Plan Quality Metrics Research Articles

A Novel Pre-plan Technique for On-table Adaptation of Pancreatic Stereotactic MR-guided Online Adaptive Radiotherapy

Currently on-table adaptive radiotherapy (ART) is performed by daily re-optimization of initial planning objectives developed from the initial simulation anatomy. Eliminating the initial plan creation may reduce start time, simulation and planning resources, and improve clinical efficacy. We hypothesize that for the treatment of pancreas SBRT using ART, adapting a generic pre-plan to the anatomy of the day results in dosimetrically equivalent plan quality of adapting a patient-specific initial plan to the anatomy of the day. Thirty-eight (n = 38) patients were treated for pancreatic cancer with MR-guided online adaptive radiotherapy (MRgoART) to 50 Gy/5 fractions (fx). An initial plan was developed over 8 days. For on-table adaptation, the initial plan was adapted to the anatomy of the day for each fx and delivered. Retrospectively, we developed a single template pre-plan technique based on optimization objectives from simulation anatomy across 38 patients. To assess feasibility of the pre-plan technique, we adapted the template pre-plan to the on-table fx 3 anatomy of the day (ADAPTtemplate-to-fx3) and compared the plan quality to the patient-specific initial plan adapted to fx 3 anatomy of the day (ADAPTsim-to-fx3) for all patients. Adapted plans were single-optimized and then normalized to greatest violated OAR. RTOG plan quality metrics were evaluated for target coverage (TC) (PTV V100%/PTV vol), homogeneity index (HI) (PTV D2%/ D98%), high dose conformity (PITV), low dose conformity (D2cm), and gradient (R50%). PTV coverage of V100%, D80% were evaluated. Seventy-six (n = 76) adaptions were performed for ADAPTtemplate-to-fx3 and ADAPTsim-to-fx3. Due to plan-normalization, no OARs were violated. PTV coverage was improved for respective ADAPTtemplate-to-fx3 compared to ADAPTsim-to-fx3 for 58% of cases for D80% (45.8±5.9 vs 45.3±9.1, p = 0.8), 61% of cases for TC (0.66±0.30 vs 0.55±0.30, p = 0.1), and 66% of cases for V100% (63±27% vs 55±32% p = 0.2). Homogeneity and conformity indices between respective ADAPTtemplate-to-fx3 and ADAPTsim-to-fx3 were 0.72±0.39 vs 0.60±0.41, p = 0.2 (PITV), 1.94±0.71 vs 1.52±0.22, p<0.001 (HI), 3.14±1.45 vs 3.52±1.80, p = 0.3 (R50%), and 0.61±0.12 vs 0.58±0.22, p = 0.5 (D2cm). For online adaptation in pancreas SBRT, 66% of the single-optimized template plan had improved coverage (PTV V100) with no OAR violation compared to the patient-specific single-optimized initial plan to the anatomy of the day. Given that the plan quality was not significantly different, this works supports AI-plan automation for patients receiving online ART, potentially eliminating the need for initial plan creation and even simulation. Future investigation of the feasibility and efficacy will be launched in a Phase I/II randomized clinical trial between template plan and initial plan for MRgoART.

Comparison of Cardiopulmonary Sparing and Dose Conformality in Distal Esophageal Cancer: Intensity Modulated Proton Therapy (IMPT), MR-guided Radiotherapy (MRgRT), and Volumetric Modulated Arc Therapy (VMAT)

