers dosimetric advantages for HS-WBRT, its implementation presents significant challenges (10). The planning process is complex and time-consuming, requiring steep dose gradients around the hippocampus while maintaining adequate brain coverage (11). Plan quality depends heavily on planner expertise and optimization time, leading to potential inconsistencies across patients (12).

Automated planning (AP) approaches have shown promise in improving planning efficiency while maintaining or enhancing plan quality across various treatment sites (13). These approaches include knowledge-based planning (KBP) systems that leverage historical treatment plans to predict optimal dose distributions (14-16), multicriteria optimization (MCO) approaches that generate multiple Pareto-optimal solutions allowing interactive or automated selection based on clinical priorities (17-19), and script-based techniques that employ rule-based algorithms to automate optimization workflows (20-23). Among these, script-based approaches offer implementation flexibility without requiring extensive training datasets, making them particularly suitable for institutions initiating AP programs.

Recently, Rhee et al. (23) developed a script-based AP system for HS-WBRT, demonstrating clinical feasibility through physician quality scoring and establishing the foundation for script-based automation in this treatment site. Building upon these developments, we sought to explore alternative optimization architectures and provide comprehensive evaluation of planning efficiency and plan consistency.

In this study, we developed and evaluated a script-based AP system for HS-WBRT using VMAT, employing a modular four-stage optimization algorithm [clinical goals satisfaction → hotspot suppression → clinical goal restoration → organs-at-risk (OAR) dose reduction]. This architecture systematically addresses clinical goal satisfaction and dose homogeneity through dedicated optimization phases, incorporating feedback control to maintain previously achieved goals and proactive optimization to further reduce OAR dose beyond constraint satisfaction. We conducted systematic evaluation of dosimetric quality, planning efficiency through direct timing comparisons, and plan consistency through inter-plan variability analysis. We hypothesized that this multi-stage optimization approach would achieve superior dose homogeneity, enhanced hippocampal sparing, improved plan consistency, and greater planning efficiency compared to manual planning (MP), while maintaining equivalent target coverage. We present this article in accordance with the STROBE reporting checklist (available at https://tro.amegroups.com/article/view/10.21037/tro-25-30/rc).

Methods

Patient selection

Twelve patients who had received HS-WBRT at our institution between January 2023 and December 2024 were randomly selected. There were 10 males and 2 females aged 40 to 75 years (median age: 67.5 years). Each patient’s previously treated plan was replanned using both MP and AP techniques for the comparison of planning efficiency and dosimetric outcomes.

Replanning with MP was conducted rather than using the original clinical plans for two primary reasons. First, planning time data had not been collected during the original clinical planning process, which was essential for evaluating planning efficiency. Second, replanning ensured that both MP and AP techniques would follow the same dose constraints criteria specified in the RTOG 0933 protocol, thereby enabling a fair and standardized comparison between the two planning approaches.

The MP replanning was performed by two senior medical physicists, each with over 2 years of experience in radiation therapy planning, to ensure high-quality treatment plans aligned with clinical standards.

Target delineation

All patients underwent computed tomography (CT) simulation scans using a GE CT scanner (Discovery RT; GE Healthcare, Waukesha, WI, USA) with a slice thickness of 2.5 mm. Each patient also underwent T1-weighted magnetic resonance imaging (MRI). The MRI scans were rigidly registered with the simulation CT images for accurate delineation of target volumes and OAR. For this retrospective study, all anatomical structures, including the bilateral hippocampi, had been previously delineated by radiation oncologists during the original treatment planning process, following the RTOG 0933 atlas definition (8). Hippocampal avoidance regions (HARs) were created by adding a 5 mm volumetric margin around the hippocampi to account for setup uncertainty and patient motion.

The clinical target volume (CTV) was defined as the whole brain soft tissue, and the planning target volume (PTV) was generated by adding a 2 mm margin from the CTV and subtracting the HARs. Additional OARs, including the lenses, eyes, optic nerves, optic chiasm, and both cochleae, were also contoured for dose evaluation and constraint purposes.

Dose prescription and OAR constraints

Dose prescription and OAR dose constraints followed the RTOG 0933 protocol compliance criteria as detailed in Table 1. Additionally, dose limitations for other OARs followed the ALARA (as low as reasonably achievable) principle, with particular attention to minimizing dose to the eyes, lenses, and cochleae.

