Background Knowledge-based planning (KBP) improves radiotherapy efficiency and consistency by using machine learning models trained on prior high-quality plans. Most breast KBP studies focus on whole- or partial-breast treatments without nodal coverage, and limited attention has been given to laterality-specific chest wall volumetric modulated arc therapy (VMAT) plans. This study develops left-sided, right-sided, and combined KBP models for chest wall and regional nodal irradiation and compares their performance with clinical plans. Materials and methods KBP models were created using 47 left-sided and 44 right-sided chest wall patients involving regional lymph nodes. A combined model incorporating all cases was also developed. Optimization objectives were iteratively refined using model-predicted, manual, and normal tissue objectives (NTOs). Model performance was evaluated using the coefficient of determination (R²), chi-square (χ²), and mean squared error (MSE). For validation, 10 left-sided, 10 right-sided, and 20 combined KBP plans were compared with their corresponding clinical plans using paired t-tests (p < 0.05). Dosimetric endpoints included target coverage, conformity index, homogeneity index, organ-at-risk (OAR) dose-volume metrics, and delivery efficiency factor (intensity modulated radiation therapy factor). KBP plans were generated without planner intervention, with additional optimization using a monitor unit objective when required. All plans underwent blinded physician review for clinical acceptability and preference. Results Left-sided KBP plans significantly reduced mean doses to the esophagus (-499 ± 147 cGy, p < 0.05) and thyroid (-296 ± 96 cGy, p < 0.05), with small increases in spinal cord maximum dose (20 ± 82 cGy, p > 0.05) and V15% of the contralateral lung (10.1 ± 1.9, p < 0.05). Right-sided KBP plans reduced mean doses to the thyroid (-220 ± 86 cGy, p < 0.05) and heart (-46.1 ± 9.3 cGy, p < 0.05), with minimal impact on coverage of the planning target volume of the internal mammary nodes (PTV_IMN). The combined model demonstrated similar dosimetric patterns. KBP plans improved conformity in six of 10 cases and homogeneity in 50% of cases, achieving higher dose values corresponding to 98% of the prescription dose (D98%) in 60-65% of cases for laterality-specific and combined KBP models, respectively, while maintaining the maximum dose (Dmax) and the dose received by 105% of the prescription dose (D105%) within constraints. Delivery efficiency remained comparable. A blinded physician review favored KBP in the majority of cases, with equivalence noted in others. Conclusions Laterality-specific and combined KBP models for VMAT chest wall and nodal irradiation generate plans that are dosimetrically comparable or superior to manual plans, providing consistent target coverage and improved or maintained OAR sparing. KBP offers a reproducible and efficient strategy for complex breast and chest wall cases, supports workflow standardization, and requires only minimal refinement in selected situations.

Introduction In this work, we aimed to create and assess the performance of a knowledge-based planning (KBP) model for optimizing intensity-modulated proton therapy (IMPT) in the treatment of prostate cancer involving pelvic lymph nodes (LNs). Materials and methods Fifty patients previously treated with IMPT to the prostate/prostate bed, including LNs and optional gross tumor volume (GTV) boost, were used for the training of a KBP model. The model was iteratively refined by replanning a subset of 20 of these patients. For validation, 20 patients not included in the model training set were used. Treatment plans were optimized using the objective list predicted by the model. Plan quality was evaluated using dosimetric metrics for both target and organs at risk (OARs), and the results were compared with manually generated plans using paired t-tests (p < 0.05). Results Eighteen out of 20 plans generated by the model were deemed to be clinically acceptable without the need for additional adjustments. The plans produced by the model demonstrated comparable robustness in clinical target volume (CTV) coverage. Significant improvements in OAR sparing were achieved for the rectum (V40Gy = -4.26 ± 3.00%), bladder (V40Gy = -6.36 ± 4.34%), and penile bulb (Dmean = -1.61 ± 9.76 Gy) when using the KBP model, compared to the manual plans. Other significant differences include slightly higher doses to the cauda equina (D0.03cc = 3.44 ± 6.09 Gy) and the left femur (D0.03cc = 2.50 ± 3.69 Gy) when compared to manual plans. No statistically significant differences were found for other OARs. Conclusions This study demonstrated that the KBP model produced plans comparable to manually generated clinical plans, and these plans are clinically acceptable. The iterative tuning process improved the quality of plans generated by the KBP model.

