Script-based implementation of automatic grid placement for lattice stereotactic body radiation therapy.

Background and purpose: Spatially fractionated radiation therapy (SFRT) has demonstrated promising clinical response in treating large tumors with heterogeneous dose distributions. Lattice stereotactic body radiation therapy (SBRT) is an SFRT technique that leverages inverse optimization to precisely localize regions of high and lose dose within disease. The aim of this study was to evaluate an automated heuristic approach to sphere placement in lattice SBRT treatment planning. Materials and methods: A script-based algorithm for sphere placement in lattice SBRT based on rules described by protocol was implemented within a treatment planning system. The script was applied to 22 treated cases and sphere distributions were compared with manually placed spheres in terms of number of spheres, number of protocol violations, and time required to place spheres. All cases were re-planned using script-generated spheres and plan quality was compared with clinical plans. Results: The mean number of spheres placed excluding those that violate rules was greater using the script (13.8) than that obtained by either dosimetrist (10.8 and 12.0, p < 0.001 and p = 0.003) or physicist (12.7, p = 0.061). The mean time required to generate spheres was significantly less using the script (2.5 min) compared to manual placement by dosimetrists (25.0 and 29.9 min) and physicist (19.3 min). Plan quality indices were similar in all cases with no significant differences, and OAR constraints remained met on all plans except two. Conclusion: A script placed spheres for lattice SBRT according to institutional protocol rules. The script-produced placement was superior to that of manually-specified spheres, as characterized by sphere number and rule violations.

A quality assurance framework for routine monitoring of deep learning cardiac substructure computed tomography segmentation models in radiotherapy.

BACKGROUND: For autosegmentation models, the data used to train the model (e.g., public datasets and/or vendor-collected data) and the data on which the model is deployed in the clinic are typically not the same, potentially impacting the performance of these models by a process called domain shift. Tools to routinely monitor and predict segmentation performance are needed for quality assurance. Here, we develop an approach to perform such monitoring and performance prediction for cardiac substructure segmentation. PURPOSE: To develop a quality assurance (QA) framework for routine or continuous monitoring of domain shift and the performance of cardiac substructure autosegmentation algorithms. METHODS: A benchmark dataset consisting of computed tomography (CT) images along with manual cardiac substructure delineations of 241 breast cancer radiotherapy patients were collected, including one "normal" image domain of clean images and five "abnormal" domains containing images with artifact (metal, contrast), pathology, or quality variations due to scanner protocol differences (field of view, noise, reconstruction kernel, and slice thickness). The QA framework consisted of an image domain shift detector which operated on the input CT images and a shape quality detector on the output of an autosegmentation model, and a regression model for predicting autosegmentation model performance. The image domain shift detector was composed of a trained denoising autoencoder (DAE) and two hand-engineered image quality features to detect normal versus abnormal domains in the input CT images. The shape quality detector was a variational autoencoder (VAE) trained to estimate the shape quality of the auto-segmentation results. The output from the image domain shift and shape quality detectors was used to train a regression model to predict the per-patient segmentation accuracy, measured by Dice coefficient similarity (DSC) to physician contours. Different regression techniques were investigated including linear regression, Bagging, Gaussian process regression, random forest, and gradient boost regression. Of the 241 patients, 60 were used to train the autosegmentation models, 120 for training the QA framework, and the remaining 61 for testing the QA framework. A total of 19 autosegmentation models were used to evaluate QA framework performance, including 18 convolutional neural network (CNN)-based and one transformer-based model. RESULTS: When tested on the benchmark dataset, all abnormal domains resulted in a significant DSC decrease relative to the normal domain for CNN models ( p < 0.001 $p < 0.001$ ), but only for some domains for the transformer model. No significant relationship was found between the performance of an autosegmentation model and scanner protocol parameters ( p = 0.42 $p = 0.42$ ) except noise ( p = 0.01 $p = 0.01$ ). CNN-based autosegmentation models demonstrated a decreased DSC ranging from 0.07 to 0.41 with added noise, while the transformer-based model was not significantly affected (ANOVA, p = 0.99 $p=0.99$ ). For the QA framework, linear regression models with bootstrap aggregation resulted in the highest mean absolute error (MAE) of 0.041 ± 0.002 $0.041 \pm 0.002$ , in predicted DSC (relative to true DSC between autosegmentation and physician). MAE was lowest when combining both input (image) detectors and output (shape) detectors compared to output detectors alone. CONCLUSIONS: A QA framework was able to predict cardiac substructure autosegmentation model performance for clinically anticipated "abnormal" domain shifts.