Treatment planning

Both MPs and APs were generated using the VMAT technique in the RayStation treatment planning system for delivery on a Varian Edge linear accelerator (Varian Medical Systems, Palo Alto, CA, USA), utilizing 6 MV flattening filter-free (FFF) beams with a maximum dose rate of 1,400 monitor units (MU)/min. The VMAT plan was based on the split-arc partial-field VMAT (sapf-VMAT) technique described by Yuen et al. (24), with two major modifications. First, instead of split arcs, we used four full coplanar arcs. Fields A and B were set with collimator angles of 15° and 345°, respectively, while fields C and D utilized perpendicular collimator angles of 90°. The gantry rotation alternated between clockwise (CW; 181° to 179°) and counterclockwise (CCW; 179° to 181°) directions. Second, we modified the jaw settings for the fields with perpendicular collimators. For fields C and D, asymmetric jaw settings were employed where field C targeted the superior part of the target with X1 jaw positioned -2 cm from isocenter, while field D covered the inferior part of the target with X2 jaw set +2 cm from isocenter, creating a 4-cm overlapping region (Figure 1). These modifications were implemented to enhance hippocampal sparing while maintaining adequate target coverage. The isocenter was positioned at the geometric center between the bilateral hippocampus. All treatment plans were calculated using the collapsed cone (CC) algorithm with a dose calculation grid size of 2 mm.

Figure 1 BEV of the four-arc VMAT field arrangement for HS-WBRT. Fields A and B use angled collimators (15°/345°) while fields C and D use perpendicular collimators (90°). The jaw positions are adjusted for each field to ensure target coverage while maintaining hippocampal sparing. Blue and yellow structures represent the right and left hippocampus, orange structures indicate the HARs, and magenta structures show the PTV. BEV, beam’s eye view; CCW, counterclockwise; CW, clockwise; HAR, hippocampal avoidance region; HS-WBRT, hippocampal-sparing whole brain radiotherapy; PTV, planning target volume; VMAT, volumetric modulated arc therapy.

Automatic HS-WBRT planning

Automated plans were generated using a custom Python script developed in-house and executed within the RayStation 10B scripting environment. The AP process was implemented through a script-based approach consisting of three main phases: patient modeling, plan design, and plan optimization (Figure 2). The AP workflow required specific manual prerequisites before script execution: (I) image registration between CT and MRI; and (II) manual delineation of all anatomical structures (PTV, bilateral hippocampi, and OARs) performed by radiation oncologists following RTOG 0933 atlas definitions. Once these prerequisites were completed, the automated script executed without manual intervention from plan creation through final dose calculation.

Figure 2 Flowchart of automatic HS-WBRT planning process. HS-WBRT, hippocampal-sparing whole brain radiotherapy; OAR, organs-at-risk.

In the patient modeling phase, the script automatically created auxiliary structures for optimization control. A ring structure was generated by applying a 2-cm uniform expansion to the PTV and subtracting the original PTV volume. This ring structure was used to encourage rapid dose fall-off outside the PTV through dose fall-off objectives (Table 2; high dose level: 3,000 cGy, low dose level: 2,100 cGy, distance: 1.2 cm).

The plan design phase comprised three sequential steps: (I) treatment plan creation; (II) beam setup configuration; and (III) adjustment of jaw position. The treatment plan creation step involved the automated generation of a photon-based treatment plan with the VMAT technique, along with the assignment of the prescription dose and number of fractions. The beam configuration and jaw positioning were automatically performed according to the previously described specifications.

The plan optimization phase employed a four-stage iterative optimization strategy: (I) clinical goal satisfaction; (II) hotspot suppression; (III) clinical goal restoration; and (IV) OAR dose reduction.

Stage 1: clinical goal satisfaction

The script began by adding clinical goals based on RTOG 0933 protocol constraints (Table 1) and setting initial optimization objectives (Table 2). The maximum number of iterations per optimization cycle was set to 80. Following initial optimization, the script automatically evaluated whether OAR clinical goals were achieved.

If any OAR clinical goals remained unmet, the script implemented an iterative weight adjustment strategy: objective weights for unmet OAR goals were automatically increased by five, and optimization was repeated. This iterative process was limited to a maximum of four attempts. If OAR clinical goals remained unachieved after four attempts, the script proceeded to the next stage.