To investigate the tradeoff of esophageal treatment between real-time breath hold (BH) MR-guided radiotherapy (MRgRT), free breathing (FB) CT-based intensity modulated proton therapy (IMPT), and FB VMAT. We hypothesize that improved cardiopulmonary sparing would be achieved with BH MRgRT or FB IMPT compared to FB VMAT. We retrospectively evaluated differences in heart/lung dose and treatment plan conformality among 28 patients with distal esophageal cancer who were each treated with either IMPT (n = 10), MRgRT on an MR-guided Linac (n = 11), or VMAT (n = 7). All were prescribed 50.4 Gy/28 fractions. IMPT and VMAT patients received a 4DCT simulation scan to quantify the internal target volume margin (ITV) for respiratory motion. MRgRT patients were simulated and treated in an inspiration BH with a 3 mm CTV to PTV margin. In lieu of a PTV margin for IMPT cases, robust optimization for range and setup uncertainty was performed on iCTV (i.e., CTV with respiratory ITV). RTOG plan quality metrics were used to evaluate target coverage (TC) (PTV V100%/PTV vol), homogeneity index (HI) (PTV D2%/ D98%), high dose conformity (PITV), low dose conformity (D2cm), and gradient (R50%). For all coverage metrics, the CTV was used. Mean cardiac sparing among all patients between respective IMPT, MRgRT, and VMAT was 65.3cc, 73.0cc, and 141.5cc (V30Gy); 45.4cc, 38.5cc, and 84.7cc (V40Gy), 19.7cc, 12.5cc, 39.0cc, (V50Gy), 6.4cc, 15.0cc, 19.2cc (mean dose). Lung and liver sparing between respective IMPT, MRgRT and VMAT plans was 86.8cc, 299.7cc, and 909.0cc (lung, V20), 4.2Gy, 6.8Gy, and 15.6Gy (lung, mean dose), 3.8Gy, 19.5Gy, 16.2Gy (liver, mean dose). Homogeneity and conformity indices between IMPT, MRgRT, and VMAT were 1.6, 1.3, 1.5 (PITV), 1.1, 1.2, 1.1 (HI), 5.2, 5.2, 6.3 (R50%), and 45.3%, 56.6%, 54.1%(D2cm), respectively. CTV coverages for IMPT and MRgRT plans were 50.7, 50.5, 51.1 (D90%) and 0.99, 0.99, 0.98 (TC), respectively. Statistically significant differences in heart and lung dose are demonstrated in the Table below; no statistical difference was observed for coverage or conformality. FB IMPT and BH MRgRT both achieve significant lung and/or heart sparing compared to FB VMAT while maintaining excellent plan conformality. The varying extent of cardiopulmonary sparing achieved across these advanced radiation therapy modalities warrants prospective evaluation.Abstract 3498; TableHeart (V30)Heart (V40)Heart (V50)Heart (Mean)Lung (Mean)Lung (V20)Liver (Mean)MRgRT vs IMPTp = 0.695p = 0.602p = 0.792p = 0.0001p = 0.079p = 0.001p = 0.0001MRgRT vs VMATp = 0.048p = 0.058p = 0.053p = 0.101p = 0.0001p = 0.001p = 0.258IMPT vs VMATp = 0.031p = 0.134p = 0.218p = 0.0001p = 0.0001p = 0.0002p = 0.001 Open table in a new tab

Dosimetric feasibility of hypofractionation for SBRT treatment of lymph node oligometastases on the 1.5T MR-linac

PurposeAt our department, MR-guided stereotactic body radiation therapy (SBRT) using the 1.5T MR-linac system (Unity, Elekta AB, Stockholm, Sweden) has been initiated for patients with lymph node oligometastases. Superior soft tissue contrast and the possibility for online plan adaptation on the Unity may allow for hypofractionated treatment. The purpose of this study was to investigate the dosimetric feasibility and compare the plan quality of different hypofractionated schemes. Methods and materialsData was used from 12 patients with single lymph node oligometastases (10 pelvic, 2 para-aortic), which were all treated on the Unity with a prescribed dose of 5x7 Gy to 95% of the PTV. Hypofractionation was investigated for 3x10 Gy and 1x20 Gy schemes (all 60 Gy BED α/β = 10). The pre-treatment plans were evaluated based on dose criteria and plan quality. If all criteria were met, the number of online adapted plans which also met all dose criteria was investigated. For pre-treatment plans meeting the criteria for all three fractionation schemes, the plan quality after online adaptation was compared using the four parameters described in the NRG-BR001 phase 1 trial. ResultsPre-treatment plans met all clinical criteria for the three different fractionation schemes in 10, 9 and 6 cases. 50/50, 45/45 17/30 of the corresponding online adapted plans met all criteria, respectively. Violations were primarily caused by surrounding organs at risk overlapping or adjacent to the PTV. The 1x20 Gy treatment plans were, in general, of lesser quality than the 5x7 Gy and 3x10 Gy plans. ConclusionHypofractionated radiotherapy for lymph node oligometastases on the 1.5T MR-linac is feasible based on dose criteria and plan quality metrics. The location of the target relative to critical structures should be considered in choosing the most suitable fractionation scheme. Especially for single fraction treatment, meeting all dose criteria in the pre-treatment situation does not guarantee that this also applies during online treatment.

IntuitivePlan inverse planning performance evaluation for Gamma Knife radiosurgery of AVMs.

PurposeTo compare planning indices achieved using manual and inverse planning approaches for Gamma Knife radiosurgery of arterio‐venous malformations (AVMs).Methods and materialsFor a series of consecutive AVM patients, treatment plans were manually created by expert planners using Leksell GammaPlan (LGP). Patients were re‐planned using a new commercially released inverse planning system, IntuitivePlan. Plan quality metrics were calculated for both groups of plans and compared.ResultsOverall, IntuitivePlan created treatment plans of similar quality to expert planners. For some plan quality metrics statistically significant higher scores were achieved for the inversely generated plans (Coverage 96.8% vs 96.3%, P = 0.027; PCI 0.855 vs 0.824, P = 0.042), but others did not show statistically significant differences (Selectivity 0.884 vs 0.856, P = 0.071; GI 2.85 vs 2.76, P = 0.096; Efficiency Index 47.0% vs 48.1%, P = 0.242; Normal Brain V12(cc) 5.81 vs 5.79, P = 0.497). Automatic inverse planning demonstrated significantly shorter planning times over manual planning (3.79 vs 11.58 min, P < 10−6) and greater numbers of isocentres (40.4 vs 10.8, P < 10−6), with an associated cost of longer treatment times (57.97 vs 49.52 min, P = 0.009). When planning and treatment time were combined, there was no significant difference in the overall time between the two methods (61.76 vs 61.10, P = 0.433).ConclusionsIntuitivePlan can offer savings on the labor of treatment planning. In many cases, it achieves higher quality indices than those achieved by an “expert planner”.

