Advertisement

Implementation of automatic plan optimization in Italy: Status and perspectives

Published:December 04, 2021DOI:https://doi.org/10.1016/j.ejmp.2021.11.013

      Highlights

      • Online survey of Italian medical physicists working in radiotherapy.
      • Extensive clinical use of automated planning is still limited.
      • Most physicists consider the use of automated techniques to be beneficial.
      • Users show a prevalently positive attitude.

      Abstract

      Purpose

      To investigate and report on the diffusion and clinical use of automated radiotherapy planning systems in Italy and to assess the perspectives of the community of Italian medical physicists involved in radiotherapy on the use of these tools.

      Materials and Methods

      A survey of medical physicists (one per Institute) of 175 radiotherapy centers in Italy was conducted between February 21st and April 1st, 2021. The information collected included the institute’s characteristics, plan activity, availability/use of automatic tools and related issues regarding satisfaction, criticisms, expectations, and perceived professional modifications. Responses were analysed, including the impact of a few variables such as the institute type and experience.

      Results

      125 of the centers (71%) answered the survey, with regional variability (range: 47%–100%); among these, 49% have a TPS with some automatic option. Clinical use of automatic planning is present in 33% of the centers, with 13% applying it in >50% of their plans.
      Among the 125 responding centres the most used systems are Pinnacle (16%), Raystation (9%) and Eclipse (4%). The majority of participants consider the use of automated techniques to be beneficial, while only 1% do not see any advantage; 83% of respondents see the possibility of enriching their professional role as a potential benefit, while 3% see potential threats.

      Conclusions

      Our survey shows that 49% of the responding centres have an automatic planning solution although clinically used in only 33% of the cases. Most physicists consider the use of automated techniques to be beneficial and show a prevalently positive attitude.

      Keywords

      1. Introduction

      In recent years there have been important technological advancements in the field of radiation therapy, leading to more accurate radiotherapy treatments [
      • Teoh M.
      • Clark C.H.
      • Wood K.
      • Whitaker S.
      • Nisbet A.
      Volumetric modulated arc therapy: a review of current literature and clinical use in practice.
      ]. Intensity modulated radiation therapy (IMRT), volumetric modulated arc therapy (VMAT) and image-guided radiotherapy (IGRT), are widespread practices that represent indispensable prerequisites for any personalized therapeutic approach that can lead to better disease control. In this context, numerous technological aspects may contribute, but radiotherapy treatment planning undoubtedly plays a central role. VMAT or IMRT plans are generated using an inverse planning approach where objectives and constraints for the planning target volume (PTV) and for organs at risk (OAR) are set. Planners pilot the optimizer by selecting appropriate cost functions that are iteratively adjusted following a try-and-error approach.
      Treatment plan optimization is a multi-criteria problem, where the skills, experience, and time at the disposal of planners are key parameters that may influence the plan quality and consequently, ultimately affect the clinical outcome. The long planning time required to optimize an IMRT or VMAT plan can limit the access for all patients to “the best possible” treatment plan. In addition to these issues, even if all planners had high experience and no time constraints, this would not ensure that the generated plan would achieve the optimal trade-off between objectives, nor that dose to OARs could not be further reduced.
      To improve the efficiency and quality of plan optimization, in recent years automated planning systems were developed, now available in several commercial platforms [

      Hussein M, Heijmen BJM, Verellen D, Nisbet A. Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations. Br J Radiol 2018;91. doi:10.1259/bjr.20180270.

      ].
      Three main and diverse solutions have been proposed: knowledge-based planning (KBP), protocol-based planning (PBP) and multi-criteria optimisation (MCO) [

      Hussein M, Heijmen BJM, Verellen D, Nisbet A. Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations. Br J Radiol 2018;91. doi:10.1259/bjr.20180270.

      ]. Although each of these approaches has different characteristics, numerous papers published in the literature [

      Hussein M, Heijmen BJM, Verellen D, Nisbet A. Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations. Br J Radiol 2018;91. doi:10.1259/bjr.20180270.

      ] generally report results that are in line with expectations in terms of time savings, increased efficiency, standardization, and high/improved quality of plans. What is unknown is the actual deployment of these systems and how they have been received by the medical physics community or impacted on clinical routine. Beyond economic aspects related to the willingness of administrators to invest in automated planning systems, the human factor plays a central role in facilitating the introduction of new technologies in hospital settings. In this context, medical physicists play a strategic role since, due to their training, skills, and experience, they are the major actors leading the process of radiotherapy treatment planning.
      The introduction of automated planning in clinical practice may, in principle, represent a huge change in the activities of medical physicists involved in radiotherapy planning.
      Together with the availability and adoption of these systems, the attitude of medical physicists with respect to automated planning can play a pivotal role in their proliferation and, further, in the clinical application of such systems.
      The aim of this paper is to assess the availability of automated radiotherapy planning systems in Italy and to evaluate how the community of Italian medical physicists involved in radiotherapy is embracing these tools. In this analysis, we consider automated approaches also those that apply some manual fine-tuning to automatically generated plans.
      For this purpose, a digital survey was administered to one medical physicist per radiotherapy facility of all 175 radiotherapy centers between February 21st and April 1st, 2021.
      First, a brief introduction on the main approaches used in radiotherapy treatment automation, including advantages and major limitations, is reported for the sake of completeness.

      2. Materials and methods

      2.1 Automated planning optimization solutions

      Systems for automatic planning implemented in the main treatment planning systems (TPSs) can be divided into three main families: multi-criteria optimization (MCO), protocol-based optimization, and knowledge-based (KB) optimization [

      Hussein M, Heijmen BJM, Verellen D, Nisbet A. Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations. Br J Radiol 2018;91. doi:10.1259/bjr.20180270.

      ]. The most diffused approaches implemented by TPSs mentioned in this survey are described in the following paragraphs.

      2.1.1 Raystation

      RayStation platform (RaySearch Medical Laboratories AB, Stockholm, Sweden) implements an “a posteriori'' MCO approach in which a Pareto-frontier of plans is first generated for next selection of the user's preferred plan [

      Hussein M, Heijmen BJM, Verellen D, Nisbet A. Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations. Br J Radiol 2018;91. doi:10.1259/bjr.20180270.

      ].
      In the “a posteriori” MCO the Pareto frontier is automatically populated by a discrete set of plans that are Pareto-optimal and deliverable with respect to the user-specified set of trade-off objectives and constraints. They are generated giving different emphasis to the planning objectives: anchor plans, representing the extreme trade-offs, are generated by optimizing each objective individually, a balanced plan is obtained setting equal weights for all the objectives, and, finally, auxiliary plans, obtained by interpolation sandwiching the Pareto frontier, are added to achieve optimal sampling [
      • Craft D.L.
      • Halabi T.F.
      • Shih H.A.
      • Bortfeld T.R.
      Approximating convex Pareto surfaces in multiobjective radiotherapy planning.
      ]. A limited number of N + 1 plans, where N is the number of objectives, is sufficient to obtain a good approximation of the Pareto frontier objectives, with reasonable requirements of computational resources [
      • Craft D.
      • Bortfeld T.
      How many plans are needed in an IMRT multi-objective plan database?.
      ]. As Pareto optimality is a necessary, but not necessarily sufficient condition for clinical optimality, in the “a posteriori“ approach the Pareto frontier is interactively navigated by the decision maker to choose a clinically optimal plan by means of a suitable navigation graphical user interface [
      • Craft D.
      • Halabi T.
      • Shih H.A.
      • Bortfeld T.
      An approach for practical multiobjective IMRT treatment planning.
      ,
      • Monz M.
      • Küfer K.H.
      • Bortfeld T.R.
      • Thieke C.
      Pareto navigation – algorithmic foundation of interactive multi-criteria IMRT planning.
      ]. It allows the operator to change the preferences on the criteria showing, in real time, the adaptation of the dose distribution and DVH calculated by interpolation of the plans on the Pareto frontier. This interactive workflow is the strength and the weakness of the method: on one hand it has the potential to avoid the human iteration loop between physician and planner achieving optimal clinical plans tailored to the specific patient care needs; by the other, the quality of the solution is dependent on the decision maker’s experience and ability, especially when many objectives are involved. Nevertheless, MCO has been proven to be beneficial to less experienced operators, assisting them in improving the plan quality even in the most challenging scenarios [

      Kierkels RGJ, Visser R, Bijl HP, Langendijk JA, van ’t Veld AA, Steenbakkers RJHM, 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. doi: 10.1186/s13014-015-0385-9.

