Is it possible for knowledge-based planning to improve intensity modulated radiation therapy plan quality for planners with different planning experiences in left-sided breast cancer patients?

Knowledge-based versus deep learning based treatment planning for breast radiotherapy

Background and Purpose

To improve radiotherapy (RT) planning efficiency and plan quality, knowledge-based planning (KBP) and deep learning (DL) solutions have been developed. We aimed to make a direct comparison of these models for breast cancer planning using the same training, validation, and testing sets.

Materials and Methods

Two KBP models were trained and validated with 90 RT plans for left-sided breast cancer with 15 fractions of 2.6 Gy. The versions either used the full dataset (non-clean model) or a cleaned dataset (clean model), thus eliminating geometric and dosimetric outliers. Results were compared with a DL U-net model (previously trained and validated with the same 90 RT plans) and manually produced RT plans, for the same independent dataset of 15 patients. Clinically relevant dose volume histogram parameters were evaluated according to established consensus criteria.

Results

Both KBP models underestimated the mean heart and lung dose equally 0.4 Gy (0.3-1.1 Gy) and 1.4 Gy (1.1-2.8 Gy) compared to the clinical plans 0.8 Gy (0.5-1.8 Gy) and 1.7 Gy (1.3-3.2 Gy) while in the final calculations the mean lung dose was higher 1.9-2.0 Gy (1.5-3.5 Gy) for both KPB models. The U-Net model resulted in a mean planning target volume dose of 40.7 Gy (40.4-41.3 Gy), slightly higher than the clinical plans 40.5 Gy (40.1-41.0 Gy).

Conclusions

Only small differences were observed between the estimated and final dose calculation and the clinical results for both KPB models and the DL model. With a good set of breast plans, the data cleaning module is not needed and both KPB and DL models lead to clinically acceptable results.

Evaluation of a clinically introduced deep learning model for radiotherapy treatment planning of breast cancer

Deep learning (DL) models are increasingly studied to automate the process of radiotherapy treatment planning. This study evaluates the clinical use of such a model for whole breast radiotherapy. Treatment plans were automatically generated, after which planners were allowed to manually adapt them. Plans were evaluated based on clinical goals and DVH parameters. Thirty-seven of 50plans did fulfill all clinical goals without adjustments. Thirteen of these 37 plans were still adjusted but did not improve mean heart or lung dose. These results leave room for improvement of both the DL model as well as education on clinically relevant adjustments.

Can knowledge based treatment planning of VMAT for post-mastectomy locoregional radiotherapy involving internal mammary chain and supraclavicular fossa improve performance efficiency?

Introduction

To validate and evaluate the performance of knowledge-based treatment planning for Volumetric Modulated Arc Radiotherapy for post-mastectomy loco-regional radiotherapy.

Material and methods

Two knowledge-based planning (KBP) models for different dose prescriptions were built using the Eclipse RapidPlanTM v 16.1 (Varian Medical Systems, Palo Alto, USA) utilising the plans of previously treated patients with left-sided breast cancer who had undergone irradiation of the left chest wall, internal mammary nodal (IMN) region and supra-clavicular fossa (SCF). Plans of 60 and 73 patients were used to generate the KBP models for the prescriptions of 40 Gy in 15 fractions and 26 Gy in 5 fractions, respectively. A blinded review of all the clinical plans (CLI) and KBPs was done by two experienced radiation oncology consultants. Statistical analysis of the two groups was also done using the standard two-tailed paired t-test or Wilcoxon signed rank test, and p<0.05 was considered significant.

Results

A total of 20 metrics were compared. The KBPs were found to be either better (6/20) or comparable (10/20) to the CLIs for both the regimens. Dose to heart, contralateral breast,contralateral lung were either better or comparable in the KBP plans except of ipsilateral lung. Mean dose (Gy) for the ipsilateral lung are significantly (p˂0.001) higher in KBP though the values were acceptable clinically. Plans were of similar quality as per the result of the blinded review which was conducted by slice-by-slice evaluation of dose distribution for target coverage, overdose volume and dose to the OARs. However, it was also observed that treatment times in terms of monitoring units (MUs) and complexity indices are more in CLIs as compared with KBPs (p<0.001).

Discussion

KBP models for left-sided post-mastectomy loco-regional radiotherapy were developed and validated for clinical use. These models improved the efficiency of treatment delivery as well as work flow for VMAT planning involving both moderately hypo fractionated and ultra-hypo fractionated radiotherapy regimens.

Three-dimensional dose and LETD prediction in proton therapy using artificial neural networks

PURPOSE

Challenges in proton therapy include identifying patients most likely to benefit; ensuring consistent, high-quality plans as its adoption becomes more widespread; and recognizing biological uncertainties that may be related to increased relative biologic effectiveness driven by linear energy transfer (LET). Knowledge-based planning (KBP) is a domain that may help to address all three.

METHODS

Artificial neural networks were trained using 117 unique treatment plans and associated dose and dose-weighted LET (LET_D ) distributions. The data set was split into training (n = 82), validation (n = 17), and test (n = 18) sets. Model performance was evaluated on the test set using dose- and LET_D -volume metrics in the clinical target volume (CTV) and nearby organs at risk and Dice similarity coefficients (DSC) comparing predicted and planned isodose lines at 50%, 75%, and 95% of the prescription dose.

RESULTS

Dose-volume metrics significantly differed (α = 0.05) between predicted and planned dose distributions in only one dose-volume metric, D_2% to the CTV. The maximum observed root mean square (RMS) difference between corresponding metrics was 4.3 Gy_RBE (8% of prescription) for D_1cc to optic chiasm. DSC were 0.90, 0.93, and 0.88 for the 50%, 75%, and 95% isodose lines, respectively. LET_D -volume metrics significantly differed in all but one metric, L_0.1cc of the brainstem. The maximum observed difference in RMS differences for LET_D metrics was 1.0 keV/μm for L_0.1cc to brainstem.

CONCLUSIONS

We have devised the first three-dimensional dose and LET_D -prediction model for cranial proton radiation therapy has been developed. Dose accuracy compared favorably with that of previously published models in other treatment sites. The agreement in LET_D supports future investigations with biological doses in mind to enable the full potential of KBP in proton therapy.

A practical method to improve the performance of knowledge-based VMAT planning for endometrial and cervical cancer

PURPOSE

The aim of this work was to demonstrate a practical and effective method to improve the performance of RapidPlan (RP) model.

METHODS

203 consecutive clinical VMAT plans (P₀) for cervical and endometrial cancer were used to train an RP model (M₀). The plans were then reoptimized by M₀ to generate 203 new plans (P₁). Compared with P₀, 150 plans with a lower mean dose (MD) of bladder, rectum and PBM were selected from P₁ to configure a new RP model (M₁). A final RP model (M₂) was trained using plans in M₁ and the remaining 53 plans from P₁ (excluding OARs with worse MD) and the corresponding plans from P₀ (only including OARs with better MD). The models were validated on the mentioned 53 plans (closed-loop set) and 46 patient cohorts outside the training library (open-loop set). p < 0.05 was considered statistically significant.

RESULTS

For closed-loop validation, the difference of D_2%, D_98% and CI_95% between groups was of no statistical significance, the homogeneity index (HI) was lower in the groups of RP models (p < 0.05). The MD of all OARs decreased monotonically in the sequence of the clinical group, group M₀, M₁ and M₂, except the MD of bowel in M₁ and MD of LFH in M₂. Similarly, for open-loop validation, there was no significant difference in D_2%, D_98% and HI between groups, but CI_95% was larger in the clinical group (p < 0.05). The MD of all OARs decreased monotonically in the sequence of the clinical group, group M₀, M₁ and M₂, with the exception of bowel in M₁.

CONCLUSION

The practical method of incorporating plan data of better-sparing OARs from both the clinical VMAT plans and the re-optimized plans could further improve the performance of the RP model.

Knowledge-based automatic plan optimization for left-sided whole breast tomotherapy

Background/Purpose

Tomotherapy may deliver high-quality whole breast irradiation at static angles. The aim of this study was to implement Knowledge-Based (KB) automatic planning for left-sided whole breast using this modality.

Materials/Methods

Virtual volumetric plans were associated to the dose distributions of 69 Tomotherapy (TT) clinical plans of previously treated patients, aiming to train a KB-model using a commercial tool completely implemented in our treatment planning system. An individually optimized template based on the resulting KB-model was generated for automatic plan optimization. Thirty patients of the training set and ten new patients were considered for internal/external validation. Fully-automatic plans (KB-TT) were generated and compared using the same geometry/number of fields of the corresponding clinical plans.

Results

KB-TT plans were successfully generated in 26/30 and 10/10 patients of the internal/external validation sets; for 4 patients whose original plans used only two fields, the manual insertion of one/two fields before running the automatic template was sufficient to obtain acceptable plans. Concerning internal validation, planning target volume V_95%/D_1%/dose distribution standard deviation improved by 0.9%/0.4Gy/0.2Gy (p < 0.05) against clinical plans; Organs at risk mean doses were also slightly improved (p < 0.05) by 0.07/0.4/0.2/0.01 Gy for left lung/heart/right breast/right lung respectively. Similarly satisfactory results were replicated in the external validation set. The resulting treatment duration was 8 ± 1 min, consistent with our clinical experience. The active planner time per patient was 5–10 minutes.

