iCycle: Integrated, multicriterial beam angle, and profile optimization for generation of coplanar and noncoplanar IMRT plans

Artificial Intelligence in Radiation Therapy

Artificial intelligence (AI) has great potential to transform the clinical workflow of radiotherapy. Since the introduction of deep neural networks, many AI-based methods have been proposed to address challenges in different aspects of radiotherapy. Commercial vendors have started to release AI-based tools that can be readily integrated to the established clinical workflow. To show the recent progress in AI-aided radiotherapy, we have reviewed AI-based studies in five major aspects of radiotherapy including image reconstruction, image registration, image segmentation, image synthesis, and automatic treatment planning. In each section, we summarized and categorized the recently published methods, followed by a discussion of the challenges, concerns, and future development. Given the rapid development of AI-aided radiotherapy, the efficiency and effectiveness of radiotherapy in the future could be substantially improved through intelligent automation of various aspects of radiotherapy.

TBS-BAO: fully automated beam angle optimization for IMRT guided by a total-beam-space reference plan

Properly selected beam angles contribute to the quality of radiotherapy treatment plans. However, the beam angle optimization (BAO) problem is difficult to solve to optimality due to its non-convex discrete nature with many local minima. In this study, we propose TBS-BAO, a novel approach for solving the BAO problem, and test it for non-coplanar robotic CyberKnife radiotherapy for prostate cancer. First, an ideal Pareto-optimal reference dose distribution is automatically generated usinga priorimulti-criterial fluence map optimization (FMO) to generate a plan that includes all candidate beams (total-beam-space, TBS). Then, this ideal dose distribution is reproduced as closely as possible in a subsequent segmentation/beam angle optimization step (SEG/BAO), while limiting the number of allowed beams to a user-selectable preset value. SEG/BAO aims at a close reproduction of the ideal dose distribution. For each of 33 prostate SBRT patients, 18 treatment plans with different pre-set numbers of allowed beams were automatically generated with the proposed TBS-BAO. For each patient, the TBS-BAO plans were then compared to a plan that was automatically generated with an alternative BAO method (Erasmus-iCycle) and to a high-quality manually generated plan. TBS-BAO was able to automatically generate plans with clinically feasible numbers of beams (∼25), with a quality highly similar to corresponding 91-beam ideal reference plans. Compared to the alternative Erasmus-iCycle BAO approach, similar plan quality was obtained for 25-beam segmented plans, while computation times were reduced from 10.7 hours to 4.8/1.5 hours, depending on the applied pencil-beam resolution in TBS-BAO. 25-beam TBS-BAO plans had similar quality as manually generated plans with on average 48 beams, while delivery times reduced from 22.3 to 18.4/18.1 min. TBS reference plans could effectively steer the discrete non-convex BAO.

Comparing Multi-Objective Local Search Algorithms for the Beam Angle Selection Problem

In intensity-modulated radiation therapy, treatment planners aim to irradiate the tumour according to a medical prescription while sparing surrounding organs at risk as much as possible. Although this problem is inherently a multi-objective optimisation (MO) problem, most of the models in the literature are single-objective ones. For this reason, a large number of single-objective algorithms have been proposed in the literature to solve such single-objective models rather than multi-objective ones. Further, a difficulty that one has to face when solving the MO version of the problem is that the algorithms take too long before converging to a set of (approximately) non-dominated points. In this paper, we propose and compare three different strategies, namely random PLS (rPLS), judgement-function-guided PLS (jPLS) and neighbour-first PLS (nPLS), to accelerate a previously proposed Pareto local search (PLS) algorithm to solve the beam angle selection problem in IMRT. A distinctive feature of these strategies when compared to the PLS algorithms in the literature is that they do not evaluate their entire neighbourhood before performing the dominance analysis. The rPLS algorithm randomly chooses the next non-dominated solution in the archive and it is used as a baseline for the other implemented algorithms. The jPLS algorithm first chooses the non-dominated solution in the archive that has the best objective function value. Finally, the nPLS algorithm first chooses the solutions that are within the neighbourhood of the current solution. All these strategies prevent us from evaluating a large set of BACs, without any major impairment in the obtained solutions’ quality. We apply our algorithms to a prostate case and compare the obtained results to those obtained by the PLS from the literature. The results show that algorithms proposed in this paper reach a similar performance than PLS and require fewer function evaluations.

Implementation of automatic plan optimization in Italy: Status and perspectives

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.

A data-driven approach to optimal beam/arc angle selection for liver stereotactic body radiation therapy treatment planning

BACKGROUND

Stereotactic body radiation therapy (SBRT) for liver cancer has shown promising therapeutic effects. Effective treatment relies not only on the precise delivery provided by image-guided radiation therapy (IGRT) but also high dose gradient formed around the treatment volume to spare functional liver tissue, which is highly dependent on the beam/arc angle selection. In this study, we aim to develop a decision support model to learn human planner's beam navigation approach for beam angle/arc angle selection for liver SBRT.

METHODS

A total of 27 liver SBRT/HIGRT patients (10 IMRT, 17 VMAT/DCA) were included in this study. A dosimetric budget index was defined for each beam angle/control point considering dose penetration through the patient body and liver tissue. Optimal beam angle setting (beam angles for IMRT and start/terminal angle for VMAT/DCA) was determined by minimizing the loss function defined as the sum of total dosimetric budget index and beam span penalty function. Leave-one-out validation was exercised on all 27 cases while weighting coefficients in the loss function was tuned in nested cross validation. To compare the efficacy of the model, a model plan was generated using automatically generated beam setting while retaining the original optimization constraints in the clinical plan. Model plan was normalized to the same planning target volume (PTV) V100% as the clinical plans. Dosimetric endpoints including PTV D98%, D2%, liver V20Gy and total MU were compared between two plan groups. Wilcoxon Signed-Rank test was performed with the null hypothesis being that no difference exists between two plan groups.

RESULTS

Beam setting prediction was instantaneous. Mean PTV D98% was 91.3% and 91.3% (P=0.566), while mean PTV D2% was 107.9% and 108.1% (P=0.164) for clinical plan and model plan respectively. Liver V20Gy showed no significant difference (P=0.590) with 23.3% for clinical plan and 23.4% for the model plan. Total MU is comparable (P=0.256) between the clinical plan (avg. 2,389.6) and model plan (avg. 2,319.6).

CONCLUSIONS

The evidence driven beam setting model yielded similar plan quality as hand-crafted clinical plan. It is capable of capturing human's knowledge in beam selection decision making. This model could facilitate decision making for beam angle selection while eliminating lengthy trial-and-error process of adjusting beam setting during liver SBRT treatment planning.

Conic formulation of fluence map optimization problems

The convexity of objectives and constraints in fluence map optimization (FMO) for radiation therapy has been extensively studied. Next to convexity, there is another important characteristic of optimization functions and problems, which has thus far not been considered in FMO literature: conic representation. Optimization problems that are conically representable using quadratic, exponential and power cones are solvable with advanced primal-dual interior-point algorithms. These algorithms guarantee an optimal solution in polynomial time and have good performance in practice. In this paper, we construct conic representations for most FMO objectives and constraints. This paper is the first that shows that FMO problems containing multiple biological evaluation criteria can be solved in polynomial time. For fractionation-corrected functions for which no exact conic reformulation is found, we provide an accurate approximation that is conically representable. We present numerical results on the TROTS data set, which demonstrate very stable numerical performance for solving FMO problems in conic form. With ongoing research in the optimization community, improvements in speed can be expected, which makes conic optimization a promising alternative for solving FMO problems.

Evaluation of alternative parameter settings for dose restoration and full plan adaptation in IMPT for prostate cancer

BACKGROUND/PURPOSE

Intensity-modulated proton therapy is highly sensitive to anatomical variations. A dose restoration method and a full plan adaptation method have been developed earlier, both requiring several parameter settings. This study evaluates the validity of the previously selected settings by systematically comparing them to alternatives.

MATERIALS/METHODS

The dose restoration method takes a prior plan and uses an energy-adaptation followed by a spot-intensity re-optimization to restore the plan to its initial state. The full adaptation method uses an energy-adaptation followed by the addition of new spots and a spot-intensity optimization to fit the new anatomy. We varied: 1) The margins and robustness settings of the prior plan, 2) the spot-addition sample size, i.e. the number of added spots, 3) the spot-addition stopping criterion, and 4) the spot-intensity optimization approach. The last three were evaluated only for the full plan adaptation. Evaluations were done on 88 CT scans of 11 prostate cancer patients. Dose was prescribed as 55 Gy(RBE) to the lymph nodes and seminal vesicles with a boost to 74 Gy(RBE) to the prostate.

RESULTS

For the dose restoration method, changing the applied CTV-to-PTV margins and plan robustness in the prior plans yielded insufficient target coverage or increased OAR doses. For the full plan adaptation, more spot-addition iterations and using a different optimization approach resulted in lower OAR doses compared to the default settings while maintaining target coverage. However, the calculation times increased by up to 20 times, making these variations infeasible for online-adaptation.

CONCLUSION

We recommend maintaining the default setting for the dose restoration approach. For the full plan adaptation we recommend to focus on fine-tuning the optimization-parameters, and apart from this using the default settings.

How should we model and evaluate breathing interplay effects in IMPT?