Stage 2: hotspot suppression

Once initial clinical goals were satisfied (or maximum attempts reached), the script automatically identified hotspot regions defined as volumes receiving >105% of the prescription dose (>31.5 Gy). The script created a new structure encompassing these hotspot regions and added maximum dose objectives set at 31.5 Gy with a weight of 1. Optimization was then re-executed to suppress high-dose regions and improve PTV dose homogeneity.

Stage 3: clinical goal restoration

After hotspot suppression, the script re-evaluated whether previously achieved clinical goals remained satisfied. If any goals were compromised during hotspot optimization, the script automatically increased the weights for these specific objectives by five and repeated the optimization. This iterative process continued until all initial clinical goals were restored or a maximum of four attempts was reached, ensuring that improvements in dose homogeneity did not compromise target coverage or OAR sparing.

Stage 4: OAR dose reduction

In the final optimization stage, the script implemented a “push for better” strategy to further reduce OAR doses while maintaining achieved target coverage. For selected OARs (bilateral hippocampi, lenses, eyes, and cochleae), the script retrieved their current achieved dose values and automatically created new objectives set at 80% of these values. Maximum dose objectives were applied to most OARs, while mean dose objectives [maximum equivalent uniform dose (EUD) with parameter a=1] were used for bilateral cochleae. These low-weighted objectives (weight =0.1) enabled proactive dose reduction beyond constraint satisfaction without compromising target coverage or previously achieved goals. A final optimization was executed to produce the completed treatment plan.

MP

The MP process followed the same three-phase approach but required human intervention and decision-making at each step. In the patient modeling phase, planners manually created auxiliary structures based on their clinical experience, including ring structures for dose conformity control with dimensions adjusted according to individual patient anatomy.

During the plan design phase, planners manually configured all beam parameters, including beam arrangement, collimator angles, and jaw positions. The optimization process in MP relied heavily on the planners’ experience and iterative trial-and-error approach. Planners manually adjusted optimization objectives and weights based on intermediate dose distributions and dose-volume histogram (DVH) evaluation, requiring multiple optimization iterations with subjective decision-making between each cycle to achieve satisfactory results.

The key distinction between MP and AP lies in the standardization and automation of decision-making. While AP follows a predetermined algorithmic approach across all cases, MP outcomes depend on individual planner experience and preferences, with optimization strategies varying between planners and cases.

Plan quality analysis

Plan quality was assessed using multiple dosimetric parameters. For the PTV, dose to 98% of the volume (D_98%) and dose to 2% of the volume (D_2%) were evaluated, while hippocampal dosimetry included dose to 100% of the volume (D_100%) and maximum point dose (D_max). Additionally, maximum or mean doses to various OARs were analyzed. Dose homogeneity was quantified using the homogeneity index (HI) as defined in the International Commission on Radiation Units (ICRU) and Report 83:

HI=D2%−D98%D50%[1]

Lower HI values indicate more homogeneous dose distributions, with an ideal value of 0 representing perfect homogeneity.

To further evaluate plan homogeneity, we also analyzed volume receiving 105% of the prescribed dose (V_105%) of the PTV, which represents the percentage of PTV volume receiving at least 105% of the prescribed dose. This parameter provides complementary information to HI by specifically quantifying hot spot volumes, as these two metrics assess different aspects of dose distribution quality.

In addition to dosimetric evaluation, we recorded the total planning time for each technique from initiation to completion of clinically acceptable plans. This total planning time included beam setup, plan optimization, and final dose calculation. To evaluate treatment delivery efficiency, we also compared the total MU required for both planning approaches, as this metric correlates with beam-on time and delivery complexity.

Statistical analysis

Statistical comparisons between the two treatment planning approaches were performed using the Wilcoxon signed-rank test for paired samples. All analyses were conducted using IBM SPSS Statistics software version 26.0 (IBM Corp., Armonk, NY, USA). Statistical significance was defined as a two-tailed P value <0.05.

Ethical considerations

This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Chang Gung Medical Foundation (IRB No. 202500595B0) and individual consent for this retrospective analysis was waived due to the use of de-identified clinical data.

Results

Dosimetric comparison of PTV parameters

The dosimetric parameters for the PTV were evaluated for both MP and AP techniques. As shown in Table 3, most PTV coverage parameters demonstrated comparable performance between the two planning approaches. The mean D_98% for MP and AP were 29.52±0.51 and 29.31±0.57 Gy, respectively (P=0.07), indicating that target coverage was adequately maintained with both techniques. Similarly, no significant differences were observed in the percentage of volume receiving 30 Gy (V_{30 Gy}) (96.44%±0.89% vs. 96.48%±0.76%, P=0.51), D_max (33.40±0.38 vs. 33.19±0.35 Gy, P=0.31), or HI (0.09±0.02 for both techniques, with no difference observed).

