What is plan quality in radiotherapy? The importance of evaluating dose metrics, complexity, and robustness of treatment plans

Treatment plan complexity quantification for predicting gamma passing rates in patient-specific quality assurance for stereotactic volumetric modulated arc therapy

PURPOSE

To investigate the beam complexity of stereotactic Volumetric Modulated Arc Therapy (VMAT) plans quantitively and predict gamma passing rates (GPRs) using machine learning.

METHODS

The entire dataset is exclusively made of stereotactic VMAT plans (301 plans with 594 beams) from Varian Edge LINAC. The GPRs were analyzed using Varian's portal dosimetry with 2%/2 mm criteria. A total of 27 metrics were calculated to investigate the correlation between metrics and GPRs. Random forest and gradient boosting models were developed and trained to predict the GPRs based on the extracted complexity features. The threshold values of complexity metric were obtained to predict a given beam to pass or fail from ROC curve analysis.

RESULTS

The three moderately significant values of Spearman's rank correlation to GPRs were 0.508 (p < 0.001), 0.445 (p < 0.001), and -0.416 (p < 0.001) for proposed metric LAAM, the ratio of the average aperture area over jaw area (AAJA) and index of modulation, respectively. The random forest method achieved 98.74% prediction accuracy with mean absolute error of 1.23% using five-fold cross-validation, and 98.71% with 1.25% for gradient boosting regressor method, respectively. LAAM, leaf travelling distance (LT), AAJA, LT modulation complexity score (LTMCS) and index of modulation, were the top five most important complexity features. The LAAM metric showed the best performance with AUC value of 0.801, and threshold value of 0.365.

CONCLUSIONS

The calculated metrics were effective in quantifying the complexity of stereotactic VMAT plans. We have demonstrated that the GPRs could be accurately predicted using machine learning methods based on extracted complexity metrics. The quantification of complexity and machine learning methods have the potential to improve stereotactic treatment planning and identify the failure of QA results promptly.

A joint ESTRO and AAPM guideline for development, clinical validation and reporting of artificial intelligence models in radiation therapy

BACKGROUND AND PURPOSE

Artificial Intelligence (AI) models in radiation therapy are being developed with increasing pace. Despite this, the radiation therapy community has not widely adopted these models in clinical practice. A cohesive guideline on how to develop, report and clinically validate AI algorithms might help bridge this gap.

METHODS AND MATERIALS

A Delphi process with all co-authors was followed to determine which topics should be addressed in this comprehensive guideline. Separate sections of the guideline, including Statements, were written by subgroups of the authors and discussed with the whole group at several meetings. Statements were formulated and scored as highly recommended or recommended.

RESULTS

The following topics were found most relevant: Decision making, image analysis, volume segmentation, treatment planning, patient specific quality assurance of treatment delivery, adaptive treatment, outcome prediction, training, validation and testing of AI model parameters, model availability for others to verify, model quality assurance/updates and upgrades, ethics. Key references were given together with an outlook on current hurdles and possibilities to overcome these. 19 Statements were formulated.

CONCLUSION

A cohesive guideline has been written which addresses main topics regarding AI in radiation therapy. It will help to guide development, as well as transparent and consistent reporting and validation of new AI tools and facilitate adoption.

Robustness evaluation of pencil beam scanning proton therapy treatment planning: A systematic review

Compared to conventional radiotherapy using X-rays, proton therapy, in principle, allows better conformity of the dose distribution to target volumes, at the cost of greater sensitivity to physical, anatomical, and positioning uncertainties. Robust planning, both in terms of plan optimization and evaluation, has gained high visibility in publications on the subject and is part of clinical practice in many centers. However, there is currently no consensus on the methods and parameters to be used for robust optimization or robustness evaluation. We propose to overcome this deficiency by following the modified Delphi consensus method. This method first requires a systematic review of the literature. We performed this review using the PubMed and Web Of Science databases, via two different experts. Potential conflicts were resolved by a third expert. We then explored the different methods before focusing on clinical studies that evaluate robustness on a significant number of patients. Many robustness assessment methods are proposed in the literature. Some are more successful than others and their implementation varies between centers. Moreover, they are not all statistically or mathematically equivalent. The most sophisticated and rigorous methods have seen more limited application due to the difficulty of their implementation and their lack of widespread availability.

Incorporating plan complexity into the statistical process control of volumetric modulated arc therapy pre-treatment verifications

BACKGROUND

Statistical process control (SPC) is a powerful statistical tool for process monitoring that has been highly recommended in healthcare applications, including radiation therapy quality assurance (QA). The AAPM TG-218 report described the clinical implementation of SPC for Volumetric Modulated Arc Therapy (VMAT) pre-treatment verifications, pointing out the need to adjust tolerance limits based on plan complexity. However, the quantification of plan complexity and its integration into SPC remains an unresolved challenge.

PURPOSE

The primary aim of this study is to investigate the incorporation of plan complexity into the SPC framework for VMAT pre-treatment verifications. The study explores and evaluates various strategies for this incorporation, discussing their merits and limitations, and provides recommendations for clinical application.

METHODS

A retrospective analysis was conducted on 309 VMAT plans from diverse anatomical sites using the PTW OCTAVIUS 4D device for QA measurements. Gamma Passing Rates (GPR) were obtained, and lower control limits were computed using both the conventional Shewhart method and three heuristic methods (scaled weighted variance, weighted standard deviations, and skewness correction) to accommodate non-normal data distributions. The 'Identify-Eliminate-Recalculate' method was employed for robust analysis. Eight complexity metrics were analyzed and two distinct strategies for incorporating plan complexity into SPC were assessed. The first strategy focused on establishing control limits for different treatment sites, while the second was based on the determination of control limits as a function of individual plan complexity. The study extensively examines the correlation between control limits and plan complexity and assesses the impact of complexity metrics on the control process.

RESULTS

The control limits established using SPC were strongly influenced by the complexity of treatment plans. In the first strategy, a clear correlation was found between control limits and average plan complexity for each site. The second approach derived control limits based on individual plan complexity metrics, enabling tailored tolerance limits. In both strategies, tolerance limits inversely correlated with plan complexity, resulting in all highly complex plans being classified as in control. In contrast, when plans were collectively analyzed without considering complexity, all the out-of-control plans were highly complex.

CONCLUSIONS

Incorporating plan complexity into SPC for VMAT verifications requires meticulous and comprehensive analysis. To ensure overall process control, we advocate for stringent control and minimization of plan complexity during treatment planning, especially when control limits are adjusted based on plan complexity.

Impact of beam complexity on plan delivery accuracy verification of a transmission detector in volumetric modulated arc therapy

OBJECTIVE

To study the effect of beam complexity on VMAT delivery accuracy evaluated by means of a transmission detector, together with the possibility of scoring plan complexity.

METHODS

43 clinical VMAT plans delivered by a TrueBeam linear accelerator to both Delta4 Discover and Delta4 Phantom+ for patient-specific quality assurance were evaluated. Global Dose-γ analysis, MLC-γ analysis, percentage of leaves with a deviation between planned and measured leaf tip position lower than 1 mm (LD) were computed. Modulation complexity score (MCSv), average leaf travel (LT), a multiplicative combination of LT and MCSv (LTMCS), percentage of leaves with speed lower than 5 mm/s (LS), from 5 to 20 mm/s (MS), higher than 20 mm/s (HS) and the average value of leaf speed (MLCSav) were evaluated by means of an home-made Matlab script.

RESULTS

Dose-γ passing rate showed a moderate correlation with MCSv, LT, MLCSav, LS and HS, while a stronger positive correlation was found with LTMCS. A strong correlation was observed between LD and both LT and leaves speed, while a weak correlation was observed with MCSv. A correlation between MLC-γ pass rate and plan complexity parameters was found except for MCSv; a moderate correlation with LS was observed, while all other parameters showed weak correlations.

CONCLUSIONS

The study confirmed the possibility to establish correlations between plan complexity indices versus dose distribution and MLC parameters measured by a transmissive detector. Further investigation is necessary to define specific values of the complexity indices to evaluate whether a VMAT plan is deliverable as intended.

Human-like intelligent automatic treatment planning of head and neck cancer radiation therapy

Objective.Automatic treatment planning of radiation therapy (RT) is desired to ensure plan quality, improve planning efficiency, and reduce human errors. We have proposed an Intelligent Automatic Treatment Planning framework with a virtual treatment planner (VTP), an artificial intelligence robot built using deep reinforcement learning, autonomously operating a treatment planning system (TPS). This study extends our previous successes in relatively simple prostate cancer RT planning to head-and-neck (H&N) cancer, a more challenging context even for human planners due to multiple prescription levels, proximity of targets to critical organs, and tight dosimetric constraints.Approach.We integrated VTP with a real clinical TPS to establish a fully automated planning workflow guided by VTP. This integration allowed direct model training and evaluation using the clinical TPS. We designed the VTP network structure to approach the decision-making process in RT planning in a hierarchical manner that mirrors human planners. The VTP network was trained via theQ-learning framework. To assess the effectiveness of VTP, we conducted a prospective evaluation in the 2023 Planning Challenge organized by the American Association of Medical Dosimetrists (AAMD). We extended our evaluation to include 20 clinical H&N cancer patients, comparing the plans generated by VTP against their clinical plans.Main results.In the prospective evaluation for the AAMD Planning Challenge, VTP achieved a plan score of 139.08 in the initial phase evaluating plan quality, and 15 min of planning time with the first place ranking in the adaptive phase competing for planning efficiency while meeting all plan quality requirements. For clinical cases, VTP-generated plans achieved an average VTP score of125.33±11.12, which outperformed the corresponding clinical plans with an average score of117.76±13.56.Significance.We successfully integrated VTP with the clinical TPS to achieve a fully automated treatment planning workflow. The compelling performance of VTP demonstrated its potential in automating H&N cancer RT planning.

Dose-toxicity surface histogram-based prediction of radiation dermatitis severity and shape

Objective.This study aimed to develop a new approach to predict radiation dermatitis (RD) by using the skin dose distribution in the actual area of RD occurrence to determine the predictive dose by grade.Approach.Twenty-three patients with head and neck cancer treated with volumetric modulated arc therapy were prospectively and retrospectively enrolled. A framework was developed to segment the RD occurrence area in skin photography by matching the skin surface image obtained using a 3D camera with the skin dose distribution. RD predictive doses were generated using the dose-toxicity surface histogram (DTH) calculated from the skin dose distribution within the segmented RD regions classified by severity. We then evaluated whether the developed DTH-based framework could visually predict RD grades and their occurrence areas and shapes according to severity.Main results.The developed framework successfully generated the DTH for three different RD severities: faint erythema (grade 1), dry desquamation (grade 2), and moist desquamation (grade 3); 48 DTHs were obtained from 23 patients: 23, 22, and 3 DTHs for grades 1, 2, and 3, respectively. The RD predictive doses determined using DTHs were 28.9 Gy, 38.1 Gy, and 54.3 Gy for grades 1, 2, and 3, respectively. The estimated RD occurrence area visualized by the DTH-based RD predictive dose showed acceptable agreement for all grades compared with the actual RD region in the patient. The predicted RD grade was accurate, except in two patients.Significance. The developed DTH-based framework can classify and determine RD predictive doses according to severity and visually predict the occurrence area and shape of different RD severities. The proposed approach can be used to predict the severity and shape of potential RD in patients and thus aid physicians in decision making.