Conclusion

Automatic TT left-sided breast KB-plans are comparable to or slightly better than clinical plans and can be obtained with limited planner time. The approach is currently under clinical implementation.

The development of a deep reinforcement learning network for dose-volume-constrained treatment planning in prostate cancer intensity modulated radiotherapy

Although commercial treatment planning systems (TPSs) can automatically solve the optimization problem for treatment planning, human planners need to define and adjust the planning objectives/constraints to obtain clinically acceptable plans. Such a process is labor-intensive and time-consuming. In this work, we show an end-to-end study to train a deep reinforcement learning (DRL) based virtual treatment planner (VTP) that can behave like a human to operate a dose-volume constrained treatment plan optimization engine following the parameters used in Eclipse TPS for high-quality treatment planning. We considered the prostate cancer IMRT treatment plan as the testbed. The VTP took the dose-volume histogram (DVH) of a plan as input and predicted the optimal strategy for constraint adjustment to improve the plan quality. The training of VTP followed the state-of-the-art Q-learning framework. Experience replay was implemented with epsilon-greedy search to explore the impacts of taking different actions on a large number of automatically generated plans, from which an optimal policy can be learned. Since a major computational cost in training was to solve the plan optimization problem repeatedly, we implemented a graphical processing unit (GPU)-based technique to improve the efficiency by 2-fold. Upon the completion of training, the established VTP was deployed to plan for an independent set of 50 testing patient cases. Connecting the established VTP with the Eclipse workstation via the application programming interface, we tested the performance the VTP in operating Eclipse TPS for automatic treatment planning with another two independent patient cases. Like a human planner, VTP kept adjusting the planning objectives/constraints to improve plan quality until the plan was acceptable or the maximum number of adjustment steps was reached under both scenarios. The generated plans were evaluated using the ProKnow scoring system. The mean plan score (± standard deviation) of the 50 testing cases were improved from 6.18 ± 1.75 to 8.14 ± 1.27 by the VTP, with 9 being the maximal score. As for the two cases under Eclipse dose optimization, the plan scores were improved from 8 to 8.4 and 8.7 respectively by the VTP. These results indicated that the proposed DRL-based VTP was able to operate the in-house dose-volume constrained TPS and Eclipse TPS to automatically generate high-quality treatment plans for prostate cancer IMRT.

Evaluation of auto-planning in VMAT for locally advanced nasopharyngeal carcinoma

The aim of this study is to demonstrate the feasibility of a commercially available Auto-Planning module for the radiation therapy treatment planning for locally advanced nasopharyngeal carcinoma (NPC). 22 patients with locally advanced NPC were included in this study. For each patient, volumetric modulated arc therapy (VMAT) plans were generated both manually by an experienced physicist and automatically by the Auto-Planning module. The dose distribution, dosimetric parameters, monitor units and planning time were compared between automatic plans (APs) and manual plans (MPs). Meanwhile, the overall stage of disease was factored into the evaluation. The target dose coverage of APs was comparable to that of MPs. For the organs at risk (OARs) except spinal cord, the dose parameters of APs were superior to that of MPs. The D_max and V₅₀ of brainstem were statistically lower by 1.0 Gy and 1.32% respectively, while the D_max of optic nerves and chiasm were also lower in the APs (p < 0.05). The APs provided a similar or superior quality to MPs in most cases, except for several patients with stage IV disease. The dose differences for most OARs were similar between the two types of plans regardless of stage while the APs provided better brainstem sparing for patients with stage III and improved the sparing of the parotid glands for stage IV patients. The total monitor units and planning time were significantly reduced in the APs. Auto-Planning is feasible for the VMAT treatment planning for locally advanced NPC.

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.

Performance of a knowledge-based planning model for optimizing intensity-modulated radiotherapy plans for partial breast irradiation

Purpose

To evaluate a knowledge‐based (KB) planning model for RapidPlan, generated using a five‐field intensity‐modulated radiotherapy (IMRT) class solution beam strategy and rigorous dosimetric constraints for accelerated partial breast irradiation (APBI).

Materials and methods

The RapidPlan model was configured using 64 APBI treatment plans and validated for 120 APBI patients who were not included in the training dataset. KB plan dosimetry was compared to clinical plan dosimetry, the clinical planning constraints, and the constraints used in phase III APBI trials. Dosimetric differences between clinical and KB plans were evaluated using paired two‐tailed Wilcoxon signed‐rank tests.

Results

KB planning was able to produce IMRT‐based APBI plans in a single optimization without manual intervention that are comparable or better than the conventionally optimized, clinical plans. Comparing KB plans to clinical plans, differences in PTV, heart, contralateral breast, and ipsilateral lung dose–volume metrics were not clinically significant. The ipsilateral breast volume receiving at least 50% of the prescription dose was statistically and clinically significantly lower in the KB plans.

Conclusion

KB planning for IMRT‐based APBI provides equivalent or better dosimetry compared to conventional inverse planning. This model may be reliably applied in clinical practice and could be used to transfer planning expertise to ensure consistency in APBI plan quality.

Artificial intelligence based treatment planning of radiotherapy for locally advanced breast cancer

Background and purpose

Treatment planning of radiotherapy for locally advanced breast cancer patients can be a time consuming process. Artificial intelligence based treatment planning could be used as a tool to speed up this process and maintain plan quality consistency. The purpose of this study was to create treatment plans for locally advanced breast cancer patients using a Convolutional Neural Network (CNN).

Materials and methods

Data of 60 patients treated for left-sided breast cancer was used with a training, validation and test split of 36/12/12, respectively. The in-house built CNN model was a hierarchically densely connected U-net (HD U-net). The inputs for the HD U-net were 2D distance maps of the relevant regions of interest. Dose predictions, generated by the HD U-net, were used for a mimicking algorithm in order to create clinically deliverable plans.

Results

Dose predictions were generated by the HD U-net and mimicked using a commercial treatment planning system. The predicted plans fulfilling all clinical goals while showing small (≤0.5 Gy) statistically significant differences (p < 0.05) in the doses compared to the manual plans. The mimicked plans show statistically significant differences in the average doses for the heart and lung of ≤0.5 Gy and a reduced D2% of all PTVs. In total, ten of the twelve mimicked plans were clinically acceptable.

Conclusions

We created a CNN model which can generate clinically acceptable plans for left-sided locally advanced breast cancer patients. This model shows great potential to speed up the treatment planning process while maintaining consistent plan quality.

Artificial intelligence applications in intensity modulated radiation treatment planning: an overview

Artificial intelligence (AI) refers to methods that improve and automate challenging human tasks by systematically capturing and applying relevant knowledge in these tasks. Over the past decades, a number of approaches have been developed to address different types and needs of system intelligence ranging from search strategies to knowledge representation and inference to robotic planning. In the context of radiation treatment planning, multiple AI approaches may be adopted to improve the planning quality and efficiency. For example, knowledge representation and inference methods may improve dose prescription by integrating and reasoning about the domain knowledge described in many clinical guidelines and clinical trials reports. In this review, we will focus on the most studied AI approach in intensity modulated radiation therapy (IMRT)/volumetric modulated arc therapy (VMAT)-machine learning (ML) and describe our recent efforts in applying ML to improve the quality, consistency, and efficiency of IMRT/VMAT planning. With the available high-quality data, we can build models to accurately predict critical variables for each step of the planning process and thus automate and improve its outcomes. Specific to the IMRT/VMAT planning process, we can build models for each of the four critical components in the process: dose-volume histogram (DVH), Dose, Fluence, and Human Planner. These models can be divided into two general groups. The first group focuses on encoding prior experience and knowledge through ML and more recently deep learning (DL) from prior clinical plans and using these models to predict the optimal DVH (DVH prediction model), or 3D dose distribution (dose prediction model), or fluence map (fluence map model). The goal of these models is to reduce or remove the trial-and-error process and guarantee consistently high-quality plans. The second group of models focuses on mimicking human planners' decision-making process (planning strategy model) during the iterative adjustments/guidance of the optimization engine. Each critical step of the IMRT/VMAT treatment planning process can be improved and automated by AI methods. As more training data becomes available and more sophisticated models are developed, we can expect that the AI methods in treatment planning will continue to improve accuracy, efficiency, and robustness.

Cardiac Dose Control and Optimization Strategy for Left Breast Cancer Radiotherapy With Non-Uniform VMAT Technology

Purpose: A novel in-house technology "Non-Uniform VMAT (NU-VMAT)" was developed for automated cardiac dose reduction and treatment planning optimization in the left breast radiotherapy. Methods: The NU-VMAT model based on I_GM (gantry MLC Movement coefficient index) was established to optimize the volumetric modulated arc therapy (VMAT) MLC movement and modulation intensity in certain gantry angles. The ESAPI embedded in Eclipse® was employed to connect TPS and the optimization program via I/O relevant DICOM RT files. The adjuvant whole-breast radiotherapy of 14 patients with left breast cancer was replanned using our NU-VMAT technology in comparison with VMAT and IMRT technology. Dosimetric parameters including D_1%, D_99%, and D_mean of PTV, V₅, V₁₀, and V₂₀ of ipisilateral lung, V₅, D₂₀, D₃₀, and D_mean of heart, monitor units (MUs), and delivery time derived from IMRT, VMAT, and NU-VMAT plans were evaluated for plan quality and delivery efficiency. The quality assurance (QA) was conducted using both point-dose and planar-dose measurements for all treatment plans. Results: The I_GM-NU-VMAT curves with plan optimization (range from 50% to 147%) were converged more significantly than I_GM-_VMAT curves (range from 0% to 297%). The dose distribution requirements of the target and normal tissues could be met using IMRT, VMAT, or NU-VMAT; the lowest D_mean was achieved in NU-VMAT plans (5.38 ± 0.46 Gy vs 5.63 ± 0.61 Gy in IMRT and 7.95 ± 0.52 Gy in VMAT plans). Statistically significant differences were found in terms of delivery time and MU when comparing IMRT with VMAT and NU-VMAT plans (P < .05). In comparison with IMRT plans, the MU and delivery time in NU-VMAT plans dramatically decreased by 69.8% and 28.4%, respectively. Moreover, NU-VMAT plans showed a high gamma passing rate (96.5% ± 1.11) in plane dose verification and minimal dose difference (2.4% ± 0.19) in point absolute dose verification. Conclusion: Our non-uniform VMAT facilitated the treatment strategy optimization for left breast cancer radiotherapy with dosimetric advantage in cardiac dose reduction and delivery efficiency in comparison with the conventional VMAT and IMRT.

