RapidPlan hippocampal sparing whole brain model version 2—how far can we reduce the dose?

IntroductionBrain metastasis, the most common form of brain cancer, is approximately 10-fold more common than primary brain tumors and affects nearly 170,000 Americans annually.Epidemiology of brain metastases and leptomeningeal disease.,Khuntia D Brown P Li J Mehta MP Whole-brain radiotherapy in the management of brain metastasis. Roughly 20% of all patients with solid tumors will develop brain metastases over the course of their disease, with primary lung cancers being the most frequent source.Epidemiology of brain metastases and leptomeningeal disease.,Barnholtz-Sloan JS Sloan AE Davis FG Vigneau FD Lai P Sawaya RE Incidence proportions of brain metastases in patients diagnosed (1973 to 2001) in the Metropolitan Detroit Cancer Surveillance System.Schouten LJ Rutten J Huveneers HAM Twijnstra A Incidence of brain metastases in a cohort of patients with carcinoma of the breast, colon, kidney, and lung and melanoma.Brown PD Ahluwalia MS Khan OH Asher AL Wefel JS Gondi V Whole-brain radiotherapy for brain metastases: Evolution or revolution?. For the past 70 years, whole-brain radiotherapy (WBRT) has been the standard palliative treatment for patients with brain metastases due to its effectiveness, availability, and ease of administration.Brown PD Ahluwalia MS Khan OH Asher AL Wefel JS Gondi V Whole-brain radiotherapy for brain metastases: Evolution or revolution?. However, with further improvements in systemic therapy and overall survival, recent efforts have been aimed at reducing cognitive toxicities associated with WBRT.Brown PD Ahluwalia MS Khan OH Asher AL Wefel JS Gondi V Whole-brain radiotherapy for brain metastases: Evolution or revolution?.The hippocampal dentate gyrus, a complex temporal lobe structure that houses an exquisitely radiosensitive compartment of neural stem cells, plays an important role in learning and memory.Hippocampus in health and disease: An overview. Significant WBRT-induced toxicities can be attributed to damage in this region, leading to impairment in cognitive function and quality of life.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.Monje ML Mizumatsu S Fike JR Palmer TD Irradiation induces neural precursor-cell dysfunction.Gondi V Hermann BP Mehta MP Tomé WA. Hippocampal dosimetry predicts neurocognitive function impairment after fractionated stereotactic radiotherapy for benign or low-grade adult brain tumors. A prospective multi-institutional randomized phase III trial, NRG-CC001 (ClinicalTrials.gov identifier: NCT02360215), found therapeutic radiation doses to the whole brain while limiting hippocampal radiation doses (D100% ≤ 9 Gy and Dmax ≤ 16 Gy) along with twice-daily memantine, significantly reduced cognitive toxicity at 4 and 6 months.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.Despite strong clinical evidence that hippocampal avoidance during WBRT enhances clinical outcomes, cognitive function as a result of hippocampal dose gradient remains unknown. However, there has been a steep dose-response relationship between higher maximum dose to the bilateral hippocampus and deterioration in short-term memory.Redmond KJ Grimm J Robinson C et al.steep dose-response relationship between maximum hippocampal dose and memory deficits following hippocampal avoidance whole brain radiation therapy (HA-WBRT) for brain metastases: A secondary analysis of NRG/RTOG 0933. In clinical trials, the lowest point dose goal for hippocampal sparing is generally ≤9 Gy driven by feasibility of sparing.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.,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.,Martinage G Hong AM Fay M et al.Quality assurance analysis of hippocampal avoidance in a melanoma whole brain radiotherapy randomized trial shows good compliance. In clinical practice, treatment planning is labor intensive and has high inter-planner variability for parameters, such as organ-at-risk (OAR) sparing, homogeneity index, and conformality index.Nelms BE Robinson G Markham J et al.Variation in external beam treatment plan quality: An inter-institutional study of planners and planning systems.RapidPlanTM is a commercial implementation of Knowledge Based Planning (KBP) (Varian Medical Systems, Palo Alto, CA). KBP software utilizes geometric features (i.e., OAR distance to the planning target volume (PTV)) and OAR overlap volume histogram to search for the best prior case(s) from a database or to construct a dose prediction model (i.e., machine learning, statistical model). In 2016, the Hippocampal Sparing Whole Brain Model Version 1.0 (HSWBv1) was made publicly available.Magliari V Magliari A Foster R hippocampal sparing whole brain: Rapid Plan ModelFollowing the NRG-CC001 Protocol. HSWBv1 utilized a complex recursive model creation process and leveraged an intricate scorecard to assess plan quality. This scorecard, based on NRG-CC001, contained additional metrics to further reduce dose to the hippocampus and included additional OAR objectives (Table 1).Rusu I Roeske J Solanki A Kang H Fully automated planning and delivery of hippocampal-sparing whole brain irradiation.,Momin S Fu Y Lei Y et al.Knowledge-based radiation treatment planning: A data-driven method survey.