Breathing interplay effects in Intensity Modulated Proton Therapy (IMPT) arise from the interaction between target motion and the scanning beam. Assessing the detrimental effect of interplay and the clinical robustness of several mitigation techniques requires statistical evaluation procedures that take into account the variability of breathing during dose delivery. In this study, we present such a statistical method to model intra-fraction respiratory motion based on breathing signals and assess clinical relevant aspects related to the practical evaluation of interplay in IMPT such as how to model irregular breathing, how small breathing changes affect the final dose distribution, and what is the statistical power (number of different scenarios) required for trustworthy quantification of interplay effects. First, two data-driven methodologies to generate artificial patient-specific breathing signals are compared: a simple sinusoidal model, and a precise probabilistic deep learning model generating very realistic samples of patient breathing. Second, we investigate the highly fluctuating relationship between interplay doses and breathing parameters, showing that small changes in breathing period result in large local variations in the dose. Our results indicate that using a limited number of samples to calculate interplay statistics introduces a bigger error than using simple sinusoidal models based on patient parameters or disregarding breathing hysteresis during the evaluation. We illustrate the power of the presented statistical method by analyzing interplay robustness of 4DCT and Internal Target Volume (ITV) treatment plans for a 8 lung cancer patients, showing that, unlike 4DCT plans, even 33 fraction ITV plans systematically fail to fulfill robustness requirements.

Enhancing Radiotherapy for Locally Advanced Non-Small Cell Lung Cancer Patients with iCE, a Novel System for Automated Multi-Criterial Treatment Planning Including Beam Angle Optimization

In this study, the novel iCE radiotherapy treatment planning system (TPS) for automated multi-criterial planning with integrated beam angle optimization (BAO) was developed, and applied to optimize organ at risk (OAR) sparing and systematically investigate the impact of beam angles on radiotherapy dose in locally advanced non-small cell lung cancer (LA-NSCLC). iCE consists of an in-house, sophisticated multi-criterial optimizer with integrated BAO, coupled to a broadly used commercial TPS. The in-house optimizer performs fluence map optimization to automatically generate an intensity-modulated radiotherapy (IMRT) plan with optimal beam angles for each patient. The obtained angles and dose-volume histograms are then used to automatically generate the final deliverable plan with the commercial TPS. For the majority of 26 LA-NSCLC patients, iCE achieved improved heart and esophagus sparing compared to the manually created clinical plans, with significant reductions in the median heart D_mean (8.1 vs. 9.0 Gy, p = 0.02) and esophagus D_mean (18.5 vs. 20.3 Gy, p = 0.02), and reductions of up to 6.7 Gy and 5.8 Gy for individual patients. iCE was superior to automated planning using manually selected beam angles. Differences in the OAR doses of iCE plans with 6 beams compared to 4 and 8 beams were statistically significant overall, but highly patient-specific. In conclusion, automated planning with integrated BAO can further enhance and individualize radiotherapy for LA-NSCLC.

Outcome-based multiobjective optimization of lymphoma radiation therapy plans

At its core, radiation therapy (RT) requires balancing therapeutic effects against risk of adverse events in cancer survivors. The radiation oncologist weighs numerous disease and patient-level factors when considering the expected risk-benefit ratio of combined treatment modalities. As part of this, RT plan optimization software is used to find a clinically acceptable RT plan delivering a prescribed dose to the target volume while respecting pre-defined radiation dose-volume constraints for selected organs at risk. The obvious limitation to the current approach is that it is virtually impossible to ensure the selected treatment plan could not be bettered by an alternative plan providing improved disease control and/or reduced risk of adverse events in this individual. Outcome-based optimization refers to a strategy where all planning objectives are defined by modeled estimates of a specific outcome's probability. Noting that various adverse events and disease control are generally incommensurable, leads to the concept of a Pareto-optimal plan: a plan where no single objective can be improved without degrading one or more of the remaining objectives. Further benefits of outcome-based multiobjective optimization are that quantitative estimates of risks and benefit are obtained as are the effects of choosing a different trade-off between competing objectives. Furthermore, patient-level risk factors and combined treatment modalities may be integrated directly into plan optimization. Here, we present this approach in the clinical setting of multimodality therapy for malignant lymphoma, a malignancy with marked heterogeneity in biology, target localization, and patient characteristics. We discuss future research priorities including the potential of artificial intelligence.

Variations in Head and Neck Treatment Plan Quality Assessment Among Radiation Oncologists and Medical Physicists in a Single Radiotherapy Department

Background

Agreement between planners and treating radiation oncologists (ROs) on plan quality criteria is essential for consistent planning. Differences between ROs and planning medical physicists (MPs) in perceived quality of head and neck cancer plans were assessed.

Materials and Methods

Five ROs and four MPs scored 65 plans for in total 15 patients. For each patient, the clinical (CLIN) plan and two or four alternative plans, generated with automated multi-criteria optimization (MCO), were included. There was always one MCO plan aiming at maximally adhering to clinical plan requirements, while the other MCO plans had a lower aimed quality. Scores were given as follows: 1-7 and 1-2, not acceptable; 3-5, acceptable if further planning would not resolve perceived weaknesses; and 6-7, straightway acceptable. One MP and one RO repeated plan scoring for intra-observer variation assessment.

Results

For the 36 unique observer pairs, the median percentage of plans for which the two observers agreed on a plan score (100% = 65 plans) was 27.7% [6.2, 40.0]. In the repeat scoring, agreements between first and second scoring were 52.3% and 40.0%, respectively. With a binary division between unacceptable (scores 1 and 2) and acceptable (3-7) plans, the median inter-observer agreement percentage was 78.5% [63.1, 86.2], while intra-observer agreements were 96.9% and 86.2%. There were no differences in observed agreements between RO-RO, MP-MP, and RO-MP pairs. Agreements for the highest-quality, automatically generated MCO plans were higher than for the CLIN plans.

Conclusions

Inter-observer differences in plan quality scores were substantial and could result in inconsistencies in generated treatment plans. Agreements among ROs were not better than between ROs and MPs, despite large differences in training and clinical role. High-quality automatically generated plans showed the best score agreements.

MR-Linac Radiotherapy - The Beam Angle Selection Problem

Background

With the large-scale introduction of volumetric modulated arc therapy (VMAT), selection of optimal beam angles for coplanar static-beam IMRT has increasingly become obsolete. Due to unavailability of VMAT in current MR-linacs, the problem has re-gained importance. An application for automated IMRT treatment planning with integrated, patient-specific computer-optimization of beam angles (BAO) was used to systematically investigate computer-aided generation of beam angle class solutions (CS) for replacement of computationally expensive patient-specific BAO. Rectal cancer was used as a model case.

Materials and Methods

23 patients treated at a Unity MR-linac were included. BAO_x plans (x=7-12 beams) were generated for all patients. Analyses of BAO₁₂ plans resulted in CS_x class solutions. BAO_x plans, CS_x plans, and plans with equi-angular setups (EQUI_x, x=9-56) were mutually compared.

Results

For x>7, plan quality for CS_x and BAO_x was highly similar, while both were superior to EQUI_x. E.g. with CS₉, bowel/bladder D_mean reduced by 22% [11%, 38%] compared to EQUI₉ (p<0.001). For equal plan quality, the number of EQUI beams had to be doubled compared to BAO and CS.

Conclusions

Computer-generated beam angle CS could replace individualized BAO without loss in plan quality, while reducing planning complexity and calculation times, and resulting in a simpler clinical workflow. CS and BAO largely outperformed equi-angular treatment. With the developed CS, time consuming beam angle re-optimization in daily adaptive MR-linac treatment could be avoided. Further systematic research on computerized development of beam angle class solutions for MR-linac treatment planning is warranted.

Comparison of non-coplanar optimization of static beams and arc trajectories for intensity-modulated treatments of meningioma cases

Two methods for non-coplanar beam direction optimization, one for static beams and another for arc trajectories, were proposed for intracranial tumours. The results of the beam angle optimizations were compared with the beam directions used in the clinical plans. Ten meningioma cases already treated were selected for this retrospective planning study. Algorithms for non-coplanar beam angle optimization (BAO) and arc trajectory optimization (ATO) were used to generate the corresponding plans. A plan quality score, calculated by a graphical method for plan assessment and comparison, was used to guide the beam angle optimization process. For each patient, the clinical plans (CLIN), created with the static beam orientations used for treatment, and coplanar VMAT approximated plans (VMAT) were also generated. To make fair plan comparisons, all plan optimizations were performed in an automated multicriteria calculation engine and the dosimetric plan quality was assessed. BAO and ATO plans presented, on average, moderate global plan score improvements over VMAT and CLIN plans. Nevertheless, while BAO and CLIN plans assured a more efficient OARs sparing, the ATO and VMAT plans presented a higher coverage and conformity of the PTV. Globally, all plans presented high-quality dose distributions. No statistically significant quality differences were found, on average, between BAO, ATO and CLIN plans. However, automated plan solution optimizations (BAO or ATO) may improve plan generation efficiency and standardization. In some individual patients, plan quality improvements were achieved with ATO plans, demonstrating the possible benefits of this automated optimized delivery technique.

Fully automated noncoplanar radiation therapy treatment planning

PURPOSE

To perform fully automated noncoplanar (NC) treatment planning, we propose a method called NC-POPS to produce NC plans using the Pareto optimal projection search (POPS) algorithm.

METHODS

NC radiation therapy treatment planning has the potential to improve dosimetric quality as compared to traditional coplanar techniques. Likewise, automated treatment planning algorithms can reduce a planner's active treatment planning time and remove inter-planner variability. Our NC-POPS algorithm extends the original POPS algorithm to the NC setting with potential applications to both intensity-modulated radiation therapy (IMRT) and volumetric modulated arc therapy (VMAT). The proposed algorithm consists of two main parts: (1) NC beam angle optimization (BAO) and (2) fully automated inverse planning using the POPS algorithm.