In silico trial of simulation-free hippocampal-avoidance whole brain adaptive radiotherapy.

Background and Purpose: Hippocampal-avoidance whole brain radiotherapy (HA-WBRT) can be a time-consuming process compared to conventional whole brain techniques, thus potentially limiting widespread utilization. Therefore, we evaluated the in silico clinical feasibility, via dose-volume metrics and timing, by leveraging a computed tomography (CT)-based commercial adaptive radiotherapy (ART) platform and workflow in order to create and deliver patient-specific, simulation-free HA-WBRT. Materials and methods: Ten patients previously treated for central nervous system cancers with cone-beam computed tomography (CBCT) imaging were included in this study. The CBCT was the adaptive image-of-the-day to simulate first fraction on-board imaging. Initial contours defined on the MRI were rigidly matched to the CBCT. Online ART was used to create treatment plans at first fraction. Dose-volume metrics of these simulation-free plans were compared to standard-workflow HA-WBRT plans on each patient CT simulation dataset. Timing data for the adaptive planning sessions were recorded. Results: For all ten patients, simulation-free HA-WBRT plans were successfully created utilizing the online ART workflow and met all constraints. The median hippocampi D100% was 7.8 Gy (6.6-8.8 Gy) in the adaptive plan vs 8.1 Gy (7.7-8.4 Gy) in the standard workflow plan. All plans required adaptation at first fraction due to both a failing hippocampal constraint (6/10 adaptive fractions) and sub-optimal target coverage (6/10 adaptive fractions). Median time for the adaptive session was 45.2 min (34.0-53.8 min). Conclusions: Simulation-free HA-WBRT, with commercially available systems, was clinically feasible via plan-quality metrics and timing, in silico.

Prospective In Silico Evaluation of Cone-Beam Computed Tomography-Guided StereoTactic Adaptive Radiation Therapy (CT-STAR) for the Ablative Treatment of Ultracentral Thoracic Disease.

Purpose: We conducted a prospective, in silico study to evaluate the feasibility of cone-beam computed tomography (CBCT)-guided stereotactic adaptive radiation therapy (CT-STAR) for the treatment of ultracentral thoracic cancers (NCT04008537). We hypothesized that CT-STAR would reduce dose to organs at risk (OARs) compared with nonadaptive stereotactic body radiation therapy (SBRT) while maintaining adequate tumor coverage. Methods and Materials: Patients who were already receiving radiation therapy for ultracentral thoracic malignancies underwent 5 additional daily CBCTs on the ETHOS system as part of a prospective imaging study. These were used to simulate CT-STAR, in silico. Initial, nonadaptive plans (PI) were created based on simulation images and simulated adaptive plans (PA) were based on study CBCTs. 55 Gy/5 fractions was prescribed, with OAR constraint prioritization over PTV coverage under a strict isotoxicity approach. PI were applied to patients' anatomy of the day and compared with daily PA using dose-volume histogram metrics, with selection of superior plans for simulated delivery. Feasibility was defined as completion of the end-to-end adaptive workflow while meeting strict OAR constraints in ≥80% of fractions. CT-STAR was performed under time pressures to mimic clinical adaptive processes. Results: Seven patients were accrued, 6 with intraparenchymal tumors and 1 with a subcarinal lymph node. CT-STAR was feasible in 34 of 35 simulated fractions. In total, 32 dose constraint violations occurred when the PI was applied to anatomy-of-the-day across 22 of 35 fractions. These violations were resolved by the PA in all but one fraction, in which the proximal bronchial tree dose was still numerically improved through adaptation. The mean difference between the planning target volume and gross total volume V100% in the PI and the PA was -0.24% (-10.40 to 9.90) and -0.62% (-11.00 to 8.00), respectively. Mean end-to-end workflow time was 28.21 minutes (18.02-50.97). Conclusions: CT-STAR widened the dosimetric therapeutic index of ultracentral thorax SBRT compared with nonadaptive SBRT. A phase 1 protocol is underway to evaluate the safety of this paradigm for patients with ultracentral early-stage NSCLC.