Limiting treatment plan complexity by applying a novel commercial tool.

PurposeA recently introduced commercial tool is tested to assess whether it is able to reduce the complexity of a treatment plan and improve deliverability without compromising overall quality.MethodsTen prostate and ten oropharynx plans of previously treated patients were reoptimized using the aperture shape controller (ASC) tool recently introduced in Eclipse TPS (Varian Medical Systems, Palo Alto, CA). The performance of ASC was assessed in terms of the overall plan quality using a plan quality metric, the reduction in plan complexity through the analysis of 14 of the most common plan complexity metrics, and the change in plan deliverability through 3D dosimetric measurements. Similarly, plans optimized limiting the total number of delivered monitor units was assessed and compared. The two strategies were also combined to assess their potential combination.ResultsThe plans optimized by exploiting the ASC generally show a reduced number of total Monitor Units, a more constant gantry rotation and a MLC modulation characterized by larger and less complicated shapes with leaves traveling shorter overall lengths.ConclusionsThis first experience suggests that the ASC is an effective tool to reduce the unnecessary complexity of a plan. This turns into an increased plan deliverability with no loss of plan quality.

Personalized setting of plan parameters using feasibility dose volume histogram for auto-planning in Pinnacle system.

PurposeThe personalized setting of plan parameters in the Auto‐Planning module of the Pinnacle treatment planning system (TPS) using the PlanIQ feasibility tool was evaluated for lung cancer conventional fractionated radiotherapy (CFRT).Materials and methodWe reviewed the records of ten patients with lung cancer who were treated with volumetric modulated arc therapy (VMAT). Three plans were designed for each patient: the clinically accepted manual plan (MP) and two automatic plans including one generated using the generic plan parameters in technique script (AP1) and the other generated using personalized plan parameters derived based on feasibility dose volume histogram (FDVH) in PlanIQ (AP2). The plans were assessed according to the dosimetric parameters, monitor units, and planning time. A plan quality metric (PQM) was defined according to the clinical requirements for plan assessment.ResultsAP2 achieved better lung sparing than AP1 and MP. The PQM value of AP2 (52.5 ± 14.3) was higher than those of AP1 (49.2 ± 16.2) and MP (44.8 ± 16.9) with P < 0.05. The monitor units of AP2 (585.9 ± 142.9 MU) was higher than that of AP1 (511.1 ± 136.5 MU) and lower than that of MP (632.8 ± 143.8 MU) with p < 0.05. The planning time of AP2 (33.2 ± 4.8 min) was slightly higher than that of AP1 (28.2 ± 4.0 min) and substantially lower than that of MP (72.9 ± 28.5 min) with P < 0.05.ConclusionsThe Auto‐Planning module of the Pinnacle system using personalized plan parameters suggested by the PlanIQ Feasibility tool provides superior quality for lung cancer plans, especially in terms of lung sparing. The time consumption of Auto‐Planning was slightly higher with the personalized parameters compared to that with the generic parameters, but significantly lower than that for the manual plan.

Development and evaluation of machine learning models for voxel dose predictions in online adaptive magnetic resonance guided radiation therapy.

PurposeDaily online adaptive plan quality in magnetic resonance imaging guided radiation therapy (MRgRT) is difficult to assess in relation to the fully optimized, high quality plans traditionally established offline. Machine learning prediction models developed in this work are capable of predicting 3D dose distributions, enabling the evaluation of online adaptive plan quality to better inform adaptive decision‐making in MRgRT.MethodsArtificial neural networks predicted 3D dose distributions from input variables related to patient anatomy, geometry, and target/organ‐at‐risk relationships in over 300 treatment plans from 53 patients receiving adaptive, linac‐based MRgRT for abdominal cancers. The models do not include any beam related variables such as beam angles or fluence and were optimized to balance errors related to raw dose and specific plan quality metrics used to guide daily online adaptive decisions.ResultsAveraged over all plans, the dose prediction error and the absolute error were 0.1 ± 3.4 Gy (0.1 ± 6.2%) and 3.5 ± 2.4 Gy (6.4 ± 4.3%) respectively. Plan metric prediction errors were −0.1 ± 1.5%, −0.5 ± 2.1%, −0.9 ± 2.2 Gy, and 0.1 ± 2.7 Gy for V95, V100, D95, and Dmean respectively. Plan metric prediction absolute errors were 1.1 ± 1.1%, 1.5 ± 1.5%, 1.9 ± 1.4 Gy, and 2.2 ± 1.6 Gy. Approximately 10% (25) of the plans studied were clearly identified by the prediction models as inferior quality plans needing further optimization and refinement.ConclusionMachine learning prediction models for treatment plan 3D dose distributions in online adaptive MRgRT were developed and tested. Clinical integration of the models requires minimal effort, producing 3D dose predictions for a new patient’s plan using only target and OAR structures as inputs. These models can enable improved workflows for MRgRT through more informed plan optimization and plan quality assessment in real time.