Phase I quality control framework for monitoring organ-at-risk dose

Objective.The aim of this work was to develop a Phase I control chart framework for the recently proposed multivariate risk-adjusted Hotelling'sT2chart. Although this control chart alone can identify most patients receiving extreme organ-at-risk (OAR) dose, it is restricted by underlying distributional assumptions, making it sensitive to extreme observations in the sample, as is typically found in radiotherapy plan quality data such as dose-volume histogram (DVH) points. This can lead to slightly poor-quality plans that should have been identified as out-of-control (OC) to be signaled in-control (IC).Approach. We develop a robust iterative control chart framework to identify all OC patients with abnormally high OAR dose and improve them via re-optimization to achieve an IC sample prior to establishing the Phase I control chart, which can be used to monitor future treatment plans.Main Results. Eighty head-and-neck patients were used in this study. After the first iteration, P14, P67, and P68 were detected as OC for high brainstem dose, warranting re-optimization aimed to reduce brainstem dose without worsening other planning criteria. The DVH and control chart were updated after re-optimization. On the second iteration, P14, P67, and P68 were IC, but P40 was identified as OC. After re-optimizing P40's plan and updating the DVH and control chart, P40 was IC, but P14* (P14's re-optimized plan) and P62 were flagged as OC. P14* could not be re-optimized without worsening target coverage, so only P62 was re-optimized. Ultimately, a fully IC sample was achieved. Multiple iterations were needed to identify and improve all OC patients, and to establish a more robust control limit to monitor future treatment plans.Significance. The iterative procedure resulted in a fully IC sample of patients. With this sample, a more robust Phase I control chart that can monitor OAR doses of new plans was established.

Technical note: The migration distance - a unidirectional distance metric for region-of-interest comparisons

BACKGROUND

The radiotherapy process relies on several metrics in determining a notion of "distance" from one three-dimensional region-of-interest (ROI) to another. The majority are symmetric (or commutative) and do not contain information pertaining to directionality. Growth versus regression, for example, is not inherently distinguished by these metrics.

PURPOSE

The purpose of this work was to formalize a unidirectional distance metric, motivated by radiotherapy margin concepts, which we term the migration distance. Informally, the migration distance from ROI X $X$ to Y $Y$ is the minimum isotropic expansion of X $X$ such that Y $Y$ is completely encompassed by the expansion. If Y $Y$ is contained within X $X$ , the migration distance is negative with magnitude equal to the maximum isotropic contraction of X $X$ such that Y $Y$ remains contained within contraction. The metric is demonstrated by quantifying glioblastoma interfraction target changes.

METHODS

An explicit mathematical formulation of the migration distance is presented and contrasted with the related Hausdorff distance. The results are demonstrated for the gross tumor volume (GTV) dynamics of a glioblastoma cohort consisting of 111 patients that underwent standard chemoradiotherapy with offline MR imaging at planning, fraction 10, fraction 20, and 1-month post radiotherapy.

RESULTS

The mean ± SD of the GTV migration distance relative to planning was 5.9 ± 3.9 mm at fraction 10, 6.2 ± 4.4 mm at fraction 20, and 7.9 ± 7.1 mm at 1-month post radiotherapy. The maximum GTV migration distance across all patients at the same timepoints was 20.4, 20.7, and 45.5 mm, respectively.

CONCLUSIONS

We have proposed and demonstrated a unidirectional distance metric. The migration distance may have applications in the quantification of anatomical changes, planning target volume designs, and dosimetric radiotherapy plan assessment.

Impact of data transfer between treatment planning systems on dosimetric parameters

PURPOSE

In radiotherapy it is often necessary to transfer a patient's DICOM (Digital Imaging and COmmunications in Medicine) dataset from one system to another for re-treatment, plan-summation or registration purposes. The aim of the study is to evaluate effects of dataset transfer between treatment planning systems.

MATERIALS AND METHODS

Twenty-five patients treated in a 0.35T MR-Linac (MRidian, ViewRay) for locally-advanced pancreatic cancer were enrolled. For each patient, a nominal dose distribution was optimized on the planning MRI. Each plan was daily re-optimized if needed to match the anatomy and exported from MRIdian-TPS (ViewRay Inc.) to Eclipse-TPS (Siemens-Varian). A comparison between the two TPSs was performed considering the PTV and OARs volumes (cc), as well as dose coverages and clinical constraints.

RESULTS

From the twenty-five enrolled patients, 139 plans were included in the data comparison. The median values of percentage PTV volume variation are 10.8 % for each fraction, while percentage differences of PTV coverage have a mean value of -1.4 %. The median values of the percentage OARs volume variation are 16.0 %, 7.0 %, 10.4 % and 8.5 % for duodenum, stomach, small and large bowel, respectively. The percentage variations of the dose constraints are 41.0 %, 52.7 % and 49.8 % for duodenum, stomach and small bowel, respectively.

CONCLUSIONS

This study has demonstrated a non-negligible variation in size and dosimetric parameters when datasets are transferred between TPSs. Such variations should be clinically considered. Investigations are focused on DICOM structure algorithm employed by the TPSs during the transfer to understand the cause of such variations.

A radiotherapy community data-driven approach to determine which complexity metrics best predict the impact of atypical TPS beam modeling on clinical dose calculation accuracy

PURPOSE

To quantify the impact of treatment planning system beam model parameters, based on the actual spread in radiotherapy community data, on clinical treatment plans and determine which complexity metrics best describe the impact beam modeling errors have on dose accuracy.

METHODS

Ten beam modeling parameters for a Varian accelerator were modified in RayStation to match radiotherapy community data at the 2.5, 25, 50, 75, and 97.5 percentile levels. These modifications were evaluated on 25 patient cases, including prostate, non-small cell lung, H&N, brain, and mesothelioma, generating 1,000 plan perturbations. Differences in the mean planned dose to clinical target volumes (CTV) and organs at risk (OAR) were evaluated with respect to the planned dose using the reference (50th-percentile) parameter values. Correlation between CTV dose differences, and 18 different complexity metrics were evaluated using linear regression; R-squared values were used to determine the best metric.

RESULTS

Perturbations to MLC offset and transmission parameters demonstrated the greatest changes in dose: up to 5.7% in CTVs and 16.7% for OARs. More complex clinical plans showed greater dose perturbation with atypical beam model parameters. The mean MLC Gap and Tongue & Groove index (TGi) complexity metrics best described the impact of TPS beam modeling variations on clinical dose delivery across all anatomical sites; similar, though not identical, trends between complexity and dose perturbation were observed among all sites.

CONCLUSION

Extreme values for MLC offset and MLC transmission beam modeling parameters were found to most substantially impact the dose distribution of clinical plans and careful attention should be given to these beam modeling parameters. The mean MLC Gap and TGi complexity metrics were best suited to identifying clinical plans most sensitive to beam modeling errors; this could help provide focus for clinical QA in identifying unacceptable plans.

Multi-institution single geometry plan complexity characteristics based on IROC phantoms

BACKGROUND

Clinical intensity modulated radiation therapy plans have been described using various complexity metrics to help identify problematic radiotherapy plans. Most previous studies related to the quantification of plan complexity and their utility have relied on institution-specific plans which can be highly variable depending on the machines, planning techniques, delivery modalities, and measurement devices used. In this work, 1723 plans treating one of only four standardized geometries were simultaneously analyzed to investigate how radiation plan complexity metrics vary across four different sets of common objectives.

PURPOSE

To assess the treatment plan complexity characteristics of plans developed for Imaging and Radiation Oncology Core (IROC) phantoms. Specifically, to understand the variability in plan complexity between institutions for a common plan objective, and to evaluate how various complexity metrics differentiate relevant groups of plans.

METHODS

1723 plans treating one of four standardized IROC phantom geometries representing four different anatomical sites of treatment were analyzed. For each plan, 22 MLC-descriptive plan complexity metrics were calculated, and principal component analysis (PCA) was applied to the 22 metrics in order to evaluate differences in plan complexity between groups. Across all metrics, pairwise comparisons of the IROC phantom data were made for the following classifications of the data: anatomical phantom treated, treatment planning system (TPS), and the combination of MLC model and treatment planning system. An objective k-means clustering algorithm was also applied to the data to determine if any meaningful distinctions could be made between different subgroups. The IROC phantom database was also compared to a clinical database from the University of Wisconsin-Madison (UW) which included plans treating the same four anatomical sites as the IROC phantoms using a TrueBeam™ STx and Pinnacle3 TPS.

RESULTS

The IROC head and neck and spine plans were distinct from the prostate and lung plans based on comparison of the 22 metrics. All IROC phantom plan group complexity metric distributions were highly variable despite all plans being designed for identical geometries and plan objectives. The clusters determined by the k-means algorithm further supported that the IROC head and neck and spine plans involved similar amounts of complexity and were largely distinct from the prostate and lung plans, but no further distinctions could be made. Plan complexity in the head and neck and spine IROC phantom plans were similar to the complexity encountered in the UW clinical plans.

CONCLUSIONS

There is substantial variability in plan complexity between institutions when planning for the same objective. For each IROC anatomical phantom treated, the magnitude of variability in plan complexity between institutions is similar to the variability in plan complexity encountered within a single institution database containing several hundred unique clinical plans treating corresponding anatomies in actual patients.

Robustness analysis of dynamic trajectory radiotherapy and volumetric modulated arc therapy plans for head and neck cancer

Background and purpose

Dynamic trajectory radiotherapy (DTRT) has been shown to improve healthy tissue sparing compared to volumetric arc therapy (VMAT). This study aimed to assess and compare the robustness of DTRT and VMAT treatment-plans for head and neck (H&N) cancer to patient-setup (PS) and machine-positioning uncertainties.

Materials and methods

The robustness of DTRT and VMAT plans previously created for 46 H&N cases, prescribed 50-70 Gy to 95 % of the planning-target-volume, was assessed. For this purpose, dose distributions were recalculated using Monte Carlo, including uncertainties in PS (translation and rotation) and machine-positioning (gantry-, table-, collimator-rotation and multi-leaf collimator (MLC)). Plan robustness was evaluated by the uncertainties' impact on normal tissue complication probabilities (NTCP) for xerostomia and dysphagia and on dose-volume endpoints. Differences between DTRT and VMAT plan robustness were compared using Wilcoxon matched-pair signed-rank test (α  = 5 %).

Results

Average NTCP for moderate-to-severe xerostomia and grade ≥ II dysphagia was lower for DTRT than VMAT in the nominal scenario (0.5 %, p = 0.01; 2.1 %, p < 0.01) and for all investigated uncertainties, except MLC positioning, where the difference was not significant. Average differences compared to the nominal scenario were ≤ 3.5 Gy for rotational PS (≤ 3°) and machine-positioning (≤ 2°) uncertainties, <7 Gy for translational PS uncertainties (≤ 5 mm) and < 20 Gy for MLC-positioning uncertainties (≤ 5 mm).