Clinical application of a template-guided automated planning routine.

PURPOSE: Determine the dosimetric quality and the planning time reduction when utilizing a template-based automated planning application. METHODS: A software application integrated through the treatment planning system application programing interface, QuickPlan, was developed to facilitate automated planning using configurable templates for contouring, knowledge-based planning structure matching, field design, and algorithm settings. Validations are performed at various levels of the planning procedure and assist in the evaluation of readiness of the CT image, structure set, and plan layout for automated planning. QuickPlan is evaluated dosimetrically against 22 hippocampal-avoidance whole brain radiotherapy patients. The required times to treatment plan generation are compared for the validations set as well as 10 prospective patients whose plans have been automated by QuickPlan. RESULTS: The generations of 22 automated treatment plans are compared against a manual replanning using an identical process, resulting in dosimetric differences of minor clinical significance. The target dose to 2% volume and homogeneity index result in significantly decreased values for automated plans, whereas other dose metric evaluations are nonsignificant. The time to generate the treatment plans is reduced for all automated plans with a median difference of 9' 50″ ± 4' 33″. CONCLUSIONS: Template-based automated planning allows for reduced treatment planning time with consistent optimization structure creation, treatment field creation, plan optimization, and dose calculation with similar dosimetric quality. This process has potential expansion to numerous disease sites.

A clinical and time savings evaluation of a deep learning automatic contouring algorithm.

Automatic contouring algorithms may streamline clinical workflows by reducing normal organ-at-risk (OAR) contouring time. Here we report the first comprehensive quantitative and qualitative evaluation, along with time savings assessment for a prototype deep learning segmentation algorithm from Siemens Healthineers. The accuracy of contours generated by the prototype were evaluated quantitatively using the Sorensen-Dice coefficient (Dice), Jaccard index (JC), and Hausdorff distance (Haus). Normal pelvic and head and neck OAR contours were evaluated retrospectively comparing the automatic and manual clinical contours in 100 patient cases. Contouring performance outliers were investigated. To quantify the time savings, a certified medical dosimetrist manually contoured de novo and, separately, edited the generated OARs for 10 head and neck and 10 pelvic patients. The automatic, edited, and manually generated contours were visually evaluated and scored by a practicing radiation oncologist on a scale of 1-4, where a higher score indicated better performance. The quantitative comparison revealed high (> 0.8) Dice and JC performance for relatively large organs such as the lungs, brain, femurs, and kidneys. Smaller elongated structures that had relatively low Dice and JC values tended to have low Hausdorff distances. Poor performing outlier cases revealed common anatomical inconsistencies including overestimation of the bladder and incorrect superior-inferior truncation of the spinal cord and femur contours. In all cases, editing contours was faster than manual contouring with an average time saving of 43.4% or 11.8 minutes per patient. The physician scored 240 structures with > 95% of structures receiving a score of 3 or 4. Of the structures reviewed, only 11 structures needed major revision or to be redone entirely. Our results indicate the evaluated auto-contouring solution has the potential to reduce clinical contouring time. The algorithm's performance is promising, but human review and some editing is required prior to clinical use.

Robustness of deep learning segmentation of cardiac substructures in noncontrast computed tomography for breast cancer radiotherapy.