RESULTS

We evaluate the performance of NC-POPS by comparing between various NC and coplanar configurations. To evaluate plan quality, we compute the homogeneity index (HI), conformity index (CI), and dose-volume histogram statistics for various organs-at-risk (OARs). As compared to the evaluated coplanar baseline methods, the proposed NC-POPS method achieves significantly better OAR sparing, comparable or better dose conformity, and similar dose homogeneity.

CONCLUSIONS

Our proposed NC-POPS algorithm provides a modular approach for fully automated treatment planning of NC IMRT cases with the potential to substantially improve treatment planning workflow and plan quality.

Replacing Manual Planning of Whole Breast Irradiation With Knowledge-Based Automatic Optimization by Virtual Tangential-Fields Arc Therapy

Purpose

To implement Knowledge Based (KB) automatic planning for right and left-sided whole breast treatment through a new volumetric technique (ViTAT, Virtual Tangential-fields Arc Therapy) mimicking conventional tangential fields (TF) irradiation.

Materials and Method

A total of 193 clinical plans delivering TF with wedged or field-in-field beams were selected to train two KB-models for right(R) and left(L) sided breast cancer patients using the RapidPlan (RP) tool implemented in the Varian Eclipse system. Then, a template for ViTAT optimization, incorporating individual KB-optimized constraints, was interactively fine-tuned. ViTAT plans consisted of four arcs (6 MV) with start/stop angles consistent with the TF geometry variability within our population; the delivery was completely blocked along the arcs, apart from the first and last 20° of rotation for each arc. Optimized fine-tuned KB templates for automatic plan optimization were generated. Validation tests were performed on 60 new patients equally divided in R and L breast treatment: KB automatic ViTAT-plans (KB-ViTAT) were compared against the original TF plans in terms of OARs/PTVs dose-volume parameters. Wilcoxon-tests were used to assess the statistically significant differences.

Results

KB models were successfully generated for both L and R sides. Overall, 1(3%) and 7(23%) out of 30 automatic KB-ViTAT plans were unacceptable compared to TF for R and L side, respectively. After the manual refinement of the start/stop angles, KB-ViTAT plans well fitted TF-performances for these patients as well. PTV coverage was comparable, while PTV D_1% was improved with KB-ViTAT by R:0.4/L:0.2 Gy (p < 0.05); ipsilateral OARs D_mean were similar with a slight (i.e., few % volume) improvement/worsening in the 15–35 Gy/2–15 Gy range, respectively. KB-ViTAT better spared contralateral OARs: D_mean of contralateral OARs was 0.1 Gy lower (p < 0.05); integral dose was R:5%/L:8% lower (p < 0.05) than TF. The overall time for the automatic plan optimization and final dose calculation was 12 ± 2 minutes.

Conclusions

Fully automatic KB-optimization of ViTAT can efficiently replace manually optimized TF planning for whole breast irradiation. This approach was clinically implemented in our institute and may be suggested as a large-scale strategy for efficiently replacing manual planning with large sparing of time, elimination of inter-planner variability and of, seldomly occurring, sub-optimal manual plans.

A reinforcement learning application of a guided Monte Carlo Tree Search algorithm for beam orientation selection in radiation therapy

Current beam orientation optimization algorithms for radiotherapy, such as column generation (CG), are typically heuristic or greedy in nature because of the size of the combinatorial problem, which leads to suboptimal solutions. We propose a reinforcement learning strategy using a Monte Carlo Tree Search (MCTS) that can find a better beam orientation set in less time than CG. We utilize a reinforcement learning structure involving a supervised learning network to guide the MCTS and to explore the decision space of beam orientation selection problems. We previously trained a deep neural network (DNN) that takes in the patient anatomy, organ weights, and current beams, then approximates beam fitness values to indicate the next best beam to add. Here, we use this DNN to probabilistically guide the traversal of the branches of the Monte Carlo decision tree to add a new beam to the plan. To assess the feasibility of the algorithm, we used a test set of 13 prostate cancer patients, distinct from the 57 patients originally used to train and validate the DNN, to solve five-beam plans. To show the strength of the guided MCTS (GTS) compared to other search methods, we also provided the performances of Guided Search, Uniform Tree Search and Random Search algorithms. On average, GTS outperformed all the other methods. It found a better solution than CG in 237 s on average, compared to 360 s for CG, and outperformed all other methods in finding a solution with a lower objective function value in less than 1000 s. Using our GTS method, we could maintain planning target volume (PTV) coverage within 1% error similar to CG, while reducing the organ-at-risk mean dose for body, rectum, left and right femoral heads; the mean dose to bladder was 1% higher with GTS than with CG.

Comparison of volumetric modulated arc therapy and intensity-modulated radiotherapy for left-sided whole-breast irradiation using automated planning

Background

Published treatment technique comparisons for postoperative left-sided whole breast irradiation (WBI) with deep-inspiration breath-hold (DIBH) are scarce, small, and inconclusive. In this study, fully automated multi-criterial plan optimization, generating a single high-quality, Pareto-optimal plan per patient and treatment technique, was used to compare for a large patient cohort 1) intensity modulated radiotherapy (IMRT) with two tangential fields and 2) volumetric modulated arc therapy (VMAT) with two small tangential subarcs.

Materials and methods

Forty-eight randomly selected patients recently treated with DIBH and 16 × 2.66 Gy were included. The optimizer was configured for the clinical planning protocol. Comparisons between IMRT and VMAT included dosimetric plan parameters, estimated excess relative risks (ERR) for toxicities, delivery times, MUs, and deliverability accuracy at a linac.

Results

The automatically generated IMRT and VMAT plans applied in this study were similar or higher in quality than the manually generated clinical plans. For equal PTVin V95% (98.4 ± 0.9%), VMAT had significant advantages compared to IMRT regarding breast dose homogeneity and doses in heart and ipsilateral lung, at the cost of some minor deteriorations for contralateral breast (few cases with larger deteriorations) and lung. Conformality improved from 1.38 to 1.18 (p < 0.001). With VMAT, ERR for major coronary events and ipsilateral lung tumors were reduced by 3% (range: −1–12%) and 16% (range: −3–38%), respectively. MUs and delivery times were higher for VMAT. There were no statistical differences in γ passing rates.

Conclusion

For WBI in conservative therapy of left-sided breast patients treated with DIBH, VMAT with two tangential subarcs was generally dosimetrically superior to IMRT with two tangential static fields. Results need confirmation by robustness analyses.

Assessing the need for adaptive radiotherapy in head and neck cancer patients using an automatic planning tool

Background

Unbiased analysis of the impact of adaptive radiotherapy (ART) is necessary to evaluate dosimetric benefit and optimize clinics' workflows. The aim of the study was to assess the need for adaptive radiotherapy (ART) in head and neck (H&N) cancer patients using an automatic planning tool in a retrospective planning study.

Materials and methods

Thirty H&N patients treated with adaptive radiotherapy were analysed. Patients had a CT scan for treatment planning and a verification CT during treatment according to the clinic's protocol. Considering these images, three plans were retrospectively generated using the iCycle tool to simulate the scenarios with and without adaptation: 1) the optimized plan based on the planning CT; 2) the optimized plan based on the verification CT (ART-plan); 3) the plan obtained by considering treatment plan 1 re-calculated in the verification CT (non-ART plan). The dosimetric endpoints for both target volumes and OAR were compared between scenarios 2 and 3 and the SPIDERplan used to evaluate plan quality.

Results

The most significant impact of ART was found for the PTVs, which demonstrated decreased D98% in the non-ART plan. A general increase in the dose was observed for the OAR but only the spinal cord showed a statistical significance. The SPIDERplan analysis indicated an overall loss of plan quality in the absence of ART.

Conclusion

These results confirm the advantages of ART in H&N patients, especially for the coverage of target volumes. The usage of an automatic planning tool reduces planner-induced bias in the results, guaranteeing that the observed changes derive from the application of ART.

Fully automated treatment planning for MLC-based robotic radiotherapy

PURPOSE

To propose and validate a fully automated multicriterial treatment planning solution for a CyberKnife® equipped with an InCise^TM 2 multileaf collimator.

METHODS

The AUTO BAO plans are generated using fully automated prioritized multicriterial optimization (AUTO MCO) of pencil-beam fluence maps with integrated noncoplanar beam angle optimization (BAO), followed by MLC segment generation. Both the AUTO MCO and segmentation algorithms have been developed in-house. AUTO MCO generates for each patient a single, high-quality Pareto-optimal IMRT plan. The segmentation algorithm then accurately mimics the AUTO MCO 3D dose distribution, while considering all candidate beams simultaneously, rather than replicating the fluence maps. Pencil-beams, segment dose depositions, and final dose calculations are performed with a stand-alone version of the clinical dose calculation engine. For validation, AUTO BAO plans were generated for 33 prostate SBRT patients and compared to reference plans (REF) that were manually generated with the commercial treatment planning system (TPS), in absence of time pressure. REF plans were also compared to AUTO RB plans, for which fluence map optimization was performed for the beam angle configuration used in the REF plan, and the segmentation could use all these beams or only a subset, depending on the dosimetry.

RESULTS

AUTO BAO plans were clinically acceptable and dosimetrically similar to REF plans, but had on average reduced numbers of beams ((beams in AUTO BAO)/(beams in REF) (relative improvement): 24.7/48.3 (-49%)), segments (59.5/98.9 (-40%)), and delivery times (17.1/22.3 min. (-23%)). Dosimetry of AUTO RB and REF were also similar, but AUTO RB used on average fewer beams (38.0/48.3 (-21%)) and had on average shorter delivery times (18.6/22.3 min. (-17%)). Delivered Monitor Units (MU) were similar for all three planning approaches.