      ]. Raystation MCO has been extensively investigated and validated in the main anatomical sites [
      • Wala J.
      • Craft D.
      • Paly J.
      • Zietman A.
      • Efstathiou J.
      Maximizing dosimetric benefits of IMRT in the treatment of localized prostate cancer through multicriteria optimization planning.
      ,
      • Ghandour S.
      • Matzinger O.
      • Pachoud M.
      Volumetric-modulated arc therapy planning using multicriteria optimization for localized prostate cancer.
      ,
      • Kamran S.C.
      • Mueller B.S.
      • Paetzold P.
      • Dunlap J.
      • Niemierko A.
      • Bortfeld T.
      • et al.
      Multi-criteria optimization achieves superior normal tissue sparing in a planning study of intensity-modulated radiation therapy for RTOG 1308-eligible non-small cell lung cancer patients.
      ,
      • Craft D.L.
      • Hong T.S.
      • Shih H.A.
      • Bortfeld T.R.
      Improved Planning Time and Plan Quality Through Multicriteria Optimization for Intensity-Modulated Radiotherapy.
      ] showing significant improvements in terms of plan quality and reduction of planning time with respect to the standard manual planning.
      Another commercial solution implemented in the Raystation platform is Plan Explorer. Whereas MCO allows the operator to explore trade-offs of multiple clinical goals in one plan, Plan Explorer enables the operator to explore trade-offs between different machines, techniques, energies and beam arrangements for a given set of clinical goals. It automatically generates a large number of plans optimized using a lexicographic algorithm and provides efficient means to filer and browse among plan candidates to find the most desired one. The significant amount of clinician time, required by MCO and Plan Explorer to navigate for the optimal solution, is the main drawback for their use in the clinical routine.
      A further approach, based on artificial intelligence, has been implemented in Raystation® 8B*. The proposed machine learning automated treatment planning method, learns from historical patient and plan data, infers a 3D dose distribution of new patients and uses a dose mimicking optimization to generate an optimized deliverable treatment plan [
      • McIntosh C.
      • Welch M.
      • McNiven A.
      • Jaffray D.A.
      • Purdie T.G.
      Fully automated treatment planning for head and neck radiotherapy using a voxel-based dose prediction and dose mimicking method.
      ]. Although promising this approach is still in its clinical infancy and strongly dependent on both size and plan quality of the training dataset.
      Finally, a novel auto-planning algorithm for VMAT/IMRT called GPS (Genetic Planning System), based on genetic optimization and using Python scripts to integrate the algorithm in the RayStation TPS was recently reported [
      • Fiandra C.
      • Alparone A.
      • Gallio E.
      • Vecchi C.
      • Balestra G.
      • Bartoncini S.
      • et al.
      Automated heuristic optimization of prostate VMAT treatment planning.
      ].
      Genetic Algorithms are well-known optimization algorithms inspired by Darwin’s theory of evolution. While there is a large variation in treatment protocols among centres, it has been demonstrated that it is possible to configure the algorithm in one centre and use it for auto-planning in the others, without any centre-specific fine tuning [
      • Fiandra C.
      • Rossi L.
      • Alparone A.
      • Zara S.
      • Vecchi C.
      • Sardo A.
      • et al.
      Automatic genetic planning for volumetric modulated arc therapy: A large multi-centre validation for prostate cancer.
      ].

      2.1.2 Pinnacle

      Another way to achieve the optimal plan in an automatic way is to automate iterative adjustments of goals and constraints in the optimization function. This approach was implemented in the AutoPlanning module of Pinnacle 3 TPS [
      • Gintz D.
      • Latifi K.
      • Caudell J.
      • Nelms B.
      • Zhang G.
      • Moros E.
      • et al.
      Initial evaluation of automated treatment planning software.
      ] (Philips Radiation Oncology Systems, Fitchburg, WI).
      The starting point is a user-defined template with the target prescriptions and the goals for OAR sparing according to the required clinical protocol. Users can enter the priority used for both objectives and constraints from those that have low significance to those that become hard constraints. The optimization process continues with auto creation of dummy structures (considering the overlap between OAR and PTV), ring and other “help” structures to control target uniformity and the dose falloff in the body of the patient. Management of target uniformity and, therefore, the presence of hot/cold spots is controlled through options which can be fine-tuned by users.
      A 5-loop iterative optimization cycle gradually fine-tunes the plan to achieve the final solution: (in order) i) add target objectives; ii) add OAR objectives; iii) tune OAR objectives; iv) add hot/cold spot objectives; v) fine tune each objective. All objective dose and weight parameters are tuned using proprietary algorithms.
      In the newest Personalized Planning module, implemented in version 16.4.1 of Pinnacle3 Evolution, planners can choose to add a patient’s personalized objectives for OARs based on actual patient anatomy [
      • Xia W.
      • Han F.
      • Chen J.
      • Miao J.
      • Dai J.
      Personalized setting of plan parameters using feasibility dose volume histogram for auto-planning in Pinnacle system Auto-Planning, lung cancer, OAR sparing, planning time, plan quality.
      ,
      • Cilla S.
      • Romano C.
      • Morabito V.E.
      • Macchia G.
      • Buwenge M.
      • Dinapoli N.
      • et al.
      Personalized treatment planning automation in prostate cancer radiation oncology: a comprehensive dosimetric study.
      ]. This is done using the Feasibility module, originally developed in the PlanIQ software (Sun Nuclear Corporation, Melbourne, FL) and now integrated into the Pinnacle Personalized planning. Feasibility is a model-based calculation engine [
      • Ahmed S.
      • Nelms B.
      • Gintz D.
      • Caudell J.
      • Zhang G.
      • Moros E.G.
      • et al.
      A method for a priori estimation of best feasible DVH for organs-at-risk: validation for head and neck VMAT planning.
      ] that uses CT images, dose prescriptions and the geometric relationship between target volumes and OAR, to create dose-volume histogram (DVH) for each OAR to help planners evaluate, before starting the plan optimization, the achievability of patient specific goals. The main strength of the Pinnacle approach lies in its simplicity. Only a few parameters need to be set and fine-tuned to create a new treatment technique. On the other side, the simplicity of its approach can also become a weakness, encouraging hasty implementation that can lead to sub-optimal automatic techniques development.
      Pinnacle Autoplanning and the newest Personalized Planning have been extensively investigated and validated in the main anatomical sites [
      • Hazell I.
      • Bzdusek K.
      • Kumar P.
      • Hansen C.R.
      • Bertelsen A.
      • Eriksen J.G.
      • et al.
      Automatic planning of head and neck treatment plans.
      ,
      • Nawa K.
      • Haga A.
      • Nomoto A.
      • Sarmiento R.A.
      • Shiraishi K.
      • Yamashita H.
      • et al.
      Evaluation of a commercial automatic treatment planning system for prostate cancers.
      ,
      • Gallio E.
      • Giglioli F.R.
      • Girardi A.
      • Guarneri A.
      • Ricardi U.
      • Ropolo R.
      • et al.
      Evaluation of a commercial automatic treatment planning system for liver stereotactic body radiation therapy treatments.
      ,
      • Marrazzo L.
      • Meattini I.
      • Arilli C.
      • Calusi S.
      • Casati M.
      • Talamonti C.
      • et al.
      Auto-planning for VMAT accelerated partial breast irradiation.
      ,
      • Arilli C.
      • Zani M.
      • Marrazzo L.
      • Scoccianti S.
      • Casati M.
      • Compagnucci A.
      • et al.
      Automatic VMAT technique to treat glioblastoma: a two years’ experience.
      ].

      2.1.3 Eclipse

      The automated solution proposed in the Eclipse (Varian Inc.) TPS is RapidPlan which is the first commercial implementation of a KB approach [

      Hussein M, Heijmen BJM, Verellen D, Nisbet A. Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations. Br J Radiol 2018;91. doi:10.1259/bjr.20180270.