PURPOSE: To develop and evaluate deep learning-based autosegmentation of cardiac substructures from noncontrast planning computed tomography (CT) images in patients undergoing breast cancer radiotherapy and to investigate the algorithm sensitivity to out-of-distribution data such as CT image artifacts. METHODS: Nine substructures including aortic valve (AV), left anterior descending (LAD), tricuspid valve (TV), mitral valve (MV), pulmonic valve (PV), right atrium (RA), right ventricle (RV), left atrium (LA), and left ventricle (LV) were manually delineated by a radiation oncologist on noncontrast CT images of 129 patients with breast cancer; among them 90 were considered in-distribution data, also named as "clean" data. The image/label pairs of 60 subjects were used to train a 3D deep neural network while the remaining 30 were used for testing. The rest of the 39 patients were considered out-of-distribution ("outlier") data, which were used to test robustness. Random rigid transformations were used to augment the dataset during training. We investigated multiple loss functions, including Dice similarity coefficient (DSC), cross-entropy (CE), Euclidean loss as well as the variation and combinations of these, data augmentation, and network size on overall performance and sensitivity to image artifacts due to infrequent events such as the presence of implanted devices. The predicted label maps were compared to the ground-truth labels via DSC and mean and 90th percentile symmetric surface distance (90th-SSD). RESULTS: When modified Dice combined with cross-entropy (MD-CE) was used as the loss function, the algorithm achieved a mean DSC = 0.79 ± 0.07 for chambers and  0.39 ± 0.10 for smaller substructures (valves and LAD). The mean and 90th-SSD were 2.7 ± 1.4 and 6.5 ± 2.8 mm for chambers and 4.1 ± 1.7 and 8.6 ± 3.2 mm for smaller substructures. Models with MD-CE, Dice-CE, MD, and weighted CE loss had highest performance, and were statistically similar. Data augmentation did not affect model performances on both clean and outlier data and model robustness was susceptible to network size. For a certain type of outlier data, robustness can be improved via incorporating them into the training process. The execution time for segmenting each patient was on an average 2.1 s. CONCLUSIONS: A deep neural network provides a fast and accurate segmentation of large cardiac substructures in noncontrast CT images. Model robustness of two types of clinically common outlier data were investigated and potential approaches to improve them were explored. Evaluation of clinical acceptability and integration into clinical workflow are pending.

Comprehensive validation of halcyon 2.0 plans and the implementation of patient specific QA with multiple detector platforms.

PURPOSE: To perform a comprehensive validation of plans generated on a preconfigured Halcyon 2.0 with preloaded beam model, including evaluations of new features and implementing the patient specific quality assurance (PSQA) process with multiple detectors. METHODS: A total of 56 plans were generated in Eclipse V15.6 (Varian Medical System) with a preconfigured Halcyon treatment machine. Ten plans were developed via the AAPM TG-119 test suite with both IMRT and VMAT techniques. 34 clinically treated plans using C-arm LINAC from 24 patients were replanned on Halcyon using IMRT or VMAT techniques for a variety of sites including: brain, head and neck, lung, breast, abdomen, and pelvis. Six of those plans were breast VMAT plans utilizing the extended treatment field technique available with Halcyon 2.0. The dynamically flattened beam (DFB), another new feature on Halcyon 2.0, was also used for an AP/PA spine and four field box pelvis, as well as ten 3D breast plans. All 56 plans were measured with an ion chamber (IC), film, portal dosimetry (PD), ArcCHECK, and Delta4. Tolerance and action limits were calculated and compared to the recommendations of TG-218. RESULTS: TG-119 IC and film confidence limits met those set by the task group, except for IMRT target point dose. Forty-four of 46 clinical plans were within 3% for IC measurements. Average gamma passing rates with 3% dose difference and 2mm distance-to-agreement for IMRT/VMAT plans were: Film - 96.8%, PD - 99.9%, ArcCHECK - 99.1%, and Delta4 - 99.2%. Calculated action limits were: Film - 86.3%, PD - 98.4%, ArcCHECK - 96.1%, and Delta4 - 95.7%. Extended treatment field technique was fully validated and 3D plans with DFB had similar results to IMRT/VMAT plans. CONCLUSION: Halcyon plan deliveries were verified with multiple measurement devices. New features of Halcyon 2.0 were also validated. Traditional PSQA techniques and process specific tolerance and action limits were successfully implemented.

Technical Note: Dosimetric characterization of the dynamic beam flattening MLC sequence on a ring shaped, Jawless Linear Accelerator with double stacked MLC.