Retrospective quality metrics review of stereotactic radiosurgery plans treating multiple targets using single-isocenter volumetric modulated arc therapy.

PurposeTo characterize key plan quality metrics in multi‐target stereotactic radiosurgery (SRS) plans treated using single‐isocenter volumetric modulated arc therapy (VMAT) in comparison to dynamic conformal arc (DCA) plans treating single target. To investigate the feasibility of quality improvement in VMAT planning based on previous planning knowledge.Materials and methods97 VMAT plans of multi‐target and 156 DCA plans of single‐target treated in 2017 at a single institution were reviewed. A total of 605 targets were treated with these SRS plans. The prescription dose was normalized to 20 Gy in all plans for this analysis. Two plan quality metrics, target conformity index (CI) and normal tissue volume receiving more than 12 Gy (V12Gy), were calculated for each target. The distribution of V12Gy per target was plotted as a function of the target volume. For multi‐target VMAT plans, the number of targets being treated in the same plan and the distance between targets were calculated to evaluate their impact on V12Gy. VMAT plans that had a large deviation of V12Gy from the average level were re‐optimized to determine the possibility of reducing the variation of V12Gy in VMAT planning.ResultsConformity index of multi‐target VMAT plans were lower than that of DCA plans while the mean values of 12 Gy were comparable. The V12Gy for a target in VMAT plan did not show apparent dependence on the total number of targets or the distance between targets. The distribution of V12Gy exhibited a larger variation in VMAT plans compared to DCA plans. Re‐optimization of outlier plans reduced V12 Gy by 33.9% and resulted in the V12Gy distribution in VMAT plans more closely resembling that of DCA plans.ConclusionThe benchmark data on key plan quality metrics were established for single‐isocenter multi‐target SRS planning. It is feasible to use this knowledge to guide VMAT planning and reduce high V12Gy outliers.

A longitudinal evaluation of improvements in treatment plan quality for lung cancer with volumetric modulated arc therapy.

PurposeTo investigate planning time and number of optimizations in routine clinical lung cancer plans based on the plan quality improvements following each optimization.Materials and methodWe selected 40 patients with lung cancer who were treated with conventional fractionated radiotherapy (CFRT). The 40 plans (divided into two groups with one or two target volumes) were completed by 9 planners using volumetric modulated arc therapy (VMAT). A planning strategy, including technique script for each group and a planning process for data collection, was introduced. The total planning time, number of optimizations, and dose–volume parameters of each plan were recorded and analyzed. A plan quality metric (PQM) was defined according to the clinical constraints. Statistical analysis of parameters of each plan following each optimization was performed for evaluating improvements in plan quality.ResultsAccording to the clinical plans generated by different planners, the median number of optimizations of each group was 4, and the median planning time was approximately 1 h (68.6 min and 62.0 min for plans with one or two target volumes, respectively). The dose deposited in organs at risk (OARs) gradually decreased, and the PQM values gradually improved following each optimization. The improvements were significant only between adjacent optimizations from the first optimization (Opt1) to the third optimization (Opt3).ConclusionIncreasing the number of optimizations was associated with significantly improved sparing of OARs with slight effects on the dose coverage and homogeneity of target volume. Generally, based on the designed planning strategy, there was no significant improvement of the plan quality for more than three optimizations.

Variation in Interinstitutional Plan Quality When Adopting a Hypofractionated Protocol for Prostate Cancer External Beam Radiation Therapy

This study quantified plan quality differences across the 4 cancer centers in Alberta, Canada for plans that followed the PROstate Fractionated Irradiation Trial protocol. Prostate plans of 235 patients were retrospectively reviewed. Interinstitutional plan quality comparisons were made based on distributions of protocol-specified parameters using 1-way analysis of variance with Games-Howell post hoc analysis. Dosimetrically representative cases were selected from each center using k-medoid clustering, enabling side-by-side comparison of dose-volume histograms and dose distributions. Fourteen anatomic features were investigated to explore interinstitutional patient population differences. Anatomically representative cases were selected from each center to explore differences in planning practices. Tumor control probability (TCP), as well as rectal wall and bladder wall normal tissue complication probabilities (NTCPs), were calculated to quantify the clinical effect of the differences in plan quality. Comparing the mean value of each center to the other 3, statistically significant differences were observed for bladder wall D30% and D50%, left and right femoral heads D5%, planning target volume D99% and D1cc, and clinical target volume D99%. Dosimetrically representative cases demonstrated consistent results. Although anatomic differences were observed between the center-specific populations, an analysis using anatomically similar cases demonstrated consistent trends in the dosimetric differences, suggesting the dosimetric variation is not exclusively due to anatomic differences. Minimal differences (<1%) among the 4 centers were noted for TCP and NTCPs, suggesting the reported differences in plan quality may not have any clinical significance. Despite common guidelines, statistically significant differences in plan quality metrics occurred among the 4 investigated centers. The differences are due at least in part to variation in local planning practices. TCP and NTCP calculations suggest that the clinical significance of the differences is minimal. These results can serve as a reference for the degree of variation among centers that can be accepted when a common protocol is adopted.