      ,
      • Ge Y.
      • Wu Q.J.
      Knowledge-based planning for intensity-modulated radiation therapy: a review of data-driven approaches.
      ].
      The KB optimization is based on a library of existing clinical treatment plans used to estimate the dosimetric features expected in new patients, considering the individual anatomical characteristics of each specific patient.
      The common characteristic is that for each new patient, the achievable dosimetric features are predicted using a training dataset of clinical high-quality plans. The basic assumption is that patients with similar anatomies should have similar achievable dose distributions. For this reason, in contrast to other automatic planning techniques, the performance of any KB approach is highly dependent on the quality of the available plans managed during the modelling phase, which is representative of the clinical experience of the centre.
      Given a certain delivery/planning technique and treatment site, existing clinical treatment plans may be modelled in the form of a DVH-estimation model to individually estimate the most likely dosimetric features expected in new patients. Combining principal-component-analysis (PCA) and regression techniques, the tool predicts the achievable DVH based on the features of similar contours and quality of treatment plan used to build the KB-model. Then, plans may be automatically optimized by building a template based on the KB individually optimized constraints.
      Once a KB-model has been configured, its performances need to be verified. At least 20 plans are required by the system in order to generate the regression model. However, as largely discussed in literature, training sets with less than 40 plans need to be avoided and it has also been proved that training sets with 80–100 plans guarantee a higher robustness of models [
      • Boutilier J.J.
      • Craig T.
      • Sharpe M.B.
      • Chan T.C.Y.
      Sample size requirements for knowledge-based treatment planning.
      ,
      • Cagni E.
      • Botti A.
      • Wang Y.
      • Iori M.
      • Petit S.F.
      • Heijmen B.J.M.
      Pareto-optimal plans as ground truth for validation of a commercial system for knowledge-based DVH-prediction.
      ]. During the configuration, the model needs to be interactively fine-tuned aiming to maximize its robustness [
      • Cagni E.
      • Botti A.
      • Wang Y.
      • Iori M.
      • Petit S.F.
      • Heijmen B.J.M.
      Pareto-optimal plans as ground truth for validation of a commercial system for knowledge-based DVH-prediction.
      ,
      • Alpuche Aviles J.E.
      • Cordero Marcos M.I.
      • Sasaki D.
      • Sutherland K.
      • Kane B.
      • Kuusela E.
      Creation of knowledge-based planning models intended for large scale distribution: minimizing the effect of outlier plans.
      ]. After this fine-tuning process, the predicted DVH may be used to produce objectives for the automated planning optimization. Also, this phase required an appropriate fine-tuning to find the right compromise between the priority of PTV-coverage and OARs-sparing [
      • Alpuche Aviles J.E.
      • Cordero Marcos M.I.
      • Sasaki D.
      • Sutherland K.
      • Kane B.
      • Kuusela E.
      Creation of knowledge-based planning models intended for large scale distribution: minimizing the effect of outlier plans.
      ,

      Fogliata A, Reggiori G, Stravato A, Lobefalo F, Franzese C, Franceschini D, et al. RapidPlan head and neck model: The objectives and possible clinical benefit. Radiat Oncol 2017;12. doi:10.1186/s13014-017-0808-x.

      ,
      • Hussein M.
      • South C.P.
      • Barry M.A.
      • Adams E.J.
      • Jordan T.J.
      • Stewart A.J.
      • et al.
      Clinical validation and benchmarking of knowledge-based IMRT and VMAT treatment planning in pelvic anatomy.
      ,
      • Delaney A.R.
      • Tol J.P.
      • Dahele M.
      • Cuijpers J.
      • Slotman B.J.
      • Verbakel W.F.A.R.
      Effect of dosimetric outliers on the performance of a commercial knowledge-based planning solution.
      ,
      • Castriconi R.
      • Cattaneo G.M.
      • Mangili P.
      • Esposito P.
      • Broggi S.
      • Cozzarini C.
      • et al.
      Clinical implementation of knowledge-based automatic plan optimization for helical tomotherapy.
      ].
      The RapidPlan has been widely investigated and validated in several clinical scenarios [

      Fogliata A, Reggiori G, Stravato A, Lobefalo F, Franzese C, Franceschini D, et al. RapidPlan head and neck model: The objectives and possible clinical benefit. Radiat Oncol 2017;12. doi:10.1186/s13014-017-0808-x.

      ,
      • Hussein M.
      • South C.P.
      • Barry M.A.
      • Adams E.J.
      • Jordan T.J.
      • Stewart A.J.
      • et al.
      Clinical validation and benchmarking of knowledge-based IMRT and VMAT treatment planning in pelvic anatomy.
      ,
      • Chatterjee A.
      • Serban M.
      • Faria S.
      • Souhami L.
      • Cury F.
      • Seuntjens J.
      Novel knowledge-based treatment planning model for hypofractionated radiotherapy of prostate cancer patients.
      ,
      • Castriconi R.
      • Fiorino C.
      • Passoni P.
      • Broggi S.
      • Di Muzio N.G.
      • Cattaneo G.M.
      • et al.
      Knowledge-based automatic optimization of adaptive early-regression-guided VMAT for rectal cancer.
      ,

      Rago M, Placidi L, Polsoni M, Rambaldi G, Cusumano D, Greco F, et al. Evaluation of a generalized knowledge-based planning performance for VMAT irradiation of breast and locoregional lymph nodes—Internal mammary and/or supraclavicular regions. PLoS One 2021;16. doi:10.1371/journal.pone.0245305.

      ,
      • Visak J.
      • McGarry R.C.
      • Randall M.E.
      • Pokhrel D.
      Development and clinical validation of a robust knowledge-based planning model for stereotactic body radiotherapy treatment of centrally located lung tumors.
      ,
      • Cagni E.
      • Botti A.
      • Micera R.
      • Galeandro M.
      • Sghedoni R.
      • Orlandi M.
      • et al.
      Knowledge-based treatment planning: An inter-technique and inter-system feasibility study for prostate cancer.
      ,
      • Castriconi R.
      • Fiorino C.
      • Broggi S.
      • Cozzarini C.
      • Di Muzio N.
      • Calandrino R.
      • et al.
      Comprehensive Intra-Institution stepping validation of knowledge-based models for automatic plan optimization.
      ,
      • Costa E.
      • Richir T.
      • Robilliard M.
      • Bragard C.
      • Logerot C.
      • Kirova Y.
      • et al.
      Assessment of a conventional volumetric-modulated arc therapy knowledge-based planning model applied to the new Halcyon© O-ring linac in locoregional breast cancer radiotherapy.
      ,
      • Scaggion A.
      • Fusella M.
      • Roggio A.
      • Bacco S.
      • Pivato N.
      • Rossato M.A.
      • et al.
      Reducing inter- and intra-planner variability in radiotherapy plan output with a commercial knowledge-based planning solution.
      ,
      • Kubo K.
      • Monzen H.
      • Ishii K.
      • Tamura M.
      • Kawamorita R.
      • Sumida I.
      • et al.
      Dosimetric comparison of RapidPlan and manually optimized plans in volumetric modulated arc therapy for prostate cancer.
      ].
      In addition to the KB, an MCO a posteriori approach, named MCO based Trade-Off Exploration, has been recently implemented in the Varian Eclipse TPS. An initial optimized plan, called the balanced plan, is required at the outset. The balanced plan is used with the chosen N optimization objectives for the generation of the 3 N + 1 alternative plans for objectives. Then users can explore the trade-offs of the possible solutions in the optimization interface and select the plan that best clinically fulfils the treatment goals. The initial plan influences the subsequent Pareto frontier approximation, therefore, exploration of trade-offs around an initial optimized plan is desirable. The MCO Trade-Off Exploration has been investigated especially in combination with the RapidPlan tool [

      Miguel-Chumacero E, Currie G, Johnston A, Currie S. Effectiveness of Multi-Criteria Optimization-based Trade-Off exploration in combination with RapidPlan for head & neck radiotherapy planning. Radiat Oncol 2018;13. doi:10.1186/s13014-018-1175-y.

