Is a single isocenter sufficient for volumetric modulated arc therapy radiosurgery when multiple intracranial metastases are spatially dispersed?

Long-Term Intracranial Outcomes With Combination Dual Immune-Checkpoint Blockade and Stereotactic Radiosurgery in Patients With Melanoma and Non-Small Cell Lung Cancer Brain Metastases

PURPOSE

The intracranial benefit of offering dual immune-checkpoint inhibition (D-ICPI) with ipilimumab and nivolumab to patients with melanoma or non-small cell lung cancer (NSCLC) receiving stereotactic radiosurgery (SRS) for brain metastases (BMs) is unknown. We hypothesized that D-ICPI improves local control compared with SRS alone.

METHODS AND MATERIALS

Patients with melanoma or NSCLC treated with SRS from 2014 to 2022 were evaluated. Patients were stratified by treatment with D-ICPI, single ICPI (S-ICPI), or SRS alone. Local recurrence, intracranial progression (IP), and overall survival were estimated using competing risk and Kaplan-Meier analyses. IP included both local and distant intracranial recurrence.

RESULTS

Two hundred eighty-eight patients (44% melanoma, 56% NSCLC) with 1,704 BMs were included. Fifty-three percent of patients had symptomatic BMs. The median follow-up was 58.8 months. Twelve-month local control rates with D-ICPI, S-ICPI, and SRS alone were 94.73% (95% CI, 91.11%-96.90%), 91.74% (95% CI, 89.30%-93.64%), and 88.26% (95% CI, 84.07%-91.41%). On Kaplan-Meier analysis, only D-ICPI was significantly associated with reduced local recurrence (P = .0032). On multivariate Cox regression, D-ICPI (hazard ratio [HR], 0.4003; 95% CI, 0.1781-0.8728; P = .0239) and planning target volume (HR, 1.022; 95% CI, 1.004-1.035; P = .0059) correlated with local control. One hundred seventy-three (60%) patients developed IP. The 12-month cumulative incidence of IP was 41.27% (95% CI, 30.27%-51.92%), 51.86% (95% CI, 42.78%-60.19%), and 57.15% (95% CI, 44.98%-67.59%) after D-ICPI, S-ICPI, and SRS alone. On competing risk analysis, only D-ICPI was significantly associated with reduced IP (P = .0408). On multivariate Cox regression, D-ICPI (HR, 0.595; 95% CI, 0.373-0.951; P = .0300) and presentation with >10 BMs (HR, 2.492; 95% CI, 1.668-3.725; P < .0001) remained significantly correlated with IP. The median overall survival after D-ICPI, S-ICPI, and SRS alone was 26.1 (95% CI, 15.5-40.7), 21.5 (16.5-29.6), and 17.5 (11.3-23.8) months. S-ICPI, fractionation, and histology were not associated with clinical outcomes. There was no difference in hospitalizations or neurologic adverse events between cohorts.

CONCLUSIONS

The addition of D-ICPI for patients with melanoma and NSCLC undergoing SRS is associated with improved local and intracranial control. This appears to be an effective strategy, including for patients with symptomatic or multiple BMs.

Impact of setup errors on the robustness of linac-based single-isocenter coplanar and non-coplanar VMAT plans for multiple brain metastases

PURPOSE

Patient setup errors have been a primary concern impacting the dose delivery accuracy in radiation therapy. A robust treatment plan might mitigate the effects of patient setup errors. In this reported study, we aimed to evaluate the impact of translational and rotational errors on the robustness of linac-based, single-isocenter, coplanar, and non-coplanar volumetric modulated arc therapy treatment plans for multiple brain metastases.

METHODS

Fifteen patients were retrospectively selected for this study with a combined total of 49 gross tumor volumes (GTVs). Single-isocenter coplanar and non-coplanar plans were generated first with a prescribed dose of 40 Gy in 5 fractions or 42 Gy in 7 fractions to cover 95% of planning target volume (PTV). Next, four setup errors (+1  and +2 mm translation, and +1° and +2° rotation) were applied individually to generate modified plans. Different plan quality evaluation metrics were compared between coplanar and non-coplanar plans. 3D gamma analysis (3%/2 mm) was performed to compare the modified plans (+2 mm and +2° only) and the original plans. Paired t-test was conducted for statistical analysis.

RESULTS

After applying setup errors, variations of all plan evaluation metrics were similar (p > 0.05). The worst case for V100% to GTV was 92.07% ± 6.13% in the case of +2 mm translational error. 3D gamma pass rates were > 90% for both coplanar (+2 mm and +2°) and the +2 mm non-coplanar groups but was 87.40% ± 6.89% for the +2° non-coplanar group.

CONCLUSION

Translational errors have a greater impact on PTV and GTV dose coverage for both planning methods. Rotational errors have a greater negative impact on gamma pass rates of non-coplanar plans. Plan evaluation metrics after applying setup errors showed that both coplanar and non-coplanar plans were robust and clinically acceptable.

Single-Isocenter Linac-Based Radiosurgery for Brain Metastases with Coplanar Arcs: A Dosimetric and Clinical Analysis

The efficacy of linac-based SRS/fSRS treatments using the single-isocenter coplanar FFF-VMAT technique for both single and multiple BM was investigated. Seventy patients (129 BM) treated with 15-21 Gy in 1 (n = 59) or 27 Gy in 3 (n = 11) fractions were analyzed. For each fraction, plans involving the intra-fractional errors measured by post-treatment CBCT were recalculated. The relationships of BM size, distance-to-isocenter, and barycenter shift with the difference in target coverage were evaluated. Clinical outcomes were assessed using logistic regression and Kaplan-Meier analysis. The median delivery time was 3.78 min (range, 1.83-9.25). The median post-treatment 3D error was 0.5 mm (range, 0.1-2.7) and the maximum rotational error was 0.3° (range, 0.0-1.3). In single BM patients, the GTV D95% was never reduced by >5%, whereas PTV D95% reductions >1% occurred in only 11 cases (29%). In multiple BM patients, dose deficits >5% and >1% occurred in 2 GTV (2%) and 34 PTV (37%), respectively. The differences in target coverage showed a moderate-to-strong correlation only with barycenter shift. Local failure of at least one treated BM occurred in 13 (21%) patients and the 1-year and 2-year local control rates for all lesions were 94% and 90%, respectively. The implemented workflow ensured that the degradation of target and brain dose metrics in delivered treatments was negligible. Along with encouraging clinical outcomes, these findings warrant a reduction in the PTV margins at our institution.