CONCLUSIONS

A new, vendor-independent optimization workflow for fully automated generation of deliverable high-quality CyberKnife® plans was proposed, including BAO. Compared to manual planning with the commercial TPS, fraction delivery times were reduced by 5.3 min. (-23%) due to large reductions in beam and segment numbers.

Evaluation of fully automated a priori MCO treatment planning in VMAT for head-and-neck cancer

PURPOSE

Automated planning techniques aim to reduce manual planning time and inter-operator variability without compromising the plan quality which is particularly challenging for head-and-neck (HN) cancer radiotherapy. The objective of this study was to evaluate the performance of an a priori-multicriteria plan optimization algorithm on a cohort of HN patients.

METHODS

A total of 14 nasopharyngeal carcinoma (upper-HN) and 14 "middle-lower indications" (lower-HN) previously treated in our institution were enrolled in this study. Automatically generated plans (autoVMAT) were compared to manual VMAT or Helical Tomotherapy planning (manVMAT-HT) by assessing differences in dose delivered to targets and organs at risk (OARs), calculating plan quality indexes (PQIs) and performing blinded comparisons by clinicians. Quality control of the plans and measurements of the delivery times were also performed.

RESULTS

For the 14 lower-HN patients, with equivalent planning target volume (PTV) dosimetric criteria and dose homogeneity, significant decrease in the mean doses to the oral cavity, esophagus, trachea and larynx were observed for autoVMAT compared to manVMAT-HT. Regarding the 14 upper-HN cases, the PTV coverage was generally significantly superior for autoVMAT which was also confirmed with higher calculated PQIs on PTVs for 13 out of 14 patients, whereas PQIs calculated on OARs were generally equivalent. Number of MUs and total delivery time were significantly higher for autoVMAT compared to manVMAT. All plans were considered clinically acceptable by clinicians.

CONCLUSIONS

Overall superiority of autoVMAT compared to manVMAT-HT plans was demonstrated for HN cancer. The obtained plans were operator-independent and required no post-optimization or manual intervention.

Personalized Treatment Planning Automation in Prostate Cancer Radiation Oncology: A Comprehensive Dosimetric Study

Background

In radiation oncology, automation of treatment planning has reported the potential to improve plan quality and increase planning efficiency. We performed a comprehensive dosimetric evaluation of the new Personalized algorithm implemented in Pinnacle³ for full planning automation of VMAT prostate cancer treatments.

Material and Methods

Thirteen low-risk prostate (without lymph-nodes irradiation) and 13 high-risk prostate (with lymph-nodes irradiation) treatments were retrospectively taken from our clinical database and re-optimized using two different automated engines implemented in the Pinnacle treatment system. These two automated engines, the currently used Autoplanning and the new Personalized are both template-based algorithms that use a wish-list to formulate the planning goals and an iterative approach able to mimic the planning procedure usually adopted by experienced planners. In addition, the new Personalized module integrates a new engine, the Feasibility module, able to generate an “a priori” DVH prediction of the achievability of planning goals. Comparison between clinically accepted manually generated (MP) and automated plans generated with both Autoplanning (AP) and Personalized engines (Pers) were performed using dose-volume histogram metrics and conformity indexes. Three different normal tissue complication probabilities (NTCPs) models were used for rectal toxicity evaluation. The planning efficiency and the accuracy of dose delivery were assessed for all plans.

Results

For similar targets coverage, Pers plans reported a significant increase of dose conformity and less irradiation of healthy tissue, with significant dose reduction for rectum, bladder, and femurs. On average, Pers plans decreased rectal mean dose by 11.3 and 8.3 Gy for low-risk and high-risk cohorts, respectively. Similarly, the Pers plans decreased the bladder mean doses by 7.3 and 7.6 Gy for low-risk and high-risk cohorts, respectively. The integral dose was reduced by 11–16% with respect to MP plans. Overall planning times were dramatically reduced to about 7 and 15 min for Pers plans. Despite the increased complexity, all plans passed the 3%/2 mm γ-analysis for dose verification.

Conclusions

The Personalized engine provided an overall increase of plan quality, in terms of dose conformity and sparing of normal tissues for prostate cancer patients. The Feasibility “a priori” DVH prediction module provided OARs dose sparing well beyond the clinical objectives. The new Pinnacle Personalized algorithms outperformed the currently used Autoplanning ones as solution for treatment planning automation.

On the Importance of Individualized, Non-Coplanar Beam Configurations in Mediastinal Lymphoma Radiotherapy, Optimized With Automated Planning

Background and Purpose

Literature is non-conclusive regarding selection of beam configurations in radiotherapy for mediastinal lymphoma (ML) radiotherapy, and published studies are based on manual planning with its inherent limitations. In this study, coplanar and non-coplanar beam configurations were systematically compared, using a large number of automatically generated plans.

Material and Methods

An autoplanning workflow, including beam configuration optimization, was configured for young female ML patients. For each of 25 patients, 24 plans with different beam configurations were generated with autoplanning: 11 coplanar CP_x plans and 11 non-coplanar NCP_x plans with x = 5 to 15 IMRT beams with computer-optimized, patient-specific configurations, and the coplanar VMAT and non-coplanar Butterfly VMAT (B-VMAT) beam angle class solutions (600 plans in total).

Results

Autoplans compared favorably with manually generated, clinically delivered plans, ensuring that beam configuration comparisons were performed with high quality plans. There was no beam configuration approach that was best for all patients and all plan parameters. Overall there was a clear tendency towards higher plan quality with non-coplanar configurations (NCP_x≥12 and B-VMAT). NCP_x≥12 produced highly conformal plans with on average reduced high doses in lungs and patient and also a reduced heart Dmean, while B-VMAT resulted in reduced low-dose spread in lungs and left breast.

Conclusions

Non-coplanar beam configurations were favorable for young female mediastinal lymphoma patients, with patient-specific and plan-parameter-dependent dosimetric advantages of NCP_x≥12 and B-VMAT. Individualization of beam configuration approach, considering also the faster delivery of B-VMAT vs. NCP_x≥12, can importantly improve the treatments.

Dosimetric Impact of Intrafraction Motion in Online-Adaptive Intensity Modulated Proton Therapy for Cervical Cancer

PURPOSE

A method was recently developed for online-adaptive intensity modulated proton therapy (IMPT) in patients with cervical cancer. The advantage of this approach, relying on the use of tight margins, is challenged by the intrafraction target motion. The purpose of this study was to evaluate the dosimetric effect of intrafraction motion on the target owing to changes in bladder filling in patients with cervical cancer treated with online-adaptive IMPT.

METHODS AND MATERIALS

In 10 patients selected to have large uterus motion induced by bladder filling, the intrafraction anatomic changes were simulated for several prefraction durations for online (automated) contouring and planning. For each scenario, the coverage of the primary target was evaluated with margins of 2.5 and 5 mm.

RESULTS

Using a 5- mm planning target volume margin, median accumulated D98% was greater than 42.75 Gy_RBE1.1 (95% of the prescribed dose) in the case of a prefraction duration of 5 and 10 minutes. For a prefraction duration of 15 minutes, this parameter deteriorated to 42.6 Gy_RBE1.1. When margins were reduced to 2.5 mm, only a 5-minute duration resulted in median target D98% above 42.75 Gy_RBE1.1. In addition, smaller bladders were found to be associated with larger dose degradations compared with larger bladders.

CONCLUSIONS

This study indicates that intrafraction anatomic changes can have a substantial dosimetric effect on target coverage in an online-adaptive IMPT scenario for patients subject to large uterus motion. A margin of 5 mm was sufficient to compensate for the intrafraction motion due to bladder filling for up to 10 minutes of prefraction time. However, compensation for the uncertainties that were disregarded in this study, by using margins or robust optimization, is also required. Furthermore, a large bladder volume restrains intrafraction target motion and is recommended for treating patients in this scenario. Assuming that online-adaptive IMPT remains beneficial as long as narrow margins are used (5 mm or below), this study demonstrates its feasibility with regard to intrafraction motion.

Complementing Prostate SBRT VMAT With a Two-Beam Non-Coplanar IMRT Class Solution to Enhance Rectum and Bladder Sparing With Minimum Increase in Treatment Time

Purpose

Enhance rectum and bladder sparing in prostate SBRT with minimum increase in treatment time by complementing dual-arc coplanar VMAT with a two-beam non-coplanar IMRT class solution (CS).

Methods

For twenty patients, an optimizer for automated multi-criterial planning with integrated beam angle optimization (BAO) was used to generate dual-arc VMAT plans, supplemented with five non-coplanar IMRT beams with individually optimized orientations (VMAT+5). In all plan generations, reduction of high rectum dose had the highest priority after obtaining adequate PTV coverage. A CS with two most preferred directions in VMAT+5 and largest rectum dose reductions compared to dual-arc VMAT was then selected to define VMAT+CS. VMAT+CS was compared with automatically generated i) dual-arc coplanar VMAT plans (VMAT), ii) VMAT+5 plans, and iii) IMRT plans with 30 patient-specific non-coplanar beam orientations (30-NCP). Plans were generated for a 4 x 9.5 Gy fractionation scheme. Differences in PTV doses, healthy tissue sparing, and computation and treatment delivery times were quantified.