Machine learning for dose-volume histogram based clinical decision-making support system in radiation therapy plans for brain tumors

•

Extraction, analysis, and interpretation of historical treatment planning data is valuable but very time-consuming.

•

Proposed machine learning model classifies radiotherapy plans based on their treatment planning objectives and trade-offs.

•

Application of double nested cross-validation enabled to build a robust model that achieved 94% accuracy on a testing data.

•

Model reasoning investigated with SHAP values showed consistency with clinical observations.

Purpose

To create and investigate a novel, clinical decision-support system using machine learning (ML).

Methods and Materials

The ML model was developed based on 79 radiotherapy plans of brain tumor patients that were prescribed a total dose of 60 Gy delivered with volumetric-modulated arc therapy (VMAT). Structures considered for analysis included planning target volume (PTV), brainstem, cochleae, and optic chiasm. The model aimed to classify the target variable that included class-0 corresponding to plans for which the PTV treatment planning objective was met and class-1 that was associated with plans for which the PTV objective was not met due to the priority trade-off to meet one or more organs-at-risk constraints. Several models were evaluated using double-nested cross-validation and an area-under-the-curve (AUC) metric, with the highest performing one selected for further investigation. The model predictions were explained with Shapely additive explanation (SHAP) interaction values.

Results

The highest-performing model was Logistic Regression achieving an accuracy of 93.8 ± 4.1% and AUC of 0.98 ± 0.02 on the testing data. The SHAP analysis indicated that the ΔD_99% metric for PTV had the greatest influence on the model predictions. The least important feature was ΔD_MAX for the left and right cochleae.

Conclusions

The trained model achieved satisfactory accuracy and can be used by medical physicists in a data-driven quality assurance program as well as by radiation oncologists to support their decision-making process in terms of treatment plan approval and potential plan modifications. Model explanation analysis showed that the model relies on clinically valid logic when making predictions.

Knowledge-Based Volumetric Modulated Arc Therapy Treatment Planning for Breast Cancer

Purpose:

To create and to validate knowledge-based volumetric modulated arc therapy (VMAT) models for breast cancer treatments without lymph node irradiation.

Materials and Methods:

One hundred VMAT-based breast plans (manual plans [MP]) were selected to create two knowledge-based VMAT models (breast left and breast right) using RapidPlan™. The plans were generated on Eclipse v15.5 (Varian Medical Systems, Palo Alto, CA) with 6 MV of a Novalis Tx equipped with a high-resolution multileaf collimator. The models were verified based on goodness-of-fit statistics using the coefficients of determination (R²) and Chi-square (χ²), and the goodness-of-estimation statistics through the mean square error (MSE). Geometrical and dosimetrical constraints were identified and removed from the RP models using statistical evaluation metrics and plots. For validation, 20 plans that integrate the models and 20 plans that do not were reoptimized with RP (closed and opened validation). Dosimetrical parameters of interest were used to compare MP versus RP plans for the Heart, Homolateral_Lung, Contralateral_Lung, and Contralateral_Breast. Optimization planning time and user independency were also analyzed.

Results:

The most unfavorable results of R² in both models for the organs at risk were as follows: for Contralateral_Lung 0.51 in RP right breast (RP_RB) and for Heart 0.60 in RP left breast (RP_LB). The most unfavorable results of χ² test were: for Contralateral_Breast 1.02 in RP_RB and for Heart 1.03 in RP_LB. These goodness-of-fit results show that no overfitting occurred in either of the models. There were no unfavorable results of mean square error (MSE, all < 0.05) in any of the two models. These goodness-of-estimation results show that the models have good estimation power. For closed validation, significant differences were found in RP_RB for Homolateral_Lung (all P ≤ 0.001), and in the RP_LB differences were found for the heart (all P ≤ 0.04) and for Homolateral_Lung (all P ≤ 0.022). For open validation, no statistically significant differences were obtained in either of the models. RP models had little impact on reducing optimization planning times for expert planners; nevertheless, the result showed a 30% reduction time for beginner planners. The use of RP models generates high-quality plans, without differences from the planner experience.

Conclusion:

Two RP models for breast cancer treatment using VMAT were successfully implemented. The use of RP models for breast cancer reduces the optimization planning time and improves the efficiency of the treatment planning process while ensuring high-quality plans.

Dosimetric comparison of RapidPlan and manually optimised volumetric modulated arc therapy plans in prostate cancer

Purpose:

The aim of this study was to evaluate whether RapidPlan (RP) could generate clinically acceptable prostate volumetric modulated arc therapy (VMAT) plans.

Methods:

The in-house RP model was used to generate VMAT plans for 50 previously treated prostate cancer patients, with no additional optimisation being performed. The VMAT plans that were generated using the RP model were compared with the patients’ previous, manually optimised clinical plans (MP), none of which had been used for the development of the in-house RP prostate model. Differences between RP and MP in planning target volume (PTV) doses, organs at risk (OAR) sparing, monitor units (MU) and planning time required to produce treatment plans were analysed. Assessment of PTV doses was based on the conformation number (CN), homogeneity index (HI), D2%, D99% and the mean dose of the PTV. The OAR doses evaluated were the rectal V50 Gy, V65 Gy, V70 Gy and the mean dose, the bladder V65 Gy, V70 Gy and the mean dose, and the mean dose to both femurs.

Results:

D99% and mean dose of the PTV were lower for RP than for MP (p = 0·006 and p = 0·040, respectively).V50 Gy, V65 Gy and the mean dose to rectum were lower in RP than in MP (p < 0·001). V65 Gy, V70 Gy and the mean dose to bladder were lower in RP than in MP (p < 0·001). RP had enhanced the sparing of both femurs (p < 0·001) and significantly reduced the planning time to less than 5% of the time taken with MP. MU in RP was significantly higher than MP by an average of 52·5 MU (p < 0·001) and 46 out of the 50 RP plans were approved by the radiation oncologist.

Conclusion:

This study has demonstrated that VMAT plans generated using an in-house RP prostate model in a single optimisation for prostate patients were clinically acceptable with comparable or better plan quality compared to MP. RP can add value and improve treatment planning efficiency in a high-throughput radiotherapy department through reduced plan optimisation time while maintaining consistency in the plan quality.

Harmonization of breast cancer radiotherapy treatment planning in the Netherlands

PURPOSE

The aim was to reach consensus in The Netherlands on which parameters should be used to evaluate breast cancer radiotherapy (RT) plans.

MATERIALS AND METHODS

A Benchmark Case with delineated planning target volumes (PTVs) and Organs At Risk (OARs) was sent to all Dutch radiotherapy centres in combination with a questionnaire, with the request to generate RT plans prescribing 15 times 2.67 Gy for four different treatment indications according to the institutional irradiation technique. The plans and accompanying questionnaire answers were analysed using descriptive statistics. These results, together with a harmonisation proposal, were sent to all centres. The proposal was discussed at a meeting of the Dutch Society of Radiation Oncology breast cancer platform. Distinct parameters were accepted if consensus on them was reached.

RESULTS

19 out of 20 Dutch departments participated in this study. PTV coverage varied considerably, with D98% between 63% and 99% for the breast and between 37% and 97% for the internal mammary nodes (IMN). Also substantial OAR dose differences were observed, with e.g. mean heart doses ranging between 1.85 Gy and 5.42 Gy in case the IMN were included in the PTV. For evaluation of the PTVs D98%, D2% and Dmean were chosen to report on, with target values of ≥ 95% (90% for the PTV_IMN), ≤ 107%, and 99-101%, respectively. For OARs, consensus was reached on the parameters to be evaluated, without target values: Dmean of the heart, Dmean and V5% of the lungs, and in case of periclavicular radiotherapy V30Gy of the thyroid gland. For patients younger than 40 years a contralateral mean breast dose of ≤ 1 Gy was agreed upon.

CONCLUSION

A new Dutch consensus guideline for evaluation of breast cancer RT plans has been established.