However, significant differences were observed in certain dosimetric parameters. The maximum D_2% was significantly lower in AP plans compared to MP plans (31.99±0.05 vs. 32.34±0.30 Gy, P=0.008). More notably, despite comparable HI values between the two planning approaches (AP: 0.09±0.02 vs. MP: 0.09±0.02, no significant difference), the V_105% was substantially reduced in AP compared to MP (19.28%±1.44% vs. 34.23%±12.04%, P=0.004). This discordance between similar HI and divergent V_105% can be explained by the fundamental difference in what these metrics quantify: HI reflects the relative relationship between D_2%, D_98%, and D_50%, while V_105% depends on the spatial distribution of dose hot spots (dose over 105% of the prescription). MP may generate multiple small, localized hot spots that collectively increase V_105% without substantially affecting D_2%, as D_2% represents only the highest 2% of the PTV volume.

Figure 3 illustrates the dose distribution comparison between MP and AP for a representative patient case. The dose distribution appears more uniform in the AP plan, with better control of high-dose regions, consistent with the quantitative findings of reduced V_105% values.

Figure 3 Axial view comparison of dose distributions between MP (left) and AP (right) for a representative patient case. The green contour delineates the PTV. The blue and yellow contours delineate the right and left hippocampi, respectively. The purple isodose line represents the 105% prescription dose. AP, automated planning; MP, manual planning; PTV, planning target volume.

OAR sparing evaluation

The comprehensive dosimetric analysis of OAR revealed significant improvements in critical structure sparing with AP compared to MP. As detailed in Table 4, the minimum dose to both the left and right hippocampus (D_100%) was significantly reduced in AP plans (left: 7.90±0.23 vs. 8.43±0.34 Gy, P=0.002; right: 7.92±0.18 vs. 8.46±0.26 Gy, P<0.001). Although the maximum doses to both hippocampi (D_max) were slightly lower in AP plans, these differences did not reach statistical significance (left: 15.14±0.53 vs. 15.37±0.39 Gy, P=0.06; right: 15.22±0.53 vs. 15.40±0.39 Gy, P=0.27).

Notably, the mean doses to both the left and right cochlea were substantially reduced in AP plans compared to MP plans (16.77±0.67 vs. 21.88±4.84 Gy and 16.77±0.66 vs. 21.91±4.79 Gy, respectively, P=0.005 for both). This represents an approximate 23.4% reduction in the mean cochlear dose for both the left and right cochlea. Other OARs, including optic nerves, chiasm, lenses, and eyes, showed no statistically significant differences in dose metrics between the two planning techniques.

Efficiency analysis

One of the most notable advantages of AP over MP was the substantial reduction in planning time. As shown in Table 5, the mean planning time was reduced by approximately 62.2% with AP compared to MP (14.61±1.65 vs. 38.68±8.71 minutes, P=0.002).

The number of MU required for plan delivery was slightly higher in AP plans compared to MP plans (MU: 1,646.69±103.15 vs. 1,561.74±129.16), though this difference was not statistically significant (P=0.08). The similar MU values suggest that plan complexity and delivery efficiency are comparable between the two techniques.

Plan consistency

A notable observation across all parameters was the reduction in variability with AP plans compared to MP plans. As shown in Figure 4, the boxplots for D_2% and V_105% reveal that AP plans consistently exhibit smaller interquartile ranges (e.g., V_105% ranges from approximately 17% to 22% in AP vs. 7% to 55% in MP) and fewer outliers compared to MP plans, indicating greater consistency. For D_2%, the standard deviation (SD) in AP plans was reduced six-fold compared to MP plans (0.05 vs. 0.30 Gy). Similarly, for V_105%, the variability was substantially reduced, with the SD in AP plans being 1.44% compared to 12.04% in MP plans. Planning time also showed markedly less variability with AP (SD: 1.65 minutes) than with MP (SD: 8.71 minutes). This consistency in AP plans extended to OAR dosimetry as well, particularly for the cochlea, where the SD was reduced from approximately 4.8 Gy in MP plans to 0.67 Gy in AP plans.