      ,
      • Teichert K.
      • Currie G.
      • Küfer K.-H.
      • Miguel-Chumacero E.
      • Süss P.
      • Walczak M.
      • et al.
      Targeted multi-criteria optimisation in IMRT planning supplemented by knowledge based model creation.
      ,

      Cagni E, Botti A, Chendi A, Iori M, Spezi E. Use of knowledge based DVH predictions to enhance automated re-planning strategies in head and neck adaptive radiotherapy. Phys Med Biol 2021;66. doi:10.1088/1361-6560/ac08b0.

      ].

      2.1.4 Monaco

      Monaco TPS (Elekta AB, Stockholm, Sweden) does not currently propose any automatic planning module. To overcome this limitation and to be able to automate (at least to some extent) the planning processes, the use of robust templates (or class solutions) has been investigated by several Monaco users. A class solution is defined as a set of beam arrangements, planning objectives and penalty parameters, that are robust enough to produce clinically acceptable dose distributions regardless of patient size, anatomy, target volumes and organs at risk.
      There are several publications on this topic [
      • Wood M.
      • Fonseca A.
      • Sampson D.
      • Kovendy A.
      • Westhuyzen J.
      • Shakespeare T.
      • et al.
      Prostate intensity-modulated radiotherapy planning in seven mouse clicks: development of a class solution for cancer.
      ,
      • Marrazzo L.
      • Arilli C.
      • Pellegrini R.
      • Bonomo P.
      • Calusi S.
      • Talamonti C.
      • et al.
      Automated planning through robust templates and multicriterial optimization for lung VMAT SBRT of lung lesions.
      ,
      • Bral S.
      • Duchateau M.
      • Versmessen H.
      • Engels B.
      • Tournel K.
      • Vinh-Hung V.
      • et al.
      Toxicity and outcome results of a class solution with moderately hypofractionated radiotherapy in inoperable Stage III non-small cell lung cancer using helical tomotherapy.
      ,
      • Forde E.
      • Bromley R.
      • Kneebone A.
      • Eade T.
      A class solution for volumetric-modulated arc therapy planning in postprostatectomy radiotherapy.
      ] applied to different anatomical sites (also with other TPSs). A common conclusion of the above cited studies is that a properly built class solution can result in deliverable plans in large cohorts of patients.
      Recently Elekta has developed a new auto-planning tool [
      • Bijman R.
      • Sharfo A.W.
      • Rossi L.
      • Breedveld S.
      • Heijmen B.
      Pre-clinical validation of a novel system for fully-automated treatment planning.
      ,
      • Biston M.-C.
      • Costea M.
      • Gassa F.
      • Serre A.-A.
      • Voet P.
      • Larson R.
      • et al.
      Evaluation of fully automated a priori MCO treatment planning in VMAT for head-and-neck cancer.
      ] using the concepts of “a priori” MCO and wish-list, already implemented in Erasmus-iCycle [
      • Breedveld S.
      • Storchi P.R.M.
      • Voet P.W.J.
      • Heijmen B.J.M.
      ICycle: Integrated, multicriterial beam angle, and profile optimization for generation of coplanar and noncoplanar IMRT plans.
      ] developed by the Erasmus MC Cancer Institute in Rotterdam (The Netherlands).
      The concept of Erasmus-iCycle wish-list consists of a prioritized list of constraints and objectives that are used during the plan optimization to identify the best clinical plan among those lying in the Pareto front [
      • Breedveld S.
      • Storchi P.R.M.
      • Voet P.W.J.
      • Heijmen B.J.M.
      ICycle: Integrated, multicriterial beam angle, and profile optimization for generation of coplanar and noncoplanar IMRT plans.
      ]. Various publications on the validity of the Erasmus-iCycle optimizer were published following the clinical validation performed at Erasmus University for different anatomical sites such as Cervix [
      • Sharfo A.W.M.
      • Voet P.W.J.
      • Breedveld S.
      • Mens J.W.M.
      • Hoogeman M.S.
      • Heijmen B.J.M.
      Comparison of VMAT and IMRT strategies for cervical cancer patients using automated planning.
      ], Prostate [
      • Heijmen B.
      • Voet P.
      • Fransen D.
      • Penninkhof J.
      • Milder M.
      • Akhiat H.
      • et al.
      Fully automated, multi-criterial planning for Volumetric Modulated Arc Therapy – an international multi-center validation for prostate cancer.
      ], Head & Neck [

      Yan D, Liang J. Expected treatment dose construction and adaptive inverse planning optimization: Implementation for offline head and neck cancer adaptive radiotherapy. Med Phys 2013;40. doi:10.1118/1.4788659.

      ] and Lung [
      • Della Gala G.
      • Dirkx M.L.P.
      • Hoekstra N.
      • Fransen D.
      • Lanconelli N.
      • van de Pol M.
      • et al.
      Fully automated VMAT treatment planning for advanced-stage NSCLC patients Vollautomatische VMAT-Behandlungsplanung für Patienten mit fortgeschrittenem NSCLC.
      ].

      2.2 Survey

      An online survey using Google Forms and comprising 28 multiple-choice questions was sent via email to 175 medical physicists working in radiotherapy centres in Italy between February 21st and April 1st, 2021. Only a single spokesperson for each radiotherapy centre was contacted and asked to provide responses that were representative of the center’s planning activity. She/he was asked to provide some information on the instrumentations and working habitats in her/his centre and to share their perspectives on the use of automated planning in radiotherapy. When the reference physicist for planning activity was not known, the director of the Medical Physics Department/Unit was contacted.
      The first invitation was sent on February 21st, 2021, followed by a reminder on March 1st sent to those who did not respond to the first e-mail. A final reminder was carried out for non-responders by contacting them either by phone or by sending individual emails. The survey was closed on April 1, 2021.
      The survey (the full text of which is provided in the in Supplementary Materials) was divided into three sections: 1- Characteristics of the participating centre; 2- Opinions on the use of automatic planning and 3- Experience in using automated planning. For some questions, more than one answer was allowed. The survey’s questions have been conceived by the Authors; they have been discussed and iteratively modified during some meetings. Before circulating, the survey has been preliminarily submitted to a group of medical physicists in order to get their feedback on completeness, understandability, ease of reading and ease of filling out. To our knowledge, no similar experience was performed and published in other Countries.
      Unless otherwise stated, the percentages reported in the results are always calculated with respect to the number of responses received (total or per centre size).

      3. Results

      125 medical physicists of the 175 contacted completed the survey. A global percentage of 71 % was obtained with a response percentage per region varying from 47% to 100% (summarized in Fig. 1).
      Figure thumbnail gr1
      Fig. 1Percentage of responding centres per region.