Knowledge-based radiation treatment planning: A data-driven method survey

This paper surveys the data-driven dose prediction methods investigated for knowledge-based planning (KBP) in the last decade. These methods were classified into two major categories-traditional KBP methods and deep-learning (DL) methods-according to their techniques of utilizing previous knowledge. Traditional KBP methods include studies that require geometric or anatomical features to either find the best-matched case(s) from a repository of prior treatment plans or to build dose prediction models. DL methods include studies that train neural networks to make dose predictions. A comprehensive review of each category is presented, highlighting key features, methods, and their advancements over the years. We separated the cited works according to the framework and cancer site in each category. Finally, we briefly discuss the performance of both traditional KBP methods and DL methods, then discuss future trends of both data-driven KBP methods to dose prediction.

The status of medical physics in radiotherapy in China

PURPOSE

To present an overview of the status of medical physics in radiotherapy in China, including facilities and devices, occupation, education, research, etc. MATERIALS AND METHODS: The information about medical physics in clinics was obtained from the 9-th nationwide survey conducted by the China Society for Radiation Oncology in 2019. The data of medical physics in education and research was collected from the publications of the official and professional organizations.

RESULTS

By 2019, there were 1463 hospitals or institutes registered to practice radiotherapy and the number of accelerators per million population was 1.5. There were 4172 medical physicists working in clinics of radiation oncology. The ratio between the numbers of radiation oncologists and medical physicists is 3.51. Approximately, 95% of medical physicists have an undergraduate or graduate degrees in nuclear physics and biomedical engineering. 86% of medical physicists have certificates issued by the Chinese Society of Medical Physics. There has been a fast growth of publications by authors from mainland of China in the top international medical physics and radiotherapy journals since 2018.

CONCLUSIONS

Demand for medical physicists in radiotherapy increased quickly in the past decade. The distribution of radiotherapy facilities in China became more balanced. High quality continuing education and training programs for medical physicists are deficient in most areas. The role of medical physicists in the clinic has not been clearly defined and their contributions have not been fully recognized by the community.

Dosimetric Evaluation of Simplified Knowledge-Based Plan with an Extensive Stepping Validation Approach in Volumetric-Modulated Arc Therapy-Stereotactic Body Radiotherapy for Lung Cancer

Purpose:

We investigated the performance of the simplified knowledge-based plans (KBPs) in stereotactic body radiotherapy (SBRT) with volumetric-modulated arc therapy (VMAT) for lung cancer.

Materials and Methods:

For 50 cases who underwent SBRT, only three structures were registered into knowledge-based model: total lung, spinal cord, and planning target volume. We performed single auto-optimization on VMAT plans in two steps: 19 cases used for the model training (closed-loop validation) and 16 new cases outside of training set (open-loop validation) for TrueBeam (TB) and Halcyon (Hal) linacs. The dosimetric parameters were compared between clinical plans (CLPs) and KBPs: CLP_closed, KBP_closed-TB and KBP_closed-Hal in closed-loop validation, CLP_open, KBP_open-TB and KBP_open-Hal in open-loop validation.

Results:

All organs at risk were comparable between CLPs and KBPs except for contralateral lung: V₅ of KBPs was approximately 3%–7% higher than that of CLPs. V₂₀ of total lung for KBPs showed comparable to CLPs; CLP_closed vs. KBP_closed-TB and CLP_closed vs. KBP_closed-Hal: 4.36% ± 2.87% vs. 3.54% ± 1.95% and 4.36 ± 2.87% vs. 3.54% ± 1.94% (P = 0.54 and 0.54); CLP_open vs. KBP_open-TB and CLP_open vs. KBP_open-Hal: 4.18% ± 1.57% vs. 3.55% ± 1.27% and 4.18% ± 1.57% vs. 3.67% ± 1.26% (P = 0.19 and 0.27). CI₉₅ of KBPs with both linacs was superior to that of the CLP in closed-loop validation: CLP_closed vs. KBP_closed-TB vs. KBP_closed-Hal: 1.32% ± 0.12% vs. 1.18% ± 0.09% vs. 1.17% ± 0.06% (P < 0.01); and open-loop validation: CLP_open vs. KBP_open-TB vs. KBP_open-Hal: 1.22% ± 0.09% vs. 1.14% ± 0.04% vs. 1.16% ± 0.05% (P ≤ 0.01).

Conclusions:

The simplified KBPs with limited number of structures and without planner intervention were clinically acceptable in the dosimetric parameters for lung VMAT-SBRT planning.

Challenges in lung and heart avoidance for postmastectomy breast cancer radiotherapy: Is automated planning the answer?

Postmastectomy radiotherapy (PMRT) has been shown to improve the overall survival for invasive breast cancer patients. However, it represents a challenging treatment geometry and individualized planning strategies with complex field arrangements are usually adopted to decrease radiotoxicity to heart and lungs. Automated treatment planning has the potential to improve plan quality consistency and planning efficiency. Herein, we describe the application of the Pinnacle³ Autoplanning engine as a valuable technological resource able to allow the treatment of challenging patients theoretically unfit for radiotherapy for major cardiac and pulmonary comorbidities. Treatment was planned for three left-sided chest wall and regional lymph-nodes postmastectomy breast cancer patients. A deep inspiration breath-hold (DIBH) technique was used aiming to reduce the OARs irradiation. Three manually generated plans (hybrid-IMRT (HMRT), hybrid-VMAT (HVMAT) and full VMAT (MP-VMAT) and a fully automated plan created by the Autoplanning engine (AP-VMAT) were optimized in order to ensure a safe radiation therapy to the patients. The plans were evaluated based on planning target volumes (PTVs) coverage, dose homogeneity index (HI), conformity index (CN), dose to organs at risk (OARs) and normal tissue complication probabilities (NTCPs) of pericarditis, long term mortality and pneumonitis. Despite the use of deep moderated breath-hold, all human-driven plans failed to reach the stringent dose objectives for OARs. All plans provided an optimal coverage for chest wall and lymph-nodal area. AP-VMAT delivered the lowest mean dose to the heart (3.4 to 4.9 Gy) and ipsilateral lung (7.5 to 12.5 Gy) reporting the lowest NTCP for pneumonitis (<1%), confirming the only chance to comply the dose objectives. Moreover, AP-VMAT reported a decrease of the integral dose, which was lower by about 4-8% with respect to manual plans. AP-VMAT plan resulted in up to 58% increase of MUs with respect to manual plans, suggesting a more pronounced fluence modulation and plan complexity. A major difference was found for the planning time which was reduced to less than 30 minutes by using the Auto-Planning module. With improved planning quality and efficiency, Auto-planning is an effective tool to enable high-quality plans in challenging postmastectomy breast cancer radiotherapy.

Dose-volume Histogram Analysis of Knowledge-based Volumetric-modulated Arc Therapy Planning in Postoperative Breast Cancer Irradiation

BACKGROUND/AIM

We evaluated the dosimetric profiles of manually generated volumetric-modulated arc therapy (VMAT) plans and performance of a commercial knowledge-based planning system (KBP) in treating breast cancer.

MATERIALS AND METHODS

We defined the manually generated VMAT plan as the manual plan (MP). Twenty MPs were generated for left-sided breast cancer patients who underwent breast-conserving surgery and used to develop a KBP training set. The other five patients were used for validation. The dosimetric parameters among MPs, tangential irradiation plans (TPs), and KBP-VMAT plans (KBP-Ps) were compared.

RESULTS

D₉₅ and homogeneity of the planning target volume (PTV) were significantly higher and greater in MPs and KBP-Ps than in TPs. Lung V₂₀, V₄₀ The Dmean for the left anterior descending artery was lower in MPs and KBP-Ps than in TPs. KBP could save time in generating VMAT plans.

CONCLUSION

MPs and KBP-Ps could ensure higher dose uniformity of PTV than TPs. KBP could faster generate comparable MPs for breast cancer.

Knowledge-based planning in robotic intracranial stereotactic radiosurgery treatments

Purpose

To develop a knowledge‐based planning (KBP) model that predicts dosimetric indices and facilitates planning in CyberKnife intracranial stereotactic radiosurgery/radiotherapy (SRS/SRT).

Methods

Forty CyberKnife SRS/SRT plans were retrospectively used to build a linear KBP model which correlated the equivalent radius of the PTV (r_{eq_PTV}) and the equivalent radius of volume that receives a set of prescription dose (r_{eq_Vi}, where V_i = V_10%, V_20% … V_120%). To evaluate the model’s predictability, a fourfold cross‐validation was performed for dosimetric indices such as gradient measure (GM) and brain V_50%. The accuracy of the prediction was quantified by the mean and the standard deviation of the difference between planned and predicted values, (i.e., ΔGM = GM_pred − GM_clin and fractional ΔV_50% = (V_50%pred − V_50%clin)/V_50%clin) and a coefficient of determination, R². Then, the KBP model was incorporated into the planning for another 22 clinical cases. The training plans and the KBP test plans were compared in terms of the new conformity index (nCI) as well as the planning efficiency.

Results

Our KBP model showed desirable predictability. For the 40 training plans, the average prediction error from cross‐validation was only 0.36 ± 0.06 mm for ΔGM, and 0.12 ± 0.08 for ΔV_50%. The R² for the linear fit between r_{eq_PTV} and r_{eq_vi} was 0.985 ± 0.019 for isodose volumes ranging from V_10% to V_120%; particularly, R² = 0.995 for V_50% and R² = 0.997 for V_100%. Compared to the training plans, our KBP test plan nCI was improved from 1.31 ± 0.15 to 1.15 ± 0.08 (P < 0.0001). The efficient automatic generation of the optimization constraints by using our model requested no or little planner’s intervention.