Conclusions

DTRT and VMAT plan robustness to the investigated uncertainties depended on uncertainty direction and location of the structure-of-interest to the target. NTCP remained on average lower for DTRT than VMAT even when considering uncertainties.

A Novel Approach for Position Verification and Dose Calculation through Local MVCT Reconstruction

Traditional positioning verification using cone-beam computed tomography (CBCT) may incur errors due to potential misalignments between the isocenter of CBCT and the treatment beams in radiotherapy. This study introduces an innovative method for verifying patient positioning in radiotherapy. Initially, the transmission images from an electronic portal imaging device (EPID) are acquired from 10 distinct angles. Utilizing the ART-TV algorithm, a sparse reconstruction of local megavoltage computed tomography (MVCT) is performed. Subsequently, this MVCT is aligned with the planning CT via a three-dimensional mutual information registration technique, pinpointing any patient-positioning discrepancies and facilitating corrective adjustments to the treatment setup. Notably, this approach employs the same radiation source as used in treatment to obtain three-dimensional images, thereby circumventing errors stemming from misalignment between the isocenter of CBCT and the accelerator. The registration process requires only 10 EPID images, and the dose absorbed during this process is included in the total dose calculation. The results show that our method's reconstructed MVCT images fulfill the requirements for registration, and the registration algorithm accurately detects positioning errors, thus allowing for adjustments in the patient's treatment position and precise calculation of the absorbed dose.

Plan complexity metrics for head and neck VMAT competition plans

Previous plan competitions have largely focused on dose metric assessments. However, whether the submitted plans were realistic and reasonable from a quality assurance (QA) perspective remains unclear. This study aimed to investigate the relationship between aperture-based plan complexity metrics (PCM) in volumetric modulated arc therapy (VMAT) competition plans and clinical treatment plans verified through patient-specific QA (PSQA). In addition, the association of PCMs with plan quality was examined. A head and neck (HN) plan competition was held for Japanese institutions from June 2019 to July 2019, in which 210 competition plans were submitted. Dose distribution quality was quantified based on dose-volume histogram (DVH) metrics by calculating the dose distribution plan score (DDPS). Differences in PCMs between the two VMAT treatment plan groups (HN plan competitions held in Japan and clinically accepted HN VMAT plans through PSQA) were investigated. The mean (± standard deviation) DDPS for the 98 HN competition plans was 158.5 ± 20.6 (maximum DDPS: 200). DDPS showed a weak correlation with PCMs with a maximum r of 0.45 for monitor unit (MU); its correlation with some PCMs was "very weak." Significant differences were found in some PCMs between plans with the highest 20% DDPSs and the remaining plans. The clinical VMAT and competition plans revealed similar distributions for some PCMs. Deviations in PCMs for the two groups were comparable, indicating considerable variability among planners regarding planning skills. The plan complexity for HN VMAT competition plans increased for high-quality plans, as shown by the dose distribution. Direct comparison of PCMs between competition plans and clinically accepted plans showed that the submitted HN VMAT competition plans were realistic and reasonable from the QA perspective. This evaluation may provide a set of criteria for evaluating plan quality in plan competitions.

Detecting outliers beyond tolerance limits derived from statistical process control in patient-specific quality assurance

BACKGROUND

Tolerance limit is defined on pre-treatment patient specific quality assurance results to identify "out of the norm" dose discrepancy in plan. An out-of-tolerance plan during measurement can often cause treatment delays especially if replanning is required. In this study, we aim to develop an outlier detection model to identify out-of-tolerance plan early during treatment planning phase to mitigate the above-mentioned risks.

METHODS

Patient-specific quality assurance results with portal dosimetry for stereotactic body radiotherapy measured between January 2020 and December 2021 were used in this study. Data were divided into thorax and pelvis sites and gamma passing rates were recorded using 2%/2 mm, 2%/1 mm, and 1%/1 mm gamma criteria. Statistical process control method was used to determine six different site and criterion-specific tolerance and action limits. Using only the inliers identified with our determined tolerance limits, we trained three different outlier detection models using the plan complexity metrics extracted from each treatment field-robust covariance, isolation forest, and one class support vector machine. The hyperparameters were optimized using the F1-score calculated from both the inliers and validation outliers' data.

RESULTS

308 pelvis and 200 thorax fields were used in this study. The tolerance (action) limits for 2%/2 mm, 2%/1 mm, and 1%/1 mm gamma criteria in the pelvis site are 99.1% (98.1%), 95.8% (91.1%), and 91.7% (86.1%), respectively. The tolerance (action) limits in the thorax site are 99.0% (98.7%), 97.0% (96.2%), and 91.5% (87.2%). One class support vector machine performs the best among all the algorithms. The best performing model in the thorax (pelvis) site achieves a precision of 0.56 (0.54), recall of 1.0 (1.0), and F1-score of 0.72 (0.70) when using the 2%/2 mm (2%/1 mm) criterion.

CONCLUSION

The model will help the planner to identify an out-of-tolerance plan early so that they can refine the plan further during the planning stage without risking late discovery during measurement.

Evaluating dose coverage and conformity in stereotactic ablative body radiotherapy (SABR) plans

PURPOSE

Accepted conformity metrics in stereotactic ablative body radiotherapy (SABR) have significant limitations. This work aimed to develop a spatial assessment methodology that improves and automates checks of dose prescription and dose gradient from planning target volume (PTV) edge.

METHODS

A Python-based script was developed to determine linear distances from the PTV edge to specified isodose, every 15 degrees on all axial slices and along the central axis in the coronal plane. A new "Internal PTV contour" distance metric is introduced as a size and shape indicator. 134 previously treated SABR patients stratified by anatomical site and PTV volume were analysed to establish baselines and tolerances for automation acceptability.

RESULTS

In the axial plane, median distance (MD) from PTV edge to the 100 % isodose was 0.13 mm (range: -0.67 to 0.53 mm), and for the 90 % isodose was 2.37 mm (1.36 to 3.40 mm). Lung and non-Lung dose gradient criteria was established by fitting a second order polynomial to the MD as a function of "Internal PTV contour". This resulted in acceptability criteria of MD + 1 mm for 80 % isodose and MD + 2 mm for the 50 % isodose. For the coronal plane, MD to the 100 % isodose was 0.49 mm (-1.24 to 2.14 mm) and for the 90 % was 1.73 mm (-0.49 to 4.13 mm).

CONCLUSIONS

Our in-house script enables a high-quality spatial assessment of PTV dose coverage and gradient, with the new 'Internal PTV contour' distance metric correlating well with dose gradient.

Multi-institutional investigation into the robustness of intra-cranial multi-target stereotactic radiosurgery plans to delivery errors

BACKGROUND

The use of modulated techniques for intra-cranial stereotactic radiosurgery (SRS) results in highly modulated fields with small apertures, which may be susceptible to uncertainties in the delivery device.

PURPOSE

This study aimed to quantify the impact of simulated delivery errors on treatment plan dosimetry and how this is affected by treatment planning system (TPS), plan geometry, delivery technique, and plan complexity. A beam modelling error was also included as context to the dose uncertainties due to treatment delivery errors.

METHODS

Delivery errors were assessed for multiple-target brain SRS plans obtained through the Trans-Tasman Radiation Oncology Group (TROG) international treatment planning challenge (2018). The challenge dataset consisted of five intra-cranial targets, each with a prescription of 20 Gy. Of the final dataset of 54 plans, 51 were created using the volumetric modulated arc therapy (VMAT) technique and three used intensity modulated arc therapy (IMRT). Thirty-five plans were from the Varian Eclipse TPS, 17 from Elekta Monaco TPS, and one plan each from RayStation and Philips Pinnacle TPS. The errors introduced included: monitor unit calibration errors, multi-leaf collimator (MLC) bank offset, single MLC leaf offset, couch rotations, and collimator rotations. Dosimetric leaf gap (DLG) error was also included as a beam modelling error. Dose to targets was assessed via dose covering 98% of planning target volume (PTV) (D98%), dose covering 2% of PTV (D2%), and dose covering 99% of gross tumor volume (GTV) (D99%). Dose to organs at risk (OARs) was assessed using the volume of normal brain receiving 12 Gy (V12Gy), mean dose to normal brain, and maximum dose covering 0.03cc brainstem (D0.03cc). Plan complexity was also assessed via edge metric, modulation complexity score (MCS), mean MLC gap, mean MLC speed, and plan modulation (PM).

RESULTS

PTV D98% showed high robustness on average to most errors with the exception of a bank shift of 1.0 mm and large rotational errors ≥1.0° for either the couch or collimator. However, in some cases, errors close to or within generally accepted machine tolerances resulted in clinically relevant impacts. The greatest impact upon normal brain V12Gy, mean dose to normal brain, and D0.03cc brainstem was found for DLG error in alignment with other recent studies. All delivery errors had on average a minimal impact across these parameters. Comparing plans from the Monaco TPS and the Eclipse TPS, showed a lesser increase to V12Gy, mean dose to normal brain, and D0.03cc brainstem for Monaco plans (p < 0.01) when DLG error was simulated. Monaco plans also correlated to lower plan complexity. Using Spearman's correlation coefficient (r) a strong negative correlation (r ≤ -0.8) was found between the mean MLC gap and dose to OARs for DLG errors.

CONCLUSIONS

Reducing MLC complexity and using larger mean MLC gaps is recommended to improve plan robustness and reduce sensitivity to delivery and modelling errors. For cases in which the calculated dose distribution or dose indices are close to the clinically acceptable limits, this is especially important.

A robust evaluation of 49 high-dose-rate prostate brachytherapy treatment plans including all major uncertainties

BACKGROUND

Uncertainties in radiotherapy cause deviation from the planned dose distribution and may result in delivering a treatment that fails to meet clinical objectives. The impact of uncertainties is unique to the patient anatomy and the needle locations in HDR prostate brachytherapy. Evaluating this impact during treatment planning is not common practice, relying on margins around the target or organs-at-risk to account for uncertainties.

PURPOSE

A robust evaluation framework for HDR prostate brachytherapy treatment plans was evaluated on 49 patient plans, measuring the range of possible dosimetric outcomes to the patient due to 14 major uncertainties.

METHODS

Patient plans were evaluated for their robustness to uncertainties by simulating probable uncertainty scenarios. Five-thousand probabilistic and 1943 worst-case scenarios per patient were simulated by changing the position and size of structures and length of dwell times from their nominal values. For each uncertainty scenario, the prostate D90 and maximum doses to the urethra, D0.01cc , and rectum, D0.1cc , were calculated.

RESULTS

The D90 was an average 1.16 ± 0.51% (mean ± SD) below nominal values for the probabilistic scenarios; the D0.01cc metric was 2.24 ± 0.90% higher; and D0.1cc was greater by 0.48 ± 0.30%. The D0.01cc and D90 metrics were more sensitive to uncertainties than D0.1cc , with a median of 79.0% and 84.9% of probabilistic scenarios passing the constraints, compared to 96.5%. The median pass-rate for scenarios that passed all three metrics simultaneously was 63.4%.

CONCLUSIONS

Assessing treatment plan robustness improves plan quality assurance, is achievable in less than 1-min, and identifies treatment plans with poor robustness, allowing re-optimization before delivery.