Gantry triggered x-ray verification during single-isocenter stereotactic radiosurgery: Increased certainty for a no-margin strategy

BACKGROUND

Single-isocenter linac-based stereotactic radiosurgery (SRS) has emerged as a dedicated treatment option for multiple brain metastases. Consequently, image-guidance for patient positioning and motion management has become very important. The purpose of this study was to analyze intra-fraction errors measured with stereoscopic x-rays and their impact on the dose distribution.

MATERIALS AND METHODS

Treatments were planned with non- coplanar dynamic conformal arcs for 33 patients corresponding to 127 brain lesions and 356 arcs. Intra-arc positioning errors were measuredusing stereoscopic x-rays (ExacTrac Dynamic, Brainlab), triggered during arc delivery. Couch corrections above 0.7 mm and 0.5° were always applied. Intra-arc positioning data was analyzed. The dose impact was evaluated by applying the measured errors to the dose given in each arc.

RESULTS

Median residual errors were 0.10 mm, 0.13 mm and 0.08 mm for the lateral, longitudinal and vertical directions and 0.10°, 0.08° and 0.13° for the pitch, roll and yaw angles respectively. 90% of the treatment arcs showed shifts of less than 0.4 mm and 0.4°in all directions. Dosimetric impact of motion showed the largest losses in coverage on small targets. All targets achieved at least 95% of the prescription dose to 95% of their volume, even when planned without margins.

CONCLUSIONS

Intra-fractional errors measured during beam delivery were found to be notably low with a dose impact that showed acceptable target coverage when applying these intra-arc errors to the dose distributions of the individual treatment arcs. Using an adequate immobilization and intra-fraction imaging prior to and during irradiation, no margins need to be added to compensate for intra-fraction motion.

Radiosurgery for Five to Fifteen Brain Metastases: A Single Centre Experience and a Review of the Literature

Purpose

Stereotactic radiosurgery (SRS) is now mainstream for patients with 1-4 brain metastases however the management of patients with 5 or more brain metastases remains controversial. Our aim was to evaluate the clinical outcomes of patients with 5 or more brain metastases and to compare with published series as a benchmarking exercise.

Methods

Patients with 5 or more brain metastases treated with a single isocentre dynamic conformal arc technique on a radiosurgery linac were identified from the institutional database. Endpoints were local control, distant brain failure, leptomeningeal disease and overall survival. Dosimetric data were extracted from the radiosurgery plans. Series reporting outcomes following SRS for multiple brain metastases were identified by a literature search.

Results

36 patients, of whom 35 could be evaluated, received SRS for 5 or more brain metastases between February 2015 and October 2021. 25 patients had 5-9 brain metastases (group 1) and 10 patients had 10-15 brain metastases (group 2). The mean number of brain metastases in group 1 was 6.3 (5-9) and 12.3 (10-15) in group 2. The median cumulative irradiated volume was 4.6 cm³ (1.25-11.01) in group 1 and 7.2 cm³ (2.6-11.1) in group 2. Median follow-up was 12 months. At last follow-up, local control rates per BM were 100% and 99.8% as compared with a median of 87% at 1 year in published series. Distant brain failure was 36% and 50% at a median interval of 5.2 months and 7.4 months after SRS in groups 1 and 2 respectively and brain metastasis velocity at 1 year was similar in both groups (9.7 and 11). 8/25 patients received further SRS and 7/35 patients received whole brain radiotherapy. Median overall survival was 10 months in group 1 and 15.7 months in group 2, which compares well with the 7.5 months derived from the literature. There was one neurological death in group 2, leptomeningeal disease was rare (2/35) and there were no cases of radionecrosis.

Conclusion

With careful patient selection, overall survival following SRS for multiple brain metastases is determined by the course of the extracranial disease. SRS is an efficacious and safe modality that can achieve intracranial disease control and should be offered to patients with 5 or more brain metastases and a constellation of good prognostic factors.

Comparison of dosimetric impact of intra-fractional setup discrepancy between multiple- and single-isocenter approaches in linac-based stereotactic radiotherapy of multiple brain metastases

Introduction

Treatment of multiple brain metastases by linac‐based stereotactic radiotherapy (SRT) can employ either a multiple‐isocenter (MI) or single‐isocenter (SI) approach. The purposes of this study were to evaluate the dosimetric results of MI and SI approaches and compare the impacts of intra‐fractional setup discrepancies on the robustness of respective approaches using isocenter shifts, whether the same magnitude of translational and rotational effects could lead to a significant difference between the two approaches.

Methods

Twenty‐two patients with multiple brain metastases treated by linac‐based SRT were recruited. Treatment plans were computed with both the MI and SI approaches. For the MI approach, the isocenter was located at the geometric center of each planning target volumes (PTVs), whereas the isocenter of the SI approach was located midway between the PTV centroids. To simulate the intra‐fractional errors, isocenter displacements including translational and rotational shifts were hypothetically applied. Apart from the dosimetric outcomes of the two approaches, the impact of the isocenter shifts on PTVs and organs at risk (OARs) were recorded in terms of the differences (δ) in dose parameters relative to the reference plan and was then compared between the MI and SI approaches.

Results

Both MI and SI plans met the plan acceptance criteria. The mean Paddick conformity index (Paddick CI) and D_max of most OARs between MI and SI plans did not show a significant difference, except that higher doses to the left optic nerve and optic chiasm were found in SI plans (p = 0.03). After the application of the isocenter shifts, δCI increased with an increase in the magnitude of the isocenter shift. When comparing between MI and SI plans, the δCIs were similar (p > 0.05) for all extents of translational shifts, but δCIs were significantly higher in SI plans after application of all rotations particularly ±1.5° and ±2.0° shifts. Despite the result that the majority of δD_Max of OARs were higher in the SI plans, only the differences in the left optic nerve and chiasm showed generally consistent significance after both translational ≥±1 mm and rotational shifts of ≥±1∘.

Conclusion

Both MI and SI approaches could produce clinically acceptable plans. However, isocenter shifts brought dosimetric impacts to both MI and SI approaches and the effects increased with the increase of the shift magnitude. Although similar impacts were shown in plans of both approaches after translational isocenter shift, SI plans were relatively more vulnerable than MI plans to rotational shifts.

Improving efficiency in the radiation management of multiple brain metastases using a knowledge-based planning solution for single-isocentre volumetric modulated arc therapy (VMAT) technique

INTRODUCTION

This study aimed to develop a single-isocentre volumetric modulated arc therapy (si-VMAT) technique for multiple brain metastases using knowledge-based planning software, comparing it with a multiple-isocentre stereotactic radiosurgery (mi-SRS) planning approach.