Conclusion

We demonstrated a linear KBP based on PTV volumes that accurately predicts CyberKnife SRS/SRT planning dosimetric indices and greatly helps achieve superior plan quality and planning efficiency.

Evaluation of a generalized knowledge-based planning performance for VMAT irradiation of breast and locoregional lymph nodes-Internal mammary and/or supraclavicular regions

PURPOSE

To evaluate the performance of eleven Knowledge-Based (KB) models for planning optimization (RapidPlantm (RP), Varian) of Volumetric Modulated Arc Therapy (VMAT) applied to whole breast comprehensive of nodal stations, internal mammary and/or supraclavicular regions.

METHODS AND MATERIALS

Six RP models have been generated and trained based on 120 VMAT plans data set with different criteria. Two extra-structures were delineated: a PTV for the optimization and a ring structure. Five more models, twins of the previous models, have been created without the need of these structures.

RESULTS

All models were successfully validated on an independent cohort of 40 patients, 30 from the same institute that provided the training patients and 10 from an additional institute, with the resulting plans being of equal or better quality compared with the clinical plans. The internal validation shows that the models reduce the heart maximum dose of about 2 Gy, the mean dose of about 1 Gy and the V20Gy of 1.5 Gy on average. Model R and L together with model B without optimization structures ensured the best outcomes in the 20% of the values compared to other models. The external validation observed an average improvement of at least 16% for the V5Gy of lungs in RP plans. The mean heart dose and for the V20Gy for lung IPSI were almost halved. The models reduce the maximum dose for the spinal canal of more than 2 Gy on average.

CONCLUSIONS

All KB models allow a homogeneous plan quality and some dosimetric gains, as we saw in both internal and external validation. Sub-KB models, developed by splitting right and left breast cases or including only whole breast with locoregional lymph nodes, have shown good performances, comparable but slightly worse than the general model. Finally, models generated without the optimization structures, performed better than the original ones.

Rolling out RapidPlan: What we've learnt

INTRODUCTION

RapidPlan (RP), a knowledge-based planning system, aims to consistently improve plan quality and efficiency in radiotherapy. During the early stages of implementation, some of the challenges include knowing how to optimally train a model and how to integrate RP into a department. We discuss our experience with the implementation of RP into our institution.

METHODS

We reviewed all patients planned using RP over a 7-month period following inception in our department. Our primary outcome was clinically acceptable plans (used for treatment) with secondary outcomes including model performance and a comparison of efficiency and plan quality between RP and manual planning (MP).

RESULTS

Between November 2017 and May 2018, 496 patients were simulated, of which 217 (43.8%) had an available model. RP successfully created a clinically acceptable plan in 87.2% of eligible patients. The individual success of the 24 models ranged from 50% to 100%, with more than 90% success in 15 (62.5%) of the models. In 40% of plans, success was achieved on the 1st optimisation. The overall planning time with RP was reduced by up to 95% compared with MP times. The quality of the RP plans was at least equivalent to historical MP plans in terms of target coverage and organ at risk constraints.

CONCLUSION

While initially time-consuming and resource-intensive to implement, plans optimised with RP demonstrate clinically acceptable plan quality, while significantly improving the efficiency of a department, suggesting RP and its application is a highly effective tool in clinical practice.

Comparative Study of Auto Plan and Manual Plan for Nasopharyngeal Carcinoma Intensity-Modulated Radiation Therapy

Purpose and Objective

Auto planning might reduce the manual time required for the optimization and could also potentially improve the overall plan quality. The aim of this study is to demonstrate the statistical comparison of automatic (AU) and manually (MA) generated nasopharyngeal carcinoma (NPC) intensity-modulated radiation therapy (IMRT) plans.

Materials and Methods

The study included 105 nasopharyngeal carcinoma patients, admitted to our hospital. The patients underwent IMRT treatments. The clinically delivered plans were performed with Eclipse (Version 11.0) using manual optimization. The same plans were optimized successively in Pinnacle^TM3 (version 9.10) treatment planning system using the auto plan software package module. D95 (dose of 95% volume) and D98 (dose of 98% volume) were calculated for the targets and maximum dose (Dmax) and mean dose (Dmean) for the organ at risks (OARs); moreover, the average doses of each target and OARs for 105 patients were evaluated.

Results

There is no significant difference in the homogeneity of the target between AU and MA treatment plans, while a significant difference is observed for what is concerning the OARs or most of OARs in 105 patients, OAR doses were significantly reduced in AU plan. For OARs which have no significant difference between AU and MA plans are highlighted, the mean dose of OARs in AU plans was at least not higher than MA plans.

Conclusion

Nasopharyngeal carcinoma IMRT plans made by an automatic planning tool met the clinical requirements for target prescription dose; moreover, the dose of normal tissues was lower than in MA plans. Clinical physicists' time can be saved and the influence of factors such as the lack of experience in treatment planning can be avoided.

Cautiously optimistic: A survey of radiation oncology professionals' perceptions of automation in radiotherapy planning

INTRODUCTION

While there is evidence to show the positive effects of automation, the impact on radiation oncology professionals has been poorly considered. This study examined radiation oncology professionals' perceptions of automation in radiotherapy planning.

METHOD

An online survey link was sent to the chief radiation therapists (RT) of all Australian radiotherapy centres to be forwarded to RTs, medical physicists (MP) and radiation oncologists (RO) within their institution. The survey was open from May-July 2019.

RESULTS

Participants were 204 RTs, 84 MPs and 37 ROs (response rates ∼10% of the overall radiation oncology workforce). Respondents felt automation resulted in improvement in consistency in planning (90%), productivity (88%), quality of planning (57%), and staff focus on patient care (49%). When asked about perceived impact of automation, the responses were; will change the primary tasks of certain jobs (66%), will allow staff to do the remaining components of their job more effectively (51%), will eliminate jobs (20%), and will not have an impact on jobs (6%). 27% of respondents believe automation will reduce job satisfaction. 71% of respondents strongly agree/agree that automation will cause a loss of skills, while only 25% strongly agree/agree that the training and education tools in their department are sufficient.

CONCLUSION

Although the effect of automation is perceived positively, there are some concerns on loss of skillsets and the lack of training to maintain this. These results highlight the need for continued education to ensure that skills and knowledge are not lost with automation.

Automated Intensity Modulated Radiation Therapy Treatment Planning for Cervical Cancer Based on Convolution Neural Network

Purpose:

To develop and evaluate an automatic intensity-modulated radiation therapy (IMRT) program for cervical cancer, including a Convolution Neural Network (CNN)-based prediction model and an automated optimization strategy.

Methods:

A CNN deep learning model was trained to predict a patient-specify set of IMRT objectives based on overlap volume histograms (OVH) and high-quality plan of previous patients. A total of 140 cervical cancer patients were enrolled in this study, including 100 patients in the training set, 20 patients in the validation set and 20 patients in the testing set. The input of this model was OVH data and the output were values of IMRT plan objectives. For patients in the testing set, the set of planning objectives were predicted by the CNN model and used to automatically generate IMRT plans. Meanwhile, manual plans of these patients were generated by 1 beginner planner and 1 senior planner respectively. Finally, dose distribution, dosimetric parameters and planning time were analyzed. In addition, the 3 types of plans were blinded compared and ranked by an experienced oncologist.

Results:

There were almost no statistically differences among these 3 types of plans in target coverage and dose conformity. Dose homogeneity were slightly decreased while the average dose and parameters for most organs-at-risk (OARs) were decreased in automatic plans. Especially in comparison with manual plans by the beginner planner, V₄₀ of bladder and rectum decreased 6.3% and 12.3%, while mean dose of rectum and marrow were 1.1 Gy and 1.8 Gy lower with automatic plans (either P < 0.017). In the blinded comparison, automatic plans were chosen as best plan in 14 cases.

Conclusions:

For cervical cancer, automatic IMRT plans optimized from the CNN generated objectives have superior dose sparing without compromising of target dose. It significantly reduced the planning time.

Virtual Tangential-fields Arc Therapy (ViTAT) for whole breast irradiation: Technique optimization and validation

PURPOSE

To test the performances of a volumetric arc technique named ViTAT (Virtual Tangential-fields Arc Therapy) mimicking tangential field irradiation for whole breast radiotherapy.

METHODS

ViTAT plans consisted in 4 arcs whose starting/ending position were established based on gantry angle distribution of clinical plans for right and left-breast. The arcs were completely blocked excluding the first and last 20°. Different virtual bolus densities and thicknesses were preliminarily evaluated to obtain the best plan performances. For 40 patients with tumor laterality equally divided between right and left sides, ViTAT plans were optimized considering the clinical DVHs for OARs (resulting from tangential field manual planning) to constrain them: ViTAT plans were compared with the clinical tangential-fields in terms of DVH parameters for both PTV and OARs.

RESULTS

Distal angle values were suggested in the ranges [220°,240°] for the right-breast and [115°,135°] for the left-breast cases; medial angles were [60°,40°] for the right side and [295°,315°] for the left side, limiting the risk of collision. The optimal virtual bolus had -500 HU density and 1.5 cm thickness. ViTAT plans generated dose distributions very similar to the tangential-field plans, with significantly improved PTV homogeneity. The mean doses of ipsilateral OARs were comparable between the two techniques with minor increase of the low-dose spread in the range 2-15 Gy (few % volume); contralateral OARs were slightly better spared with ViTAT.