Multi-institutional generalizability of a plan complexity machine learning model for predicting pre-treatment quality assurance results in radiotherapy

Background and purpose

Treatment plans in radiotherapy are subject to measurement-based pre-treatment verifications. In this study, plan complexity metrics (PCMs) were calculated per beam and used as input features to develop a predictive model. The aim of this study was to determine the robustness against differences in machine type and institutional-specific quality assurance (QA).

Material and methods

A number of 567 beams were collected, where 477 passed and 90 failed the pre-treatment QA. Treatment plans of different anatomical regions were included. One type of linear accelerator was represented. For all beams, 16 PCMs were calculated. A random forest classifier was trained to distinct between acceptable and non-acceptable beams. The model was validated on other datasets to investigate its robustness. Firstly, plans for another machine type from the same institution were evaluated. Secondly, an inter-institutional validation was conducted on three datasets from different centres with their associated QA.

Results

Intra-institutionally, the PCMs beam modulation, mean MLC gap, Q1 gap, and Modulation Complexity Score were the most informative to detect failing beams. Eighty-tree percent of the failed beams (15/18) were detected correctly. The model could not detect over-modulated beams of another machine type. Inter-institutionally, the model performance reached higher accuracy for centres with comparable equipment both for treatment and QA as the local institute.

Conclusions

The study demonstrates that the robustness decreases when major differences appear in the QA platform or in planning strategies, but that it is feasible to extrapolate institutional-specific trained models between centres with similar clinical practice. Predictive models should be developed for each machine type.

Recommendation on the technical and dosimetric data to be included in stereotactic body radiation therapy clinical trial publications based on a systematic review

The results of phase II and III trials on Stereotactic Body Radiation Therapy (SBRT) increased adoption of SBRT worldwide. The ability to replicate clinical trial outcomes in routine practice depends on the capability to reproduce technical and dosimetric procedures used in the clinical trial. In this systematic review, we evaluated if peer-reviewed publications of clinical trials in SBRT reported sufficient technical data to ensure safe and robust implementation in real world clinics. Twenty papers were selected for inclusion, and data was extracted by a working group of medical physicists created following the ESTRO 2021 physics workshop. A large variability in technical and dosimetric data were observed, with frequent lack of required information for reproducing trial procedures. None of the evaluated studies were judged completely reproducible from a technical perspective. A list of recommendations has been provided by the group, based on the analysis and consensus process, to ensure an adequate reproducibility of technical parameters in primary SBRT clinical trials. Future publications should consider these recommendations to assist transferability of the clinical trial in real world practice.

A Beam Angle Selection Method to Improve Plan Robustness Against Position Error in Intensity-Modulated Radiotherapy for Left-Sided Breast Cancer

PURPOSE

We report a dosimetric study in whole breast irradiation (WBI) of plan robustness evaluation against position error with two radiation techniques: tangential intensity-modulated radiotherapy (T-IMRT) and multi-angle IMRT (M-IMRT).

METHODS

Ten left-sided patients underwent WBI were selected. The dosimetric characteristics, biological evaluation and plan robustness were evaluated. The plan robustness quantification was performed by calculating the dose differences (Δ) of the original plan and perturbed plans, which were recalculated by introducing a 3-, 5-, and 10-mm shift in 18 directions.

RESULTS

M-IMRT showed better sparing of high-dose volume of organs at risk (OARs), but performed a larger low-dose irradiation volume of normal tissue. The greater shift worsened plan robustness. For a 10-mm perturbation, greater dose differences were observed in T-IMRT plans in nearly all directions, with higher ΔD98%, ΔD95%, and ΔDmean of CTV Boost and CTV. A 10-mm shift in inferior (I) direction induced CTV Boost in T-IMRT plans a 1.1 (ΔD98%), 1.1 (ΔD95%), and 1.7 (ΔDmean) times dose differences greater than dose differences in M-IMRT plans. For CTV Boost, shifts in the right (R) and I directions generated greater dose differences in T-IMRT plans, while shifts in left (L) and superior (S) directions generated larger dose differences in M-IMRT plans. For CTV, T-IMRT plans showed higher sensitivity to a shift in the R direction. M-IMRT plans showed higher sensitivity to shifts in L, S, and I directions. For OARs, negligible dose differences were found in V20 of the lungs and heart. Greater ΔDmax of the left anterior descending artery (LAD) was seen in M-IMRT plans.

CONCLUSION

We proposed a plan robustness evaluation method to determine the beam angle against position uncertainty accompanied by optimal dose distribution and OAR sparing.

Automatic dose prediction using deep learning and plan optimization with finite-element control for intensity modulated radiation therapy

BACKGROUND

Automatic solutions for generating radiotherapy treatment plans using deep learning (DL) have been investigated by mimicking the voxel's dose. However, plan optimization using voxel-dose features has not been extensively studied.

PURPOSE

This study aims to investigate the efficiency of a direct optimization strategy with finite elements (FEs) after DL dose prediction for automatic intensity-modulated radiation therapy (IMRT) treatment planning.

METHODS

A double-UNet DL model was adapted for 220 cervical cancer patients (200 for training and 20 for testing), who underwent IMRT between 2016 and 2020 at our clinic. The model inputs were computed tomography (CT) slices, organs at risk (OARs), and planning target volumes (PTVs), and the outputs were dose distributions of uniformly generated high-dose region-controlled plans. The FEs were discretized into equal intervals of the dose prediction value within the [OARs avoid PTV(O-P)] and [body avoids OARs & PTV(B-OP)] regions in the test cohort and used to define the objectives for IMRT plan optimization. The plans were optimized using a two-step process. In the beginning, the plans of two extra cases with and without low-dose region control were compared to pursue robust and optimal dose adjustment degree pattern of FEs. In the first step, the mean dose of O-P FEs were constrained to differing degrees according to the pattern. The further the FEs from the PTV, the tighter the constraints. In the second step, the mean dose of O-P FEs from first step were constrained again but weakly and the dose of the B-OP FEs from dose prediction and PTV were tightly regulated. The dosimetric parameters of the OARs and PTV were evaluated and compared using an interstep approach. In another 10 cases, the plans optimized via the aforementioned steps (method 1) were compared with those directly generated by the double-UNet dose prediction model trained by low and high region-controlled plans (method 2).

RESULTS

The mean differences in dose metrics between the UNet-predicted dose and the clinical plans were: 0.47 Gy for bladder D50% ; 0.62 Gy for rectum D50% ; 0% for small intestine V30Gy ; 1% for small intestine V40Gy ; 4% for left femoral head V30Gy ; and 6% for right femoral head V30Gy . The reductions in mean dose (p < 0.001) after FE-based optimization were: 4.0, 1.9, 2.8, 5.9, and 5.7 Gy for the bladder, rectum, small intestine, left femoral head, and right femoral head, respectively, with flat PTV homogeneity and conformity. Method 1 plans produced lower mean doses than those of method 2 for the bladder (0.7 Gy), rectum (1.0 Gy), and small intestine (0.6 Gy), while maintaining  PTV homogeneity and conformity.

CONCLUSION

FE-based direct optimization produced lower OAR doses and adequate PTV doses after DL prediction. This solution offers rapid and automatic plan optimization without manual adjustment, particularly in low-dose regions.

Impact of ovary-sparing treatment planning on plan quality, treatment time and gamma passing rates in intensity-modulated radiotherapy for stage I/II cervical cancer

BACKGROUND

This study aimed to investigate the impact of ovary-sparing intensity-modulated radiotherapy (IMRT) on plan quality, treatment time, and gamma passing rates for stage I/II cervical cancer patients.

METHODS

Fifteen stage I/II cervical cancer patients were retrospectively enrolled, and a pair of clinically suitable IMRT plans were designed for each patient, with (Group A) and without (Group B) ovary-sparing. Plan factors affecting plan quality, treatment time, and gamma passing rates, including the number of segments, monitor units, percentage of small-area segments (field area < 20 cm2), and percentage of small-MU segments (MU < 10), were compared and statistically analyzed. Key plan quality indicators, including ovarian dose, target dose coverage (D98%, D95%, D50%, D2%), conformity index, and homogeneity index, were evaluated and statistically assessed. Treatment time and gamma passing rates collected by IBA MatriXX were also compared.

RESULTS

The median ovarian dose in Group A and Group B was 7.61 Gy (range 6.71-8.51 Gy) and 38.52 Gy (range 29.84-43.82 Gy), respectively. Except for monitor units, all other plan factors were significantly lower in Group A than in Group B (all P < .05). Correlation coefficients between plan factors, treatment time, and gamma passing rates that were statistically different were all negative. Both Groups of plans met the prescription requirement (D95% ≥ 45.00 Gy) for clinical treatment. D98% was smaller for Group A than for Group B (P < .05); D50% and D2% were larger for Group A than for Group B (P < .05, P < .05). Group A plans had worse conformity index and homogeneity index than Group B plans (P < .05, P < .05). Treatment time did not differ significantly (P > .05). Gamma passing rates in Group A were higher than in Group B with the criteria of 2%/3 mm (P < .05) and 3%/2 mm (P < .05).

CONCLUSION

Despite the slightly decreased quality of the treatment plans, the ovary-sparing IMRT plans exhibited several advantages including lower ovarian dose and plan complexity, improved gamma passing rates, and a negligible impact on treatment time.

Evaluation of complexity and deliverability of IMRT treatment plans for breast cancer

This study aimed to predict the outcome of patient specific quality assurance (PSQA) in IMRT for breast cancer using complexity metrics, such as MU factor, MAD, CAS, MCS. Several breast cancer plans were considered, including LBCS, RBCS, LBCM, RBCM, left breast, right breast and the whole breast for both Edge and TrueBeam LINACS. Dose verification was completed by Portal Dosimetry (PD). The receiver operating characteristic (ROC) curve was employed to determine whether the treatment plans pass or failed. The area under the curve (AUC) was used to assess the classification performance. The correlation of PSQA and complexity metrics was examined using Spearman's rank correlation coefficient (Rs). For LINACS, the most suitable complexity metric was found to be the MU factor (Edge Rs = - 0.608, p < 0.01; TrueBeam Rs = - 0.739, p < 0.01). Regarding the specific breast cancer categories, the optimal complexity metrics were as follows: MAD (AUC = 0.917) for LBCS, MCS (AUC = 0.681) for RBCS, MU factor (AUC = 0.854) for LBCM and MAD (AUC = 0.731) for RBCM. On the Edge LINAC, the preferable method for breast cancers was MCS (left breast, AUC = 0.938; right breast, AUC = 0.813), while on the TrueBeam LINAC, it became MU factor (left breast, AUC = 0.950) and MCS (right breast, AUC = 0.806), respectively. Overall, there was no universally suitable complexity metric for all types of breast cancers. The choice of complexity metric depended on different cancer types, locations and treatment LINACs. Therefore, when utilizing complexity metrics to predict PSQA outcomes in IMRT for breast cancer, it was essential to select the appropriate metric based on the specific circumstances and characteristics of the treatment.