Results

For equal PTV coverage, VMAT+CS, consisting of dual-arc VMAT supplemented with two fixed, non-coplanar IMRT beams with fixed Gantry/Couch angles of 65°/30° and 295°/-30°, significantly reduced OAR doses and the dose bath, compared to dual-arc VMAT. Mean relative differences in rectum D_mean, D_1cc, V_40GyEq and V_60GyEq were 19.4 ± 10.6%, 4.2 ± 2.7%, 34.9 ± 20.3%, and 39.7 ± 23.2%, respectively (all p<0.001). There was no difference in bladder D_1cc, while bladder D_mean reduced by 17.9 ± 11.0% (p<0.001). Also, the clinically evaluated urethra D_5%, D_10%, and D_50% showed small, but statistically significant improvements. All patient V_X with X = 2, 5, 10, 20, and 30 Gy were reduced with VMAT+CS, with a maximum relative reduction for V_10Gy of 19.0 ± 7.3% (p<0.001). Total delivery times with VMAT+CS only increased by 1.9 ± 0.7 min compared to VMAT (9.1 ± 0.7 min). The dosimetric quality of VMAT+CS plans was equivalent to VMAT+5, while optimization times were reduced by a factor of 25 due to avoidance of individualized BAO. Compared to VMAT+CS, the 30-NCP plans were only favorable in terms of dose bath, at the cost of much enhanced optimization and delivery times.

Conclusions

The proposed two-beam non-coplanar class solution to complement coplanar dual-arc VMAT resulted in substantial plan quality improvements for OARs (especially rectum) and reduced irradiated patient volumes with minor increases in treatment delivery times.

Knowledge Models as Teaching Aid for Training Intensity Modulated Radiation Therapy Planning: A Lung Cancer Case Study

Purpose: Artificial intelligence (AI) employs knowledge models that often behave as a black-box to the majority of users and are not designed to improve the skill level of users. In this study, we aim to demonstrate the feasibility that AI can serve as an effective teaching aid to train individuals to develop optimal intensity modulated radiation therapy (IMRT) plans.

Methods and Materials: The training program is composed of a host of training cases and a tutoring system that consists of a front-end visualization module powered by knowledge models and a scoring system. The current tutoring system includes a beam angle prediction model and a dose-volume histogram (DVH) prediction model. The scoring system consists of physician chosen criteria for clinical plan evaluation as well as specially designed criteria for learning guidance. The training program includes six lung/mediastinum IMRT patients: one benchmark case and five training cases. A plan for the benchmark case is completed by each trainee entirely independently pre- and post-training. Five training cases cover a wide spectrum of complexity from easy (2), intermediate (1) to hard (2). Five trainees completed the training program with the help of one trainer. Plans designed by the trainees were evaluated by both the scoring system and a radiation oncologist to quantify planning quality.

Results: For the benchmark case, trainees scored an average of 21.6% of the total max points pre-training and improved to an average of 51.8% post-training. In comparison, the benchmark case's clinical plans score an average of 54.1% of the total max points. Two of the five trainees' post-training plans on the benchmark case were rated as comparable to the clinically delivered plans by the physician and all five were noticeably improved by the physician's standards. The total training time for each trainee ranged between 9 and 12 h.

Conclusion: This first attempt at a knowledge model based training program brought unexperienced planners to a level close to experienced planners in fewer than 2 days. The proposed tutoring system can serve as an important component in an AI ecosystem that will enable clinical practitioners to effectively and confidently use KBP.

Three-dimensional MRI-based treatment planning approach for non-invasive ocular proton therapy

PURPOSE

To develop a high-resolution three-dimensional (3D) magnetic resonance imaging (MRI)-based treatment planning approach for uveal melanomas (UM) in proton therapy.

MATERIALS/METHODS

For eight patients with UM, a segmentation of the gross tumor volume (GTV) and organs-at-risk (OARs) was performed on T1- and T2-weighted 7 Tesla MRI image data to reconstruct the patient MR-eye. An extended contour was defined with a 2.5-mm isotropic margin derived from the GTV. A broad beam algorithm, which we have called πDose, was implemented to calculate relative proton absorbed doses to the ipsilateral OARs. Clinically favorable gazing angles of the treated eye were assessed by calculating a global weighted-sum objective function, which set penalties for OARs and extreme gazing angles. An optimizer, which we have named OPT'im-Eye-Tool, was developed to tune the parameters of the functions for sparing critical-OARs.

RESULTS

In total, 441 gazing angles were simulated for every patient. Target coverage including margins was achieved in all the cases (V95%  > 95%). Over the whole gazing angles solutions space, maximum dose (Dmax ) to the optic nerve and the macula, and mean doses (Dmean ) to the lens, the ciliary body and the sclera were calculated. A forward optimization was applied by OPT'im-Eye-Tool in three different prioritizations: iso-weighted, optic nerve prioritized, and macula prioritized. In each, the function values were depicted in a selection tool to select the optimal gazing angle(s). For example, patient 4 had a T2 equatorial tumor. The optimization applied for the straight gazing angle resulted in objective function values of 0.46 (iso-weighted situation), 0.90 (optic nerve prioritization) and 0.08 (macula prioritization) demonstrating the impact of that angle in different clinical approaches.

CONCLUSIONS

The feasibility and suitability of a 3D MRI-based treatment planning approach have been successfully tested on a cohort of eight patients diagnosed with UM. Moreover, a gaze-angle trade-off dose optimization with respect to OARs sparing has been developed. Further validation of the whole treatment process is the next step in the goal to achieve both a non-invasive and a personalized proton therapy treatment.

Comprehensive Comparison of Progressive Optimization Algorithm Based Automatic Plan and Manually Planned Treatment Technique for Cervical-Thoracic Esophageal Cancers

Purpose:

The purpose of the present study was first to apply the progressive optimization algorithm based automatic volumetric modulated arc therapy (POA-VMAT) technology to accelerate and improve the radiotherapy of cervicothoracic esophageal cancer (CTEC). We comprehensive analyze the feasibility, normal tissue complication probability (NTCP) and dosimetric results of POA-VMAT, manual based VMAT and step-shoot intensity-modulated radiation therapy (IMRT) plans in the treatment of CTEC.

Methods:

Sixty patients with CTEC with or without concomitant chemotherapy at our institution between 2017 and 2019 were retrospectively identified. The manual 7field-IMRT (7f-IMRT), Single-arc-VMAT and Double-arc-VMAT (Single-Arc/Double-Arc) plans were generated in all cases. The POA-VMAT was designed using the automatic dual-arc VMAT technology of Pinnacle³ 9.10 planning system based on progressive optimization algorithm. Specially, it includes the selection of treatment techniques, the running of automated planning scripts, and the evaluation of the final radiotherapy regimen. Subsequently, quantitative evaluation of plans was performed by means of standard dose–volume histograms, homogeneity index (HI) and conformity index (CI).

Results:

Target dose conformity of the 7f-IMRT plan was inferior to all plans, whereas the Double-Arc plan was slightly inferior to the POA-VMAT but superior to the Single-Arc and 7f-IMRT plan. The HI for 7f-IMRT, Single-Arc, Double-Arc and POA-VMAT were 0.17 ± 0.08, 0.28 ± 0.06, 0.29 ± 0.06 and 0.28 ± 0.03, respectively. For the NTCP results, there was significant statistical difference among POA-VMAT, IMRT and VMAT plans. The total MU was reduced by 48.3% and 42.1% in Single-Arc and POA-VMAT plans compare to IMRT plans.

Conclusions:

By comprehensive consideration, POA-VMAT efficiently generate acceptable treatment plans for CTEC without dose escalation to OARs and overall superior to manual planning which is a good option for treating CTEC.

Offline Quality Assurance for Intensity Modulated Radiation Therapy Treatment Plans for NRG-HN001 Head and Neck Clinical Trial Using Knowledge-Based Planning

PURPOSE

This study aimed to investigate whether a disease site-specific, multi-institutional knowledge based-planning (KBP) model can improve the quality of intensity modulated radiation therapy treatment planning for patients enrolled in the head and neck NRG-HN001clinical trial and to establish a threshold of improvements of treatment plans submitted to the clinical trial.

METHODS AND MATERIALS

Fifty treatment plans for patients enrolled in the NRG-HN001 clinical trial were used to build a KBP model; the model was then used to reoptimize 50 other plans. We compared the dosimetric parameters of the submitted and KBP reoptimized plans. We compared differences between KBP and submitted plans for single- and multi-institutional treatment plans.

RESULTS

Mean values for the dose received by 95% of the planning target volume (PTV_6996) and for the maximum dose (D0.03cc) of PTV_6996 were 0.5 Gy and 2.1 Gy higher in KBP plans than in the submitted plans, respectively. Mean values for D0.03cc to the brain stem, spinal cord, optic nerve_R, optic nerve_L, and chiasm were 2.5 Gy, 1.9 Gy, 6.4 Gy, 6.6 Gy, and 5.7 Gy lower in the KBP plans than in the submitted plans. Mean values for Dmean to parotid_R and parotid_L glands were 2.2 Gy and 3.8 Gy lower in KBP plans, respectively. In 33 out of 50 KBP plans, we observed improvements in sparing of at least 7 organs at risk (OARs) (brain stem, spinal cord, optic nerves (R & L), chiasm, and parotid glands [R & L]). A threshold of improvement of OARs sparing of 5% of the prescription dose was established for providing the quality assurance results back to the treating institution.

CONCLUSIONS

A disease site-specific, multi-institutional, clinical trial-based KBP model improved sparing of OARs in a large number of reoptimized plans submitted to the NRG-HN001 clinical trial, and the model is being used as an offline quality assurance tool.