METHODS

Twenty-six si-VMAT plans were created and uploaded into RapidPlanTM (RP) to create a si-VMAT model. Ten patients, with 2 to 6 metastases (mets), were planned with a si-VMAT technique utilising RP, and a mi-SRS technique on Brainlab iPlan. Paddick Conformity Index (PCI) was used to compare conformity. The volumes of the brain receiving 15Gy, 12Gy, 10Gy, 7.5Gy and 3Gy were also compared. Retrospective treatment times from the last eight patients treated were averaged for pre-imaging and beam on time to calculate treatment times for both techniques.

RESULTS

There was a significant difference in the PCI scores for the mi-SRS plans (M = 0.667, SD = 0.114) and si-VMAT plans (M = 0.728, SD = 0.088), with PCI values suggesting better prescription dose conformity with the si-VMAT technique (P = 0.014). Percentage of total brain volume receiving low-dose wash at four of the five different dose levels was significantly less (P < 0.05) with mi-SRS. Average time to treat a single met with current mi-SRS technique is 25.7 min, with each additional met requiring this same amount of time. The average time to treat 2-3 mets using si-VMAT would be 25.3 min and 4+ metastases 33.5 min.

CONCLUSION

A knowledge-based si-VMAT approach was efficient in planning and treating multi metastases while achieving clinically acceptable dosimetry with respect to dose conformity and low-dose fall off.

Evaluation of parameters affecting gamma passing rate in patient-specific QAs for multiple brain lesions IMRS treatments using ray-station treatment planning system

Purpose

Using intensity‐modulated radiosurgery (IMRS) with single isocenter for the treatment of multiple brain lesions has gained acceptance in recent years. One of the challenges of this technique is conducting a patient‐specific quality assurance (QA), involving accurate gamma passing rate (GPR) calculations for small and wide spread‐out targets. We evaluated effects of parameters such as dose grid and energy on GPR using our clinical IMRS plans.

Methods

Ten patients with total of 40 volumetric modulated arc therapy (VMAT) plans were created in Raystation (V.8A) treatment planning system (TPS) for the Varian Edge Linac using 6 and 10 flattening filter‐free (FFF) beams and planned dose grids of 1 mm and 2 mm resulting in four plans with 6–10 targets per patient. All parameters and objectives except dose grid and energy were kept the same in all plans. Next, patient‐specific QAs were measured evaluating GPR with 10% threshold, 3%/3 mm objective, and an acceptance criterion of 95%. Modulation factors (MF) and confidence intervals were calculated. Two modes of measurements, standard density (SD) and high density (HD), were used.

Results

Generally, plans computed with 1 mm dose grid have higher GPRs than those with 2 mm dose grid for both energies used. The GPRs of 6 FFF plans were higher than those of 10 FFF plans. GPR showed no noticeable difference between HD and SD measurements. Negative correlation between MF and GPR was observed. The HD pass rates fall within the confidence interval of SD.

Conclusion

Calculated dose grid should be less than or equal to one‐third of distance to agreement, thus 1 mm planned dose grid is recommended to reduce artifacts in gamma calculation. GPR of SD and HD measurement modes is almost the same, which indicates that SD mode is clinically preferable for performing patient‐specific QAs. According to our results, using 6 FFF beams with 1 mm planned dose grid is more accurate and reliable for dose calculation of IMRS plans.

Longitudinal Grouping of Target Volumes for Volumetric-Modulated Arc Therapy of Multiple Brain Metastases

Purpose

Treatment of multiple brain metastases with single-isocenter volumetric modulated arc therapy causes unnecessary exposure to normal brain tissue. In this study, a longitudinal grouping method was developed to reduce such unnecessary exposure.

Materials and Methods

This method has two main aspects: grouping brain lesions longitudinally according to their longitudinal projection positions in beam’s eye view, and rotating the collimator to 90° to make the multiple leaf collimator leaves conform to the targets longitudinally group by group. For 11 patients with multiple (5–30) brain metastases, two single-isocenter volumetric modulated arc therapy plans were generated using a longitudinal grouping strategy (LGS) and the conventional strategy (CVS). The prescription dose was 52 Gy for 13 fractions. Dose normalization to 100% of the prescription dose in 95% of the planning target volume was adopted. For plan quality comparison, Paddick conformity and the gradient index of the planning target volume, and the mean dose, the V_100%, V_50%, V_25%, and V_10% volumes of normal brain tissue were calculated.

Results

There were no significant differences between the LGS and CVS plans in Paddick conformity (p = 0.374) and the gradient index (p = 0.182) of the combined planning target volumes or for V_100% (p = 0.266) and V_50% (p = 0.155) of the normal brain. However, the V_25% and V_10% of the normal brain which represented the low-dose region were significantly reduced in the LGS plans (p = 0.004 and p = 0.003, respectively). Consistently, the mean dose of the entire normal brain was 12.04 and 11.17 Gy in the CVS and LGS plans, respectively, a significant reduction in the LGS plans (p = 0.003).

Conclusions

The longitudinal grouping method can decrease unnecessary exposure and reduces the low-dose range in normal brain tissue.

Single isocenter SRS using CAVMAT offers improved robustness to commissioning and treatment delivery uncertainty compared to VMAT

Purpose

In this study, we evaluate and compare single isocenter multiple target VMAT (SIMT) and Conformal Arc Informed VMAT (CAVMAT) radiosurgery's sensitivity to uncertainties in dosimetric leaf gap (DLG) and treatment delivery. CAVMAT is a novel planning technique that uses multiple target conformal arcs as the starting point for limited inverse VMAT optimization.

Methods

All VMAT and CAVMAT plans were recalculated with DLG values of 0.4, 0.8, and 1.2 mm. DLG effect on V_6Gy[cc], V_12Gy[cc], and V_16Gy[cc], and target dose was evaluated. Plans were delivered to a Delta⁴ (ScandiDos, Madison, WI) phantom and gamma analysis performed with varying criteria. Log file analysis was performed to evaluate MLC positional error. Sixteen targets were delivered to a SRS MapCHECK (Sun Nuclear Corp., Melbourne, FL) to evaluate VMAT and CAVMAT's dose difference (DD) as a function of DLG.