Comparative Study of Plan Robustness for Breast Radiotherapy: Volumetric Modulated Arc Therapy Plans with Robust Optimization versus Manual Flash Approach

A previous study investigated robustness of manual flash (MF) and robust optimized (RO) volumetric modulated arc therapy plans for breast radiotherapy based on five patients in 2020 and indicated that the RO was more robust than the MF, although the MF is still current standard practice. The purpose of this study was to compare their plan robustness in terms of dose variation to clinical target volume (CTV) and organs at risk (OARs) based on a larger sample size. This was a retrospective study involving 34 female patients. Their plan robustness was evaluated based on measured volume/dose difference between nominal and worst scenarios (ΔV/ΔD) for each CTV and OARs parameter, with a smaller difference representing greater robustness. Paired sample t-test was used to compare their robustness values. All parameters (except CTV ΔD98%) of the RO approach had smaller ΔV/ΔD values than those of the MF. Also, the RO approach had statistically significantly smaller ΔV/ΔD values (p < 0.001-0.012) for all CTV parameters except the CTV ΔV95% and ΔD98% and heart ΔDmean. This study's results confirm that the RO approach was more robust than the MF in general. Although both techniques were able to generate clinically acceptable plans for breast radiotherapy, the RO could potentially improve workflow efficiency due to its simpler planning process.

Impact of the gradient in gantry-table rotation on dynamic trajectory radiotherapy plan quality

BACKGROUND

To improve organ at risk (OAR) sparing, dynamic trajectory radiotherapy (DTRT) extends VMAT by dynamic table and collimator rotation during beam-on. However, comprehensive investigations regarding the impact of the gantry-table (GT) rotation gradient on the DTRT plan quality have not been conducted.

PURPOSE

To investigate the impact of a user-defined GT rotation gradient on plan quality of DTRT plans in terms of dosimetric plan quality, dosimetric robustness, deliverability, and delivery time.

METHODS

The dynamic trajectories of DTRT are described by GT and gantry-collimator paths. The GT path is determined by minimizing the overlap of OARs with planning target volume (PTV). This approach is extended to consider a GT rotation gradient by means of a maximum gradient of the path (G m a x ${G}_{max}$ ) between two adjacent control points (G = | Δ table angle / Δ gantry angle | $G = | \Delta {{\mathrm{table\ angle}}/\Delta {\mathrm{gantry\ angle}}} |$ ) and maximum absolute change of G (Δ G m a x ${{\Delta}}{G}_{max}$ ). Four DTRT plans are created with different maximum G&∆G:G m a x ${G}_{max}$ &Δ G m a x ${{\Delta}}{G}_{max}$  = 0.5&0.125 (DTRT-1), 1&0.125 (DTRT-2), 3&0.125 (DTRT-3) and 3&1‍(DTRT-4), including 3-4 dynamic trajectories, for three clinically motivated cases in the head and neck and brain region (A, B, and C). A reference VMAT plan for each case is created. For all plans, plan quality is assessed and compared. Dosimetric plan quality is evaluated by target coverage, conformity, and OAR sparing. Dosimetric robustness is evaluated against systematic and random patient-setup uncertainties between± 3 mm $ \pm 3\ {\mathrm{mm}}$ in the lateral, longitudinal, and vertical directions, and machine uncertainties between± 4 ∘ $ \pm 4^\circ \ $ in the dynamically rotating machine components (gantry, table, collimator rotation). Delivery time is recorded. Deliverability and delivery accuracy on a TrueBeam are assessed by logfile analysis for all plans and additionally verified by film measurements for one case. All dose calculations are Monte Carlo based.

RESULTS

The extension of the DTRT planning process with user-definedG m a x & Δ G m a x ${G}_{max}\& {{\Delta}}{G}_{max}$ to investigate the impact of the GT rotation gradient on plan quality is successfully demonstrated. With increasingG m a x & Δ G m a x ${G}_{max}\& {{\Delta}}{G}_{max}$ , slight (case C,D m e a n , p a r o t i d l . ${D}_{mean,\ parotid\ l.}$ : up to‍-1‍Gy) and substantial (case A,D 0.03 c m 3 , o p t i c n e r v e r . ${D}_{0.03c{m}^3,\ optic\ nerve\ r.}$ : up to -9.3 Gy, case‍B,D m e a n , b r a i n $\ {D}_{mean,\ brain}$ : up to -4.7‍Gy) improvements in OAR sparing are observed compared to VMAT, while maintaining similar target coverage. All plans are delivered on the TrueBeam. Expected and actual machine position values recorded in the logfiles deviated by <0.2° for gantry, table and collimator rotation. The film measurements agreed by >96% (2%‍global/2 mm Gamma passing rate) with the dose calculation. With increasingG m a x & Δ G m a x ${G}_{max}\& {{\Delta}}{G}_{max}$ , delivery time is prolonged by <2 min/trajectory (DTRT-4) compared to VMAT and DTRT-1. The DTRT plans for case A and B and the VMAT plan for case C plan reveal the best dosimetric robustness for the considered uncertainties.

CONCLUSION

The impact of the GT rotation gradient on DTRT plan quality is comprehensively investigated for three cases in the head and neck and brain region. Increasing freedom in this gradient improves dosimetric plan quality at the cost of increased delivery time for the investigated cases. No clear dependency of GT rotation gradient on dosimetric robustness is observed.

Scorecards: Quantifying Dosimetric Plan Quality in Pancreatic Ductal Adenocarcinoma Stereotactic Body Radiation Therapy

Purpose

A scoring mechanism called the scorecard that objectively quantifies the dosimetric plan quality of pancreas stereotactic body radiation therapy treatment plans is introduced.

Methods and Materials

A retrospective analysis of patients with pancreatic ductal adenocarcinoma receiving stereotactic body radiation therapy at our institution between November 2019 and November 2020 was performed. Ten patients were identified. All patients were treated to 36 Gy in 5 fractions, and organs at risk (OARs) were constrained based on Alliance A021501. The scorecard awarded points for OAR doses lower than those cited in Alliance A021501. A team of 3 treatment planners and 2 radiation oncologists, including a physician resident without plan optimization experience, discussed the relative importance of the goals of the treatment plan and added additional metrics for OARs and plan quality indexes to create a more rigorous scoring mechanism. The scorecard for this study consisted of 42 metrics, each with a unique piecewise linear scoring function which is summed to calculate the total score (maximum possible score of 365). The scorecard-guided plan, the planning and optimization for which were done exclusively by the physician resident with no prior plan optimization experience, was compared with the clinical plan, the planning and optimization for which were done by expert dosimetrists, using the Sign test.

Results

Scorecard-guided plans had, on average, higher total scores than those clinically delivered for each patient, averaging 280.1 for plans clinically delivered and 311.7 for plans made using the scorecard (P = .003). Additionally, for most metrics, the average score of each metric across all 10 patients was higher for scorecard-guided plans than for clinically delivered plans. The scorecard guided the planner toward higher coverage, conformality, and OAR sparing.

Conclusions

A scorecard tool can help clarify the goals of a treatment plan and provide an objective method for comparing the results of different plans. Our study suggests that a completely novice treatment planner can use a scorecard to create treatment plans with enhanced coverage, conformality, and improved OAR sparing, which may have significant effects on both tumor control and toxicity. These tools, including the scorecard used in this study, have been made freely available.

Online adaptive radiotherapy: Assessment of planning technique and its impact on longitudinal plan quality robustness in pancreatic cancer

BACKGROUND AND PURPOSE

Planning on a static dataset that reflects the simulation day anatomy is routine for SBRT. We hypothesize the quality of on-table adaptive plans is similar to the baseline plan when delivering stereotactic MR-guided adaptive radiotherapy (SMART) for pancreatic cancer (PCa).

MATERIALS AND METHODS

Sixty-seven inoperable PCa patients were prescribed 50 Gy/5-fraction SMART. Baseline planning included: 3-5 mm gastrointestinal (GI) PRV, 50 Gy optimization target (PTVopt) based on GI PRV, conformality rings, and contracted GTV to guide the hotspot. For each adaptation, GI anatomy was re-contoured, followed by re-optimization. Plan quality was evaluated for target coverage (TC = PTVopt V100%/volume), PTV D90% and D80%, homogeneity index (HI = PTVopt D2%/D98%), prescription isodose/target volume (PITV), low-dose conformity (D2cm = maximum dose at 2 cm from PTVopt/Rx dose), and gradient index (R50%=50% Rx isodose volume/PTVopt volume).A novel global planning metric, termed the Pancreas Adaptive Radiotherapy Score (PARTS), was developed and implemented based on GI OAR sparing, PTV/GTV coverage, and conformality. Adaptive robustness (baseline to fraction 1) and stability (difference between two fractions with highest GI PRV variation) were quantified.

RESULTS

OAR constraints were met on all baseline (n = 67) and adaptive (n = 318) plans. Coverage for baseline/adaptive plans was mean ± SD at 44.9 ± 5.8 Gy/44.3 ± 5.5 Gy (PTV D80%), 50.1 ± 4.2 Gy/49.1 ± 4.7 Gy (PTVopt D80%), and 80%±18%/74%±18% (TC), respectively. Mean homogeneity and conformality for baseline/adaptive plans were 0.87 ± 0.25/0.81 ± 0.30 (PITV), 3.81 ± 1.87/3.87 ± 2.0 (R50%), 1.53 ± 0.23/1.55 ± 0.23 (HI), and 58%±7%/59%±7% (D2cm), respectively. PARTS was found to be a sensitive metric due to its additive influence of geometry changes on PARTS' sub-metrics. There were no statistical differences (p > 0.05) for stability, except for PARTS (p = 0.04, median difference -0.6%). Statistical differences for robustness when significant were small for most metrics (<2.0% median). Median adaptive re-optimizations were 2.

CONCLUSION

We describe a 5-fraction ablative SMART planning approach for PCa that is robust and stable during on-table adaption, due to gradients controlled by a GI PRV technique and the use of rings. These findings are noteworthy given that daily interfraction anatomic GI OAR differences are routine, thus necessitating on-table adaptation. This work supports feasibility towards utilizing a patient-independent, template on-table adaptive approach.

A generalized fit index for evaluating treatment plans of multiple target volumes with different prescribed dose: Generalized dose distribution fit index

BACKGROUND AND PURPOSE

The differential fit index (dFI) and cumulative fit index (cFI) were defined in our previous study to evaluate the fit of isodose surfaces to the target volume. They were only applicable to plans for a single target volume. Therefore, this study aimed to generalize these indices for evaluating plans for multiple target volumes and different prescribed doses.

MATERIALS AND METHODS

dFI was redefined as the ratio of the integral dose of the volume occupied by an isodose surface to that of the union of all target volumes. cFI was defined as the integral of dFI from a certain dose level of interest to the prescribed dose to be evaluated. To evaluate the performance of the generalized fit index, brain metastasis, head and neck, lung cancer, liver cancer, and cervical cancer cases were selected. For each case, a pair of plans was designed, with one plan having a better fitting dose distribution. The dose fit of these plans was investigated using cFI, the dose gradient index (GI), and the conformity index (CI).

RESULTS

In total, 26 pairs of evaluations were performed. The correct evaluation rates for cFI, GI, and CI were 96%, 26.92%, and 92.31%, respectively, illustrating that GI was not valid for evaluating complex plans.