Using deep learning to predict beam-tunable Pareto optimal dose distribution for intensity-modulated radiation therapy

PURPOSE

Many researchers have developed deep learning models for predicting clinical dose distributions and Pareto optimal dose distributions. Models for predicting Pareto optimal dose distributions have generated optimal plans in real time using anatomical structures and static beam orientations. However, Pareto optimal dose prediction for intensity-modulated radiation therapy (IMRT) prostate planning with variable beam numbers and orientations has not yet been investigated. We propose to develop a deep learning model that can predict Pareto optimal dose distributions by using any given set of beam angles, along with patient anatomy, as input to train the deep neural networks. We implement and compare two deep learning networks that predict with two different beam configuration modalities.

METHODS

We generated Pareto optimal plans for 70 patients with prostate cancer. We used fluence map optimization to generate 500 IMRT plans that sampled the Pareto surface for each patient, for a total of 35 000 plans. We studied and compared two different models, Models I and II. Although they both used the same anatomical structures - including the planning target volume (PTV), organs at risk (OARs), and body - these models were designed with two different methods for representing beam angles. Model I directly uses beam angles as a second input to the network as a binary vector. Model II converts the beam angles into beam doses that are conformal to the PTV. We divided the 70 patients into 54 training, 6 validation, and 10 testing patients, thus yielding 27 000 training, 3000 validation, and 5000 testing plans. Mean square loss (MSE) was taken as the loss function. We used the Adam optimizer with a default learning rate of 0.01 to optimize the network's performance. We evaluated the models' performance by comparing their predicted dose distributions with the ground truth (Pareto optimal) dose distribution, in terms of dose volume histogram (DVH) plots and evaluation metrics such as PTV D98 , D95 , D50 , D2 , Dmax , Dmean , Paddick Conformation Number, R50, and Homogeneity index.

RESULTS

Our deep learning models predicted voxel-level dose distributions that precisely matched the ground truth dose distributions. The DVHs generated also precisely matched the ground truth. Evaluation metrics such as PTV statistics, dose conformity, dose spillage (R50), and homogeneity index also confirmed the accuracy of PTV curves on the DVH. Quantitatively, Model I's prediction error of 0.043 (confirmation), 0.043 (homogeneity), 0.327 (R50), 2.80% (D95), 3.90% (D98), 0.6% (D50), and 1.10% (D2) was lower than that of Model II, which obtained 0.076 (confirmation), 0.058 (homogeneity), 0.626 (R50), 7.10% (D95), 6.50% (D98), 8.40% (D50), and 6.30% (D2). Model I also outperformed Model II in terms of the mean dose error and the max dose error on the PTV, bladder, rectum, left femoral head, and right femoral head.

CONCLUSIONS

Treatment planners who use our models will be able to use deep learning to control the trade-offs between the PTV and OAR weights, as well as the beam number and configurations in real time. Our dose prediction methods provide a stepping stone to building automatic IMRT treatment planning.

Predicting patient specific Pareto fronts from patient anatomy only

PURPOSE

To demonstrate the feasibility of predicting the patient-specific treatment planning Pareto front (PF) for prostate cancer patients based only on delineations of PTV, rectum and body.

MATERIAL/METHODS

Our methodology consists of four steps. First, using Erasmus-iCycle, the Pareto fronts of 112 prostate cancer patients were constructed by generating per patient 42 Pareto optimal treatment plans with different priorities. Dose parameters associated to homogeneity, conformity and dose to rectum were extracted. Second, a 3D convex function representing the PF spanned by the 42 plans was fitted for each patient using three patient-specific parameters. Third, ten features were extracted from the, aforementioned, structures to train a linear-regressor prediction algorithm to predict these three patient-specific parameters. Fourth, the quality of the predictions was assessed by calculating the average and maximum distances of the predicted PF to the 42 plans for patients in the validation cohort.

RESULTS

The prediction model was able to predict the clinically relevant PF within 2 Gy for 90% of the patients with a median average distance of 0.6 Gy.

CONCLUSIONS

We demonstrate the feasibility of fast, accurate predictions of the patient-specific PF for prostate cancer patients based only on delineations of PTV, rectum and body.

Reducing the Risk of Secondary Lung Cancer in Treatment Planning of Accelerated Partial Breast Irradiation

Purpose: Adjuvant accelerated partial breast irradiation (APBI) results in low local recurrence risks. However, the survival benefit of adjuvant radiotherapy APBI for low-risk breast cancer might partially be offset by the risk of radiation-induced lung cancer. Reducing the lung dose mitigates this risk, but this could result in higher doses to the ipsilateral breast. Different external beam APBI techniques are equally conformal and homogenous, but the intermediate to low dose distribution differs. Thus, the risk of toxicity is different. The purpose of this study is to quantify the trade-off between secondary lung cancer risk and breast dose in treatment planning and to compare an optimal coplanar and non-coplanar technique. Methods: A total of 440 APBI treatment plans were generated using automated treatment planning for a coplanar VMAT beam-setup and a non-coplanar robotic stereotactic radiotherapy beam-setup. This enabled an unbiased comparison of two times 11 Pareto-optimal plans for 20 patients, gradually shifting priority from maximum lung sparing to maximum ipsilateral breast sparing. The excess absolute risks of developing lung cancer and breast fibrosis were calculated using the Schneider model for lung cancer and the Avanzo model for breast fibrosis. Results: Prioritizing lung sparing reduced the mean lung dose from 2.2 Gy to as low as 0.3 Gy for the non-coplanar technique and from 1.9 Gy to 0.4 Gy for the coplanar technique, corresponding to a 7- and 4-fold median reduction of secondary lung cancer risk, respectively, compared to prioritizing breast sparing. The increase in breast dose resulted in a negligible 0.4% increase in fibrosis risk. The use of non-coplanar beams resulted in lower secondary cancer and fibrosis risks (p < 0.001). Lung sparing also reduced the mean heart dose for both techniques. Conclusions: The risk of secondary lung cancer of external beam APBI can be dramatically reduced by prioritizing lung sparing during treatment planning. The associated increase in breast dose did not lead to a relevant increase in fibrosis risk. The use of non-coplanar beams systematically resulted in the lowest risks of secondary lung cancer and fibrosis. Prioritizing lung sparing during treatment planning could increase the overall survival of early-stage breast cancer patients by reducing mortality due to secondary lung cancer and cardiovascular toxicity.

Online-adaptive versus robust IMPT for prostate cancer: How much can we gain?

BACKGROUND/PURPOSE

Intensity-modulated proton therapy (IMPT) is highly sensitive to anatomical variations which can cause inadequate target coverage during treatment. Available mitigation techniques include robust treatment planning and online-adaptive IMPT. This study compares a robust planning strategy to two online-adaptive IMPT strategies to determine the benefit of online adaptation.

MATERIALS/METHODS

We derived the robustness settings and safety margins needed to yield adequate target coverage (V_95%≥98%) for >90% of 11 patients in a prostate cancer cohort (88 repeat CTs). For each patient, we also adapted a non-robust prior plan using a simple restoration and a full adaptation method. The restoration uses energy-adaptation followed by a fast spot-intensity re-optimization. The full adaptation uses energy-adaptation followed by the addition of new spots and a range-robust spot-intensity optimization. Dose was prescribed as 55 Gy(RBE) to the low-dose target (lymph nodes and seminal vesicles) with a boost to 74 Gy(RBE) to the high-dose target (prostate). Daily patient set-up was simulated using implanted intra-prostatic markers.

RESULTS

Margins of 4 and 8 mm around the high- and low-dose target regions, a 6 mm setup error and a 3% range error were found to obtain adequate target coverage for all repeat CTs of 10/11 patients (94.3% of all 88 repeat CTs). Both online-adaptive strategies yielded V_95%≥98% and better OAR sparing in 11/11 patients. Median OAR improvements up to 11%-point and 16%-point were observed when moving from robust planning to respectively restoration and full adaption.

CONCLUSION

Both full plan adaptation and simple dose restoration can increase OAR sparing besides better conforming to the target criteria compared to robust treatment planning.

First system for fully-automated multi-criterial treatment planning for a high-magnetic field MR-Linac applied to rectal cancer

Background and purpose: In this study we developed a workflow for fully-automated generation of deliverable IMRT plans for a 1.5 T MR-Linac (MRL) based on contoured CT scans, and we evaluated automated MRL planning for rectal cancer.Methods: The Monte Carlo dose calculation engine used in the clinical MRL TPS (Monaco, Elekta AB, Stockholm, Sweden), suited for high accuracy dose calculations in a 1.5 T magnetic field, was coupled to our in-house developed Erasmus-iCycle optimizer. Clinically deliverable plans for 23 rectal cancer patients were automatically generated in a two-step process, i.e., multi-criterial fluence map optimization with Erasmus-iCycle followed by a conversion into a deliverable IMRT plan in the clinical TPS. Automatically generated plans (AUTOplans) were compared to plans that were manually generated with the clinical TPS (MANplans).Results: With AUTOplanning large reductions in planning time and workload were obtained; 4-6 h mainly hands-on planning for MANplans vs ∼1 h of mainly computer computation time for AUTOplans. For equal target coverage, the bladder and bowel bag D_mean was reduced in the AUTOplans by 1.3 Gy (6.9%) on average with a maximum reduction of 4.5 Gy (23.8%). Dosimetric measurements at the MRL demonstrated clinically acceptable delivery accuracy for the AUTOplans.Conclusions: A system for fully automated multi-criterial planning for a 1.5 T MR-Linac was developed and tested for rectal cancer patients. Automated planning resulted in major reductions in planning workload and time, while plan quality improved. Negative impact of the high magnetic field on the dose distributions could be avoided.