      3.1 Characteristics of the participating centre

      Centres were stratified into large, medium, small, and very small according to the number of patients treated per year: more than 2000, between 1000 and 2000, between 500 and 1000, and less than 500 respectively.
      The distribution among the 125 responding centres was 10%, 26%, 47% and 17% for large, medium, small, and very small centres, respectively. Small/medium-sized centres are prevalent and correspond to 73% of the total answers received. The responding centre's equipment, split into the four categories, is represented in the box plot of Fig. 2A. Small and very small centres show the same distribution even if in very small centres a larger number of “other” radiation units and a lower number of linac is present. GammaKnife and Cyberknife are mainly installed in medium and large centres. Twenty-four out of a total of 29 Tomotherapy are installed in medium/large centres, with a concentration of 9 installations in three centres. The mean number of dedicated planners per day is 5.6, 2.9, 2.1, 1.7 in large, medium, small, and very small centres, respectively. Two planners per day are employed in 66% of small centres (Fig. 2B) while 1 to 6 planners per day are employed in the remaining 34% of small centres (6 planners are employed in a proton therapy centre). For small centres median, first and third quartile values are equal to 2. The equal amount of the three values explains the compressed appearance of the box and the presence of some outliers in Fig. 2b.
      Figure thumbnail gr2
      Fig. 2A) Radiation units installed in answering centres of different size; B) Number of planners employed each day in large, medium, small, and very small centres; C) Percentage of IMRT/VMAT plans (over the total number of plans) planned by large, medium, small, and very small centres; D) Percentage of daily IGRT performed in large, medium, small, and very small centres.
      58% of medium-sized centres employ more than two planners per day and 54% of large-sized centres employ more than six planners per day with a maximum of ten planners reported in one case.
      Planning activity is performed by medical physicists alone in 80% of centres, while in 20% of centres medical physicists receive support from dosimetrists. A linear trend between the average number of planners and the average number of radiation units per centre (R2 = 0.95) is observed.
      Intensity-modulated (IMRT, helical IMRT and VMAT) treatments are provided to more than 40% of patients in 92%, 79%, 78%, and 65% of large, medium, small, and very small centres, respectively, thus showing that larger centres are more oriented toward complex treatments (Fig. 2C). Only one of the 13 large centres claims to plan between 20% and 40% of their patients with an intensity-modulated approach. None of the small, medium, and large centres plan less than 10% of their patients using intensity-modulation techniques.
      Very small centres have very diverse work habits: more than 70% of IMRT/VMAT treatments are planned by 45% of centres and less than 10% by 15%. Different use of daily IGRT among centres of varying sizes is reported in Fig. 2D. More than 70% of patients undergo a daily IGRT in 77%, 36%, 25% and 40% of large, medium, small, and very small centres. Less than 10% of patients undergo a daily IGRT in 0%, 3%, 10% and 40% of large, medium, small, and very small centres.
      In Fig. 3 the percentage distribution of the TPSs is shown. The most popular TPS is Eclipse (36%). In 49% of cases, the centres have a TPS with some automatic option, but only in 33% of these the systems are in clinical use. A percentage of 2% of users consider Monaco robust templates as an automatic option. Among the 125 responding centres the most widely used TPS with an automatic option in clinical use is Pinnacle (16%) followed by Raystation (9%) and Eclipse (4%).
      Figure thumbnail gr3
      Fig. 3Percentage distribution of TPS. O = others, M = Monaco, P = Pinnacle, E = Eclipse, R = Raystation with automatic option, in clinical use Other systems include TomoPlanning, CyberKnife TPS, Gammaplan, Masterplan, ViewRay TPS and other old TPS systems. The automated systems in the category other are MBM Elements (Brainlab) and Ethos (Varian).
      The total amount of users having automated systems but not using them clinically are 17% of responders. Among them, 10% declare to be in the process of validation, 3% state a lack of training, skills, or resources, while 3% are not interested in clinical implementation.

      3.2 Opinions on the use of automatic planning

      Users’ perception of advantages and disadvantages, organizational and professional changes deriving from the adoption of automated planning in clinical practice are reported in Table 1.
      Table 1Users’ perception of advantages and disadvantages of using automated planning in clinical practice.
      Benefits from automated planning (multiple answers allowed)Work organization for medical physicistsMedical physicist roleLoss of knowledge and skills
      1% none3% changes in resources allocation job reduction2% diminished4% don't know
      33% improved quality of treatment plans15% no organizational changes5% no impact6% yes
      67% increased planning consistency with reduced operator dependency82% more resources for activities other than planning10% don't know30% probably yes
      74% reduced planning time83% enriched30% probably not
      30% no
      Most participants consider the use of automated techniques to be beneficial, while only 1% do not see any advantage. More than 81% of respondents see in the use of automated systems the possibility of freeing up resources and enriching their professional role, while 3% see automated systems as a potential threat in terms of job reductions. A percentage of 36% of responders are prone to thinking that the inclusion of such techniques will involve a loss of knowledge and skills while 60% do not think so.
      Among those who do not own an automated system (51% of responders), 22% feel that it is not a priority over other needs.
      Among the reasons for not having an automated strategy, respondents reported the lack of economic-structural availability (33%), the lack of automated options in the available TPS (6%), and the absence of an adequate implementation (2%). In 28% of cases, those who do not use automated planning in their clinical practice declare they are planning to acquire it in the following 1–2 years.

      3.3 Experience using automated planning

      The surveyed centres use automated planning in widely varying percentages: 13% of participating centres (39% of centres having automated plan in clinical use) plan more than 50% of their plans using automated solutions.
      Different percentages of plans made using an automated approach split according to the size of participating centres are shown in Fig. 4A.
      Figure thumbnail gr4
      Fig. 4A) Percentage of plans obtained with automated approach, B) implementation of automated planning in different treatment sites for different TPSs, C) distribution of automated TPS in clinical use in large, medium, small, and very small centres, D) number of centres that state the need to further optimize the automated plans never, sometimes, often, very often and always, split according to the employed TPS. The automated systems in the category other are MBM Elements (Brainlab) and Ethos (Varian). In C, 4D and 4D data are normalized to the total number of centres with automated option in clinical use.
      In large centres, percentages of plans performed with automated solutions <5%, in the range of 10%-20%, and in the range of 20%-50% are reported with the same frequency. More than 50% of plans are planned with an automated approach in 22%, 55% and 50% of medium, small, and very small centres, respectively. Centres that plan more than 50% of their patients with an automated approach employ Pinnacle, Monaco, and Raystation in 75%, 13%, and 13% of cases, respectively. It is worth mentioning that all Raystation users make use of the advanced scripting GPS system (an in-house developed system that uses a genetic algorithm) developed by the same group [
      • Fiandra C.
      • Alparone A.
      • Gallio E.
      • Vecchi C.
      • Balestra G.
      • Bartoncini S.
      • et al.
      Automated heuristic optimization of prostate VMAT treatment planning.
      ,
      • Fiandra C.
      • Rossi L.
      • Alparone A.
      • Zara S.
      • Vecchi C.
      • Sardo A.
      • et al.
      Automatic genetic planning for volumetric modulated arc therapy: A large multi-centre validation for prostate cancer.
      ]. Only two users also have the Raystation MCO option in clinical use but only one of them use it extensively.
      In Fig. 4B the use of automation for planning different treatment sites is represented, sorted according to TPS. Most centres (88%) use automated planning in more than one anatomical site. Automated solutions are applied to head/head&neck treatments in 83% of cases, to abdominal lesions in 54%, to breast treatment in 46%, to thoracic tumours in 59% and to pelvic lesions in 88%.
      Among TPS implementing an automated approach, Eclipse is present only in large and medium centres while Raystation and Pinnacle are installed in centres of all sizes (Fig. 4C).
      Fig. 4D shows how often users need to manually adjust automated plans, divided by TPS type.
      The percentage of the average time spent on setting-up and fine-tuning the automatic planning shows significant variability, spanning from a few days (15%) or a few weeks (56%) to a few months (29%). There is no correlation between the time spent tuning the system and the need to adjust the automated plan manually.
      In all cases, the automatic option was validated before introducing it in clinical daily use by comparing automated plans with manual clinical plans. Users report a variable reduction in planning time: less than10% in 7% of cases, between 10% and 50% in 66%, and greater than 50% in 27%.
      Fig. 5 shows the percentage of users that declare very high, high, and low satisfaction with automated systems’ performance. For example, among the users having an automated system in clinical use, Pinnacle and Raystation users declare to be very satisfied (dark green) in 30% and 7% of the cases respectively.
      Figure thumbnail gr5
      Fig. 5Level of satisfaction with the automated system’s performance, divided by TPS type. Data are normalized to the total number of centres with automated option in clinical use.
      A high percentage of users (93%) do not think that automated plans need more thorough quality control than manual plans, while the remaining 7% of users observe that automated plans may present a higher modulation degree and thus may require more in-depth quality control than manual plans.
      Training and education received for the commissioning and the clinical deployment of automated systems is considered sufficient by 76% of users.