CONCLUSIONS

The generalized fit index proved effective for evaluating the dose fit of plans for multiple target volumes with different prescribed doses.

Plan quality and treatment efficiency assurance of two VMAT optimization for cervical cancer radiotherapy

To investigate the difference of the fluence map optimization (FMO) and Stochastic platform optimization (SPO) algorithm in a newly-introduced treatment planning system (TPS).

METHODS

34 cervical cancer patients with definitive radiation were retrospectively analyzed. Each patient has four plans: FMO with fixed jaw plans (FMO-FJ) and no fixed jaw plans (FMO-NFJ); SPO with fixed jaw plans (SPO-FJ) and no fixed jaw plans (SPO-NFJ). Dosimetric parameters, Modulation Complexity Score (MCS), Gamma Pass Rate (GPR) and delivery time were analyzed among the four plans.

RESULTS

For target coverage, SPO-FJ plans are the best ones (P ≤ 0.00). FMO plans are better than SPO-NFJ plans (P ≤ 0.00). For OARs sparing, SPO-FJ plans are better than FMO plans for mostly OARs (P ≤ 0.04). Additionally, SPO-FJ plans are better than SPO-NFJ plans (P ≤ 0.02), except for rectum V45Gy. Compared to SPO-NFJ plans, the FMO plans delivered less dose to bladder, rectum, colon V40Gy and pelvic bone V40Gy (P ≤ 0.04). Meanwhile, the SPO-NFJ plans showed superiority in MU, delivery time, MCS and GPR in all plans. In terms of delivery time and MCS, the SPO-FJ plans are better than FMO plans. FMO-FJ plans are better than FMO-NFJ plans in delivery efficiency. MCSs are strongly correlated with PCTV length, which are negatively with PCTV length (P ≤ 0.03). The delivery time and MUs of the four plans are strongly correlated (P ≤ 0.02). Comparing plans with fixed or no fixed jaw in two algorithms, no difference was found in FMO plans in target coverage and minor difference in Kidney_L Dmean, Mu and delivery time between PCTV width≤15.5 cm group and >15.5 cm group. For SPO plans, SPO-FJ plans showed more superiority in target coverage and OARs sparing than the SPO-NFJ plans in the two groups.

CONCLUSIONS

SPO-FJ plans showed superiority in target coverage and OARs sparing, as well as higher delivery efficiency in the four plans.

Plan robustness analysis for threshold determination of SGRT-based intrafraction motion control in 3DCRT breast cancer radiation therapy

PURPOSE

The goal of this study was to obtain maximum allowed shift deviations from planning position in six degrees of freedom (DOF), that can serve as threshold values in surface guided radiation therapy (SGRT) of breast cancer patients.

METHODS

The robustness of conformal treatment plans of 50 breast cancer patients against 6DOF shifts was investigated. For that, new dose distributions were calculated on shifted computed tomography scans and evaluated with respect to target volume and spinal cord dose. Maximum allowed shift values were identified by imposing dose constraints on the target volume dose coverage for 1DOF, and consecutively, for 6DOF shifts using an iterative approach and random sampling.

RESULTS

Substantial decreases in target dose coverage and increases of spinal cord dose were observed. Treatment plans showed highly differing robustness for different DOFs or treated area. The sensitivity was particularly high if clavicular lymph nodes were irradiated, for shifts in lateral, vertical, roll or yaw direction, and showed partly pronounced asymmetries. Threshold values showed similar properties with an absolute value range of 0.8 mm to 5 mm and 1.4° to 5°.

CONCLUSION

The robustness analysis emphasized the necessity of taking differences between DOFs and asymmetrical sensitivities into account when evaluating the dosimetric impact of position deviations. It also highlighted the importance of rotational shifts, especially if clavicular lymph nodes were irradiated. A practical approach of determining 6DOF shift limits was introduced and a set of threshold values applicable for SGRT based patient motion control was identified.

EviGUIDE - a tool for evidence-based decision making in image-guided adaptive brachytherapy for cervical cancer

PURPOSE

To develop a novel decision-support system for radiation oncology that incorporates clinical, treatment and outcome data, as well as outcome models from a large clinical trial on magnetic resonance image-guided adaptive brachytherapy (MR-IGABT) for locally advanced cervical cancer (LACC).

METHODS

A system, called EviGUIDE, was developed that combines dosimetric information from the treatment planning system, patient and treatment characteristics, and established tumor control probability (TCP), and normal tissue complication probability (NTCP) models, to predict clinical outcome of radiotherapy treatment of LACC. Six Cox Proportional Hazards models based on data from 1341 patients of the EMBRACE-I study have been integrated. One TCP model for local tumor control, and five NTCP models for OAR morbidities.

RESULTS

EviGUIDE incorporates TCP-NTCP graphs to help users visualize the clinical impact of different treatment plans and provides feedback on achievable doses based on a large reference population. It enables holistic assessment of the interplay between multiple clinical endpoints and tumour and treatment variables. Retrospective analysis of 45 patients treated with MR-IGABT showed that there exists a sub-cohort of patients (20%) with increased risk factors, that could greatly benefit from the quantitative and visual feedback.

CONCLUSION

A novel digital concept was developed that can enhance clinical decision- making and facilitate personalized treatment. It serves as a proof of concept for a new generation of decision support systems in radiation oncology, which incorporate outcome models and high-quality reference data, and aids the dissemination of evidence-based knowledge about optimal treatment and serve as a blueprint for other sites in radiation oncology.

Being certain about uncertainties: a robust evaluation method for high-dose-rate prostate brachytherapy treatment plans including the combination of uncertainties

In high-dose-rate (HDR) prostate brachytherapy the combined effect of uncertainties cause a range of possible dose distributions deviating from the nominal plan, and which are not considered during treatment plan evaluation. This could lead to dosimetric misses for critical structures and overdosing of organs at risk. A robust evaluation method to assess the combination of uncertainties during plan evaluation is presented and demonstrated on one HDR prostate ultrasound treatment plan retrospectively. A range of uncertainty scenarios are simulated by changing six parameters in the nominal plan and calculating the corresponding dose distribution. Two methods are employed to change the parameters, a probabilistic approach using random number sampling to evaluate the likelihood of variation in dose distributions, and a combination of the most extreme possible values to access the worst-case dosimetric outcomes. One thousand probabilistic scenarios were run on the single treatment plan with 43.2% of scenarios passing seven of the eight clinical objectives. The prostate D90 had a standard deviation of 4.4%, with the worst case decreasing the dose by up to 27.2%. The urethra D10 was up to 29.3% higher than planned in the worst case. All DVH metrics in the probabilistic scenarios were found to be within acceptable clinical constraints for the plan under statistical tests for significance. The clinical significance of the results from the robust evaluation method presented on any individual treatment plan needs to be compared in the context of a historical data set that contains patient outcomes with robustness analysis data to ascertain a baseline acceptance.

Novel in-house knowledge-based automated planning system for lung cancer treated with intensity-modulated radiotherapy

PURPOSE

The goal of this study was to propose a knowledge-based planning system which could automatically design plans for lung cancer patients treated with intensity-modulated radiotherapy (IMRT).

METHODS AND MATERIALS

From May 2018 to June 2020, 612 IMRT treatment plans of lung cancer patients were retrospectively selected to construct a planning database. Knowledge-based planning (KBP) architecture named αDiar was proposed in this study. It consisted of two parts separated by a firewall. One was the in-hospital workstation, and the other was the search engine in the cloud. Based on our previous study, A‑Net in the in-hospital workstation was used to generate predicted virtual dose images. A search engine including a three-dimensional convolutional neural network (3D CNN) was constructed to derive the feature vectors of dose images. By comparing the similarity of the features between virtual dose images and the clinical dose images in the database, the most similar feature was found. The optimization parameters (OPs) of the treatment plan corresponding to the most similar feature were assigned to the new plan, and the design of a new treatment plan was automatically completed. After αDiar was developed, we performed two studies. The first retrospective study was conducted to validate whether this architecture was qualified for clinical practice and involved 96 patients. The second comparative study was performed to investigate whether αDiar could assist dosimetrists in improving the quality of planning for the patients. Two dosimetrists were involved and designed plans for only one trial with and without αDiar; 26 patients were involved in this study.

RESULTS

The first study showed that about 54% (52/96) of the automatically generated plans would achieve the dosimetric constraints of the Radiation Therapy Oncology Group (RTOG) and about 93% (89/96) of the automatically generated plans would achieve the dosimetric constraints of the National Comprehensive Cancer Network (NCCN). The second study showed that the quality of treatment planning designed by junior dosimetrists was improved with the help of αDiar.

CONCLUSIONS

Our results showed that αDiar was an effective tool to improve planning quality. Over half of the patients' plans could be designed automatically. For the remaining patients, although the automatically designed plans did not fully meet the clinical requirements, their quality was also better than that of manual plans.

Generating deliverable DICOM RT treatment plans for prostate VMAT by predicting MLC motion sequences with an encoder-decoder network

BACKGROUND

Deep learning-based auto-planning is an active research field; however, for some tasks a treatment planning system (TPS) is still required.

PURPOSE

To introduce a deep learning-based model generating deliverable DICOM RT treatment plans that can be directly irradiated by a linear accelerator (LINAC). The model was based on an encoder-decoder network and can predict multileaf collimator (MLC) motion sequences for prostate VMAT radiotherapy.

METHODS

A total of 619 treatment plans from 460 patients treated for prostate cancer with single-arc VMAT were included in this study. An encoder-decoder network was trained using 465 clinical treatment plans and validated on 77 plans. The performance was analyzed on a separate test set of 77 treatment plans. Separate L1 losses were computed for the leaf and jaw positions as well as the monitor units, with the leaf loss being weighted by a factor of 100 before being added to the other losses. The generated treatment plans were recalculated in a treatment planning system and the dose-volume metrics and gamma passing rates were compared to the original dose.

RESULTS

All generated treatment plans showed good agreement with the original data, with an average gamma passing rate (3%/3 mm) of 91.9 ± 7.1%. However, the coverage of the PTVs. was slightly lower for the generated plans (D98%  = 92.9 ± 2.6%) in comparison to the original plans (D98%  = 95.7 ± 2.2%). There was no significant difference in mean dose to the bladder between the predicted and original plan (Dmean of 28.0 ± 13.5 vs. 28.1 ± 13.3% of prescribed dose) or rectum (Dmean of 42.3 ± 7.4 vs. 42.6 ± 7.5%). The maximum dose to bladder was only slightly higher in the predicted plans (D2% of 100.7 ± 5.3 vs. 99.8 ± 4.0%) and for the rectum it was even lower (D2% of 100.5 ± 3.7 vs. 100.1 ± 4.3).

CONCLUSIONS

The deep learning-based model could predict MLC motion sequences in prostate VMAT plans, eliminating the need for sequencing inside a TPS, thus revolutionizing autonomous treatment planning workflows. This research completes the loop in deep learning-based treatment planning processes, enabling more efficient workflows for real-time or online adaptive radiotherapy.