Automated Radiotherapy Planning for Patient-Specific Exploration of the Trade-Off Between Tumor Dose Coverage and Predicted Radiation-Induced Toxicity-A Proof of Principle Study for Prostate Cancer

Background: Currently, radiation-oncologists generally evaluate a single treatment plan for each patient that is possibly adapted by the planner prior to final approval. There is no systematic exploration of patient-specific trade-offs between planning aims, using a set of treatment plans with a-priori defined (slightly) different balances. To this purpose, we developed an automated workflow and explored its use for prostate cancer. Materials and Methods: For each of the 50 study patients, seven plans were generated, including the so-called clinical plan, with currently clinically desired ≥99% dose coverage for the low-dose planning target volume (PTVLow). The six other plans were generated with different, reduced levels of PTVLow coverage, aiming at reductions in rectum dose and consequently in predicted grade≥2 late gastro-intestinal (GI) normal tissue complication probabilities (NTCPs), while keeping other dosimetric differences small. The applied NTCP model included diabetes as a non-dosimetric predictor. All plans were generated with a clinically applied, in-house developed algorithm for automated multi-criterial plan generation. Results: With diabetes, the average NTCP reduced from 24.9 ± 4.5% for ≥99% PTVLow coverage to 17.3 ± 2.6% for 90%, approaching the NTCP (15.4 ± 3.0%) without diabetes and full PTVLow coverage. Apart from intended differences in PTVLow coverage and rectum dose, other differences between the clinical plan and the six alternatives were indeed minor. Obtained NTCP reductions were highly patient-specific (ranging from 14.4 to 0.1%), depending on patient anatomy. Even for patients with equal NTCPs in the clinical plan, large differences were found in NTCP reductions. Conclusions: A clinically feasible workflow has been proposed for systematic exploration of patient-specific trade-offs between various treatment aims. For each patient, automated planning is used to generate a limited set of treatment plans with well-defined variations in the balances between the aims. For prostate cancer, trade-offs between PTVLow coverage and predicted GI NTCP were explored. With relatively small coverage reductions, significant NTCP reductions could be obtained, strongly depending on patient anatomy. Coverage reductions could also make up for enhanced NTCPs related to diabetes as co-morbidity, again dependent on the patient. The proposed system can play an important role in further personalization of patient care.

A collision prediction framework for noncoplanar radiotherapy planning and delivery

PURPOSE

Noncoplanar radiotherapy can provide significant dosimetric benefits. However, clinical implementation of such techniques is not fully realized, partially due to the absence of a collision prediction tool integrated into the clinical workflow. In this work, the feasibility of developing a collision prediction system (CPS) suitable for integration into clinical practice has been investigated.

METHODS

The CPS is based on a geometric model of the Linear Accelerator (Linac), and patient morphology acquired at the simulator using a combination of the planning CT scan and 3-D vision camera (Microsoft, Kinect) data. Physical dimensions of Linac components were taken to construct a geometric model. The Linac components include the treatment couch, gantry, and imaging devices. The treatment couch coordinates were determined based on a correspondence among the CT couch top, Linac couch, and the treatment isocenter location. A collision is predicted based on dot products between vectors denoting points in Linac components and patient morphology. Collision test cases were simulated with the CPS and experimentally verified using ArcCheck and Rando phantoms to simulate a patient.

RESULTS

For 111 collision test cases, the sensitivity and specificity of the CPS model were calculated to be 0.95 and 1.00, respectively. The CPS predicted collision states that left conservative margins, as designed, relative to actual collision locations. The average difference between the predicted and measured collision states was 2.3 cm for lateral couch movements. The predicted couch rotational position for a collision between the gantry and a patient analog differed from actual values on average by 3.8°. The magnitude of these differences is sufficient to account for interfractional patient positioning variations during treatment.

CONCLUSION

The feasibility of developing a CPS using geometric models and standard vector algebra has been investigated. This study outlines a framework for potential clinical implementation of a CPS for noncoplanar radiotherapy.

Shortening delivery times for intensity-modulated proton therapy by reducing the number of proton spots: an experimental verification

Delivery times of intensity-modulated proton therapy (IMPT) can be shortened by reducing the number of spots in the treatment plan, but this may affect clinical plan delivery. Here, we assess the experimental deliverability, accuracy and time reduction of spot-reduced treatment planning for a clinical case, as well as its robustness. For a single head-and-neck cancer patient, a spot-reduced plan was generated and compared with the conventional clinical plan. The number of proton spots was reduced using the iterative 'pencil beam resampling' technique. This involves repeated inverse optimization, while adding in each iteration a small sample of randomly selected spots and subsequently excluding low-weighted spots until plan quality deteriorates. Field setup was identical for both plans and comparable dosimetric quality was a prerequisite. Both IMPT plans were delivered on PSI Gantry 2 and measured in water, while delivery log-files were used to extract delivery times and reconstruct the delivered dose via Monte-Carlo dose calculations. In addition, robustness simulations were performed to assess sensitivity to machine inaccuracies and errors in patient setup and proton range. The number of spots was reduced by 96% (from 33 855 to 1510 in total) without compromising plan quality. The spot-reduced plan was deliverable on our gantry in standard clinical mode and resulted in average delivery times per field being shortened by 46% (from 51.2 to 27.6 s). For both plans, differences between measured and calculated dose were within clinical tolerance for patient-specific verifications and Monte-Carlo dose reconstructions were in accordance with clinical experience. The spot-reduced plan was slightly more sensitive to machine inaccuracies, but more robust against setup and range errors. In conclusion, for an example head-and-neck case, spot-reduced IMPT planning provided a deliverable treatment plan and enabled considerable shortening of the delivery time (∼50%) without compromising plan quality or delivery accuracy, and without substantially affecting robustness.

Automatic genetic planning for volumetric modulated arc therapy: A large multi-centre validation for prostate cancer

PURPOSE

The first clinical genetic autoplanning algorithm (Genetic Planning Solution, GPS) was validated in ten radiotherapy centres for prostate cancer VMAT by comparison with manual planning (Manual).

METHODS

Although there were large differences among centres in planning protocol, GPS was tuned with the data of a single centre and then applied everywhere without any centre-specific fine-tuning. For each centre, ten Manual plans were compared with autoGPS plans, considering dosimetric plan parameters and the Clinical Blind Score (CBS) resulting from blind clinician plan comparisons. AutoGPS plans were used as is, i.e. there was no patient-specific fine-tuning.

RESULTS

For nine centres, all ten plans were clinically acceptable. In the remaining centre, only one plan was acceptable. For the 91% acceptable plans, differences between Manual and AutoGPS in target coverage were negligible. OAR doses were significantly lower in AutoGPS plans (p < 0.05); rectum D_15% and D_mean were reduced by 8.1% and 17.9%, bladder D_25% and D_mean by 5.9% and 10.3%. According to clinicians, 69% of the acceptable AutoGPS plans were superior to the corresponding Manual plan. In case of preferred Manual plans (31%), perceived advantages compared to autoGPS were minor. QA measurements demonstrated that autoGPS plans were deliverable. A quick configuration adjustment in the centre with unacceptable plans rendered 100% of plans acceptable.

CONCLUSION

A novel, clinically applied genetic autoplanning algorithm was validated in 10 centres for in total 100 prostate cancer patients. High quality plans could be generated at different centres without centre-specific algorithm tuning.

Automating proton treatment planning with beam angle selection using Bayesian optimization

PURPOSE

To present a fully automated treatment planning process for proton therapy including beam angle selection using a novel Bayesian optimization approach and previously developed constrained hierarchical fluence optimization method.

METHODS

We adapted our in-house automated intensity modulated radiation therapy (IMRT) treatment planning system, which is based on constrained hierarchical optimization and referred to as ECHO (expedited constrained hierarchical optimization), for proton therapy. To couple this to beam angle selection, we propose using a novel Bayesian approach. By integrating ECHO with this Bayesian beam selection approach, we obtain a fully automated treatment planning framework including beam angle selection. Bayesian optimization is a global optimization technique which only needs to search a small fraction of the search space for slowly varying objective functions (i.e., smooth functions). Expedited constrained hierarchical optimization is run for some initial beam angle candidates and the resultant treatment plan for each beam configuration is rated using a clinically relevant treatment score function. Bayesian optimization iteratively predicts the treatment score for not-yet-evaluated candidates to find the best candidate to be optimized next with ECHO. We tested this technique on five head-and-neck (HN) patients with two coplanar beams. In addition, tests were performed with two noncoplanar and three coplanar beams for two patients.

RESULTS

For the two coplanar configurations, the Bayesian optimization found the optimal beam configuration after running ECHO for, at most, 4% of all potential configurations (23 iterations) for all patients (range: 2%-4%). Compared with the beam configurations chosen by the planner, the optimal configurations reduced the mandible maximum dose by 6.6 Gy and high dose to the unspecified normal tissues by 3.8 Gy, on average. For the two noncoplanar and three coplanar beam configurations, the algorithm converged after 45 iterations (examining <1% of all potential configurations).

CONCLUSIONS

A fully automated and efficient treatment planning process for proton therapy, including beam angle optimization was developed. The algorithm automatically generates high-quality plans with optimal beam angle configuration by combining Bayesian optimization and ECHO. As the Bayesian optimization is capable of handling complex nonconvex functions, the treatment score function which is used in the algorithm to evaluate the dose distribution corresponding to each beam configuration can contain any clinically relevant metric.

A plan template‐based automation solution using a commercial treatment planning system

Purpose

The purpose of this study was to develop an auto‐planning platform to be interfaced with a commercial treatment planning system (TPS). The main goal was to obtain robust and high‐quality plans for different anatomic sites and various dosimetric requirements.