CONCLUSION

ViTAT dose distributions were similar to tangential-fields. ViTAT should allow automatic plan optimization by developing knowledge-based DVH prediction models of patients treated with tangential-fields.

Urethra-Sparing Stereotactic Body Radiation Therapy for Prostate Cancer: Quality Assurance of a Randomized Phase 2 Trial

PURPOSE

To present the radiation therapy quality assurance results from a prospective multicenter phase 2 randomized trial of short versus protracted urethra-sparing stereotactic body radiation therapy (SBRT) for localized prostate cancer.

METHODS AND MATERIALS

Between 2012 and 2015, 165 patients with prostate cancer from 9 centers were randomized and treated with SBRT delivered either every other day (arm A, n = 82) or once a week (arm B, n = 83); 36.25 Gy in 5 fractions were prescribed to the prostate with (n = 92) or without (n = 73) inclusion of the seminal vesicles (SV), and the urethra planning-risk volume received 32.5 Gy. Patients were treated either with volumetric modulated arc therapy (VMAT; n = 112) or with intensity modulated radiation therapy (IMRT; n = 53). Deviations from protocol dose constraints, planning target volume (PTV) homogeneity index, PTV Dice similarity coefficient, and number of monitor units for each treatment plan were retrospectively analyzed. Dosimetric results of VMAT versus IMRT and treatment plans with versus without inclusion of SV were compared.

RESULTS

At least 1 major protocol deviation occurred in 51 patients (31%), whereas none was observed in 41. Protocol violations were more frequent in the IMRT group (P < .001). Furthermore, the use of VMAT yielded better dosimetric results than IMRT for urethra planning-risk volume D_98% (31.1 vs 30.8 Gy, P < .0001), PTV D_2% (37.9 vs 38.7 Gy, P < .0001), homogeneity index (0.09 vs 0.10, P < .0001), Dice similarity coefficient (0.83 vs 0.80, P < .0001), and bladder wall V_50% (24.5% vs 33.5%, P = .0001). To achieve its goals volumetric modulated arc therapy required fewer monitor units than IMRT (2275 vs 3378, P <.0001). The inclusion of SV in the PTV negatively affected the rectal wall V_90% (9.1% vs 10.4%, P = .0003) and V_80% (13.2% vs 15.7%, P = .0003).

CONCLUSIONS

Protocol deviations with potential impact on tumor control or toxicity occurred in 31% of patients in this prospective clinical trial. Protocol deviations were more frequent with IMRT. Prospective radiation therapy quality assurance protocols should be strongly recommended for SBRT trials to minimize potential protocol deviations.

Evaluation of automatic VMAT plans in locally advanced nasopharyngeal carcinoma

OBJECTIVE

This study aimed to evaluate the quality of locally advanced nasopharyngeal carcinoma (NPC) radiotherapy plans generated by the automated planning module of a commercial treatment planning system (TPS).

METHODS

Data of 30 patients with locally advanced NPC were retrospectively investigated. For each patient, volumetric modulated arc therapy (VMAT) plans with double arcs were generated manually by experienced physicists and automatically in the Pinnacle³ Auto-Planning module (Philips Medical Systems, Fitchburg, WI, USA). The anatomic distance between the second clinical target volume (CTV₂) and the pons of the brainstem, and the T category of disease were factored into the evaluation. Dosimetric verification was evaluated in terms of gamma pass rate. Target coverage, sparing of organs at risk (OARs), and monitor units were evaluated and compared between the manual and automatic VMAT plans.

RESULTS

Not all treatment plans fully met the dose objectives for planning target volumes (PTVs) and OARs, particularly in T4 patients. Overall, automatic VMAT provides a comparable or superior plan quality to manual VMAT in most cases. In stratified analysis, plan quality is mainly independent on T category but is also affected by anatomic distance. If the anatomic distance is less than 5 mm, the automatic VMAT plan quality is equal or even inferior to manual VMAT performed by experienced physicists. Conversely, if the anatomic distance is greater than 5 mm, the automatic VMAT plan quality is superior to manual VMAT. Gamma pass rates for quality assurance are similar between manual and automatic VMAT plans for the former case, but significantly higher in automatic VMAT for the latter.

CONCLUSION

The selection of manual versus automatic VMAT planning in locally advanced NPC should be made individually based on the anatomic distance, rather than blindly and habitually, since automatic VMAT is not good enough to completely replace manual VMAT.

Migration of treatment planning system using existing commissioned planning system

Introduction:

Commissioning of a new planning system involves extensive data acquisition which can be onerous involving significant clinic downtime. This could be circumvented by extracting data from existing treatment planning system (TPS) to speed up the process.

Material and methods:

In this study, commissioning beam data was obtained from a clinically commissioned TPS (Pinnacle™) using Matlab™ generated Pinnacle™ executable scripts to commission an independent 3D dose verification TPS (Eclipse™). Profiles and output factors for commissioning as required by Eclipse™ were computed on a 50 × 50 × 50 cm3 water phantom at a dose grid resolution of 2 mm3. Verification doses were computed and compared to clinical TPS dose profiles based on TG-106 guidelines. Standard patient plans from Pinnacle™ including intensity modulated radiation therapy and volumetric modulated arc therapy were re-computed on Eclipse™ TPS while maintaining the same monitor units. Computed dose was exported back to Pinnacle for comparison with the original plans. This methodology enabled us to alleviate all ambiguities that arise in such studies.

Results:

Profile analysis using in-house software showed that for all field sizes including small multi-leaf collimator-generated fields, >95% of infield and penumbra data points of Eclipse™ match Pinnacle™ generated and measured profiles with 2%/2 mm gamma criteria. Excellent agreement was observed in the penumbra regions, with >95% of the data points passing distance to agreement criteria for complex C-shaped and S-shaped profiles. Dose volume histograms and isodose lines of patient plans agreed well to within a 0·5% for target coverage.

Findings:

Migration of TPS is possible without compromising accuracy or enduring the cumbersome measurement of commissioning data. Economising time for commissioning such a verification system or for migration of TPS can add great QA value and minimise downtime.

Adjuvant breast inversely planned intensity-modulated radiotherapy with simultaneous integrated boost for early stage breast cancer : Results from a phase II trial

PURPOSE

To report early toxicity and 5‑year clinical outcomes of adjuvant breast inversely planned intensity-modulated radiotherapy with simultaneously integrated boost (IMRT-SIB) after breast-conserving surgery for early stage breast cancer patients.

PATIENTS AND METHODS

In all, 467 patients including 406 invasive breast cancer and 61 ductal carcinoma in situ (DCIS) were enrolled in a single institutional phase II trial. All patients underwent IMRT-SIB treatment to irradiate the whole breast and the tumor bed. Doses to whole breast and surgical bed were 45 and 60 Gy, respectively, delivered in 25 fractions over 5 weeks. The grade of maximum acute skin toxicity during treatment was recorded. Lung toxicity was noted within 6 months and patient-reported cosmetic outcomes were recorded at the 12 month follow-up after the end of radiotherapy. Clinical outcomes were assessed during follow-up.

RESULTS

Median follow-up time was 5.46 years. Median age was 46 years old (range 22-70 years old). No patient with DCIS had a local recurrence or distant metastasis. Among 406 patients with invasive breast cancer, the unadjusted 5‑year actuarial rate of locoregional control was 98.7% (95% confidence interval [CI] 97.5-100), and distant metastasis-free survival 98.7% (95% CI 97.4-100), respectively. Acute skin toxicity was recorded at grade 0-1 in 76.5% of patients, and grade 2 in 23.5% of patients. None of these patients had grade 3 or more than grade 3 skin toxicity. Grade 1 pneumonitis was found in 25.3% of patients. Assessment of patient reported cosmetic outcomes at the 12 month follow-up showed good or excellent outcome in 86.5% of cases.

CONCLUSIONS

The use of inversely planned IMRT-SIB as part of breast-conserving therapy results in optimal 5‑year tumor control and minor early toxicities.

Validation of in-house knowledge-based planning model for advance-stage lung cancer patients treated using VMAT radiotherapy

OBJECTIVES

Radiotherapy plan quality may vary considerably depending on planner's experience and time constraints. The variability in treatment plans can be assessed by calculating the difference between achieved and the optimal dose distribution. The achieved treatment plans may still be suboptimal if there is further scope to reduce organs-at-risk doses without compromising target coverage and deliverability. This study aims to develop a knowledge-based planning (KBP) model to reduce variability of volumetric modulated arc therapy (VMAT) lung plans by predicting minimum achievable lung volume-dose metrics.

METHODS

Dosimetric and geometric data collected from 40 retrospective plans were used to develop KBP models aiming to predict the minimum achievable lung dose metrics via calculating the ratio of the residual lung volume to the total lung volume. Model accuracy was verified by replanning 40 plans. Plan complexity metrics were calculated using locally developed script and their effect on treatment delivery was assessed via measurement.

RESULTS

The use of KBP resulted in significant reduction in plan variability in all three studied dosimetric parameters V5, V20 and mean lung dose by 4.9% (p = 0.007, 10.8 to 5.9%), 1.3% (p = 0.038, 4.0 to 2.7%) and 0.9 Gy (p = 0.012, 2.5 to 1.6Gy), respectively. It also increased lung sparing without compromising the overall plan quality. The accuracy of the model was proven as clinically acceptable. Plan complexity increased compared to original plans; however, the implication on delivery errors was clinically insignificant as demonstrated by plan verification measurements.