      3. Discussion

      Automation in planning is a growing field in which both industry companies and users are investing resources. The amount of literature on the use and set-up of automated planning is in line with the amount of literature on the availability of such systems and the algorithms behind their operation. Yet there is a lack of information on the diffusion of such systems among radiotherapy centres and their use in clinical practice as well as on the perception of automated planning tools by medical physicists involved in planning activity. In this context the idea of the present survey originated. To our knowledge, no similar experience was performed and published in other Countries.
      The high percentage of responding centres (71%) demonstrates that automated planning is perceived as a relevant topic for the community of medical physicists working in radiotherapy.
      On the other hand, there is a percentage (29% over the 175 contacted centres) of centres who opted not to participate, despite three reminders. This could be attributed to a lack of interest, knowledge, or awareness of automated planning or to a mistrust of our initiative.
      In addition to collecting information on the use of automation in planning, our survey achieved the goal of capturing a snapshot of Italian radiotherapy equipment and the job habits of medical physicists in large, medium, small, and very small Italian centres.
      Among the responses, 73% come from small and medium-sized centres that report intensive use of intensity-modulated techniques. More than 40% of patients are treated with intensity-modulated techniques in 79% and 78% of medium and small centres, respectively. This percentage rises to 92% in large centres, thus showing the widespread use of this planning and delivery technique. Among the very small centres there is a large heterogeneity in the use of intensity modulated techniques: 15% and 45% of very small centres plan less than 10% and more than 70% of treatments, respectively, with an intensity-modulated approach. These results suggest a bi-modal distribution of highly and poorly specialized very small centres, confirmed by the different rates of daily IGRT performed.
      The most popular TPS is Eclipse (36%), followed by Monaco (31%), Pinnacle (27%) and Raystation (22%). However, looking at the distribution of TPS with an automated module in clinical use, the distribution changes. The most used TPS is Pinnacle (16%) followed by Raystation (9%), Eclipse (4%) and Monaco (2%). It is worth noting that 3% of users consider Monaco robust templates as an automatic option. In the current commercial implementation, Monaco does not have a truly automated system. However, since using robust planning templates the workload is considerably reduced, as is the planner intervention, some users consider Monaco among the systems capable of automating the planning process.
      Conversely, RayStation has three different systems already commercially available, but only one center (out of the 11 clinically using it) claims to use extensively MCO, while the others claim to use the advanced scripting GPS system.
      The total number of centres using an automated approach in clinical practice will increase shortly since another 17% of users have an automated option in their TPS and about 10% are in the validation phase. In addition, there seems to be an emerging interest among those who do not currently own such systems. Among the reasons cited for the unavailability of automated options is the lack of economic resources (33%); expected acquisition in the next 1–2 years was reported by 28%. Only 22% state that automated system acquisition has a low priority with respect to other needs, and only two respondents are not interested in current automated system implementations.
      The use of automatic planning is considered beneficial by all but one participant. The favorable opinion is shared both by those who do not own automated systems and those who already use them. Among the greatest benefits, users see a reduced planning time and improved efficiency of the planning process and an increased planning consistency with reduced operator dependency.
      Interestingly, those responding to the survey do not feel threatened by the spread of automated planning systems even though radiation therapy planning is a core activity of the medical physics profession. Among all participants, 82% acknowledge the potential of automatic systems to free up resources that can be profitably dedicated to other activities. In contrast, the perception of losing skills is not as sharp: only 60% of respondents hold that the use of automated planning techniques will not involve a loss of knowledge and skills, while 36% think the opposite. The same evaluation was obtained by restricting the analysis to automated planning users. These differing views can likely be attributable to different users’ participation in the set-up of the automated system as well as on the type of automatic planning implementation used in their centre. A physicist more involved in the commissioning and setting phase of an automated system can perceive this task as requiring and guaranteeing a high level of knowledge and skills in planning.
      A limitation of our analysis with regard to operator perceptions is that only one medical physicist per center was asked to respond to the survey. Thus, it must be kept in mind that the picture we captured may not be fully representative.
      It is interesting to compare the percentage of users who have an automatic option compared to those who have already put it into clinical use (Fig. 3). The percentages are: 10% vs. 4% for Eclipse, 15% vs 9% for Raystation and 21% vs 16% for Pinnacle. As for Monaco, all its users could create robust templates, but this is not a widespread practice, or, at least, it is not perceived as a real automatic option for all users and was, therefore, not declared in the survey.
      Most centres use automated planning on several anatomical sites with a higher frequency for head/head&neck and pelvis, regardless of the TPS used.
      The approach to automation followed by different types of TPS could be responsible for the differences in the spread of implementations in clinical use. A knowledge-based approach requires a large amount of training of the system, while, for example, a protocol-based optimization algorithm is generally expected to be easier to set up: for this reason, automatic planning modules offered by Pinnacle and Raystation are probably easier to implement even for less experienced users. This fact has probably had a favorable impact on the spread of this type of system, as also confirmed by the degree of user’s satisfaction.
      We think that this could be a major explanation for the surprisingly higher fraction of fully automatic plans, mostly based on Pinnacle and RayStation template-based approaches, especially in small and very-small centers. Moreover, centres that use automated planning intensively are likely to have a very standardized clinical practice. On the contrary, in large and medium-size centres, where patient's retreatments are frequent and there is a large variety of protocols and treatment sites, automatic planning can have a more limited application.
      On the other hand, a rapid massive replacement of manual planning with automatic planning should also be considered with attention and caution, especially in centers with less human resources and experience. Participation in plan audits and inter-Institute plan comparison studies could be a relevant tool for assessing one’s planning capacity, helping to make appropriate choices in terms of “safe” and gradual implementation of automatic planning approaches in all centers.
      Training and education are very important aspects that can have a considerable impact on the deployment of these systems. This survey reveals that 24% of users do not feel they have had adequate training for setting up the automated planning solutions they work with.
      The average time spent on fine-tuning automatic systems is variable and spans from a few days to a few months. The need for the manual adjustment of a plan obtained with an automated approach is also variable, and it seems not to be related to the time spent on system implementation. The need for a manual adjustment may depend on a combination of the fine-tuning of automated systems and on the approach and experience of the radiotherapist who judges and approves the plan.
      All users agree that automated systems enable a reduction in planning time. Most users consider increasing the number of pretreatment checks of automated plans unnecessary. This issue and the number of current installations (the first in 2014 and now numbering 41) show the growing confidence of Italian users towards automated planning systems.

      4. Conclusions

      Our survey shows that 49% of the responding centres have an automatic planning solution although clinically used in only 33% of the cases. A generally positive attitude toward automatic planning was reported, with a prevalence of expected benefit and enrichment for the medical physics profession.
      Physicists are the key component in guaranteeing safe and optimal translation from manual to automatic planning in a rapidly increasing proportion of patients, gradually shifting their role from repetitive planning activities to expert management and customization of advanced automatic planning solutions. On the other hand, this field is still in its infancy and large margins for improving efficiency and the quality of plan optimization through automatic approaches are evident and still require dedicated efforts and investments in terms of developing skills and adequate human resources.

      Acknowledgements

      The authors would like to thank Prof. Russo for stimulating this paper with his questions. The authors would also like to thank Dr Strigari, Dr Paiusco and Dr Moretti for the preliminary revision of the questionnaire and all the participating centres for their availability, interest and time.

      Appendix A. Supplementary data

      The following are the Supplementary data to this article:

      References

        • Teoh M.
        • Clark C.H.
        • Wood K.
        • Whitaker S.
        • Nisbet A.
        Volumetric modulated arc therapy: a review of current literature and clinical use in practice.
        Br J Radiol. 2011; 84: 967-996https://doi.org/10.1259/bjr/22373346
      1. Hussein M, Heijmen BJM, Verellen D, Nisbet A. Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations. Br J Radiol 2018;91. doi:10.1259/bjr.20180270.

        • Craft D.L.
        • Halabi T.F.
        • Shih H.A.
        • Bortfeld T.R.
        Approximating convex Pareto surfaces in multiobjective radiotherapy planning.
        Med Phys. 2006; 33: 3399-3407https://doi.org/10.1118/1.2335486
        • Craft D.
        • Bortfeld T.
        How many plans are needed in an IMRT multi-objective plan database?.
        Phys Med Biol. 2008; 53: 2785-2796https://doi.org/10.1088/0031-9155/53/11/002
        • Craft D.
        • Halabi T.
        • Shih H.A.
        • Bortfeld T.
        An approach for practical multiobjective IMRT treatment planning.
        Int J Radiat Oncol Biol Phys. 2007; 69: 1600-1607https://doi.org/10.1016/j.ijrobp.2007.08.019
        • Monz M.
        • Küfer K.H.
        • Bortfeld T.R.
        • Thieke C.
        Pareto navigation – algorithmic foundation of interactive multi-criteria IMRT planning.
        Phys Med Biol. 2008; 53: 985-998https://doi.org/10.1088/0031-9155/53/4/011
      2. Kierkels RGJ, Visser R, Bijl HP, Langendijk JA, van ’t Veld AA, Steenbakkers RJHM, 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. doi: 10.1186/s13014-015-0385-9.