Table 1Summary of published studies on HA-WBRT

In this study, we developed a new multi-institutional Hippocampal Sparing Whole Brain RapidPlanTM Model Version 2.0 (HSWBv2) to further improve hippocampal sparing, homogeneity, conformity, and other OAR constraints over HSWBv1. The steep dose relationship between hippocampal dose and cognitive decline suggests tighter constraints may be beneficial.Redmond KJ Grimm J Robinson C et al.steep dose-response relationship between maximum hippocampal dose and memory deficits following hippocampal avoidance whole brain radiation therapy (HA-WBRT) for brain metastases: A secondary analysis of NRG/RTOG 0933.MethodsTarget and OAR contouring and treatment planning guidelines

Fifty anonymized whole brain DICOM computed tomography (CT) and structure sets were obtained from Northwestern Medicine Cancer Center Warrenville and Washington University in St. Louis for patients treated per NRG-CC001 protocol. A planning CT protocol that encompassed the entire head was obtained for all patient cases, with maximum axial slice thickness of 2.5 mm or smaller. T1 contrast enhanced magnetic resonance imaging with similar axial slice thickness were registered to CT images. Bilateral hippocampal volumes, OARs, and optimization structures were then generated on the planning CT.

All training and validation cases used head-first supine orientation with the head in a neutral position immobilized with an Aquaplast mask. A 4-arc volumetric modulated arc therapy technique was utilized with 4 full coplanar arcs on HalcyonTM (Varian Medical Systems, Palo Alto, CA). Arcs had alternating clockwise and counterclockwise gantry rotations with collimator positions set at 315, 0, 45, and 90 degrees. The coplanar arcs had 359.8 degrees of arc rotation for each field. Arcs were positioned at a single isocenter located in the center of the target. Each arc utilized 6X Flattening Filter-Free energy at a dose rate of 800 MU per minute. HSWBv2 was also tested with various arc geometries (both coplanar and non-coplanar) on TrueBeamTM/C-series (Millennium120 MLC). Full validation with different number of arcs, geometries, energies, and dose calculation methods, including a quantification of the relative dosimetric performance of each method, can be seen in Annex A of the clinical description document included with HSWBv2.

New scorecard tool usedDosimetric scorecards were used to guide the RapidPlanTM model creation process. The PlanScoreCard Eclipse Scripting Application Programming Interface tool, available free on the Varian Innovation Center GitHub, was used to create scoring metrics, automatically generate additional optimization and evaluation structures, and score candidate plans throughout the process. These dosimetric scorecards use established scoring methodology of multiple piecewise linear score functions which measure specific plan quality metrics.Nelms BE Robinson G Markham J et al.Variation in external beam treatment plan quality: An inter-institutional study of planners and planning systems.Dosimetric scorecards as objective measure of plan qualityThe original plan quality metrics scorecard Version 1.0 (100 points in total) was developed in 2016 for HSWBv1 and was based on dose constraints from NRG-CC001. The Version 1.0 scorecard awarded points for OAR doses lower than those cited in NRG-CC001 and added additional metrics for OARs and plan quality indexes not listed in the protocol. The new Version 2.0 dosimetric scorecard used in this study was further revised from the 2016 scorecard to address perceived plan quality issues with HSWBv1 plans (142 points in total). Additional points were awarded to prior metrics and new metrics were added, in-part, via four derived structures (Fig. 1). These new structures and their associated metrics quantify reducing dose to the face, improving conformity, and reducing heterogeneity in the target.Fig 1

Fig. 1Derived structures. PTV_Brain (Brain – [Hippocamus +05]) defined in NRG-CC001, used for scorecard evaluation but not used in optimization. oPTV (Brain – [Hippocamus +07]) used as primary target during optimization. Ring ([Brain +20) – [Brain +03]) used to minimize radiation dose outside the brain. Face ([Eyes +100] – [Brain +20]) used to minimize unnecessary lower doses. BS PTV (Brainstem – [Hippocampus +08]) used as secondary target to prevent <Rx dose between Hippocampus structures. (Units: mm). PTV, planning target volume.

Further hippocampal sparing was also rewarded in the Version 2.0 scorecard. For the maximum dose goal (D0.03CC) points are earned from 8 to 17 Gy, minimum dose (D100%) 5 to 10Gy, and Dmean 6 to 12 Gy.

This Version 2.0 scorecard, once finalized, guided the model creation process by providing a singular objective measure of plan quality.Nelms BE Robinson G Markham J et al.Variation in external beam treatment plan quality: An inter-institutional study of planners and planning systems. Figure 2 shows an example Version 2.0 scorecard scoring a validation case.Fig 2

Fig. 2Example Version 2.0 scorecard; validation case score 133.56/142.