CONCLUSION

Our in-house model for VMAT lung plans led to a significant reduction in plan variability with concurrent decrease in lung dose. Our study also demonstrated that treatment delivery verifications are important prior to clinical implementation of KBP models.

ADVANCES IN KNOWLEDGE

In-house KBP models can predict minimum achievable lung dose-volume constraints for advance-stage lung cancer patients treated with VMAT. The study demonstrates that plan complexity could increase and should be assessed prior to clinical implementation.

Dose reduction to organs at risk with deep-inspiration breath-hold during right breast radiotherapy: a treatment planning study

Background

The addition of regional nodal radiation (RNI) to whole breast irradiation for high risk breast cancer improves metastases free survival and new data suggests it contributes additional benefit to overall survival. Deep inspiration breath hold (DIBH) has been shown to reduce cardiac and pulmonary dose in the context of left-sided disease treated with or without RNI, yet few studies have investigated its utility for right-breast cancer. This study investigates the potential advantages of DIBH in local and locoregional radiotherapy for right-sided breast cancer.

Methods

Free-breathing (FB) and DIBH computed tomography datasets were obtained from twenty patients who previously underwent radiotherapy for left-sided breast cancer. Ten patients were retrospectively planned for whole right breast only irradiation and ten patients were planned for irradiation to the whole breast plus ipsilateral supra-clavicular (SC) nodes, with and without irradiation of the ipsilateral internal mammary nodes (IMN). Dose-volume metrics for the clinical target volume, lungs, heart, left anterior descending artery, right coronary artery (RCA) and liver were recorded. Differences between FB and DIBH plans were analysed using Wilcoxon signed-rank tests, with P < 0.05 considered statistically significant.

Results

DIBH increased the average total lung volume compared to FB in both breast only and breast plus RNI cohorts (P = 0.001). For the breast only group, there was no significant improvement in any ipsilateral lung dose-volume metric between FB and DIBH. However, for the breast plus RNI group, there was an improvement in ipsilateral lung mean dose (18.9 ± 3.2 Gy to 15.9 ± 2.3 Gy, P = 0.002) and V20Gy (45.3 ± 13.3% to 32.9 ± 9.4%, P = 0.002). In addition, DIBH significantly reduced the maximum dose to the RCA for RNI (11.6 ± 7.2 Gy to 5.6 ± 2.9 Gy, P = 0.03). Significant reductions in the liver V20Gy and maximum dose were observed in all cohorts during DIBH compared to FB.

Conclusions

DIBH is a promising approach for right-breast radiotherapy with considerable sparing of normal tissue, particularly when the ipsilateral IMNs are also irradiated.

Prediction of dose-volume histograms in nasopharyngeal cancer IMRT using geometric and dosimetric information

A method using both patient geometric and dosimetric information was proposed to predict dose-volume histograms (DVHs) of organs at risk (OARs) for a nasopharyngeal cancer (NPC) intensity-modulated radiation therapy (IMRT) plan. A total of 106 nine-field IMRT NPC plans were used in this study. Twenty-six plans were randomly selected as testing cases, and the remaining plans were used as the training data. A method employing geometric and dosimetric information was developed for OAR DVH prediction. The dosimetric information was derived from an initial dose calculation using a simple unoptimized conformal plan. The DVHs were also predicted using only the geometric information. The DVH prediction model was a generalized regression neural network (GRNN). Mean absolute error (MAE) and R ² values were introduced to evaluate DVH prediction accuracy. Significant differences in the DVH prediction accuracy were found between the method employing the geometric and dosimetric information and the method utilizing the geometric information for the brainstem (R ², 0.98 versus 0.95, p   =  0.007; MAE, 3.52% versus 7.19%, p   =  0.002), spinal cord (R ², 0.98 versus 0.96, p   <  0.001; MAE, 2.80% versus 4.36%, p   <  0.001), left optic nerve (R ², 0.90 versus 0.77, p   =  0.014; MAE, 3.07% versus 11.29%, p   =  0.025) and other organs. On average, the R ² value increased by ~6.7% and the MAE value decreased by ~46.7% after adding the dosimetric information to the DVH prediction. We developed a method for predicting DVHs of OARs in NPC IMRT plans by using geometric and dosimetric information. Adding dosimetric information can help predict the DVHs of OARs in NPC IMRT plans.

Inter-planner variation in treatment-plan quality of plans created with a knowledge-based treatment planning system

PURPOSE

This study aimed to clarify the inter-planner variation of plan quality in knowledge-based plans created by nine planners.

METHODS

Five hypofractionated prostate-only (HPO) volumetric modulated arc therapy (VMAT) plans and five whole-pelvis (WP) VMAT plans were created by each planner using a knowledge-based planning (KBP) system. Nine planners were divided into three groups of three planners each: Senior, Junior, and Beginner. Single optimization with only priority modification for all objectives was performed to stay within the dose constraints. The coefficients of variation (CVs) for dosimetric parameters were evaluated, and a plan quality metric (PQM) was used to evaluate comprehensive plan quality.

RESULTS

Lower CVs (<0.05) were observed at dosimetric parameters in the planning target volume for both HPO and WP plans, while the CVs in the rectum and bladder for WP plans (<0.91) were greater than those for HPO plans (<0.17). The PQM values of HPO plans for Cases1-5 (average ± standard deviation) were 41.2 ± 7.1, 40.9 ± 5.6, and 39.9 ± 4.6 in the Senior, Junior, and Beginner groups, respectively. For the WP plans, the PQM values were 51.9 ± 6.3, 47.5 ± 4.3, and 40.0 ± 6.6, respectively. The number of clinically acceptable HPO and WP plans were 13/15 and 11/15 in the Senior group, 13/15 and 10/15 plans in the Junior group, and 8/15 and 2/15 plans in the Beginner group, respectively.

CONCLUSION

Inter-planner variation in the plan quality with RapidPlan remains, especially for the complicated VMAT plans, due to planners' heuristics.

RapidPlan knowledge based planning: iterative learning process and model ability to steer planning strategies

Purpose

To determine if the performance of a knowledge based RapidPlan (RP) planning model could be improved with an iterative learning process, i.e. if plans generated by an RP model could be used as new input to re-train the model and achieve better performance.

Methods

Clinical VMAT plans from 83 patients presenting with head and neck cancer were selected to train an RP model, CL-1. With this model, new plans on the same patients were generated, and subsequently used as input to train a novel model, CL-2. Both models were validated on a cohort of 20 patients and dosimetric results compared. Another set of 83 plans was realised on the same patients with different planning criteria, by using a simple template with no attempt to manually improve the plan quality. Those plans were employed to train another model, TP-1. The differences between the plans generated by CL-1 and TP-1 for the validation cohort of patients were compared with respect to the differences between the original plans used to build the two models.

Results

The CL-2 model presented an improvement relative to CL-1, with higher R² values and better regression plots. The mean doses to parallel organs decreased with CL-2, while D_1% to serial organs increased (but not significantly). The different models CL-1 and TP-1 were able to yield plans according to each original strategy.

Conclusion

A refined RP model allowed the generation of plans with improved quality, mostly for parallel organs at risk and, possibly, also the intrinsic model quality.

Applications and limitations of machine learning in radiation oncology

Machine learning approaches to problem-solving are growing rapidly within healthcare, and radiation oncology is no exception. With the burgeoning interest in machine learning comes the significant risk of misaligned expectations as to what it can and cannot accomplish. This paper evaluates the role of machine learning and the problems it solves within the context of current clinical challenges in radiation oncology. The role of learning algorithms within the workflow for external beam radiation therapy are surveyed, considering simulation imaging, multimodal fusion, image segmentation, treatment planning, quality assurance, and treatment delivery and adaptation. For each aspect, the clinical challenges faced, the learning algorithms proposed, and the successes and limitations of various approaches are analyzed. It is observed that machine learning has largely thrived on reproducibly mimicking conventional human-driven solutions with more efficiency and consistency. On the other hand, since algorithms are generally trained using expert opinion as ground truth, machine learning is of limited utility where problems or ground truths are not well-defined, or if suitable measures of correctness are not available. As a result, machines may excel at replicating, automating and standardizing human behaviour on manual chores, meanwhile the conceptual clinical challenges relating to definition, evaluation, and judgement remain in the realm of human intelligence and insight.

Knowledge-based planning for intensity-modulated radiation therapy: A review of data-driven approaches

Purpose

Intensity‐Modulated Radiation Therapy (IMRT), including its variations (including IMRT, Volumetric Arc Therapy (VMAT), and Tomotherapy), is a widely used and critically important technology for cancer treatment. It is a knowledge‐intensive technology due not only to its own technical complexity, but also to the inherently conflicting nature of maximizing tumor control while minimizing normal organ damage. As IMRT experience and especially the carefully designed clinical plan data are accumulated during the past two decades, a new set of methods commonly termed knowledge‐based planning (KBP) have been developed that aim to improve the quality and efficiency of IMRT planning by learning from the database of past clinical plans. Some of this development has led to commercial products recently that allowed the investigation of KBP in numerous clinical applications. In this literature review, we will attempt to present a summary of published methods of knowledge‐based approaches in IMRT and recent clinical validation results.