        • Wala J.
        • Craft D.
        • Paly J.
        • Zietman A.
        • Efstathiou J.
        Maximizing dosimetric benefits of IMRT in the treatment of localized prostate cancer through multicriteria optimization planning.
        Med Dosim. 2013; 38: 298-303https://doi.org/10.1016/j.meddos.2013.02.012
        • Ghandour S.
        • Matzinger O.
        • Pachoud M.
        Volumetric-modulated arc therapy planning using multicriteria optimization for localized prostate cancer.
        J Appl Clin Med Phys. 2015; 16: 258-269https://doi.org/10.1120/jacmp.v16i3.5410
        • Kamran S.C.
        • Mueller B.S.
        • Paetzold P.
        • Dunlap J.
        • Niemierko A.
        • Bortfeld T.
        • et al.
        Multi-criteria optimization achieves superior normal tissue sparing in a planning study of intensity-modulated radiation therapy for RTOG 1308-eligible non-small cell lung cancer patients.
        Radiother Oncol. 2016; 118: 515-520https://doi.org/10.1016/j.radonc.2015.12.028
        • Craft D.L.
        • Hong T.S.
        • Shih H.A.
        • Bortfeld T.R.
        Improved Planning Time and Plan Quality Through Multicriteria Optimization for Intensity-Modulated Radiotherapy.
        Int J Radiat Oncol. 2012; 82: e83-e90https://doi.org/10.1016/j.ijrobp.2010.12.007
        • McIntosh C.
        • Welch M.
        • McNiven A.
        • Jaffray D.A.
        • Purdie T.G.
        Fully automated treatment planning for head and neck radiotherapy using a voxel-based dose prediction and dose mimicking method.
        Phys Med Biol. 2017; 62: 5926-5944https://doi.org/10.1088/1361-6560/aa71f8
        • Fiandra C.
        • Alparone A.
        • Gallio E.
        • Vecchi C.
        • Balestra G.
        • Bartoncini S.
        • et al.
        Automated heuristic optimization of prostate VMAT treatment planning.
        Int J Med Physics, Clin Eng Radiat Oncol. 2018; 07: 414-425https://doi.org/10.4236/ijmpcero.2018.73034
        • Fiandra C.
        • Rossi L.
        • Alparone A.
        • Zara S.
        • Vecchi C.
        • Sardo A.
        • et al.
        Automatic genetic planning for volumetric modulated arc therapy: A large multi-centre validation for prostate cancer.
        Radiother Oncol. 2020; 148: 126-132https://doi.org/10.1016/j.radonc.2020.04.020
        • Gintz D.
        • Latifi K.
        • Caudell J.
        • Nelms B.
        • Zhang G.
        • Moros E.
        • et al.
        Initial evaluation of automated treatment planning software.
        J Appl Med Phys. 2016; 17: 331-346
        • Xia W.
        • Han F.
        • Chen J.
        • Miao J.
        • Dai J.
        Personalized setting of plan parameters using feasibility dose volume histogram for auto-planning in Pinnacle system Auto-Planning, lung cancer, OAR sparing, planning time, plan quality.
        VMAT. J Appl Clin Med Phys. 2020; 21: 119-127https://doi.org/10.1002/acm2.12897
        • Cilla S.
        • Romano C.
        • Morabito V.E.
        • Macchia G.
        • Buwenge M.
        • Dinapoli N.
        • et al.
        Personalized treatment planning automation in prostate cancer radiation oncology: a comprehensive dosimetric study.
        Front Oncol. 2021; 11https://doi.org/10.3389/fonc.2021.636529
        • Ahmed S.
        • Nelms B.
        • Gintz D.
        • Caudell J.
        • Zhang G.
        • Moros E.G.
        • et al.
        A method for a priori estimation of best feasible DVH for organs-at-risk: validation for head and neck VMAT planning.
        Med Phys. 2017; 44: 5486-5497https://doi.org/10.1002/mp.12500
        • Hazell I.
        • Bzdusek K.
        • Kumar P.
        • Hansen C.R.
        • Bertelsen A.
        • Eriksen J.G.
        • et al.
        Automatic planning of head and neck treatment plans.
        J Appl Clin Med Phys. 2016; 17: 272-282https://doi.org/10.1120/jacmp.v17i1.5901
        • Nawa K.
        • Haga A.
        • Nomoto A.
        • Sarmiento R.A.
        • Shiraishi K.
        • Yamashita H.
        • et al.
        Evaluation of a commercial automatic treatment planning system for prostate cancers.
        Med Dosim. 2017; 42: 203-209https://doi.org/10.1016/j.meddos.2017.03.004
        • Gallio E.
        • Giglioli F.R.
        • Girardi A.
        • Guarneri A.
        • Ricardi U.
        • Ropolo R.
        • et al.
        Evaluation of a commercial automatic treatment planning system for liver stereotactic body radiation therapy treatments.
        Phys Med. 2018; 46: 153-159https://doi.org/10.1016/j.ejmp.2018.01.016
        • Marrazzo L.
        • Meattini I.
        • Arilli C.
        • Calusi S.
        • Casati M.
        • Talamonti C.
        • et al.
        Auto-planning for VMAT accelerated partial breast irradiation.
        Radiother Oncol. 2019; 132: 85-92https://doi.org/10.1016/j.radonc.2018.11.006
        • Arilli C.
        • Zani M.
        • Marrazzo L.
        • Scoccianti S.
        • Casati M.
        • Compagnucci A.
        • et al.
        Automatic VMAT technique to treat glioblastoma: a two years’ experience.
        Phys Med. 2021; 90: 115-122https://doi.org/10.1016/j.ejmp.2021.09.015
        • Ge Y.
        • Wu Q.J.
        Knowledge-based planning for intensity-modulated radiation therapy: a review of data-driven approaches.
        Med Phys. 2019; 46: 2760-2775https://doi.org/10.1002/mp.13526
        • Boutilier J.J.
        • Craig T.
        • Sharpe M.B.
        • Chan T.C.Y.
        Sample size requirements for knowledge-based treatment planning.
        Med Phys. 2016; 43: 1212-1221https://doi.org/10.1118/1.4941363
        • Cagni E.
        • Botti A.
        • Wang Y.
        • Iori M.
        • Petit S.F.
        • Heijmen B.J.M.
        Pareto-optimal plans as ground truth for validation of a commercial system for knowledge-based DVH-prediction.
        Phys Med. 2018; 55: 98-106https://doi.org/10.1016/j.ejmp.2018.11.002
        • Alpuche Aviles J.E.
        • Cordero Marcos M.I.
        • Sasaki D.
        • Sutherland K.
        • Kane B.
        • Kuusela E.
        Creation of knowledge-based planning models intended for large scale distribution: minimizing the effect of outlier plans.
        J Appl Clin Med Phys. 2018; 19: 215-226https://doi.org/10.1002/acm2.12322
      3. Fogliata A, Reggiori G, Stravato A, Lobefalo F, Franzese C, Franceschini D, et al. RapidPlan head and neck model: The objectives and possible clinical benefit. Radiat Oncol 2017;12. doi:10.1186/s13014-017-0808-x.

        • Hussein M.
        • South C.P.
        • Barry M.A.
        • Adams E.J.
        • Jordan T.J.
        • Stewart A.J.
        • et al.
        Clinical validation and benchmarking of knowledge-based IMRT and VMAT treatment planning in pelvic anatomy.
        Radiother Oncol. 2016; 120: 473-479https://doi.org/10.1016/j.radonc.2016.06.022
        • Delaney A.R.
        • Tol J.P.
        • Dahele M.
        • Cuijpers J.
        • Slotman B.J.
        • Verbakel W.F.A.R.
        Effect of dosimetric outliers on the performance of a commercial knowledge-based planning solution.
        Int J Radiat Oncol Biol Phys. 2016; 94: 469-477https://doi.org/10.1016/j.ijrobp.2015.11.011
        • Castriconi R.
        • Cattaneo G.M.
        • Mangili P.
        • Esposito P.
        • Broggi S.
        • Cozzarini C.
        • et al.
        Clinical implementation of knowledge-based automatic plan optimization for helical tomotherapy.
        Pract Radiat Oncol. 2021; 11: e236-e244https://doi.org/10.1016/j.prro.2020.09.012
        • Chatterjee A.
        • Serban M.
        • Faria S.
        • Souhami L.
        • Cury F.
        • Seuntjens J.
        Novel knowledge-based treatment planning model for hypofractionated radiotherapy of prostate cancer patients.
        Phys Med. 2020; 69: 36-43https://doi.org/10.1016/J.EJMP.2019.11.023
        • Castriconi R.
        • Fiorino C.
        • Passoni P.
        • Broggi S.
        • Di Muzio N.G.
        • Cattaneo G.M.
        • et al.
        Knowledge-based automatic optimization of adaptive early-regression-guided VMAT for rectal cancer.
        Phys Med. 2020; 70: 58-64https://doi.org/10.1016/j.ejmp.2020.01.016
      4. Rago M, Placidi L, Polsoni M, Rambaldi G, Cusumano D, Greco F, et al. Evaluation of a generalized knowledge-based planning performance for VMAT irradiation of breast and locoregional lymph nodes—Internal mammary and/or supraclavicular regions. PLoS One 2021;16. doi:10.1371/journal.pone.0245305.