Figure 4 Boxplots comparing parameters between MP and AP. Left: D_2% dose (Gy) for the PTV. Right: V_105% for the PTV. The square symbols indicate the data points for case No. 10. AP, automated planning; MP, manual planning; PTV, planning target volume; V_105%, volume receiving 105% of the prescribed dose.

Discussion

This study developed and validated a script-based automated VMAT planning system for HS-WBRT that addresses critical challenges in contemporary radiotherapy practice. Our automated approach achieves superior dosimetric outcomes while substantially improving workflow efficiency and plan consistency, offering a practical solution for standardizing HS-WBRT planning in clinical practice.

Dosimetric advantages

Compared to MP, AP demonstrated superior dosimetric outcomes across multiple metrics while maintaining equivalent target coverage (D_98%: 29.31±0.57 vs. 29.52±0.51 Gy, P=0.07). Most importantly, AP substantially reduced high-dose regions, with V_105% decreasing from 34.23% to 19.28% (P=0.004) and D_2% declining from 32.34 to 31.99 Gy (P=0.008).

Enhanced OAR sparing represents another key advantage of AP. Bilateral hippocampal D_100% doses were significantly reduced (7.90 and 7.92 Gy, both P<0.001), comfortably meeting RTOG 0933 constraints while potentially offering improved neurocognitive outcomes in clinical practice (8). Furthermore, AP achieved a 23.4% reduction in mean cochlear dose (P=0.005), potentially reducing ototoxicity risk—an important consideration given the prevalence of hearing impairment.

AP produced dosimetrically superior or equivalent plans across all evaluated metrics while maintaining comparable treatment delivery complexity (MU: 1,646.69±103.15 vs. 1,561.74±129.16, P=0.08).

Efficiency and consistency improvements

Beyond dosimetric superiority, AP demonstrated remarkable improvements in planning efficiency and consistency. Planning time was reduced by 62.2% (14.61±1.65 vs. 38.68±8.71 minutes, P=0.002), representing substantial workflow optimization without compromising quality. This efficiency gain has significant clinical implications, particularly for time-sensitive cases and institutions with limited medical physics resources.

Equally important, AP exhibited markedly enhanced consistency across all dosimetric parameters. The SD of D_2% decreased six-fold (0.05 vs. 0.30 Gy), while V_105% variability reduced from 12.04% to 1.44%. This consistency extended to OAR dosimetry, with cochlear dose SD decreasing from 4.8 to 0.67 Gy. Such improvements eliminate operator- dependent variations and ensure reproducible high-quality plans regardless of planner experience—a critical advantage for maintaining treatment quality across different shifts and staff members.

Comparison with existing approaches

Previous MP studies by Rong et al. (10) and Shen et al. (11) established the dosimetric feasibility of VMAT for HS-WBRT but required extensive optimization time and introduced inter-operator variability. Our automated approach achieves comparable or superior dosimetric outcomes while reducing planning time by over 60% and eliminating operator-dependent inconsistency.

Recently, Rhee et al. (23) demonstrated the clinical feasibility of automated HS-WBRT planning through physician quality assessment, with autoplans achieving an average score of 4.9/5.0 and effectively reducing target hotspots while maintaining or improving OAR sparing. Their work established the foundation for script-based automation in this treatment site.

Building upon this work, our study provides three additional dimensions of evaluation. First, we quantify planning efficiency through direct timing comparisons—an analysis not performed in prior studies—demonstrating 62.2% time reduction (P=0.002). Second, we assess plan consistency through inter-plan variability analysis, revealing six-fold reduction in D_2% SD and substantially improved V_105% reproducibility. Third, our modular four-stage algorithm achieves superior dose homogeneity with V_105% reduced by 43.7%. These complementary analyses establish both the clinical feasibility and practical efficiency advantages of automated HS-WBRT planning.

Compared to KBP and MCO approaches, our script-based methodology offers distinct practical advantages. Unlike KBP methods (25,26), which require extensive institutional training data from high-quality historical plans, our rule-based algorithm enables immediate implementation with transparent, modifiable optimization logic. This flexibility allows institutions to customize the algorithm according to specific clinical preferences without rebuilding predictive models. While MCO approaches (17-19) enable interactive plan exploration through Pareto surface navigation, they require manual selection that may introduce operator-dependent variability. Our fully automated workflow prioritizes standardization and reproducibility, producing consistent plans without human intervention during optimization.