HSWBv2 model creation and trainingAll cases were prescribed 30 Gy in 10 fractions and satisfied the necessary contouring criteria for NRG-CC001 with additional OARs and structures. The original HSWBv1 model and Version 1.0 scorecard were used as a basis for this work. Treatment plans created from HSWBv1 with Varian HalcyonTM System were manually reoptimized to address feedback and improvements quantified using the new version 2.0 scorecard. Those manually improved, reoptimized plans became the training set for the initial HSWBv2 model. More cases were added, and a recursive model creation process was employed to ensure the final HSWBv2 training set consisted, exclusively, of plans generated from the initial HSWBv2 model. Evaluating plan scores at each step in the process informed multiple iterations of re-tuning the optimization objective set. As additional cases were added, geometric outliers became apparent. These outlier cases were omitted from the final HSWBv2 training set, resulting in 42 total cases. See Fig. 3 for process workflow details.Fig 3

Fig. 3Flowchart of HSWBv2 Model Creation, Training, and Validation......

Primary dosimetric planning goals and scorecard comparison

Treatment plan comparisons were made between HSWBv1 and HSWBv2 for the hippocampus and PTV_3000 using student t-tests. Primary dose metrics for the hippocampus include dose at 0.03CC's (D0.03CC), mean dose (Dmean), and dose at 100% volume (D100%). PTV_3000 primary metrics include dose at 98% volume (D98%), dose at 2% volume (D2%), and volume receiving 105% of prescription dose (V105%). These are included as high scoring individual metrics on the version 2.0 scorecard. The total scorecard points were also compared using student t-tests. Correlation analysis was performed to determine volume associations with hippocampal Dmean, D100%, and D0.03CC.

DiscussionKBP studies have been performed for multiple disease sites, such as prostate, lung, and head and neck.Momin S Fu Y Lei Y et al.Knowledge-based radiation treatment planning: A data-driven method survey. KBP methods have several advantages over manually optimized plans in radiotherapy. KBP can streamline the treatment planning process by standardizing plan quality, reducing human input, and decreasing time to perform treatment planning.Momin S Fu Y Lei Y et al.Knowledge-based radiation treatment planning: A data-driven method survey. For example, creating optimization objectives for a glioblastoma disease site took 2 minutes, accompanied by 5 additional minutes for optimization and dose calculation.Momin S Fu Y Lei Y et al.Knowledge-based radiation treatment planning: A data-driven method survey. Contrarily, it took 4 hours for a planner without KBP assistance to create a plan of similar quality.Chatterjee A Serban M Abdulkarim B et al.Performance of knowledge-based radiation therapy planning for the glioblastoma disease site. In a case of malignant pleural mesothelioma, planning time taken when utilizing KBP was 20 minutes vs 4 hours of manual optimization time.Dumane VA Tam J Lo Y-C Rosenzweig KE RapidPlan for knowledge-based planning of malignant pleural mesothelioma.We developed a new KBP hippocampal avoidance whole brain radiation therapy treatment model (HSWBv2) with the goal of further reducing dose to the hippocampus, increasing dose homogeneity, and improving other dosimetric parameters. The version 2.0 scorecard added more robust scoring, resulting in a more complete evaluation of dosimetric plan quality. The revised scorecard integrated feedback from 5 years of clinical use of HSWBv1. The new scorecard tool allows for metrics which automatically generate complex, dynamic structure expansion/contraction and Boolean operations. This automation expedited the creation of the additional derived structures (Fig. 1). These structures were evaluated on the scorecard and used for both optimization and dose volume histogram prediction. Using these derived structures, HSWBv2 enforced enhanced target conformation, homogeneity, and OAR sparing (Fig. 4).Fig 4

Fig. 44 Arc HalcyonTM treatment plans made with single click, RapidPlanTM. HSWBv1 (left) and HSWBv2 (right).