Results

VMAT's average maximum and minimum target dose sensitivity to DLG was 9.08 ±3.50%/mm and 9.50 ± 3.30%/mm, compared to 3.20 ± 1.60%/mm and 4.72 ± 1.60%/mm for CAVMAT. For VMAT, V_6Gy[cc], V_12Gy[cc], and V_16Gy[cc] sensitivity was 35.83 ± 9.50%/mm, 34.12 ± 6.60%/mm, and 39.23 ± 8.40%/mm. In comparison, CAVMAT's sensitivity was 23.19 ± 4.50%/mm, 22.45 ± 4.40%/mm, and 24.88 ± 4.90%/mm, respectively. Upon delivery to the Delta⁴, CAVMAT offered superior dose agreement compared to VMAT. For a 1%/1 mm gamma analysis, VMAT and CAVMAT had a passing rate of 94.53 ± 4.40% and 99.28 ± 1.70%, respectively. CAVMAT was more robust to DLG variation, with the SRS MapCHECK plans yielding an absolute average DD sensitivity of 2.99 ± 1.30%/mm compared to 5.07 ± 1.10%/mm for VMAT. Log files demonstrated minimal differences in MLC positional error for both techniques.

Conclusions

CAVMAT remains robust to delivery uncertainties while offering a target dose sensitivity to DLG less than half that of VMAT, and 65% of that of VMAT for V_6Gy[cc], V_12Gy[cc], and V_16Gy[cc]. The superior dose agreement and reduced sensitivity of CAVMAT to DLG uncertainties indicate promise as a robust alternative to VMAT for SIMT SRS.

Transition From Manual to Automated Planning and Delivery of Volumetric Modulated Arc Therapy Stereotactic Radiosurgery: Clinical, Dosimetric, and Quality Assurance Results

PURPOSE

Properly planned single isocenter volumetric modulated arc therapy (VMAT) radiosurgery plans exhibit high quality and efficiency. We report here the largest clinical experience to date, to our knowledge, comparing manual planning with a new automated platform designed to standardize and simplify radiosurgery planning and delivery processes.

METHODS

We treated 693 patients with single isocenter VMAT radiosurgical plans generated by either our conventional manual (mVMAT) or a recently implemented automated (HyperArc^TM) technique. All plans targeted the gross tumor volume without margin. Radiochromic film was used for patient-specific quality assurance (PSQA). We evaluated local control and toxicity data for a subgroup of 107 patients having 377 metastatic tumors that were treated with HyperArc.

RESULTS

The median Radiation Therapy Oncology Group (RTOG) conformity index was 1.14 and was not different between the 2 techniques. The median Paddick gradient index was 5.42 for HyperArc versus 7.09 for mVMAT (P < .001). The median mean brain doses were 4.6% and 5.1% for HyperArc and mVMAT, respectively (P = .04). The PSQA for both techniques met clinical criteria, but 97% of the HyperArc plans satisfied the gamma tolerance limit recommended by the American Association of Physicists in Medicine Task Group No. 218, compared with 94% of the mVMAT plans (P = .02). The median treatment-planning times were not significantly different. The median treatment times were 10.5 and 11.4 minutes for HyperArc and mVMAT, respectively (P < .001). The Kaplan-Meier estimate of local control was 90.1% at 1 year.

CONCLUSIONS

HyperArc produces high-quality radiosurgical plans that are at least as good as mVMAT plans created by an expert manual planner with easier planning and more efficient delivery workflow. A less experienced planner can produce very high-quality radiosurgical plans even for patients with more than 10 targets. The use of a single-isocenter technique for multiple targets with no PTV margin did not compromise clinical outcomes, and 1-year local control for treated targets remained congruent with historical series.

Radiosurgery treatment planning using conformal arc informed volumetric modulated arc therapy

Linac based radiosurgery to multiple metastases is commonly planned with volumetric modulated arc therapy (VMAT) as it effectively achieves high conformality to complex target arrangements. However, as the number of targets increases, VMAT can struggle to block between targets, which can lead to highly modulated and/or nonconformal multi-leaf collimator (MLC) trajectories that unnecessarily irradiation of healthy tissue. In this study we introduce, describe, and evaluate a treatment planning technique called Conformal Arc Informed VMAT (CAVMAT), which aims to reduce the dose to healthy tissue while generating highly conformal treatment plans. CAVMAT is a hybrid technique which combines the conformal MLC trajectories of dynamic conformal arcs with the MLC modulation and versatility of inverse optimization. CAVMAT has 3 main steps. First, targets are assigned to subgroups to maximize MLC blocking between targets. Second, arc weights are optimized to achieve the desired target dose, while minimizing MU variation between arcs. Third, the optimized conformal arc plan serves as the starting point for limited inverse optimization to improve dose conformity to each target. Twenty multifocal VMAT cases were replanned with CAVMAT with 20Gy applied to each target. The total volume receiving 2.5_Gy[cm³], 6_Gy[cm³], 12_Gy[cm³], and 16_Gy[cm³], conformity index, treatment delivery time, and the total MU were used to compare the VMAT and CAVMAT plans. In addition, CAVMAT was compared to a broad range of planning strategies from various institutions (108 linear accelerator based plans, 14 plans using other modalities) for a 5-target case utilized in a recent plan challenge. For the linear accelerator-based plans, a plan complexity metric based on aperture opening area and perimeter, total monitor units (MU), and MU for a given aperture opening was utilized in the plan challenge scoring algorithm to compare the submitted plans to CAVMAT. After re-planning the 20 VMAT cases, CAVMAT reduced the average V_2.5Gy[cm³] by 25.25 ± 19.23%, V_6Gy[cm³] by 13.68 ± 18.97%, V_12Gy[cm³] by 11.40 ± 19.44%, and V_16Gy[cm³] by 6.38 ± 19.11%. CAVMAT improved conformity by 3.81 ± 7.57%, while maintaining comparable target dose. MU for the CAVMAT plans increased by 24.35 ± 24.66%, leading to an increased treatment time of 2 minutes. For the plan challenge case, CAVMAT was 1 of 12 linac based plans that met all plan challenge scoring criteria. Compared to the average submitted VMAT plan, CAVMAT increased the V_10%Gy[%] of healthy tissue (Brain-PTV) by roughly 3.42%, but in doing so was able to reduce the V_25%Gy[%] by roughly 3.73%, while also reducing V_50%Gy[%]_, V_75%Gy[%], and V_100%Gy[%]_. The CAVMAT technique successfully eliminated insufficient MLC blocking between targets prior to the inverse optimization, leading to less complex treatment plans and improved tissue sparing. Tissue sparing, improved conformity, and decreased plan complexity at the cost of slight increase in treatment delivery time indicates CAVMAT to be a promising method to treat brain metastases.