Study limitations

Despite these substantial advantages, certain limitations warrant acknowledgment. As illustrated in Figure 4, case No. 10 demonstrated superior dosimetric outcomes with MP V_105% (7.43% vs. 19.01%) and D_2% (31.74 vs. 32.02 Gy), suggesting that predetermined algorithmic pathways may occasionally limit optimization flexibility compared to experienced planners’ extended iterative refinement. This observation indicates that while AP delivers consistent quality across most patients, its standardized algorithmic approach may limit optimization space exploration compared to MP. Experienced planners, given adequate time, can continue iterative refinement through trial-and-error to identify superior solutions in specific cases. However, the overall superior performance across the majority of cases (11 of 12) supports the clinical value of this approach, particularly given time constraints in routine practice.

Additionally, the study’s small sample size (n=12) limits the generalizability of findings. The single-institution design using a specific linear accelerator (Varian Edge) without validation across different treatment delivery platforms further constrains the applicability of results. Future work should prioritize multi-institutional validation with larger patient cohorts encompassing diverse anatomical presentations and various linear accelerator platforms to establish the algorithm’s robustness and transferability across clinical settings.

Conclusions

This study demonstrates that script-based AP for HS-WBRT generates high-quality VMAT plans with superior dosimetric outcomes and substantial efficiency gains. Compared to MP, the automated approach significantly reduced high-dose regions (V_105%: 19.28% vs. 34.23%, P=0.004), enhanced hippocampal sparing (bilateral D_100%: P<0.001), decreased cochlear doses (23.4% reduction, P=0.005), and reduced planning time by 62.2% (P=0.002), while maintaining comparable delivery complexity. The standardized workflow offers valuable clinical advantages including reduced inter-planner variability, decreased workload burden, and consistent plan quality, supporting its potential for clinical implementation.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the STROBE reporting checklist. Available at https://tro.amegroups.com/article/view/10.21037/tro-25-30/rc

Data Sharing Statement: Available at https://tro.amegroups.com/article/view/10.21037/tro-25-30/dss