Some previous studies only segmented a limited number of cranial structures which included hippocampus and 2 to 3 other OARs (i.e., optic nerves and chiasm, lens).Giaj Levra N Sicignano G Fiorentino A et al.Whole brain radiotherapy with hippocampal avoidance and simultaneous integrated boost for brain metastases: a dosimetric volumetric-modulated arc therapy study.Zieminski S Khandekar M Wang Y Assessment of multi-criteria optimization (MCO) for volumetric modulated arc therapy (VMAT) in hippocampal avoidance whole brain radiation therapy (HA-WBRT).Kazda T Vrzal M Prochazka T et al.Left hippocampus sparing whole brain radiotherapy (WBRT): A planning study.Yuen AHL Wu PM Li AKL Mak PCY Volumetric modulated arc therapy (VMAT) for hippocampal-avoidance whole brain radiation therapy: planning comparison with Dual-arc and Split-arc partial-field techniques.Nevelsky A Ieumwananonthachai N Kaidar-Person O et al.Hippocampal-sparing whole-brain radiotherapy using Elekta equipment. In comparison to published data, HSWBv2 achieved significantly lower hippocampus Dmean, D100%, and D0.03CC without compromising target coverage while also reducing hot spots substantially. Published series with a PTV prescription dose of ≥30 Gy generally achieved hippocampus D100% of approximately 8 Gy, in contrast to 5.75 Gy in our study (Table 1). Though Levra et al. report a D100% of 6.7 Gy, their study prescribed a PTV whole brain dose of 20Gy while including simultaneous integrated boost to targets greater than 30 Gy.Giaj Levra N Sicignano G Fiorentino A et al.Whole brain radiotherapy with hippocampal avoidance and simultaneous integrated boost for brain metastases: a dosimetric volumetric-modulated arc therapy study.HSWBv2 was tested on two Varian delivery platforms: HalcyonTM and TrueBeamTM of which TrueBeamTM is the most popular Varian delivery platform. Halcyon utilizes advanced dual-layer MLC which enables high modulation due to faster leaf speed and lower transmission. HSWBv2 created high quality plans on both platforms without user interaction or plan modification (“single click”), allowing for the model to be readily usable in clinics. The advantage of single click optimization is that the planner is not tied to the process. The total time taken for optimization is dependent on the speed of treatment planning system hardware and software configuration (i.e., convergence mode). For reference, HSWBv1 and HSWBv2 total optimization time on our system took 79 and 76 minutes, respectively. This HSWBv2 RapidPlanTM model is free to download to the community for all who can use such models.Varian
Whole brain with hippocampal sparing (NRG-CC001) [RapidPlan].

HSWBv2 utilized statistical and machine learning methods available in RapidPlanTM software for selecting the training set cases. The recursive method generated a KBP model that produced narrow and accurate dose volume histogram prediction bands. These well-fit estimation bands allowed for more aggressive and improved OAR sparing.

Our HSWBv2 model contains more metrics/parameters/constraints than prior hippocampal avoidance whole brain radiation therapy treatment models. Preliminary model validation with five independent cases resulted in similar dosimetric quality and plan scores compared to the testing completed on the training set cases. During development, HSWBv2 included up to 50 cases in its training set. Eight outliers were omitted due to either irregular patient head position or questionable geometric shape and/or volume of hippocampus contours. Although hippocampus volume are approximately 3.00 cc in normal adults,Hippocampal volume measurements using magnetic resonance imaging in normal young adults.Embong MF Yaacob R Abdullah MS Abdul Karim AH Ghazali AK Jalaluddin WMS MR volumetry of hippocampus in normal adult malay of age 50 years old and above.Nobis L Manohar SG Smith SM et al.Hippocampal volume across age: Nomograms derived from over 19,700 people in UK Biobank. HSWBv2 training set cases averaged 4.46 cc and cases with hippocampus volume >7.50 cc were excluded.The radiosensitive nature of the hippocampus further emphasizes the need to minimize hippocampus dose while maintaining dosimetric plan quality. In NRG-CC001, maximum doses of 14 Gy and 16 Gy were associated with a 10% and 25% risk of short-term memory deterioration at the 6-month mark.Redmond KJ Grimm J Robinson C et al.steep dose-response relationship between maximum hippocampal dose and memory deficits following hippocampal avoidance whole brain radiation therapy (HA-WBRT) for brain metastases: A secondary analysis of NRG/RTOG 0933. The hippocampus D0.03CC of 9.86 Gy and D100% of 5.75 Gy achieved in our study are significantly lower than prior studies or current existing clinical guidelines. In conclusion, we demonstrated plan quality improvements by using KBP and dosimetric scorecard to guide the process. Further clinical studies are needed to determine the impact of additional hippocampal dose reduction and its correlation to neurocognitive function.Conflict of Interest

Dr Sushil Beriwal has a leadership role as the Vice President of Medical Affairs at Varian Medical Systems, reports grant as an Elsevier consultant, and reports participation in advisory board at Xoft DSMB. Mr Ryan Clark reports employment at Varian Medical Systems. Mr Anthony Magliari reports employment at Varian Medical Systems. Dr Francisco Reynoso reports honoraria for lectures from Varian Medical Systems. Dr Heather Curry reports employment at Varian Medical Systems. Dr Patrick Kupelian reports employment at Varian Medical Systems. Dr Deepak Khuntia has a leadership role as the Senior Vice President and Chief Medical Officer at Varian Medical Systems. Mr Hefei Liu, Mr Robert Foster, Mr Matthew Schmidt, Dr Vinai Gondi, Dr Christopher Abraham have nothing to disclose.

留言 (0)

沒有登入
gif