Initial Experience With Single-Isocenter Radiosurgery to Target Multiple Brain Metastases Using an Automated Treatment Planning Software: Clinical Outcomes and Optimal Target Volume Margins Strategy

Purpose

Our purpose was to assess the clinical outcomes and target positioning accuracy of frameless linear accelerator single-isocenter multiple-target (SIMT) dynamic conformal arc (DCA) stereotactic radiosurgery (SRS) for multiple brain metastases (BM).

Methods and Materials

Between October 2016 and September 2018, 31 consecutive patients ≥18 years old with 204 BM <3 cm in maximum size receiving SIMT DCA SRS were retrospectively evaluated. All plans were created using a dedicated automated treatment planning software (Brainlab, Munich, Germany), and treatments were performed with a Truebeam STx or a Novalis Tx (Brainlab and Varian Medical Systems, CA). The accuracy of setup and interfraction patient repositioning was assessed by Brainlab ExacTrac radiograph 6-dimensional image system and the risk of compromised target dose coverage evaluated. Brain control and overall survival were estimated by Kaplan-Meier method calculated from the time of SRS.

Results

Fourteen patients were treated for 4 to 6 and 17 patients for 7 to 10 BM. The mean gross tumor volume (GTV) was 0.65 cm³ and the mean planning target volume (PTV) was 0.89 cm³. Mean V95 (the volume of the PTV covered by 95% of the prescription dose) and D95 (the prescription dose covering 95% of the PTV) were 99.5% and 21.1 Gy, respectively. With a median clinical follow-up of 11 months (range, 4-26 months), the 1-year survival was 68% and local control was 89%. As a consequence of plan isocenter residual errors, a loss of target coverage, defined as V95 < 95%, occurred in 28 PTVs (10 patients); using a 1 mm GTV-to-PTV margin, adequate dose coverage was maintained for all lesions.

Conclusions

SIMT DCA SRS represents a fast and effective approach for patients with up to 10 BM. The dosimetric effects of residual set-up and intrafraction positioning errors are modest, although a GTV-to-PTV margin of 1 mm is recommended.

Implementation and validation of an in-house geometry optimization software for SRS VMAT planning of multiple cranial metastases

Purpose

The implementation and evaluation of an in‐house developed geometry optimization (GO) software are described. The GO script provides optimal lesion clustering, isocenter placement, and collimator angle of each arc for cranial multi‐lesion stereotactic radiosurgery (SRS) volumetric modulated arc therapy (VMAT) planning.

Materials and methods

An Eclipse‐plugin program was developed to facilitate automatic plan geometry generation for multiple metastases SRS VMAT plans. A mixed, semi‐supervised exhaustive and k‐means clustering method is used to group lesions and place isocenters. The sum of squared euclidean distance (SSED) and the boundaries of lesions’ projection from beams’ eye view are used as supervised parameters to determine the optimal isocenter numbers. The collimator angle is optimized by minimizing the sum of the MLC opening area from all gantry angles for each arc. In all, 10 clinical cases treated during 2016–2017 were compared to plan quality of GO script generated plans. Paddick gradient index (GI), conformity index (CI), and local brain volume receiving 12 Gy (local V12 Gy) around each lesion were compared.

Result

For four cases, the number of isocenters was reduced in the GO plans. For four other cases, the GO plans had the same number of isocenters as their corresponding clinical plans but with different lesion grouping. The GO plans had significantly lower GI (4.1 ± 1.0 vs 4.4 ± 0.9, P < 0.0001) and local V12 Gy (5.1 ± 4.2 vs 5.5 ± 4.3 in cm³, P < 0.0001), but not significantly different mean normal brain dose or CI. The volume of normal brain receiving ≥6 Gy was significantly lower in the GO plans. The total time to run the GO script for each case was <2 min.

Conclusion

The GO software automates lesion grouping, isocenter placement, and the collimator angles for SRS VMAT planning. When tested on 10 cases, the GO script resulted in improved plan quality and shorter planning time when compared to the clinical SRS VMAT plans.

Retrospective quality metrics review of stereotactic radiosurgery plans treating multiple targets using single-isocenter volumetric modulated arc therapy

PURPOSE

To characterize key plan quality metrics in multi-target stereotactic radiosurgery (SRS) plans treated using single-isocenter volumetric modulated arc therapy (VMAT) in comparison to dynamic conformal arc (DCA) plans treating single target. To investigate the feasibility of quality improvement in VMAT planning based on previous planning knowledge.

MATERIALS AND METHODS

97 VMAT plans of multi-target and 156 DCA plans of single-target treated in 2017 at a single institution were reviewed. A total of 605 targets were treated with these SRS plans. The prescription dose was normalized to 20 Gy in all plans for this analysis. Two plan quality metrics, target conformity index (CI) and normal tissue volume receiving more than 12 Gy (V12Gy), were calculated for each target. The distribution of V12Gy per target was plotted as a function of the target volume. For multi-target VMAT plans, the number of targets being treated in the same plan and the distance between targets were calculated to evaluate their impact on V12Gy. VMAT plans that had a large deviation of V12Gy from the average level were re-optimized to determine the possibility of reducing the variation of V12Gy in VMAT planning.

RESULTS

Conformity index of multi-target VMAT plans were lower than that of DCA plans while the mean values of 12 Gy were comparable. The V12Gy for a target in VMAT plan did not show apparent dependence on the total number of targets or the distance between targets. The distribution of V12Gy exhibited a larger variation in VMAT plans compared to DCA plans. Re-optimization of outlier plans reduced V12 Gy by 33.9% and resulted in the V12Gy distribution in VMAT plans more closely resembling that of DCA plans.

CONCLUSION

The benchmark data on key plan quality metrics were established for single-isocenter multi-target SRS planning. It is feasible to use this knowledge to guide VMAT planning and reduce high V12Gy outliers.

Monte Carlo dose verification for a single-isocenter VMAT plan in multiple brain metastases

The purpose of this study was to verify the accuracy of dose calculation algorithms of a treatment planning system for a single-isocenter volumetric modulated arc therapy (VMAT) plan in multiple brain metastases, by comparing the dose distributions of treatment planning system with those of Monte Carlo (MC) simulations. We used a multitarget phantom containing 9 acrylic balls with a diameter of 15.9 mm inside a Lucy phantom measuring 17 × 17 × 17 cm³. Seven VMAT plans were created using the multitarget phantom: 1 multitarget plan (MTP) and 6 single target plans (STP). Three of the STP plans had a large jaw field setting, almost equivalent to that of the MTP, while the other plans had a jaw field setting fitted to each planning target volume. The isocenter for all VMAT plans was set to the center of the phantom. The VMAT dose distributions were calculated using the analytical anisotropic algorithm (AAA) and were also recalculated through Acuros XB (AXB) and MC simulations under the same irradiation conditions. The AAA and AXB methods tended to overestimate dosage compared with the MC method in the MTP and in STPs with large jaw field settings. The dose distribution in single-isocenter VMAT plans for multiple brain metastases was influenced by jaw field settings. Finally, we concluded that MC-VMAT dose calculations are useful for 3D dose verification of single-isocenter VMAT plans for multiple brain metastases.