Background and purpose:Liver irradiations with intensity-modulated proton therapy (IMPT) often require motion mitigation techniques that prolong treatment. A prototype spot-optimization algorithm was tested to evaluate whether plan delivery time could be reduced while preserving quality. Methods and materials:Fifteen patients previously treated with liver IMPT using breath-hold were re-planned with nominal treatment planning system (TPS) settings and using a prototype spot-optimization algorithm in which combinations of minimum Monitor Unit (MU) and layer-spacing settings were tested: 1MU/1MeV, 3MU/3MeV, 1MU/5MeV, 5MU/3MeV. Spot-optimized and nominals plans were compared using standard dose-volume histogram (DVH) metrics for targets and organs-at-risk. A Wilcoxon signed-rank test was applied (p < 0.05). Delivery time for all plans were measured by creating and delivering IMPT quality assurance (QA) plans. Gamma analyses were performed on all plans to test deliverability. Plans were considered deliverable if >90 % of points passed a gamma criterion of 3 %/3mm. Results:Minimal DVH differences were observed between nominal and spot-optimized plans. For the 3MU/3MeV setting, no DVH metrics were significantly different. Median and interquartile range (IQR) delivery times for these plans were 40 % (38 %-44 %) faster than nominal plans. 5MU/3MeV plans had median (IQR) delivery times 59 % (52 %-61 %) faster than nominal plans but had a small but significant increase in LiverEff Dmean with a median (IQR) difference of 0.2 Gy(RBE) (0.0-0.4 Gy(RBE)). QA analysis showed all spot-optimized plans were deliverable. Conclusions:The spot-optimization algorithm produced clinically deliverable plans with negligible DVH differences to nominal plans and reduced delivery time of liver IMPT by over one-third.

PURPOSE:Knowledge-based planning (KBP) aims to automate and standardize treatment planning. New KBP users are faced with many questions: How much does model size matter, and are multiple models needed to accommodate specific physician preferences? In this study, six head-and-neck KBP models were trained to address these questions.METHODS:The six models differed in training size and plan composition: The KBPFull (n = 203 plans), KBP101 (n = 101), KBP50 (n = 50), and KBP25 (n = 25) were trained with plans from two head-and-neck physicians. KBPA and KBPB each contained n = 101 plans from only one physician, respectively. An independent set of 39 patients treated to 6000-7000 cGy by a third physician was re-planned with all KBP models for validation. Standard head-and-neck dosimetric parameters were used to compare resulting plans. KBPFull plans were compared to the clinical plans to evaluate overall model quality. Additionally, clinical and KBPFull plans were presented to another physician for blind review. Dosimetric comparison of KBPFull against KBP101 , KBP50 , and KBP25 investigated the effect of model size. Finally, KBPA versus KBPB tested whether training KBP models on plans from one physician only influences the resulting output. Dosimetric differences were tested for significance using a paired t-test (p < 0.05).RESULTS:Compared to manual plans, KBPFull significantly increased PTV Low D95% and left parotid mean dose but decreased dose cochlea, constrictors, and larynx. The physician preferred the KBPFull plan over the manual plan in 20/39 cases. Dosimetric differences between KBPFull , KBP101 , KBP50 , and KBP25 plans did not exceed 187 cGy on aggregate, except for the cochlea. Further, average differences between KBPA and KBPB were below 110 cGy.CONCLUSIONS:Overall, all models were shown to produce high-quality plans. Differences between model outputs were small compared to the prescription. This indicates only small improvements when increasing model size and minimal influence of the physician when choosing treatment plans for training head-and-neck KBP models.