Dosimetric comparison of iPlanⓇ Pencil Beam (PB) and Monte Carlo (MC) algorithms in stereotactic radiosurgery/radiotherapy (SRS/SRT) plans of intracranial arteriovenous malformations

Stereotactic radiosurgery/radiotherapy (SRS/SRT) is a hypofractionated treatment where accurate dose calculation is of prime importance. The accuracy of the dose calculation depends on the treatment planning algorithm. This study is a retrospective dosimetric comparison of iPlanⓇ Monte Carlo (MC) and Pencil Beam (PB) algorithms in SRS/SRT plans of cranial arteriovenous malformations (AVMs). PB plans of 60 AVM patients who were already treated using 6 MV photons from a linear accelerator were selected and divided into 2 groups. Group-I consists of 30 patients who have undergone embolization procedure with high density OnyxⓇ prior to radiosurgery whereas Group-II had 30 patients who did not have embolization. These plans were recalculated with MC algorithm while keeping parameters like beam orientation, multileaf collimator (MLC) positions, MLC margin, prescription dose, and monitor units constant. Several treatment coverage parameters, isodose volumes, plan quality metrics, dose to organs at risk, and integral dose were used for comparing the 2 algorithms. The isodose distribution generated by the 2 algorithms was also compared with gamma analysis using 1%/1 mm criterion. The difference between the 2 groups as well as the differences in dose calculation by PB and MC algorithms were tested for significance using independent t-test and paired t-test respectively at 5% level of significance. The results of the independent t-test showed that there is no significant difference between the Group-I and Group-II patients for PB as well as MC algorithm due to the presence of high density embolization material. However, results of the paired t-test showed that the differences between the PB and MC algorithms were significant for several parameters analyzed in both groups of patients. The gamma analysis results also showed differences in the dose calculated by the 2 algorithms especially in the low dose regions. The significant differences between the 2 algorithms are probably due to the incorrect representation of the loss of lateral charged particle equilibrium and lateral broadening of small photon beams by PB algorithm. MC algorithms are generally considered not essential for dose calculations for target volumes located in the brain. This study demonstrates PB algorithm may not be sufficiently accurate to predict dose distributions for small fields where there is loss of LCPE. The lateral broadening due to the loss of LCPE as predicted by the MC algorithm could be the main reason for significant differences in the parameters compared. Hence, an accurate MC algorithm if available may prove valuable for intracranial SRS treatment planning of such benign lesions where the long life expectancy of patients makes accurate dosimetry critical.

Dosimetric performance of the Elekta Unity MR-linac system: 2D and 3D dosimetry in anthropomorphic inhomogeneous geometry

Following the clinical introduction of the Elekta Unity MR-linac, there is an urgent need for development of dosimetry protocols and tools, not affected by the presence of a magnetic field. This work presents a benchmarking methodology comprising 2D/3D passive dosimetry and involving on-couch adaptive treatment planning, a unique step in MR-linac workflows.Two identical commercially available 3D-printed head phantoms (featuring realistic bone anatomy and MR/CT contrast) were employed. One phantom incorporated a film dosimetry insert, while the second was filled with polymer gel. Gel dose-response characteristics were evaluated under the Unity irradiation and read-out conditions, using vials and a cubic container filled with gel from the same batch.Treatment plan for the head phantoms involved a hypothetical large C-shape brain lesion, partly surrounding the brainstem. An IMRT step-and-shoot 7-beam plan was employed. Pre-treatment on-couch MR-images were acquired in order for the treatment planning system to calculate the virtual couch shifts and perform adaptive planning. Absolute 2D and relative 3D measurements were compared against calculations related to both adapted and original plans. Real-time dose accumulation monitoring in the gel-filled phantom was also performed.Results from the vials and cubic container suggest that gel dose-response is linear in the dose range investigated and signal integrity is mature at the read-out timings considered. Head phantom 2D and 3D measurements agreed well with calculations with 3D gamma index passing rates above 90% in all cases, even with the most stringent criteria used (2 mm/2%). By exploiting the 3D information provided by the gel, comparison also involved DVHs, dose-volume and plan quality metrics, which also reflected the agreement between adapted and delivered plans within ±4%. No considerable discrepancies were detected between adapted and original plans.A novel methodology was developed and implemented, suitable for QA procedures in Unity. TPS calculations were validated within the experimental uncertainties involved.