Dosimetric impact of rotational errors on the quality of VMAT-SRS for multiple brain metastases: Comparison between single- and two-isocenter treatment planning techniques

Purpose

In the absence of a 6D couch and/or assuming considerable intrafractional patient motion, rotational errors could affect target coverage and OAR‐sparing especially in multiple metastases VMAT‐SRS cranial cases, which often involve the concurrent irradiation of off‐axis targets. This work aims to study the dosimetric impact of rotational errors in such applications, under a comparative perspective between the single‐ and two‐isocenter treatment techniques.

Methods

Ten patients (36 metastases) were included in this study. Challenging cases were only considered, with several targets lying in close proximity to OARs. Two multiarc VMAT plans per patient were prepared, involving one and two isocenters, serving as the reference plans. Different degrees of angular offsets at various orientations were introduced, simulating rotational errors. Resulting dose distributions were evaluated and compared using commonly employed dose‐volume and plan quality indices.

Results

For single‐isocenter plans and 1^⁰ rotations, plan quality indices, such as coverage, conformity index and D_95%, deteriorated significantly (>5%) for distant targets from the isocenter (at> 4–6 cm). Contrarily, for two‐isocenter plans, target distances to nearest isocenter were always shorter (≤4 cm), and, consequently, 1^⁰ errors were well‐tolerated. In the most extreme case considered (2^⁰ around all axes) conformity index deteriorated by on‐average 7.2%/cm of distance to isocenter, if one isocenter is used, and 2.6%/cm, for plans involving two isocenters. The effect is, however, strongly associated with target volume. Regarding OARs, for single‐isocenter plans, significant increase (up to 63%) in D_max and D_0.02cc values was observed for any angle of rotation. Plans that could be considered clinically unacceptable were obtained even for the smallest angle considered, although rarer for the two‐isocenter planning approach.

Conclusion

Limiting the lesion‐to‐isocenter distance to ≤4 cm by introducing additional isocenter(s) appears to partly mitigate severe target underdosage, especially for smaller target sizes. If OAR‐sparing is also a concern, more stringent rotational error tolerances apply.

Feasibility of 5-mm vs 2.5-mm width multileaf collimator in noncoplanar volumetric modulated arc stereotactic radiotherapy for multiple brain metastases

The aim of this study was to examine the feasibility of noncoplanar volumetric modulated arc stereotactic radiotherapy (VMAT-SRT) using a 5-mm multileaf collimator (MLC) for multiple brain metastases. We identified 34 multiple-target cases (3 to 19 targets in each case) with a total of 257 of targets and constructed noncoplanar VMAT-SRT plans using 5-mm and 2.5-mm MLCs with 4-arc. The prescribed dose was 36 Gy/6 fr. Plans were evaluated using the Paddick conformity indices (PCI), Paddick gradient index (PGI), and normal brain dose (NBD, equal to the mean brain dose minus gross tumor volume). There were no significant differences in PCI (median [range]: 5 mm, 0.88 [0.78 to 0.94]; 2.5 mm, 0.89 [0.78 to 0.94]; p= 0.691), PGI (median [range]: 5 mm, 3.96 [2.21 to 6.63]; 2.5 mm, 3.96 [2.24 to 6.45]; p= 0.358), or NBD (median [range]: 5 mm, 7.5 Gy [2.5 to 12.4]; 2.5 mm, 7.5 Gy [2.5 to 12.5]; p= 0.675). The performance of the 5-mm MLC was not inferior to the 2.5-mm MLC in applications of noncoplanar VMAT-SRT for multiple brain metastases with regards to dose conformity, gradient, and NBD. This study provides the necessary background for generalizing noncoplanar VMAT-SRT approaches in treating multiple brain lesions.

End-to-end testing of automatic plan optimization using RayStation scripting for hypofractionated multimetastatic brain stereotactic radiosurgery

For external beam stereotactic radiosurgery of multiple brain metastatic lesions, it is difficult to select optimal treatment isocenters because the orientation and volume of each planning target volume (PTV) and its proximity to critical structures are unique for each patient. The RayStation treatment planning system offers Python-based scripting to optimize the placement of the treatment isocenter by comparing scenario-based plans. This can improve the plan quality by reducing the dose to the normal brain and increasing planning efficiency. The purpose of the current study was to compare the isocenter-optimized plans generated by RayStation with clinical plans created by the Pinnacle treatment planning system and to validate the RayStation treatment planning and delivery with end-to-end testing. Ten patient plans were automatically regenerated using the script in RayStation. For each patient, 4 plans with 4 different types of isocenters were generated: (1) 2 separate isocenters at the PTV centroids, (2) a single isocenter at the mid-point of 2 centroids, (3) a single isocenter at PTV1, and (4) a single isocenter at PTV2. The best plans were compared with paired Pinnacle plans using plan quality parameters, including normal brain volume excluding PTVs receiving 4 Gy (V_4Gy), normal brain volume excluding PTVs receiving 12 Gy (V_12Gy), maximum dose to the brainstem, homogeneity index, conformity indices, gradient index of each PTV, and monitor units per fraction. All plans were verified with a cylindrical quality assurance phantom, and end-to-end testing was performed with an anthropomorphic head phantom with a radiochromic film. The script was executed within 5-6 minutes to generate 4 scenario-based automatic plans. The homogeneity index and conformity indices showed small but statistically significant improvement with the RayStation plans. The gradient index (3.9 ± 0.9 for Pinnacle and 3.5 ± 0.6 for RayStation, p = 0.04) was also more favorable in the RayStation plans. V_12Gy was significantly reduced by 13% and V_4Gy was reduced by 5%. The total monitor units per fraction was significantly reduced by 20% for the RayStation plans. Plan optimization time using RayStation was reduced by 64%. The measured doses at each PTV centroid agreed within 3%, and all RayStation plans passed quality assurance verification tests. Scenario-based automatic plan generation using Python scripting helps identify an optimal treatment isocenter to reduce the dose to the normal brain and improve planning efficiency. RayStation plans provided better plan quality, especially lower doses to the normal brain, than Pinnacle plans. Thus, RayStation is a suitable planning modality for hypofractionated stereotactic radiosurgery for multiple brain metastases.