Gantry beam angle selection was automated by a newly developed AI model and tested on 10 brain IMPT patients. The comparison showed that the AI often chooses similar beam angles to the human planners. When there are differences, dosimetric analysis demonstrated that plans created from AI beam angles have at least the same quality as the human ones. Results motivate further research into this approach showing the AI being a promising tool to fill a current gap in the strive for automating proton treatment planning, increasing planning efficiency and potentially quality.

BACKGROUND Dose deposition characteristics of proton radiation can be advantageous over photons. Proton treatment planning however poses additional challenges for the planners. Proton therapy is usually delivered with only a small number of beam angles, and the quality of a proton treatment plan is largely determined by the beam angles employed. Finding the optimal beam angles for a proton treatment plan requires time and experience, motivating the investigation of automatic beam angle selection methods. PURPOSE A deep learning-based approach to automatic beam angle selection is proposed for proton pencil-beam scanning treatment planning of liver lesions. METHODS We cast beam-angle selection as a multi-label classification problem. To account for angular boundary discontinuity, the underlying convolution neural network is trained with the proposed Circular Earth Mover's Distance based regularization and multi-label circular-smooth label technique. Furthermore, an analytical algorithm emulating proton treatment planners' clinical practice is employed in post-processing to improve the output of the model. Forty-nine patients that received proton liver treatments between 2017 and 2020 were randomly divided into training (n = 31), validation (n = 7), and test sets (n = 11). AI-selected beam angles were compared with those angles selected by human planners, and the dosimetric outcome was investigated by creating plans using knowledge-based treatment planning. RESULTS For 7 of the 11 cases in the test set, AI-selected beam angles agreed with those chosen by human planners to within 20 degrees (median angle difference = 10°; mean = 18.6°). Moreover, out of the total 22 beam angles predicted by the model, 15 (68%) were within 10 degrees of the human-selected angles. The high correlation in beam angles resulted in comparable dosimetric statistics between proton treatment plans generated using AI- and human-selected angles. For the cases with beam angle differences exceeding 20°, the dosimetric analysis showed similar plan quality although with different emphases on organ-at-risk sparing. CONCLUSIONS This pilot study demonstrated the feasibility of a novel deep learning-based beam angle selection technique. Testing on liver cancer patients showed that the resulting plans were clinically viable with comparable dosimetric quality to those using human-selected beam angles. In tandem with auto-contouring and knowledge-based treatment planning tools, the proposed model could represent a pathway for nearly fully automated treatment planning in proton therapy. This article is protected by copyright. All rights reserved.

PURPOSE:Our purpose was to investigate the effect of automated knowledge-based planning (KBP) on real-world clinical workflow efficiency, assess whether manual refinement of KBP plans improves plan quality across multiple disease sites, and develop a data-driven method to periodically improve KBP automated planning routines.METHODS AND MATERIALS:Using clinical knowledge-based automated planning routines for prostate, prostatic fossa, head and neck, and hypofractionated lung disease sites in a commercial KBP solution, workflow efficiency was compared in terms of planning time in a pre-KBP (n = 145 plans) and post-KBP (n = 503) patient cohort. Post-KBP, planning was initialized with KBP (KBP-only) and subsequently manually refined (KBP +human). Differences in planning time were tested for significance using a 2-tailed Mann-Whitney U test (P < .05, null hypothesis: planning time unchanged). Post-refinement plan quality was assessed using site-specific dosimetric parameters of the original KBP-only plan versus KBP +human; 2-tailed paired t test quantified statistical significance (Bonferroni-corrected P < .05, null hypothesis: no dosimetric difference after refinement). If KBP +human significantly improved plans across the cohort, optimization objectives were changed to create an updated KBP routine (KBP'). Patients were replanned with KBP' and plan quality was compared with KBP +human as described previously.RESULTS:KBP significantly reduced planning time in all disease sites: prostate (median: 7.6 hrs → 2.1 hrs; P < .001), prostatic fossa (11.1 hrs → 3.7 hrs; P = .001), lung (9.9 hrs → 2.0 hrs; P < .001), and head and neck (12.9 hrs → 3.5 hrs; P <.001). In prostate, prostatic fossa, and lung disease sites, organ-at-risk dose changes in KBP +human versus KBP-only were minimal (<1% prescription dose). In head and neck, KBP +human did achieve clinically relevant dose reductions in some parameters. The head and neck routine was updated (KBP'HN) to incorporate dose improvements from manual refinement. The only significant dosimetric differences to KBP +human after replanning with KBP'HN were in favor of the new routine.CONCLUSIONS:KBP increased clinical efficiency by significantly reducing planning time. On average, human refinement offered minimal dose improvements over KBP-only plans. In the single disease site where KBP +human was superior to KBP-only, differences were eliminated by adjusting optimization parameters in a revised KBP routine.