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

PurposeThis study aimed to clarify the inter-planner variation of plan quality in knowledge-based plans created by nine planners. MethodsFive 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. ResultsLower 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. ConclusionInter-planner variation in the plan quality with RapidPlan remains, especially for the complicated VMAT plans, due to planners’ heuristics.

Fluence-map generation for prostate intensity-modulated radiotherapy planning using a deep-neural-network

A deep-neural-network (DNN) was successfully used to predict clinically-acceptable dose distributions from organ contours for intensity-modulated radiotherapy (IMRT). To provide the next step in the DNN-based plan automation, we propose a DNN that directly generates beam fluence maps from the organ contours and volumetric dose distributions, without inverse planning. We collected 240 prostate IMRT plans and used to train a DNN using organ contours and dose distributions. After training was done, we made 45 synthetic plans (SPs) using the generated fluence-maps and compared them with clinical plans (CP) using various plan quality metrics including homogeneity and conformity indices for the target and dose constraints for organs at risk, including rectum, bladder, and bowel. The network was able to generate fluence maps with small errors. The qualities of the SPs were comparable to the corresponding CPs. The homogeneity index of the target was slightly worse in the SPs, but there was no difference in conformity index of the target, V60Gy of rectum, the V60Gy of bladder and the V45Gy of bowel. The time taken for generating fluence maps and qualities of SPs demonstrated the proposed method will improve efficiency of the treatment planning and help maintain the quality of plans.

Optimized Volumetric Modulated Arc Technique for Noninvasive Cardiac Ablation to Treat Ventricular Tachycardia

Recent publications (Cuculich et al., NEJM-2017, Robinson et al., Circulation-2019) have shown promising outcomes of radioablation using a single-fraction SBRT to treat patients with refractory, life-threatening, VT. This prompted growing interest across the world to use SBRT to treat VT. The primary goal of this study is to explore an alternative VMAT planning technique to achieve greater Non-Target Heart-Tissue (NTHT) sparing while simultaneously allowing for breath hold (BH) treatments. 13 patients were selected who previously received a single fraction dose of 35 Gy to ITV and 25 Gy to PTV utilizing 3 non-coplanar VMAT arcs of 6 MV photon beams using HD-MLC on a radiosurgery-optimized linac. To further improve NTHT sparing we retrospectively re-planned each patient using 10 ipsi-lateral arcs with optimal gantry, couch and collimator angles. Two plans were generated using 6X FFF and 10X FFF beams with dynamic jaw-tracking. Gantry arc lengths were limited to 90° (< 15s/arc) for feasibility of a potential BH treatments. Couch angles were limited to ± 15° to avoid collisions. Plans were optimized to maximize the NTHT sparing while meeting the protocol constraints for the stomach, esophagus and colon (OAR). Plan quality metrics for both clinical and reoptimized plans are summarized in Table 1. Revised plans showed adequate target coverage (V95%Rx=>95%), better dose conformity and gradient, while meeting the dose constraints of the OARs. Excellent NTHT sparing is achieved by reducing the volumes of V20Gy by 43 cc and V5Gy by 250 cc. Average mean dose reduction of 2.5 (range: 0.6 – 6.0) Gy and 2.2 (0.7 – 5.7 Gy) was observed in 6X FFF and 10X FFF plans, respectively. Revised plans showed the possibility of treating each arc in a single BH. The standardized technique used in this study showed improved NTHT sparing and better dose conformity. This result may be important for the group of patients where cardiac toxicity may play a vital role in long-term survival. BH treatments may reduce the required PTV margins resulting in further NTHT sparing.Abstract 3778; Table 1Vol. (cc)Plan TypeTarget CoverageConformityMean D (Gy)%VRx%V95%RxConformity IndexGradient MeasureITV48.1 ± 29Clinical38.4 ± 1.197.6 ± 2.499.7 ± 0.66X FFF36.9 ± 0.494.2 ± 3.199.3 ± 0.810X FFF36.8 ± 0.890.5 ± 11.698.7 ± 2.0PTV138.9 ± 67.8Clinical33.6 ± 1.199.0 ± 0.899.7 ± 0.41.14 ± 0.122.01 ± 0.476X FFF32.3 ± 0.493.8 ± 2.596.8 ± 2.21.02 ± 0.041.76 ± 0.4710X FFF32.3 ± 0.494.2 ± 2.397.1 ± 1.81.03 ± 0.031.55 ± 0.26Non-Target Heart TissueMean D (Gy)%V25%V20%V15%V10%V5%V2Clinical6.2 ± 2.10.9 ± 0.84.9 ± 2.511.2 ± 5.421.4 ± 9.140.5 ± 13.966.7 ± 17.86X FFF3.7 ± 1.10.3 ± 0.31.9 ± 2.74.9 ± 1.610.1 ± 3.622.5 ± 8.945.9 ± 15.7Vol. spared (cc)9 ± 1143.± 3590 ± 76160 ± 113250 ± 146284 ± 20910X FFF4.0 ± 1.10.4 ± 0.32.2 ± 0.75.5 ± 1.711.4 ± 3.825.5 ± 9.649.6 ± 16.5Vol. spared (cc)8 ± 1039 ± 3381 ± 75142 ± 108208 ± 111233 ± 150 Open table in a new tab