Peer Review File: Available at https://tro.amegroups.com/article/view/10.21037/tro-25-30/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tro.amegroups.com/article/view/10.21037/tro-25-30/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. The study was approved by the Institutional Review Board of Chang Gung Medical Foundation (IRB No. 202500595B0) and individual consent for this retrospective analysis was waived due to the use of de-identified clinical data.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Mehta MP, Tsao MN, Whelan TJ, et al. The American Society for Therapeutic Radiology and Oncology (ASTRO) evidence-based review of the role of radiosurgery for brain metastases. Int J Radiat Oncol Biol Phys 2005;63:37-46. [Crossref] [PubMed]
2. Nayak L, Lee EQ, Wen PY. Epidemiology of brain metastases. Curr Oncol Rep 2012;14:48-54. [Crossref] [PubMed]
3. Tsao MN, Xu W, Wong RK, et al. Whole brain radiotherapy for the treatment of newly diagnosed multiple brain metastases. Cochrane Database Syst Rev 2018;1:CD003869. [Crossref] [PubMed]
4. Chang EL, Wefel JS, Hess KR, et al. Neurocognition in patients with brain metastases treated with radiosurgery or radiosurgery plus whole-brain irradiation: a randomised controlled trial. Lancet Oncol 2009;10:1037-44. [Crossref] [PubMed]
5. Welzel G, Fleckenstein K, Schaefer J, et al. Memory function before and after whole brain radiotherapy in patients with and without brain metastases. Int J Radiat Oncol Biol Phys 2008;72:1311-8. [Crossref] [PubMed]
6. Monje ML, Mizumatsu S, Fike JR, et al. Irradiation induces neural precursor-cell dysfunction. Nat Med 2002;8:955-62. [Crossref] [PubMed]
7. Mizumatsu S, Monje ML, Morhardt DR, et al. Extreme sensitivity of adult neurogenesis to low doses of X-irradiation. Cancer Res 2003;63:4021-7.
8. Gondi V, Pugh SL, Tome WA, et al. Preservation of memory with conformal avoidance of the hippocampal neural stem-cell compartment during whole-brain radiotherapy for brain metastases (RTOG 0933): a phase II multi-institutional trial. J Clin Oncol 2014;32:3810-6. [Crossref] [PubMed]
9. Brown PD, Gondi V, Pugh S, et al. Hippocampal Avoidance During Whole-Brain Radiotherapy Plus Memantine for Patients With Brain Metastases: Phase III Trial NRG Oncology CC001. J Clin Oncol 2020;38:1019-29. [Crossref] [PubMed]
10. Rong Y, Evans J, Xu-Welliver M, et al. Dosimetric evaluation of intensity-modulated radiotherapy, volumetric modulated arc therapy, and helical tomotherapy for hippocampal-avoidance whole brain radiotherapy. PLoS One 2015;10:e0126222. [Crossref] [PubMed]
11. Shen J, Bender E, Yaparpalvi R, et al. An efficient volumetric-modulated arc therapy treatment planning approach for hippocampal-avoidance whole-brain radiation therapy (HA-WBRT). Med Dosim 2015;40:205-209. [Crossref] [PubMed]
12. Gondi V, Tolakanahalli R, Mehta MP, et al. Hippocampal-sparing whole-brain radiotherapy: a "how-to" technique using helical tomotherapy and linear accelerator-based intensity-modulated radiotherapy. Int J Radiat Oncol Biol Phys 2010;78:1244-52. [Crossref] [PubMed]
13. Hussein M, Heijmen BJM, Verellen D, et al. Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations. Br J Radiol 2018;91:20180270. [Crossref] [PubMed]
14. Chanyavanich V, Das SK, Lee WR, et al. Knowledge-based IMRT treatment planning for prostate cancer. Med Phys 2011;38:2515-22. [Crossref] [PubMed]
15. Tol JP, Delaney AR, Dahele M, et al. Evaluation of a knowledge-based planning solution for head and neck cancer. Int J Radiat Oncol Biol Phys 2015;91:612-20. [Crossref] [PubMed]
16. Sinha S, Kumar A, Maheshwari G, et al. Development and Validation of Single-Optimization Knowledge-Based Volumetric Modulated Arc Therapy Model Plan in Nasopharyngeal Carcinomas. Adv Radiat Oncol 2024;9:101311. [Crossref] [PubMed]
17. Craft DL, Hong TS, Shih HA, et al. Improved planning time and plan quality through multicriteria optimization for intensity-modulated radiotherapy. Int J Radiat Oncol Biol Phys 2012;82:e83-90. [Crossref] [PubMed]
18. Kierkels RG, Visser R, Bijl HP, et al. Multicriteria optimization enables less experienced planners to efficiently produce high quality treatment plans in head and neck cancer radiotherapy. Radiat Oncol 2015;10:87. [Crossref] [PubMed]
19. Müller BS, Shih HA, Efstathiou JA, et al. Multicriteria plan optimization in the hands of physicians: a pilot study in prostate cancer and brain tumors. Radiat Oncol 2017;12:168. [Crossref] [PubMed]
20. Yang Y, Shao K, Zhang J, et al. Automatic Planning for Nasopharyngeal Carcinoma Based on Progressive Optimization in RayStation Treatment Planning System. Technol Cancer Res Treat 2020;19:1533033820915710. [Crossref] [PubMed]
21. Gleeson I, Bolger N, Chun H, et al. Implementation of automated personalised breast radiotherapy planning techniques with scripting in Raystation. Br J Radiol 2023;96:20220707. [Crossref] [PubMed]
22. Funderud M, Hoem IS, Guleng MAD, et al. Script-based automatic radiotherapy planning for cervical cancer. Acta Oncol 2023;62:1798-807. [Crossref] [PubMed]
23. Rhee DJ, Perni S, Perrin KJ, et al. Semi-automated hippocampal avoidance whole-brain radiotherapy planning. J Appl Clin Med Phys 2025;26:e70076. [Crossref] [PubMed]
24. Yuen AHL, Wu PM, Li AKL, et al. Volumetric modulated arc therapy (VMAT) for hippocampal-avoidance whole brain radiation therapy: planning comparison with Dual-arc and Split-arc partial-field techniques. Radiat Oncol 2020;15:42. [Crossref] [PubMed]
25. Momin S, Fu Y, Lei Y, et al. Knowledge-based radiation treatment planning: A data-driven method survey. J Appl Clin Med Phys 2021;22:16-44. [Crossref] [PubMed]
26. Chung CV, Khan MS, Olanrewaju A, et al. Knowledge-based planning for fully automated radiation therapy treatment planning of 10 different cancer sites. Radiother Oncol 2025;202:110609. [Crossref] [PubMed]