Quality assurance (QA) in radiotherapy (RT) clinical trials is essential to ensure protocol compliance, patient safety and trial quality. However, protocol compliance does not necessarily ensure optimal plan generation. This study aimed to demonstrate the feasibility and impact of Knowledge-Based Planning (KBP) feedback as part of the Real Time Review (RTR) process for the TROG 1501 SPARK trial (Stereotactic Prostate Adaptive RT Utilising Kilovoltage Intrafraction Monitoring). A knowledge based dose-volume histogram (DVH) estimation model and automated planning routine were created using 34 SPARK RT plans that had previously been submitted as part of the TROG QA program. The KBP routine was applied to 5 subsequent patients pre-treatment. A feedback report comparing the KBP generated DVH versus the initial plan was collated using a customised script and sent to the site within 24 hours. Centres were asked to review the report and decide whether they would amend their clinical plan. Of the 5 patients, 4 were protocol compliant and 1 case was replanned due to a major protocol deviation. As a result of KBP feedback 2/4 (50%) cases which were originally protocol compliant, were nevertheless replanned. Protocol dose constraints for all 5 cases were calculated and an average for each metric was generated. The mean dose-volume metrics were then compared between the initial submission, resubmission and KBP generated plans. Overall, the rectum, bladder, penile bulb and urethra planning risk volume (PRV) demonstrated that an improved dose-volume relationship could be achieved compared to the initial submission and was implemented in practice for the 3 resubmitted cases. Variable results were observed for the femoral heads, demonstrating a potential dose trade off (Table 1).Abstract TU_36_3254: Table 1Organ at Risk (OAR)ConstraintInitial N=5KBP N=5Initial (Resubmitted) N=3Resubmitted N = 3RectumV18.13Gy ≤50%21.6 (15.5 - 29.3)12.7 (8.8 – 20.9)23.1 (15.5 – 29.3)15.0 (10.8 – 19.2)V32.63Gy ≤5%2.6 (0.4 - 4.9)1.7 (0.0 – 3.9)2.0 (0.4 – 4.6)1.4 (0.1 – 3.1)BladderV18.13Gy ≤50%15.6 (3.5 - 29.6)14.7 (4.4 – 32.2)21.1 (16.1 – 29.6)19.5 (9.4 – 32.4)V32.63Gy ≤10%4.7 (0.9 - 8.5)4.8 (1.0 – 10.4)5.8 (3.6 – 8.5)5.4 (2.0 – 9.1)Urethra PRVD0.1cc ≤38.78Gy37.6 (37.1 - 37.9)37.0 (36.6 – 37.7)37.4 (37.1 – 37.5)37.2 (37.0 – 37.5)Penile BulbD0.1cc ≤36.25Gy13.4 (3.7 - 25.3)12.2 (3.5 – 21.5)14.4 (4.9 – 25.3)12.1 (6.3 – 20.1)Left FemHeadD0.1cc ≤30Gy15.5 (12.1 – 17.3)15.6 (13.6 – 17.5)14.8 (12.1 – 16.5)16.7 (14.3 – 20.0)Right FemHeadD0.1cc ≤30Gy16.2 (14.3 – 18.1)16.8 (14.4 – 20.1)16.2 (14.3 – 18.1)18.4 (13.5 – 21.5) Open table in a new tab KBP feedback was successfully incorporated into the RTR process for the SPARK trial and demonstrated that both improvements to and validation of plan quality for OAR dosimetry could be achieved. Further prospective investigation of the role of KBP in TROG clinical trials is planned.