A Machine Learning Based Approach for Fast and Automated Plan Quality Evaluation for Online Adaptive Radiation Therapy

The evaluation of radiation therapy (RT) plans has been a labor-intensive subjective decision that involves the analysis of dose-volume histograms (DVHs) along with a slice-by-slice visual inspection of the dose distribution on the planning images. This demanding task is generally challenging in online adaptive RT (OART) where the planning time is limited. We developed an automated plan quality evaluation method based on machine learning (ML) and propose a plan quality metric for the evaluation of RT plans. Our proposed method starts with extracting and evaluating more than 30 commonly-used dose-volume parameters (DVPs), 60 non-conventional parameters (NCP) that specifically probe for hot and cold spots, and 320 radiomic features from the 3D dose distribution. To reduce the number of features in each class and investigate their importance, a recursive feature elimination algorithm with a Random Forest Regressor (RFR) estimator was employed. The influence of gray level discretization on NCPs and radiomic features was explored. The top ranked features from a given class were used to train an ensemble of RFRs. The prediction from each RFR was then aggregated to assign a plan quality score in the range from zero to five. The method was trained/tested using 200 dosimetric plans generated from planning and daily CTs acquired from 20 prostate cancer patients. Plans were evaluated by a physician as acceptable or not acceptable. To measure the effect of data variation on the method’s end performance, training and test sets were randomly created from the plan collection. The performance was quantified by the average area under the receiver operating characteristic curve (AUC). The machine quality scores were analyzed and compared to the physician’s evaluation. For each class of features, the method’s performance peaked before the top five ranked features were used. The NCPs and radiomic features performance were the greatest when the dose was discretized to 128 and 256 gray levels, respectively. Average AUC of 0.95 +/- 0.02, 0.87 +/- 0.04, 0.93 +/- 0.03, and 0.95 +/- 0.03 were achieved in the test sets when using DVPs, NCPs, radiomic features, and a combination of the best performing features, respectively. An ensemble of RFRs trained with DVPs assigned quality scores of 3, 4, and 5 that were selected as acceptable for treatment 77, 88, and 99% of the time, respectively. The analysis of the DVHs showed that in average plans with higher quality scores provided the most target coverage and normal tissue sparing. A trained model running on a quad-core CPU with a processor base frequency of 3.4 GHz was able to assign a new plan a quality score within 60 seconds. We have developed a novel ML method to objectively and automatically evaluate the quality of RT plans. As demonstrated in a prostate cancer dataset, this method can be implemented for quick plan evaluation in OART.

Automatic Planning of Whole Breast Radiation Therapy Using Machine Learning Models.

Purpose: To develop an automatic treatment planning system for whole breast radiation therapy (WBRT) based on two intensity-modulated tangential fields, enabling near-real-time planning.Methods and Materials: A total of 40 WBRT plans from a single institution were included in this study under IRB approval. Twenty WBRT plans, 10 with single energy (SE, 6MV) and 10 with mixed energy (ME, 6/15MV), were randomly selected as training dataset to develop the methodology for automatic planning. The rest 10 SE cases and 10 ME cases served as validation. The auto-planning process consists of three steps. First, an energy prediction model was developed to automate energy selection. This model establishes an anatomy-energy relationship based on principle component analysis (PCA) of the gray level histograms from training cases' digitally reconstructed radiographs (DRRs). Second, a random forest (RF) model generates an initial fluence map using the selected energies. Third, the balance of overall dose contribution throughout the breast tissue is realized by automatically selecting anchor points and applying centrality correction. The proposed method was tested on the validation dataset. Non-parametric equivalence test was performed for plan quality metrics using one-sided Wilcoxon Signed-Rank test.Results: For validation, the auto-planning system suggested same energy choices as clinical-plans in 19 out of 20 cases. The mean (standard deviation, SD) of percent target volume covered by 100% prescription dose was 82.5% (4.2%) for auto-plans, and 79.3% (4.8%) for clinical-plans (p > 0.999). Mean (SD) volume receiving 105% Rx were 95.2 cc (90.7 cc) for auto-plans and 83.9 cc (87.2 cc) for clinical-plans (p = 0.108). Optimization time for auto-plan was <20 s while clinical manual planning takes between 30 min and 4 h.Conclusions: We developed an automatic treatment planning system that generates WBRT plans with optimal energy selection, clinically comparable plan quality, and significant reduction in planning time, allowing for near-real-time planning.