Delivered Dose Distribution Visualized Directly With Onboard kV-CBCT: Proof of Principle

PURPOSE

To demonstrate proof of principle of visualizing delivered 3-dimensional (3D) dose distribution using kilovoltage (kv) cone beam computed tomography (CBCT) mounted onboard a linear accelerator. We apply this technique as a unique end-to-end verification of multifocal radiosurgery where the coincidence of radiation and imaging systems is quantified comprehensively at all targets.

METHODS AND MATERIALS

Dosimeters (9.5-cm diameter N-isopropylacrylamide) were prepared according to standard procedures at one facility and shipped to a second (remote) facility for irradiation. A 4-arc volumetric modulated arc therapy (VMAT) multifocal radiosurgery plan was prepared to deliver 20 Gy with 6-MV photons to 6 targets (1-cm diameter). A dosimeter was aligned via CBCT and irradiated, followed by 3 CBCT scans acquired immediately, with total time between pre-CBCT and final CBCT <30 minutes. Image processing included background subtraction and low-pass filters. A dose-volume structure was created per target with the same volume as the planned prescription dose volume, and their spatial agreement was quantified using volume centroid and the Jaccard index. For comparison, 5 diagnostic computed tomography (CT) scans were also acquired after >24 hours with the same spatial analysis applied; comparison with planned doses after absolute dose calibration also was conducted.

RESULTS

Regions of high dose were clearly visualized in the average CBCT with a contrast-to-noise ratio of 1.7 ± 0.7, which increased to 5.8 ± 0.5 after image processing, and 11.9 ± 3.7 for average diagnostic CT. Centroids of prescription isodose volumes agreed with the root mean square difference of 1.1 mm (range, 0.8-1.7 mm) for CBCT and 0.7 mm (0.4-0.8 mm) for diagnostic CT. The dose was proportional to density above 10 to 12 Gy with a 3D gamma pass rate of 94.0% and 99.5% using 5% for 1-mm and 3% for 2-mm criteria, respectively (threshold = 15 Gy, using global dose criteria).

CONCLUSIONS

This work demonstrates for the first time the potential to visualize in 3D delivered dose using onboard kV-CBCT (0.5 × 0.5 × 1 mm³ voxel size) immediately after irradiation with a sufficient contrast-to-noise ratio to measure radiation and imaging system coincidence to within 2 mm.

Dosimetric and localization accuracy of Elekta high definition dynamic radiosurgery

BACKGROUND AND PURPOSE

With the increasingly prominent role of stereotactic radiosurgery in radiation therapy, there is a clinical need for robust, efficient, and accurate solutions for targeting multiple sites with one patient setup. The end-to-end accuracy of high definition dynamic radiosurgery with Elekta treatment planning and delivery systems was investigated in this study.

MATERIALS AND METHODS

A patient-derived CT scan was used to create a radiosurgery plan to seven targets in the brain. Monaco was used for treatment planning using 5 VMAT non-coplanar arcs. Prior to delivery, 3D-printed phantoms from RTsafe were ordered including a gel phantom for 3D dosimetry, phantom with 2D film insert, and an ion chamber phantom for point dose measurement. Delivery was performed using the Elekta VersaHD, XVI cone-beam CT, and HexaPOD six degree of freedom tabletop.

RESULTS

Absolute dose accuracy was verified within 2%. 3D global gamma analysis in the film measurement revealed 3%/2 mm passing rates >95%. Gel dosimetry 3D global gamma analysis (3%/2 mm) were above 90% for all targets with the exception of one. Results were indicative of typical end-to-end accuracies (<1 mm spatial uncertainty, 2% dose accuracy) within 4 cm of isocenter. Beyond 4 cm, 2 mm accuracy was found.

CONCLUSIONS

High definition dynamic radiosurgery expands clinically acceptable stereotactic accuracy to a sphere around isocenter allowing for radiosurgery of several targets with one setup with a high degree of dosimetric precision. Gel dosimetry proved to be an essential tool for the validation of the 3D dose distributions in this technique.

Incorporating the rotational setup uncertainty into the planning target volume margin expansion for the single isocenter for multiple targets technique

PURPOSE

The single isocenter for multiple targets (SIMT) technique has become a popular treatment approach for multiple brain metastases. However, the rotational error that is introduced is usually not considered in planning target volume (PTV) expansion. We have developed a statistical model that takes into account both translational and rotational uncertainties. In this study, we incorporated the rotational error into PTV margin expansion for the clinical use of the SIMT technique.

METHODS AND MATERIALS

In the statistical model, both translational and rotational errors are assumed to follow the 3-dimensional, independent, normal distribution with a zero mean and standard deviations of σ_S and σ_R, where σ_R = 0.01424σ_D (rotational uncertainty in degree)×d_I ⇔ T (distance in mm from isocenter to target). Based on this model, we derived in this study the additional PTV margin, ∆M, that is required to maintain the same coverage probability when the rotational uncertainty is present as a function of M_S (initial PTV margin), σ_D, and d_I ⇔ T. The maximum allowable d_{I ⇔ T, C} and σ_{D, C} were also calculated as a function of user-specified ∆M_c/M_S, the fraction of M_S below which the extra PTV margins can be ignored.

RESULTS

Combined PTV margin, M_E, and additional PTV margin, ∆M, were plotted for commonly encountered clinical parameters including d_I ⇔ T, M_S, or σ_D. Unlike other reported margin recipes, ∆M is not a linear function of any of these 3 parameters. In addition, the rate of increase for ∆M is quite slow for small d_I ⇔ T and becomes more significant for larger d_I ⇔ T. Cutoff values d_{I ⇔ T, C} and σ_{D, C} were also plotted for various ∆M_c/M_S, which can be used to determine if an additional PTV margin is needed for the SIMT technique.

CONCLUSIONS

The presented data provide a convenient way for clinics to determine the appropriate PTV margin for the SIMT technique.

Restricted single isocenter for multiple targets dynamic conformal arc (RSIMT DCA) technique for brain stereotactic radiosurgery (SRS) planning

PURPOSE/OBJECTIVES

In stereotactic radiosurgery (SRS), the multiple isocenters for multiple targets dynamic conformal arc (MIMT DCA) technique is traditionally used to treat multiple brain metastases, with one isocenter for each target. The single isocenter for multiple targets (SIMT) technique has recently been adopted to reduce the treatment time at the cost of plan quality. The objective of this study was to develop a restricted single isocenter for multiple targets DCA (RSIMT DCA) technique that can significantly reduce the treatment time but still maintain similar plan quality as the MIMT DCA technique.