Advantages and robustness of partial VMAT with prone position for neoadjuvant rectal cancer evaluated by CBCT-based offline adaptive radiotherapy

BACKGROUND AND PURPOSE

This study aims to explore the advantages and robustness of the partial arc combined with prone position planning technique for radiotherapy in rectal cancer patients. Adaptive radiotherapy is recalculated and accumulated on the synthesis CT (sCT) obtained by deformable image registration between planning CT and cone beam CT (CBCT). Full and partial volume modulation arc therapy (VMAT) with the prone position on gastrointestinal and urogenital toxicity, based on the probability of normal tissue complications (NTCP) model in rectal cancer patients were evaluated.

MATERIALS AND METHODS

Thirty-one patients were studied retrospectively. The contours of different structures were outlined in 155 CBCT images. First, full VMAT (F-VMAT) and partial VMAT (P-VMAT) planning techniques were designed and calculated using the same optimization constraints for each individual patient. The Acuros XB (AXB) algorithm was used in order to generate more realistic dose distributions and DVH, considering the air cavities. Second, the Velocity 4.0 software was used to fuse the planning CT and CBCT to obtain the sCT. Then, the AXB algorithm was used in the Eclipse 15.6 software to conduct re-calculation based on the sCT to obtain the corresponding dose. Furthermore, the NTCP model was used to analyze its radiobiological side effects on the bladder and the bowel bag.

RESULTS

With a CTV coverage of 98%, when compared with F-VMAT, P-VMAT with the prone position technique can effectively reduce the mean dose of the bladder and the bowel bag. The NTCP model showed that the P-VMAT combined with the prone planning technique resulted in a significantly lower complication probability of the bladder (1.88 ± 2.08 vs 1.62 ± 1.41, P = 0.041) and the bowel bag (1.28 ± 1.70 vs 0.95 ± 1.52, P < 0.001) than the F-VMAT. In terms of robustness, P-VMAT was more robust than F-VMAT, considering that less dose and NTCP variation was observed in the CTV, bladder and bowel bag.

CONCLUSION

This study analyzed the advantages and robustness of the P-VMAT in the prone position from three aspects, based on the sCT fused by CBCT. Whether it is in regards to dosimetry, radiobiological effects or robustness, P-VMAT in the prone position has shown comparative advantages.

Characterizing the interplay of treatment parameters and complexity and their impact on performance on an IROC IMRT phantom using machine learning

AIM OF THE STUDY

To elucidate the important factors and their interplay that drive performance on IMRT phantoms from the Imaging and Radiation Oncology Core (IROC).

METHODS

IROC's IMRT head and neck phantom contains two targets and an organ at risk. Point and 2D dose are measured by TLDs and film, respectively. 1,542 irradiations between 2012-2020 were retrospectively analyzed based on output parameters, complexity metrics, and treatment parameters. Univariate analysis compared parameters based on pass/fail, and random forest modeling was used to predict output parameters and determine the underlying importance of the variables.

RESULTS

The average phantom pass rate was 92% and has not significantly improved over time. The step-and-shoot irradiation technique had significantly lower pass rates that significantly affected other treatment parameters' pass rates. The complexity of plans has significantly increased with time, and all aperture-based complexity metrics (except MCS) were associated with the probability of failure. Random forest-based prediction of failure had an accuracy of 98% on held-out test data not used in model training. While complexity metrics were the most important contributors, the specific metric depended on the set of treatment parameters used during the irradiation.

CONCLUSION

With the prevalence of errors in radiotherapy, understanding which parameters affect treatment delivery is vital to improve patient treatment. Complexity metrics were strongly predictive of irradiation failure; however, they are dependent on the specific treatment parameters. In addition, the use of one complexity metric is insufficient to monitor all aspects of the treatment plan.

Considerations for radiotherapy planning with MV photons using dose-to-medium

Background and purpose

Radiotherapy planning considerations were developed for the previous calculation algorithms yielding dose to water-in-water (D_w,w). Advanced algorithms improve accuracy, but their dose values in terms of dose to medium-in-medium (D_m,m) depend on the medium considered. This work aimed to show how mimicking D_w,w planning with D_m,m can introduce new issues.

Materials and methods

A head and neck case involving bone and metal heterogeneities outside the CTV was considered. Two different commercial algorithms were used to obtain D_m,m and D_w,w distributions. First, a plan was optimised to irradiate the PTV uniformly and get a homogeneous D_w,w distribution. Second, another plan was optimised to achieve homogeneous D_m,m. Both plans were calculated with D_w,w and D_m,m, and the differences between their dose distributions, clinical impact, and robustness were evaluated.

Results

Uniform irradiation produced D_m,m cold spots in bone (-4%) and implants (-10%). Uniform D_m,m compensated them by increasing fluence but, when recalculated in D_w,w, the fluence compensations produced higher doses that affected homogeneity. Additionally, doses were 1% higher for the target, and + 4% for the mandible, thus increasing toxicity risk. Robustness was impaired when increased fluence regions and heterogeneities mismatched.

Conclusion

Planning with D_m,m as with D_w,w can impact clinical outcome and impair robustness. In optimisation, uniform irradiation instead of homogeneous D_m,m distributions should be pursued when media with different D_m,m responses are involved. However, this requires adapting evaluation criteria or avoiding medium effects. Regardless of the approach, there can be systematic differences in dose prescription and constraints.

An effective and optimized patient-specific QA workload reduction for VMAT plans after MLC-modelling optimization

INTRODUCTION

Many complexity metrics characterize modulated plans. First, this study aimed at identify the optimal complexity metrics to reduce workload associated to patient-specific quality assurance (PSQA) for our equipment and processes. Second, it intended to optimize our MLC modelling to improve measurement and calculation agreement with expectation of further reducing PSQA workload.

METHODS

Correlation and sensitivity at specificity equals to 1 were evaluated for PSQA results and different complexity metrics. Thresholds to stop PSQA were determined. After validation of the optimal complexity metric and threshold for our equipment and process, the MLC modelling was reviewed with a recently published methodology. This method is based on measurements with a Farmer-type ionization chamber of synchronous and asynchronous sweeping gap plans. Effect on the PSQA results and the identified threshold was investigated.

RESULTS

In our center, the most appropriate complexity metric for reducing our PSQA workload was the Modulation Complexity Score for VMAT (MCSv). The optimization of the MLC modelling significantly reduced the number of controlled plans, specifically for one of our two Varian Clinac. Any plan with a MCSv >= 0.34 is treated without PSQA.

CONCLUSION

This study rationalized and reduced our PSQA workload by approximately 30%. It is a continuing work with new TPS, machine or PSQA equipment. It encourages centers to re-evaluate their MLC modelling as well as assess the benefit of complexity metrics to streamline their PSQA workflow. An easier access, at least for reporting, at best for optimizing plans, into the TPS would be beneficial for the community.

Updating a clinical Knowledge-Based Planning prediction model for prostate radiotherapy

BACKGROUND AND PURPOSE

Clinical knowledge-based planning (KBP) models dedicated to prostate radiotherapy treatment may require periodical updates to remain relevant and to adapt to possible changes in the clinic. This study proposes a paired comparison of two different update approaches through a longitudinal analysis.

MATERIALS AND METHODS

A clinically validated KBP model for moderately hypofractionated prostate therapy was periodically updated using two approaches: one was targeted at achieving the biggest library size (M_t), while the other one at achieving the highest mean sample quality (R_t). Four subsequent updates were accomplished. The goodness, robustness and quality of the outcomes were measured and compared to those of the common ancestor. Plan quality was assessed through the Plan Quality Metric (PQM) and plan complexity was monitored.

RESULTS

Both update procedures allowed for an increase in the OARs sparing between +3.9 % and +19.2 % compared to plans generated by a human planner. Target coverage and homogeneity slightly reduced [-0.2 %;-14.7 %] while plan complexity showed only minor changes. Increasing the sample size resulted in more reliable predictions and improved goodness-of-fit, while increasing the mean sample quality improved the outcomes but slightly reduced the models reliability.

CONCLUSIONS

Repeated updates of clinical KBP models can enhance their robustness, reliability and the overall quality of automatically generated plans. The periodical expansion of the model sample accompanied by the removal of the unacceptable low quality plans should maximize the benefits of the updates while limiting the associated workload.

Prediction models as decision-support tools for virtual patient-specific quality assurance of helical tomotherapy plans

Background and purpose

Prediction models may be reliable decision-support tools to reduce the workload associated with the measurement-based patient-specific quality assurance (PSQA) of radiotherapy plans. This study compared the effectiveness of three different models based on delivery parameters, complexity metrics and sinogram radiomics features as tools for virtual-PSQA (vPSQA) of helical tomotherapy (HT) plans.

Materials and methods

A dataset including 881 RT plans created with two different treatment planning systems (TPSs) was collected. Sixty-five indicators including 12 delivery parameters (DP) and 53 complexity metrics (CM) were extracted using a dedicated software library. Additionally, 174 radiomics features (RF) were extracted from the plans' sinograms. Three groups of variables were formed: A (DP), B (DP + CM) and C (DP + CM + RF). Regression models were trained to predict the gamma index passing rate P R γ (3%G, 2mm) and the impact of each group of variables was investigated. ROC-AUC analysis measured the ability of the models to accurately discriminate between 'deliverable' and 'non-deliverable' plans.

Results

The best performance was achieved by model C which allowed detecting around 16% and 63% of the 'deliverable' plans with 100% sensitivity for the two TPSs, respectively. In a real clinical scenario, this would have decreased the whole PSQA workload by approximately 35%.

Conclusions

The combination of delivery parameters, complexity metrics and sinogram radiomics features allows for robust and reliable PSQA gamma passing rate predictions and high-sensitivity detection of a fraction of deliverable plans for one of the two TPSs. Promising yet improvable results were obtained for the other one. The results foster a future adoption of vPSQA programs for HT.

Patient-specific quality assurance prediction models based on machine learning for novel dual-layered MLC linac

BACKGROUND

Patient-specific quality assurance (PSQA) is an indispensable and essential procedure in radiotherapy workflow, and several studies have been done to develop prediction models based on the conventional C-arm linac of single-layered multileaf collimator (MLC) with machine learning (ML) and deep learning techniques to predict PSQA results and improve efficiency. Recently, a newly designed O-ring gantry linac "Halcyon" equipped with unique jawless stacked-and-staggered dual-layered MLC was released. However, few studies have focused on developing PSQA prediction models for this novel dual-layered MLC system.

PURPOSE

To evaluate the performance of ML to predict PSQA results of fixed field intensity-modulated radiation therapy (FF-IMRT) plans for linac equipped with dual-layered MLC.

METHODS AND MATERIALS

A total of 213 FF-IMRT treatment plans, including 1383 beams from various treatment sites, were selected and delivered with portal dosimetry verification on Halcyon linac. Gamma analysis was performed using 1%/1, 2%/2, and 3%/2 mm criteria with a 10% threshold. The training set (TS) of ML models consisted of 1106 beams, and an independent evaluation set (ES) consisted of 277 beams. For each beam, 33 complexity metrics were extracted as input data for training models. Three ML algorithms (gradient boosting decision tree/GBDT, random forest/RF, and Poisson Lasso/PL) were utilized to build the models and predict gamma passing rates (GPRs). To improve the prediction accuracy in the rare region, a method of reweighting for TS has been performed and compared to the unweighted results. The importance of complexity metrics was studied by permuted interesting features.