Methods

In March 2018, a literature search was conducted in the NIH Medline database using the PubMed interface to identify publications that describe methods and validations related to KBP in IMRT including variations such as VMAT and Tomotherapy. The search criteria were designed to have a broad scope to capture relevant results with high sensitivity. The authors filtered down the search results according to a predefined selection criteria by reviewing the titles and abstracts first and then by reviewing the full text. A few papers were added to the list based on the references of the reviewed papers. The final set of papers was reviewed and summarized here.

Results

The initial search yielded a total of 740 articles. A careful review of the titles, abstracts, and eventually the full text and then adding relevant articles from reviewing the references resulted in a final list of 73 articles published between 2011 and early 2018. These articles described methods for developing knowledge models for predicting such parameters as dosimetric and dose‐volume points, voxel‐level doses, and objective function weights that improve or automate IMRT planning for various cancer sites, addressing different clinical and quality assurance needs, and using a variety of machine learning approaches. A number of articles reported carefully designed clinical studies that assessed the performance of KBP models in realistic clinical applications. Overwhelming majority of the studies demonstrated the benefits of KBP in achieving comparable and often improved quality of IMRT planning while reducing planning time and plan quality variation.

Conclusions

The number of KBP‐related studies has been steadily increasing since 2011 indicating a growing interest in applying this approach to clinical applications. Validation studies have generally shown KBP to produce plans with quality comparable to expert planners while reducing the time and efforts to generate plans. However, current studies are mostly retrospective and leverage relatively small datasets. Larger datasets collected through multi‐institutional collaboration will enable the development of more advanced models to further improve the performance of KBP in complex clinical cases. Prospective studies will be an important next step toward widespread adoption of this exciting technology.

Assessment of specific versus combined purpose knowledge based models in prostate radiotherapy

Knowledge‐based planning (KBP) can be used to improve plan quality, planning speed, and reduce the inter‐patient plan variability. KPB may also identify and reduce systematic variations in VMAT plans, something very important in multi‐institutional clinical trials. Training of a KBP library is a complex and difficult process, and models must be validated prior to their clinical use. The purpose of this work is to assess the quality of the treatment plans generated using a specific versus combined purpose model KBP library for prostate cancer. Seven KBP model libraries were created from a set of patients treated on various Institutional Review Board (IRB) approved protocols. All KBP libraries were validated using an independent set of twenty patients (half treated Pr: Prostate alone half treated PLN: prostate plus pelvic lymph nodes). Two models were tested on the Pr patients only, four tested on PLN patients only, and one tested on all patients. All plans were normalized such that at least 95% of the prostate planning target volume received 100% of the planned dose. The plans based on different model libraries were compared to each other and the expert clinical plan. For Pr plans there were almost no statistically significant differences (P < 0.008) between the plans types except conformity index (CI) with library plans better than the expert. For PLN plans, all model libraries in generally showed femur doses and CI better than the expert plans (P < 0.003). This study demonstrated that no large differences were observed between specific versus combined KBP model libraries in dosimetry of prostate cancer patients. This would allow for a fewer specific plans to be needed to create a model library. Further studies are needed to evaluate benefits of combined purpose model libraries for planning of complex sites such as head and neck cancer.

Automation in intensity modulated radiotherapy treatment planning-a review of recent innovations

Radiotherapy treatment planning of complex radiotherapy techniques, such as intensity modulated radiotherapy and volumetric modulated arc therapy, is a resource-intensive process requiring a high level of treatment planner intervention to ensure high plan quality. This can lead to variability in the quality of treatment plans and the efficiency in which plans are produced, depending on the skills and experience of the operator and available planning time. Within the last few years, there has been significant progress in the research and development of intensity modulated radiotherapy treatment planning approaches with automation support, with most commercial manufacturers now offering some form of solution. There is a rapidly growing number of research articles published in the scientific literature on the topic. This paper critically reviews the body of publications up to April 2018. The review describes the different types of automation algorithms, including the advantages and current limitations. Also included is a discussion on the potential issues with routine clinical implementation of such software, and highlights areas for future research.

Predicting the dose absorbed by organs at risk during intensity modulated radiation therapy for nasopharyngeal carcinoma

OBJECTIVE:

To develop a model for predicting the dose absorbed by organ at risk (OAR) during intensity modulated radiation therapy for nasopharyngeal carcinoma (NPC).

METHODS:

55 patients underwent intensity modulated radiation therapy for NPC. The OARs were divided into several suborgans, and SPSS software was used to evaluate multiple linear method for fitting the normalized volume for each suborgan, normalized mean dose (D_mean) D_nm (D_nm = D_mean/D_prescription), and normalized D_10%-D_100% values D_n10%-n100%(D_n10%-_n100% = D_10%-D_100%/D_prescription) for each OAR. Based on the Matlab software, the predicted D_n10%-n100% value was fitted to obtain the predicted DVH curve.

RESULTS:

The multiple linear fitting formulas revealed significant results for the oral cavity D_n100% (p = 0.017), the parotid gland D_n100% (p = 0.001), and the remaining OAR (all p < 0.0001). The correlation coefficients and p values indicated that the fitting formula was a good fit. The p values for the White test show that the prediction model is robust. This method was successfully used for verification cases.

CONCLUSION:

The present study provided a simple and effective model for predicting the dose absorbed by OAR for NPC.

ADVANCES IN KNOWLEDGE:

This method is a relatively simple mathematical model, just use prescription dose and V₀-V_n to predict the D_mean and D_10%-100%, which predict does not require buying new modules of treatment planning software or extracting the distance of each sampling point of the OAR with the dose information.

An interactive plan and model evolution method for knowledge-based pelvic VMAT planning

Purpose

To test if a RapidPlan DVH estimation model and its training plans can be improved interactively through a closed‐loop evolution process.

Methods and materials

Eighty‐one manual plans (P₀) that were used to configure an initial rectal RapidPlan model (M₀) were reoptimized using M₀ (closed‐loop), yielding 81 P₁ plans. The 75 improved P₁ (P₁₊) and the remaining 6 P₀ were used to configure model M₁. The 81 training plans were reoptimized again using M₁, producing 23 P₂ plans that were superior to both their P₀ and P₁ forms (P₂₊). Hence, the knowledge base of model M₂ composed of 6 P₀, 52 P₁₊, and 23 P₂₊. Models were tested dosimetrically on 30 VMAT validation cases (P_v) that were not used for training, yielding P_v(M₀), P_v(M₁), and P_v(M₂) respectively. The 30 P_v were also optimized by M_{2_new} as trained by the library of M₂ and 30 P_v(M₀).

Results

Based on comparable target dose coverage, the first closed‐loop reoptimization significantly (P < 0.01) reduced the 81 training plans’ mean dose to femoral head, urinary bladder, and small bowel by 2.65 Gy/15.63%, 2.06 Gy/8.11%, and 1.47 Gy/6.31% respectively, which were further reduced significantly (P < 0.01) in the second closed‐loop reoptimization by 0.04 Gy/0.28%, 0.18 Gy/0.77%, 0.22 Gy/1.01% respectively. However, open‐loop VMAT validations displayed more complex and intertwined plan quality changes: mean dose to urinary bladder and small bowel decreased monotonically using M₁ (by 0.34 Gy/1.47%, 0.25 Gy/1.13%) and M₂ (by 0.36 Gy/1.56%, 0.30 Gy/1.36%) than using M₀. However, mean dose to femoral head increased by 0.81 Gy/6.64% (M₁) and 0.91 Gy/7.46% (M₂) than using M₀. The overfitting problem was relieved by applying model M_{2_new}.

Conclusions

The RapidPlan model and its constituent plans can improve each other interactively through a closed‐loop evolution process. Incorporating new patients into the original training library can improve the RapidPlan model and the upcoming plans interactively.

A new strategy for volumetric-modulated arc therapy planning using AutoPlanning based multicriteria optimization for nasopharyngeal carcinoma

BACKGROUND

A new strategy for making the appropriate choice of the representative optimization parameters in planning processes and accurate selection criteria during Pareto surface navigation for general multicriteria optimization (MCO) was recommended in the study. The purpose was to combine both benefits of AutoPlanning optimization and MCO (APMCO) for achieving an individual volumetric-modulated arc therapy (VMAT) plan according to the clinically achieved patient-specific tradeoff among conflicting priorities. The preclinical investigation of this optimization approach for nasopharyngeal carcinoma (NPC) radiotherapy was performed and compared to general MCO VMAT.

METHODS

A total of 60 NPC patients with various stages were enrolled in this study. General MCO and APMCO plans were generated for each patient on the treatment planning system. The differences between two planning schemes were evaluated and compared.

RESULTS

All plans were capable of achieving the prescription requirement. The planning target volume coverage and conformation number were remarkably similar between general MCO and APMCO plans. There were no significant differences in most of organs at risk (OARs) sparing. However, in APMCO plans, relatively remarkable decreases were observed in the mean dose (D_mean) to the glottic larynx and pharyngeal constrictor muscles. The reductions of average D_mean to the two OARs were 10.5% (p < 0.0001) and 8.4% (p < 0.0001), respectively. APMCO technique was found to increase the planning time for an average of approximately 5 h and did not lead to a significant increase of monitor units compared to general MCO.

CONCLUSIONS

The potential of the APMCO strategy is best realized with a clinical implementation that exploits individual generation of Pareto surface representations without manual interaction. It also assists physicians to ensure navigation in a more efficient and straightforward manner.