MATERIALS AND METHODS

Treating multiple brain metastases with a single isocenter poses a challenge to SRS planning using DCA beams that are intrinsically 3D and do not modulate the beam intensity to spare the normal tissue between targets. To address this obstacle, we have developed a RSIMT DCA technique and used it to treat SRS patients with multiple brain metastases since February 2015. This planning approach is similar to the SIMT technique except that the number of targets for each isocenter is restricted and the distance between the isocenter and target is limited. In this technique, the targets are first split into batches so that all targets in a batch are within a chosen distance (e.g., 7 cm) of each other. All targets in a batch are combined into one target and the geometric center of the combined target is the isocenter for the group of DCA beams associated with that batch. Each DCA group typically consists of 3-4 DCA beams to irradiate 1-3 targets. For each DCA beam, the collimator angle is adjusted to minimize the exposure of normal tissue between targets. The dose of each treatment group is normalized so that the maximal point dose to the combined target is 125% of the prescription dose, which is equivalent to normalize the prescription dose to 80% isodose line. If the maximal point dose of a target is <123%, an additional beam is used to boost the maximal point dose of that target to 125%. To evaluate the plan quality, we randomly selected 10 cases planned with the RSIMT DCA technique, and re-planned them using the MIMT DCA technique. There were in total 38 PTVs, and 22 isocenters were used to treat all of these targets. The prescription for each target was 20 Gy with a maximal point dose of 25 Gy. Plan quality indexes were calculated and compared. Paired sample t-test was performed to determine if the mean normalized difference, (RSIMT-MIMT)/MIMT of each plan index was statistically significantly (p-value < 5%) larger than 0.

RESULTS

Satisfactory PTV coverage (V20Gy>95% and V19Gy=100%) was achieved for all plans using either technique. Most PTVs have a maximal point dose between 24.9 and 25.1 Gy, with 2 PTVs between 24.5 and 24.9 Gy. Overall, the plan quality was slightly better for the MIMT DCA technique and the normalized difference was statistically significantly larger than 0 for all investigated dose quality indexes. The normalized difference of body mean dose and conformity index (CI) between the RSIMT and MIMT techniques was respectively 4.2% (p=0.002) and 9.4% (p=0.001), indicating similar plan quality globally and in the high dose area. The difference was more pronounced for the mid-to-low dose spillage with the ratios of V12Gy and V10Gy/VPTV being 13.9% (p=3.8×10^-6) and 14.9% (p=1.3×10^-5), respectively. The treatment time was reduced by 30%-50% with the RSIMT DCA technique.

CONCLUSION

The RSIMT DCA technique can produce satisfactory SRS plans for treating multiple targets and can significantly reduce the treatment time.

Single fraction stereotactic radiosurgery for multiple brain metastases

Introduction

Due to the neurocognitive side effects of whole brain radiation therapy (WBRT), stereotactic radiosurgery (SRS) is being used with increasing frequency. The use of SRS is expanding for patients with multiple (>4) brain metastases (BM). This study summarizes our institutional experience with single-fraction, linear-accelerator-based SRS for multiple BM.

Methods and materials

All patients who were treated between January 1, 2013, and September 30, 2015, with single-fraction SRS for ≥4 BM were included in this institutional review board–approved, retrospective, single-institution study. Patients were treated with linear accelerator–based image guided SRS.

Results

A total of 59 patients with ≥4 BM were treated with single-fraction SRS. The median follow-up was 15.2 months, and the median overall survival for the entire cohort was 5.8 months. The median number of treated lesions per patient was 5 (range, 4-23). Per patient, the median planning target volume (PTV) was 4.8 cc (range, 0.7-28.8 cc). The prescribed dose across all 380 BM for the 59 patients ranged from 7 to 20 Gy. The median of the mean dose to the total PTV was 19.5 Gy. Although the number of treated lesions (4-5 vs ≥6) did not influence survival, better survival was noted for a total PTV <10 cc versus ≥10 cc (7.1 vs 4.2 months, respectively; P = .0001). A mean dose of ≥19 Gy to the entire PTV was also associated with increased survival (6.6 vs 5.0 months, respectively; P = .0172). Patients receiving a dose of >12 Gy to ≥10 cc of normal brain had worse survival (5.1 vs 8.6 months, respectively; P = .0028).

Conclusion

In single-fraction SRS for patients with multiple BM, smaller total tumor volume, higher total dose, and lower volume of normal brain receiving >12 Gy were associated with increased survival. These data suggest that using SRS for the treatment of multiple BM is efficacious and that outcomes may be affected more by total tumor volume than by the number of lesions.

Technical Note: Using k-means clustering to determine the number and position of isocenters in MLC-based multiple target intracranial radiosurgery

Purpose

To present the k‐means clustering algorithm as a tool to address treatment planning considerations characteristic of stereotactic radiosurgery using a single isocenter for multiple targets.

Methods

For 30 patients treated with stereotactic radiosurgery for multiple brain metastases, the geometric centroids and radii of each met were determined from the treatment planning system. In‐house software used this as well as weighted and unweighted versions of the k‐means clustering algorithm to group the targets to be treated with a single isocenter, and to position each isocenter. The algorithm results were evaluated using within‐cluster sum of squares as well as a minimum target coverage metric that considered the effect of target size. Both versions of the algorithm were applied to an example patient to demonstrate the prospective determination of the appropriate number and location of isocenters.

Results

Both weighted and unweighted versions of the k‐means algorithm were applied successfully to determine the number and position of isocenters. Comparing the two, both the within‐cluster sum of squares metric and the minimum target coverage metric resulting from the unweighted version were less than those from the weighted version. The average magnitudes of the differences were small (−0.2 cm² and 0.1% for the within cluster sum of squares and minimum target coverage, respectively) but statistically significant (Wilcoxon signed‐rank test, P < 0.01).

Conclusions

The differences between the versions of the k‐means clustering algorithm represented an advantage of the unweighted version for the within‐cluster sum of squares metric, and an advantage of the weighted version for the minimum target coverage metric. While additional treatment planning considerations have a large influence on the final treatment plan quality, both versions of the k‐means algorithm provide automatic, consistent, quantitative, and objective solutions to the tasks associated with SRS treatment planning using a single isocenter for multiple targets.