PURPOSE: To characterize the dosimetric features and limitations of the dynamic beam flattening (DBF) on the Halcyon 2.0 linear accelerator (Varian Medical Systems). METHODS: A predefined multi-leaf collimator (MLC) sequence was introduced and used to flatten the 6 MV flattening filter free (FFF) beam on the Halcyon 2.0. Dosimetric characterizations of the flattened beams, including beam flatness, symmetry, percent depth dose (PDD), output factor and MU linearity, were investigated. Flatness and symmetry were obtained from profile measurements with both radiographic films (EDR2) and a two dimensional ion-chamber array (IC Profiler, Sun Nuclear Corporation). MU linearity, output factors, and PDDs were measured in a water tank with a CC13 ion chamber (Scanditronix Wellhöfer, Nuremburg, Germany). In addition, the effect of the DBF sequence on 3D plan quality was evaluated by creating DBF plans for a 4-field box rectum and an AP/PA spine plan. Patient specific QA was performed on these plans. RESULTS: At 100 cm SSD and 10 cm depth, a flatness of <3% was observed on both transversal and radial profiles for all square field sizes ≥10 cm with DBF. For both larger and smaller field sizes the flatness showed a tendency to increase as the fields got bigger or smaller, respectively. Similar trends in flatness were observed at all depths measured. All measured output factors for square field sizes ≥5 cm were within 1% of the TPS prediction. Linearity was ≤2.02% for all measurements. For both treatment sites, the MD judged the plans created for the Halcyon without the use of DBF not to be clinically acceptable, however considered both the TrueBeam plan and the Halcyon plan with the DBF sequence to be clinically acceptable. CONCLUSIONS: The DBF sequence on the Halcyon and its characteristics were investigated. The analysis indicates that the DBF sequence can be used on the Halcyon to generate clinically acceptable 3D treatment plans.

Field-in-field breast planning for a jawless, double-stack MLC LINAC using flattening-filter-free beams.

BACKGROUND: This study intends to develop an efficient field-in-field (FiF) planning technique with the Eclipse treatment planning system (TPS) to determine the feasibility of using the Halcyon treatment delivery system for 3D treatment of breast cancer. METHODS: Ten treatment plans were prepared on the Halcyon treatment planning system and compared to the same patients' clinically delivered TrueBeam plans which used flattened 6 MV and 10 MV beams. Patients selected for this study were treated via simple, tangential breast irradiation and did not receive radiotherapy of the supraclavicular or internal mammary lymph nodes. Planning target volumes (PTV) volumes ranged from 519 cc to 1211 cc with a mean target volume of 877 cc. Several planning techniques involving collimator, gantry rotation, and number of FiF segments were investigated as well as the use of the dynamically flattened beam (DFB) - a predefined MLC pattern that is designed to provide a flattened beam profile at 10 cm depth on a standard water phantom. For comparison, the clinically delivered TrueBeam plans remained unaltered except for normalization of the target coverage to more readily compare the two treatment delivery techniques. RESULTS: Using the physician defined PTV, normalized such that 98% of the volume was covered by 95% of the prescribed dose, the Halcyon plans were deemed clinically acceptable and comparable to the TrueBeam plans by the radiation oncologist. Resulting average global maximum doses in the test patients were identical between the TrueBeam and Halcyon plans (108% of Rx) and a mean PTV dose of 102.5% vs 101.6%, respectively. CONCLUSIONS: From this study a practical and efficient planning method for delivering 3D conformal breast radiotherapy using the Halcyon linear accelerator has been developed. When normalized to the clinically desired coverage, hot spots were maintained to acceptable levels and overall plan quality was comparable to plans delivered on conventional C-arm LINACs.

Radiation Therapy Workflow and Dosimetric Analysis from a Phase 1/2 Trial of Noninvasive Cardiac Radioablation for Ventricular Tachycardia.