        • Visak J.
        • McGarry R.C.
        • Randall M.E.
        • Pokhrel D.
        Development and clinical validation of a robust knowledge-based planning model for stereotactic body radiotherapy treatment of centrally located lung tumors.
        J Appl Clin Med Phys. 2021; 22: 146-155https://doi.org/10.1002/acm2.v22.110.1002/acm2.13120
        • Cagni E.
        • Botti A.
        • Micera R.
        • Galeandro M.
        • Sghedoni R.
        • Orlandi M.
        • et al.
        Knowledge-based treatment planning: An inter-technique and inter-system feasibility study for prostate cancer.
        Phys Med. 2017; 36: 38-45https://doi.org/10.1016/j.ejmp.2017.03.002
        • Castriconi R.
        • Fiorino C.
        • Broggi S.
        • Cozzarini C.
        • Di Muzio N.
        • Calandrino R.
        • et al.
        Comprehensive Intra-Institution stepping validation of knowledge-based models for automatic plan optimization.
        Phys Med. 2019; 57: 231-237https://doi.org/10.1016/j.ejmp.2018.12.002
        • Costa E.
        • Richir T.
        • Robilliard M.
        • Bragard C.
        • Logerot C.
        • Kirova Y.
        • et al.
        Assessment of a conventional volumetric-modulated arc therapy knowledge-based planning model applied to the new Halcyon© O-ring linac in locoregional breast cancer radiotherapy.
        Phys Med. 2021; 86: 32-43https://doi.org/10.1016/j.ejmp.2021.05.022
        • Scaggion A.
        • Fusella M.
        • Roggio A.
        • Bacco S.
        • Pivato N.
        • Rossato M.A.
        • et al.
        Reducing inter- and intra-planner variability in radiotherapy plan output with a commercial knowledge-based planning solution.
        Phys Med. 2018; 53: 86-93https://doi.org/10.1016/j.ejmp.2018.08.016
        • Kubo K.
        • Monzen H.
        • Ishii K.
        • Tamura M.
        • Kawamorita R.
        • Sumida I.
        • et al.
        Dosimetric comparison of RapidPlan and manually optimized plans in volumetric modulated arc therapy for prostate cancer.
        Phys Med. 2017; 44: 199-204https://doi.org/10.1016/j.ejmp.2017.06.026
      5. Miguel-Chumacero E, Currie G, Johnston A, Currie S. Effectiveness of Multi-Criteria Optimization-based Trade-Off exploration in combination with RapidPlan for head & neck radiotherapy planning. Radiat Oncol 2018;13. doi:10.1186/s13014-018-1175-y.

        • Teichert K.
        • Currie G.
        • Küfer K.-H.
        • Miguel-Chumacero E.
        • Süss P.
        • Walczak M.
        • et al.
        Targeted multi-criteria optimisation in IMRT planning supplemented by knowledge based model creation.
        Oper Res Heal Care. 2019; 23: 100185https://doi.org/10.1016/j.orhc.2019.04.003
      6. Cagni E, Botti A, Chendi A, Iori M, Spezi E. Use of knowledge based DVH predictions to enhance automated re-planning strategies in head and neck adaptive radiotherapy. Phys Med Biol 2021;66. doi:10.1088/1361-6560/ac08b0.

        • Wood M.
        • Fonseca A.
        • Sampson D.
        • Kovendy A.
        • Westhuyzen J.
        • Shakespeare T.
        • et al.
        Prostate intensity-modulated radiotherapy planning in seven mouse clicks: development of a class solution for cancer.
        Reports Pract Oncol Radiother. 2016; 21: 567-570https://doi.org/10.1016/j.rpor.2016.09.005
        • Marrazzo L.
        • Arilli C.
        • Pellegrini R.
        • Bonomo P.
        • Calusi S.
        • Talamonti C.
        • et al.
        Automated planning through robust templates and multicriterial optimization for lung VMAT SBRT of lung lesions.
        J Appl Clin Med Phys. 2020; 21: 114-120https://doi.org/10.1002/acm2.v21.610.1002/acm2.12872
        • Bral S.
        • Duchateau M.
        • Versmessen H.
        • Engels B.
        • Tournel K.
        • Vinh-Hung V.
        • et al.
        Toxicity and outcome results of a class solution with moderately hypofractionated radiotherapy in inoperable Stage III non-small cell lung cancer using helical tomotherapy.
        Int J Radiat Oncol Biol Phys. 2010; 77: 1352-1359https://doi.org/10.1016/j.ijrobp.2009.06.075
        • Forde E.
        • Bromley R.
        • Kneebone A.
        • Eade T.
        A class solution for volumetric-modulated arc therapy planning in postprostatectomy radiotherapy.
        Med Dosim. 2014; 39: 261-265https://doi.org/10.1016/j.meddos.2014.04.002
        • Bijman R.
        • Sharfo A.W.
        • Rossi L.
        • Breedveld S.
        • Heijmen B.
        Pre-clinical validation of a novel system for fully-automated treatment planning.
        Radiother Oncol. 2021; 158: 253-261https://doi.org/10.1016/J.RADONC.2021.03.003
        • Biston M.-C.
        • Costea M.
        • Gassa F.
        • Serre A.-A.
        • Voet P.
        • Larson R.
        • et al.
        Evaluation of fully automated a priori MCO treatment planning in VMAT for head-and-neck cancer.
        Phys Med. 2021; 87: 31-38https://doi.org/10.1016/j.ejmp.2021.05.037
        • Breedveld S.
        • Storchi P.R.M.
        • Voet P.W.J.
        • Heijmen B.J.M.
        ICycle: Integrated, multicriterial beam angle, and profile optimization for generation of coplanar and noncoplanar IMRT plans.
        Med Phys. 2012; 39: 951-963https://doi.org/10.1118/1.3676689
        • Sharfo A.W.M.
        • Voet P.W.J.
        • Breedveld S.
        • Mens J.W.M.
        • Hoogeman M.S.
        • Heijmen B.J.M.
        Comparison of VMAT and IMRT strategies for cervical cancer patients using automated planning.
        Radiother Oncol. 2015; 114: 395-401https://doi.org/10.1016/j.radonc.2015.02.006
        • Heijmen B.
        • Voet P.
        • Fransen D.
        • Penninkhof J.
        • Milder M.
        • Akhiat H.
        • et al.
        Fully automated, multi-criterial planning for Volumetric Modulated Arc Therapy – an international multi-center validation for prostate cancer.
        Radiother Oncol. 2018; 128: 343-348https://doi.org/10.1016/j.radonc.2018.06.023
      7. Yan D, Liang J. Expected treatment dose construction and adaptive inverse planning optimization: Implementation for offline head and neck cancer adaptive radiotherapy. Med Phys 2013;40. doi:10.1118/1.4788659.

        • Della Gala G.
        • Dirkx M.L.P.
        • Hoekstra N.
        • Fransen D.
        • Lanconelli N.
        • van de Pol M.
        • et al.
        Fully automated VMAT treatment planning for advanced-stage NSCLC patients Vollautomatische VMAT-Behandlungsplanung für Patienten mit fortgeschrittenem NSCLC.
        Strahlentherapie Und Onkol. 2017; 193: 402-409https://doi.org/10.1007/s00066-017-1121-1