Automated 4π radiotherapy treatment planning with evolving knowledge‐base

Non-coplanar 4π radiotherapy generalizes intensity modulated radiation therapy (IMRT) to automate beam geometry selection but requires complicated hyperparameter tuning to attain superior plan quality, which can be tedious and inconsistent. In this study, a fully automated 4π treatment planning was developed using evolving knowledge-base (EKB) planning guided by dose prediction. Twenty 4π lung and twenty 4π head and neck (HN) cases were included. A statistical voxel dose learning model was initially trained on low-quality plans created using generic hyperparameter templates without manual tuning. To improve the automated plan quality without being limited by the training data quality, a new 4π optimization problem was formulated to include a one-sided penalty on the organ-at-risk (OAR) dose deviation from the predicted dose. This directional OAR penalty encourages superior OAR sparing. The fast iterative shrinkage-thresholding algorithm (FISTA) was used to solve the large-scale beam orientation optimization problem. With the improved plans, new predictions were created to guide the next loop of EKB planning for a total of 10 loops. Plan quality was evaluated using a plan quality metric (PQM) points system based on clinical dose constraints and compared with automated planning approaches guided by manual high-quality plans using all non-coplanar beams, automated plans using individually evolved targeted dose, and manually created 4π plans. For the lung cases, the final EKB plans had significantly higher PQM than manually created 4π (+2.60%). The improvements plateaued after the third loop. The final HN EKB plans and manually created 4π plans had comparable PQMs, but had lower PQM compared to automated plans using a high-quality training set (-3.00% and -4.44%, respectively). The PQM consistently increased up to the sixth loop. Individually evolved plans were able to improve the plan quality from initial condition due to the one-sided cost function but the 60% of them were trapped in undesired local minima that were substantially worse than their corresponding EKB plans. Evolving knowledge-base planning is a novel automated planning technique guided by the predicted three-dimensional dose distribution, which can evolve from low-quality plans. EKB allows new beams to be used in the automated planning workflow for superior plan quality.

Integration of biological factors in the treatment plan evaluation in breast cancer radiotherapy.

Tumor biology and patient smoking status have clear effects on the benefit of breast radiotherapy. This study developed treatment evaluation strategies that integrated dosimetry, tumor aggressiveness and smoking status for patients undergoing hypo-fractionated whole breast irradiation with simultaneous integrated boost. The evaluation method Plan Quality Metrics (PQM) was adapted for breast cancer. Radiotherapy (RT) benefit was assessed for three levels of tumor aggressiveness; RT risk was estimated using mean dose to organs at risk and published Excess Relative Risk per Gy data for lung cancer and cardiac mortality for smokers and non-smokers. Risk for contralateral breast cancer was also evaluated. PQM and benefit/risk was applied to four patient groups (n = 10 each). Plans using 3D conformal radiotherapy (3DCRT), 3DCRT plus intensity-modulated radiation therapy (IMRT), 3DCRT plus volumetric modulated arc therapy (VMAT) and VMAT were evaluated for each patient. 3DCRT-IMRT hybrid planning resulted in higher PQM score (median 87.0 vs. 3DCRT 82.4, p < 0.01), better dose conformity, lower doses to the heart, lungs and contralateral breast. Survival benefit was most predominant for patients with high-risk breast cancer (>7% and >4.5% gain for non-smokers and smokers). For smokers with intermediate- or low-risk breast cancer, RT induced mortality risk dominated for all techniques. When considering the risk of local recurrence, RT benefitted also smokers (>5% and >2% for intermediate- and low-risk cancer). PQM methodology was suggested for breast cancer radiotherapy evaluation. Further validation is needed. RT was beneficial for all patients with high risk of recurrence. A survival benefit for smokers with low or intermediate risk of recurrence could not be confirmed.

Multi-center evaluation of dose conformity in stereotactic body radiotherapy.

Background and purposeStereotactic body radiotherapy (SBRT) is an emerging technique for treating oligometastases, but limited data is available on what plan quality is achievable for a range of modalities and clinical sites.MethodsSBRT plans for lung, spine, bone, adrenal, liver and node sites from 17 participating centers were reviewed. Centers used various delivery techniques including static and rotational intensity-modulation and multiple non-coplanar beams. Plans were split into lung and other body sites and evaluated with different plan quality metrics, including two which are independent of target coverage; “prescription dose spillage” (PDS) and “modified gradient index” (MGI). These were compared to constraints from the ROSEL and RTOG 0813 clinical trials.ResultsPlanning target volume (PTV) coverage was compromised (PTV V100% < 90%) in 29% of patient plans in order to meet organ-at-risk (OAR) tolerances, supporting the use of plan quality metrics which are independent of target coverage. Both lung (n = 48) and other body (n = 99) site PDS values agreed well with ROSEL constraints on dose spillage, but RTOG 0813 values were too high to detect sub-optimal plans. MGI values for lung plans were mis-matched to both sets of previous constraints, with ROSEL values too high and RTOG 0813 values too low. MGI values were lower for other body plans as expected, though this was only statistically significant for PTV volumes <20 cm3.ConclusionsUpdated guidance for lung and other body site SBRT plan quality using the PDS and MGI metrics is presented.