Methods

Monaco (Elekta, St. Louis, US) was the TPS in this work. All input parameters for inverse planning could be defined in a plan template inside Monaco. A software tool called Robot Framework was used to launch auto‐planning trials with updated plan templates. The template modifier external to Monaco was the major component of our auto‐planning platform. For current implementation, it was a rule‐based system that mimics the trial‐and‐error process of an experienced planner. A template was automatically updated by changing the optimization constraints based on dosimetric evaluation of the plan obtained in the previous trial, along with the data of the iterative optimization extracted from Monaco. Treatment plans generated by Monaco with all plan evaluation criteria satisfied were considered acceptable, and such plans would be saved for further evaluation by clinicians. The auto‐planning platform was validated for 10 prostate and 10 head‐and‐neck cases in comparison with clinical plans generated by experienced planners.

Results

The performance and robustness of our auto‐planning platform was tested with clinical cases of prostate and head and neck treatment. For prostate cases, automatically generated plans had very similar plan quality with the clinical plans, and the bladder volume receiving 62.5 Gy, 50 Gy, and 40 Gy in auto‐plans was reduced by 1%, 3%, and 5%, respectively. For head and neck cases, auto‐plans had better conformity with reduced dose to the normal structures but slightly higher dose inhomogeneity in the target volume. Remarkably, the maximum dose in the spinal cord and brain stem was reduced by more than 3.5 Gy in auto‐plans. Fluence map optimization only with less than 30 trials was adequate to generate acceptable plans, and subsequent optimization for final plans was completed by Monaco without further intervention. The plan quality was weakly dependent on the parameter selection in the initial template and the choices of the step sizes for changing the constraint values.

Conclusion

An automated planning platform to interface with Monaco was developed, and our reported tests showed preliminary results for prostate and head and neck cases.

Multiobjective, Multidelivery Optimization for Radiation Therapy Treatment Planning

Purpose

To introduce multiobjective, multidelivery optimization (MODO), which generates alternative patient-specific plans emphasizing dosimetric trade-offs and conformance to quasi-constrained (QC) conditions for multiple delivery techniques.

Methods and Materials

For M delivery techniques and N organs at risk (OARs), MODO generates M (N + 1) alternative treatment plans per patient. For 30 locally advanced lung cancer cases, the algorithm was investigated based on dosimetric trade-offs to 4 OARs: each lung, heart, and esophagus (N = 4) and 4 delivery techniques (4-field coplanar intensity modulated radiation therapy [IMRT], 9-field coplanar IMRT, 27-field noncoplanar IMRT, and noncoplanar arc IMRT) and conformance to QC conditions, including dose to 95% (D95) of the planning target volume (PTV), maximum dose (Dmax) to PTV (PTV-Dmax), and spinal cord Dmax. The MODO plan set was evaluated for conformance to QC conditions while simultaneously revealing dosimetric trade-offs. Statistically significant dosimetric trade-offs were defined such that the coefficient of determination was >0.8 with dosimetric indices that varied by at least 5 Gy.

Results

Plans varied mean dose by >5 Gy to ipsilateral lung for 24 of 30 patients, contralateral lung for 29 of 30 patients, esophagus for 29 of 30 patients, and heart for 19 of 30 patients. In the 600 plans, average PTV-D95 = 67.6 ± 2.1 Gy, PTV-Dmax = 79.8 ± 5.2 Gy, and spinal cord Dmax among all plans was 51.4 Gy. Statistically significant dosimetric trade-offs reducing OAR mean dose by >5 Gy were evident in 19 of 30 patients, including multiple OAR trade-offs of at least 5 Gy in 7 of 30 cases. The most common statistically significant trade-off was increasing PTV-Dmax to reduce dose to OARs (15 of 30). The average 4-field plan reduced total lung V20 by 10.4% ± 8.3% compared with 9-field plans, 7.7% ± 7.9% compared with 27-field noncoplanar plans, and 11.7% ± 10.3% compared with 2-arc noncoplanar plans, with corresponding increases in PTV-Dmax of 5.3 ± 5.9 Gy, 4.6 ± 5.6 Gy, and 9.3 ± 7.3 Gy.

Conclusions

The proposed optimization method produces clinically relevant treatment plans that meet QC conditions and demonstrate variations in OAR doses.

Automated planning through robust templates and multicriterial optimization for lung VMAT SBRT of lung lesions

Purpose

To develop and validate a robust template for VMAT SBRT of lung lesions, using the multicriterial optimization (MCO) of a commercial treatment planning system.

Methods

The template was established and refined on 10 lung SBRT patients planned for 55 Gy/5 fr. To improve gradient and conformity a ring structure around the planning target volume (PTV) was set in the list of objectives. Ideal fluence optimization was conducted giving priority to organs at risk (OARs) and using the MCO, which further pushes OARs doses. Segmentation was conducted giving priority to PTV coverage. Two different templates were produced with different degrees of modulation, by setting the Fluence Smoothing parameter to Medium (MFS) and High (HFS).

Each template was applied on 20 further patients. Automatic and manual plans were compared in terms of dosimetric parameters, delivery time, and complexity. Statistical significance of differences was evaluated using paired two‐sided Wilcoxon signed‐rank test.

Results

No statistically significant differences in PTV coverage and maximum dose were observed, while an improvement was observed in gradient and conformity. A general improvement in dose to OARs was seen, which resulted to be significant for chest wall V_30 Gy, total lung V_20 Gy, and spinal cord D_0.1 cc. MFS plans are characterized by a higher modulation and longer delivery time than manual plans. HFS plans have a modulation and a delivery time comparable to manual plans, but still present an advantage in terms of gradient.

Conclusion

The automation of the planning process for lung SBRT using robust templates and MCO was demonstrated to be feasible and more efficient.

A fast deep learning approach for beam orientation optimization for prostate cancer treated with intensity-modulated radiation therapy

PURPOSE

Beam orientation selection, whether manual or protocol-based, is the current clinical standard in radiation therapy treatment planning, but it is tedious and can yield suboptimal results. Many algorithms have been designed to optimize beam orientation selection because of its impact on treatment plan quality, but these algorithms suffer from slow calculation of the dose influence matrices of all candidate beams. We propose a fast beam orientation selection method, based on deep learning neural networks (DNN), capable of developing a plan comparable to those developed by the state-of-the-art column generation (CG) method. Our model's novelty lies in its supervised learning structure (using CG to teach the network), DNN architecture, and ability to learn from anatomical features to predict dosimetrically suitable beam orientations without using dosimetric information from the candidate beams. This may save hours of computation.

METHODS

A supervised DNN is trained to mimic the CG algorithm, which iteratively chooses beam orientations one-by-one by calculating beam fitness values based on Karush-Kush-Tucker optimality conditions at each iteration. The DNN learns to predict these values. The dataset contains 70 prostate cancer patients - 50 training, 7 validation, and 13 test patients - to develop and test the model. Each patient's data contains 6 contours: PTV, body, bladder, rectum, and left and right femoral heads. Column generation was implemented with a GPU-based Chambolle-Pock algorithm, a first-order primal-dual proximal-class algorithm, to create 6270 plans. The DNN trained over 400 epochs, each with 2500 steps and a batch size of 1, using the Adam optimizer at a learning rate of 1 × 10-5 and a sixfold cross-validation technique.

RESULTS

The average and standard deviation of training, validation, and testing loss functions among the six folds were 0.62 ± 0.09%, 1.04 ± 0.06%, and 1.44 ± 0.11%, respectively. Using CG and supervised DNN, we generated two sets of plans for each scenario in the test set. The proposed method took at most 1.5 s to select a set of five beam orientations and 300 s to calculate the dose influence matrices for 5 beams and finally 20 s to solve the fluence map optimization (FMO). However, CG needed around 15 h to calculate the dose influence matrices of all beams and at least 400 s to solve both the beam orientation selection and FMO problems. The differences in the dose coverage of PTV between plans generated by CG and by DNN were 0.2%. The average dose differences received by organs at risk were between 1 and 6 percent: Bladder had the smallest average difference in dose received (0.956 ± 1.184%), then Rectum (2.44 ± 2.11%), Left Femoral Head (6.03 ± 5.86%), and Right Femoral Head (5.885 ± 5.515%). The dose received by Body had an average difference of 0.10 ± 0.1% between the generated treatment plans.

CONCLUSIONS

We developed a fast beam orientation selection method based on a DNN that selects beam orientations in seconds and is therefore suitable for clinical routines. In the training phase of the proposed method, the model learns the suitable beam orientations based on patients' anatomical features and omits time intensive calculations of dose influence matrices for all possible candidate beams. Solving the FMO to get the final treatment plan requires calculating dose influence matrices only for the selected beams.

Artificial Intelligence in Radiotherapy Treatment Planning: Present and Future

Treatment planning is an essential step of the radiotherapy workflow. It has become more sophisticated over the past couple of decades with the help of computer science, enabling planners to design highly complex radiotherapy plans to minimize the normal tissue damage while persevering sufficient tumor control. As a result, treatment planning has become more labor intensive, requiring hours or even days of planner effort to optimize an individual patient case in a trial-and-error fashion. More recently, artificial intelligence has been utilized to automate and improve various aspects of medical science. For radiotherapy treatment planning, many algorithms have been developed to better support planners. These algorithms focus on automating the planning process and/or optimizing dosimetric trade-offs, and they have already made great impact on improving treatment planning efficiency and plan quality consistency. In this review, the smart planning tools in current clinical use are summarized in 3 main categories: automated rule implementation and reasoning, modeling of prior knowledge in clinical practice, and multicriteria optimization. Novel artificial intelligence-based treatment planning applications, such as deep learning-based algorithms and emerging research directions, are also reviewed. Finally, the challenges of artificial intelligence-based treatment planning are discussed for future works.