RESULTS

The GBDT model had the best performance in this study. In ES, the minimal mean prediction error for unweighted results was 1.93%, 1.16%, 0.78% under 1%/1, 2%/2, and 3%/2 mm criteria, respectively, from GBDT model. Comparing to the unweighted results, the models after reweighting gained up to 30% improvement in the rare region, whereas the overall prediction error was slightly worse depending on the kind of models. For feature importance, 2 tree-based models (GBDT and RF) had in common the top 10 most important metrics as well as the same metric with the largest impact.

CONCLUSION

For linac equipped with novel dual-layered MLC, the ML model based on GBDT algorithm shows a certain degree of accuracy for GPRs prediction. The specific ML model for dual-layered MLC configuration could be a useful tool for physicists detecting PSQA measurement failures.

Machine learning-based automated planning for hippocampal avoidance prophylactic cranial irradiation

PURPOSE

Design and evaluate a knowledge-based model using commercially available artificial intelligence tools for automated treatment planning to efficiently generate clinically acceptable hippocampal avoidance prophylactic cranial irradiation (HA-PCI) plans in patients with small-cell lung cancer.

MATERIALS AND METHODS

Data from 44 patients with different grades of head flexion (range 45°) were used as the training datasets. A Rapid Plan knowledge-based planning (KB) routine was applied for a prescription of 25 Gy in 10 fractions using two volumetric modulated arc therapy (VMAT) arcs. The 9 plans used to validate the initial model were added to generate a second version of the RP model (Hippo-MARv2). Automated plans (AP) were compared with manual plans (MP) according to the dose-volume objectives of the PREMER trial. Optimization time and model quality were assessed using 10 patients who were not included in the first 44 datasets.

RESULTS

A 55% reduction in average optimization time was observed for AP compared to MP. (15 vs 33 min; p = 0.001).Statistically significant differences in favor of AP were found for D98% (22.6 vs 20.9 Gy), Homogeneity Index (17.6 vs 23.0) and Hippocampus D mean (11.0 vs 11.7 Gy). The AP met the proposed objectives without significant deviations, while in the case of the MP, significant deviations from the proposed target values were found in 2 cases.

CONCLUSION

The KB model allows automated planning for HA-PCI. Automation of radiotherapy planning improves efficiency, safety, and quality and could facilitate access to new techniques.

Prediction of portal dosimetry quality assurance results using log files-derived errors and machine learning techniques

Objective

This work aims to use machine learning models to predict gamma passing rate of portal dosimetry quality assurance with log file derived features. This allows daily treatment monitoring for patients and reduce wear and tear on EPID detectors to save cost and prevent downtime.

Methods

578 VMAT trajectory log files selected from prostate, lung and spine SBRT were used in this work. Four machine learning models were explored to identify the best performing regression model for predicting gamma passing rate within each sub-site and the entire unstratified data. Predictors used in these models comprised of hand-crafted log file-derived features as well as modulation complexity score. Cross validation was used to evaluate the model performance in terms of R² and RMSE.

Result

Using gamma passing rate of 1%/1mm criteria and entire dataset, LASSO regression has a R² of 0.121 ± 0.005 and RMSE of 4.794 ± 0.013%, SVM regression has a R² of 0.605 ± 0.036 and RMSE of 3.210 ± 0.145%, Random Forest regression has a R² of 0.940 ± 0.019 and RMSE of 1.233 ± 0.197%. XGBoost regression has the best performance with a R² and RMSE value of 0.981 ± 0.015 and 0.652 ± 0.276%, respectively.

Conclusion

Log file-derived features can predict gamma passing rate of portal dosimetry with an average error of less than 2% using the 1%/1mm criteria. This model can potentially be applied to predict the patient specific QA results for every treatment fraction.

Robustness of VMAT to setup errors in postmastectomy radiotherapy of left-sided breast cancer: Impact of bolus thickness

BACKGROUND

Volumetric modulated arc therapy (VMAT) with varied bolus thicknesses has been employed in postmastectomy radiotherapy (PMRT) of breast cancer to improve superficial target coverage. However, impact of bolus thickness on plan robustness remains unclear.

METHODS

The study enrolled ten patients with left-sided breast cancer who received radiotherapy using VMAT with 5 mm and 10 mm bolus (VMAT-5B and VMAT-10B). Inter-fractional setup errors were simulated by introducing a 3 mm shift to isocenter of the original plans in the anterior-posterior, left-right, and inferior-superior directions. The plans (perturbed plans) were recalculated without changing other parameters. Dose volume histograms (DVH) were collected for plan evaluation. Absolute dose differences in DVH endpoints for the clinical target volume (CTV), heart, and left lung between the perturbed plans and the original ones were used for robustness analysis.

RESULTS

VMAT-10B showed better target coverage, while VMAT-5B was superior in organs-at-risk (OARs) sparing. As expected, small setup errors of 3 mm could induce dose fluctuations in CTV and OARs. The differences in CTV were small in VMAT-5B, with a maximum difference of -1.05 Gy for the posterior shifts. For VMAT-10B, isocenter shifts in the posterior and right directions significantly decreased CTV coverage. The differences were -1.69 Gy, -1.48 Gy and -1.99 Gy, -1.69 Gy for ΔD95% and ΔD98%, respectively. Regarding the OARs, only isocenter shifts in the posterior, right, and inferior directions increased dose to the left lung and the heart. Differences in VMAT-10B were milder than those in VMAT-5B. Specifically, mean heart dose were increased by 0.42 Gy (range 0.10 ~ 0.95 Gy) and 0.20 Gy (range -0.11 ~ 0.72 Gy), and mean dose for the left lung were increased by 1.02 Gy (range 0.79 ~ 1.18 Gy) and 0.68 Gy (range 0.47 ~ 0.84 Gy) in VMAT-5B and VMAT-10B, respectively. High-dose volumes in the organs were increased by approximate 0 ~ 2 and 1 ~ 3 percentage points, respectively. Nevertheless, most of the dosimetric parameters in the perturbed plans were still clinically acceptable.

CONCLUSIONS

VMAT-5B appears to be more robust to 3 mm setup errors than VMAT-10B. VMAT-5B also resulted in better OARs sparing with acceptable target coverage and dose homogeneity. Therefore 5 mm bolus is recommended for PMRT of left-sided breast cancer using VMAT.

Automation of pencil beam scanning proton treatment planning for intracranial tumours

PURPOSE

To evaluate the feasibility of comprehensive automation of an intra-cranial proton treatment planning.

MATERIALS AND METHODS

Class solution (CS) beam configuration selection allows the user to identify predefined beam configuration based on target localization; automatic CS (aCS) will then explore all the possible CS beam geometries. Ten patients, already used for the evaluation of the automatic selection of the beam configuration, have been also employed to training an algorithm based on the computation of a benchmark dose exploit automatic general planning solution (GPS) optimization with a wish list approach for the planning optimization. An independent cohort of ten patients has been then used for the evaluation step between the clinical and the GPS plan in terms of dosimetric quality of plans and the time needed to generate a plan.

RESULTS

The definition of a beam configuration requires on average 22 min (range 9-29 min). The average time for GPS plan generation is 18 min (range 7-26 min). Median dose differences (GPS-Manual) for each OAR constraints are: brainstem -1.60 Gy, left cochlea -1.22 Gy, right cochlea -1.42 Gy, left eye 0.55 Gy, right eye -2.33 Gy, optic chiasm -1.87 Gy, left optic nerve -4.45 Gy, right optic nerve -2.48 Gy and optic tract -0.31 Gy. Dosimetric CS and aCS plan evaluation shows a slightly worsening of the OARs values except for the optic tract and optic chiasm for both CS and aCS, where better results have been observed.

CONCLUSION

This study has shown the feasibility and implementation of the automatic planning system for intracranial tumors. The method developed in this work is ready to be implemented in a clinical workflow.

Evaluation of measures related to dosimetric uncertainty of VMAT plans

Dosimetric uncertainty is most often not included in the process of creating and selecting plans for treatment. Treatment planning and the physician's choice of treatment plan is instead often based only on evaluation of clinical goals of the calculated absorbed dose distribution. Estimation of the dosimetric uncertainty could potentially have impact in the process of comparing and selecting volumetric modulated arc therapy (VMAT) plans. In this study, different measures for estimation of dosimetric uncertainty based on treatment plan parameters for plans with similar dose distributions were evaluated. VMAT plans with similar dose distributions but with different treatment plan designs were created using three different optimization methods for each of ten patient cases (tonsil and prostate cancer). Two plans were optimized in Eclipse, one with and one without the use of aperture shape controller, and one plan was optimized in RayStation. The studied measures related to dosimetric uncertainty of treatment plans were aperture-based complexity metric analysis, investigation of modulation level of multi leaf collimator leaves, gantry speed and dose rate, quasi-3D measurements and evaluation of simulations of realistic delivery variations. The results showed that there can be variations in dosimetric uncertainty for treatment plans with similar dose distributions. Dosimetric uncertainty assessment could therefore have impact on the choice of plan to be used for treatment and lead to a decrease in the uncertainty level of the delivered absorbed dose distribution. This study showed that aperture shape complexity had a larger impact on dosimetric uncertainty compared to modulation level of MLC, gantry speed and dose rate.

Quantitative assessment of helical tomotherapy plans complexity

PURPOSE

An unnecessary amount of complexity in radiotherapy plans affects the efficiency of the treatments, increasing the uncertainty of dose deposition and its susceptibility to anatomical changes or setup errors. To date, tools for quantitatively assessing the complexity of tomotherapy plans are still limited. In this study, new metrics were developed to characterize different aspects of helical tomotherapy (HT) plans, and their actual effectiveness was investigated.

METHODS

The complexity of 464 HT plans delivered on a Radixact platform was evaluated. A new set of metrics was devised to assess beam geometry, leaf opening time (LOT) variability, and modulation over space and time. Sixty-five complexity metrics were extracted from the dataset using the newly in-house developed software library TCoMX: 29 metrics already proposed in the literature and 36 newly developed metrics. Their reciprocal relation is discussed. Their effectiveness was evaluated through correlation analyses with patient-specific quality assurance (PSQA) results.

RESULTS

An inverse linear relation was found between the average number of closed leaves and the average number of MLC openings and closures as well as between the choice of the modulation factor and the discontinuity of the field, suggesting some intrinsic link between the LOT distribution and the geometrical complexity of the MLC openings. The newly proposed metrics were at least as correlated as the existing ones to the PSQA results. Metrics describing the geometrical complexity of the MLC openings showed the strongest connection to the PSQA results. Significant correlations were found between at least one of the new metrics and the γ index passing rate P R γ % ( 3 % G , 2 mm ) $P{R}_{\gamma}\%(3\%G,2\textit{mm})$ for six out of seven groups of plans considered.

CONCLUSION

The new metrics proposed were shown to be effective to characterize more comprehensively the complexity of HT plans. A software library for their automatic extraction is described and made available.