PURPOSE: A prospective phase 1/2 trial for electrophysiologic guided noninvasive cardiac radioablation treatment of ventricular tachycardia (ENCORE-VT) demonstrating efficacy for arrhythmia control has recently been reported. The treatment workflow, report dose-volume metrics, and overall process improvements are described here. METHODS AND MATERIALS: Patients receiving 25 Gy in a single fraction to the cardiac ventricular tachycardia substrate (identified on presimulation multimodality imaging) on the phase 1/2 trial were included for analysis. Planning target volume (PTV), R50, monitor unit ratio, and gradient measure values were compared over time using statistical process control. Outlier values in the dose-volume histogram (DVH) for PTV and organs at risk were identified by calculating inner fences based on the interquartile range. Median heart substructure doses are also reported. RESULTS: For the 16 trial patients included, the median target volumes for the gross "target" volumes, internal target volumes, and PTVs were 25.1 cm3 (minimum: 11.5 cm3, maximum: 54.9 cm3), 30.1 cm3 (17.7, 81.6), and 97.9 cm3 (66, 208.5), respectively. On statistical process control analysis, there was a significant decrease in PTV volume among the more recent cohort of cases and mean doses to the nontargeted heart (heart - PTV). Two patients had heart-minus-PTV, esophagus, and stomach DVH data significantly higher than inner fence, and 3 patients had spinal cord DVH data higher than the inner fence, but in all cases the deviations were clinically acceptable. Subjective decreases were seen in the R50, gradient measure, and treatment time from the first to last patient in this series. All plans were verified in phantom with ionization chamber measurements within 2.9% of the expected dose value. CONCLUSIONS: Over the duration of this trial, PTV volumes to the cardiac substrate target decreased significantly, and organ-at-risk constraints were met for all cases. Future directions for this clinical process will include incorporating knowledge-based planning techniques and evaluating the need for substructure optimization.

In Silico Trial of MR-Guided Midtreatment Adaptive Planning for Hypofractionated Stereotactic Radiation Therapy in Centrally Located Thoracic Tumors.

PURPOSE: Hypofractionated (>5 fraction) stereotactic radiation therapy (HSRT) may allow for ablative biologically equivalent dose to tumors with a lower risk of organ-at-risk (OAR) toxicity in central thoracic tumors. Adaptive planning may further improve OAR sparing while maintaining planning target volume (PTV) coverage. We hypothesized that midtreatment adaptive replanning would offer dosimetric advantages during HSRT for central thorax malignancies using magnetic resonance imaging (MRI)-guided radiation therapy. METHODS AND MATERIALS: Twelve patients with central thorax tumors received HSRT using MRI-guided radiation therapy. Clinically delivered regimens were 60 Gy in 12 fractions or 62.5 Gy in 10 fractions, with low-field magnetic resonance (0.35 T) volumetric setup imaging acquired at each fraction. Daily gross tumor volume (GTV) and OARs were retrospectively redefined on fraction 1, 6, and 10 MRIs, and GTV response was recorded. Simulated initial plans prescribed a dose of 60 Gy in 12 fractions based on fraction 1 MRI. Midtreatment adaptive plans were created based on fraction 6 anatomy-of-the-day. All plans were created using an isotoxicity approach with a goal of 95% PTV coverage, subject to hard OAR constraints, to represent clinically ideal OAR sparing. Plans were then compared for projected OAR sparing and PTV coverage. RESULTS: Patients demonstrated significant on-treatment MRI-defined GTV reduction (median 41.8%; range 16.7%-65.7%). At fraction 6, median reduction was 26.7%. All initial plans met OAR constraints. Initial plan application to fraction 6 and fraction 10 anatomy resulted in 8 OAR violations (5 of 13 patients) and 10 OAR violations (6 of 13 patients). All fraction 6 violations persisted at fraction 10. Midpoint adaptive planning reversed 100% of midpoint OAR violations and tended to reduce the magnitude of OAR violations incurred at fraction 10. In 40% of fractions (2 of 5) in which OAR violation resulted from initial plan application to fraction 6 anatomy, PTV coverage was increased concomitant with violation reversal. CONCLUSIONS: Midtreatment adaptive planning based on tumor response may be dosimetrically advantageous for sparing of surrounding critical structures in HSRT for central thorax malignancies and could be applied using either an online or offline paradigm.