Patient-specific quality assurance (PSQA) for VMAT is resource-intensive, and machine learning models have been proposed to streamline the process. However, previous studies trained general models across multiple disease sites, despite plan complexity being site dependent. This study investigates the potential of improving outcome prediction using site-specific models, to support the reduction of PSQA measurements and confidence in virtual PSQA. A quantile random forest (QRF) model was trained on 15,000 PSQA plan measurements to predict gamma passing rates (GPR), alongside site-specific models for gastrointestinal (GI), genitourinary (GU), head and neck (HN), lung, and prostate cases... The general model had 15 false positives (0.5%) at 95% threshold, whereas GU and prostate had none, and lung had only one (0.2%). In contrast, HN had the highest false positive rate (0.9%, 4 plans). Additionally, at a 93% threshold, 40% of the general model’s 17 false positives were HN plans. Despite improved specificity in most site-specific models (GU and prostate achieving 100%), GI and HN had lower specificity (81.82% and 80.95%) than the general model (91.23%). Learning curves revealed site-depend... Conclusions: The general QRF model achieved a sensitivity and specificity comparable to previous literature, while the site-specific models show promise in reducing false positives and improving specificity. Future work will focus on clinical validation and implementation.

Purpose
To conduct the first comprehensive survey on the current state of patient-specific quality assurance (PSQA) practice in all cancer/radiotherapy centres across Canada.

Methods
The survey was created using Research Electronic Data Capture software, consisting of 30 questions on a variety of topics, covering policies, procedures, methodologies, and devices. The centres were instructed to submit responses that reflected the collective view/opinion of the department staff. Personal follow-ups were conducted to clarify unclear responses. Data was de-identified, and analysis was performed using PivotCharts in Microsoft Excel.

Results
The survey response rate was 90%, covering 45 centres across all provinces. Only three centres, all with ≤ 5 linacs, were fully compliant in three Canadian Partnership for Quality Radiotherapy (CPQR) recommended tests. Overall, 53% had reduced their PSQA measurement frequency, with a higher reduction in high-FTE centres (68%) than low-FTE centres (42%). The most likely factors for failed or suboptimal PSQA results were measurement related issues and beam model limitations. The vast majority did not perform PSQA every fraction. 62% had performed full or partial sensitivity/specificity characterization of their PSQA system, and about 20% had performed a risk-based analysis of their PSQA program. Nearly 90% expressed interest in a national guidance document for PSQA practice. The 3D array was the most widely used PSQA detector nationwide, followed by EPID, though usage varied regionally.

Conclusions
This study provides a comprehensive overview of PSQA practices in Canada, highlighting variations in regions, correlations with centre sizes and staffing resources, as well as deviations from CPQR guidelines.

Purpose: In this study, Failure Modes and Effects Analysis (FMEA) were utilized to assess the effectiveness of different patient-specific quality assurance (PSQA) techniques with the goal of developing recommendations on appropriate PSQA usage for the Province of Ontario.

Methods: A list of failure modes (FMs) were identified from a generic RT process (with a traditional QA program but no PSQA). Nine physicists from different cancer centres in Ontario scored the FMs. To quantify a single value for each parameter (Occurrence, Severity, and Detectability), three case scenarios were considered: 1) worst and 2) average scores, 3) minimum Occurrence, maximum Severity and average Detectability scores. Two different risk profiles were used Risk Priority Number (RPN) and Action Priority (AP).

Results: In an average case scenario, the proportion of FMs with occasional failures (O>4) showed to be similar for all categories (Linac/MLC, TPS/planning stage, Documentation and Data Transfer) while a larger proportion of FMs in the Documentation and Data Transfer categories (73%) showed undetectability higher than 2% (D>4). After FMEA, there were 9 FMs selected by at least 2 criteria (RPN, AP or S>8) in the case scenario 3. The 15 top priority FMs belong mostly to the TPS (10 FMs) compared to Linac (3) and Documentation (2) categories.

Conclusion: From the perspective of improving Detectability, PSQA should be geared towards catching Documentation and Data Transfer failures while from the perspective of perceived risk, any effort to reduce occurrence or increase detectability of top FMs will add value to the traditional QA program.

Purpose: Exposure to ionizing radiation poses one of the greatest threats to astronaut health on long duration stays on the International Space Station (ISS). The radiation environment in space can depend on a large variety of factors, therefore characterizing this radiation is an important step in assessing the risk involved in a given mission. The goal of this study is to use computational simulations to characterize this radiation by recreating the space radiation environment for a specific mission duration.

Methods: The SPace ENVironment Information System (SPENVIS) was used to generate orbital coordinates for the ISS, quantifying the radiation field present. These parameters were used in the Geant4 Radiation Analysis for Space (GRAS) tool within SPENVIS to conduct Monte Carlo simulations of radiation transport through a component of the ISS. The simulation consisted of a hollow aluminum cylinder and a simple rectangular phantom inside to represent an astronaut.

Results: The absorbed dose inside the phantom was calculated to be 63.3 ± 1.1 mSv for a 6-month mission. This agreed with a dose measured in a phantom flown on the ISS during a similar period of solar activity. Key particles were identified based on their contribution to the dose, and their fluxes were scored and recorded. In addition, the equivalent linear energy transfer (LET) of the beam was calculated through a weighted sum of these particles.

Conclusions: A simulation of the space radiation environment was successfully created and validated. Fluxes and the equivalent LET from key particles were recorded to be used in future radiobiological simulations.

Purpose: To develop a quality assurance (QA) system to test the patient specific radiation targeting accuracy of radiotherapy linear accelerators (linacs) for multi-target single isocentre stereotactic radiosurgery (SRS). Purpose: To develop a quality assurance (QA) system to test the patient specific radiation targeting accuracy of radiotherapy linear accelerators (linacs) for multi-target single isocentre stereotactic radiosurgery (SRS). 

Methods: The QA system consists of the device containing a 3D array of possible positions for radiopaque targets (BBs) and the associated software. Software determines the in-phantom BB positions that most closely correspond to the patient’s lesion locations. For a given target configuration, a dynamic conformal arc (DCA) plan was generated using couch/collimator/gantry angles corresponding to the patient’s treatment plan. Images were acquired continuously (25 frames per second) throughout DCA plan delivery using the linac’s on-board electronic portal imaging device (EPID). The software detects BB positions using a Hough circle detection algorithm and determines radiation field positioning using image thresholding. This information is used to assess the 3D radiation targeting accuracy and the accuracy of the radiation field aperture.

 

Results: In a 6X DCA plan with 6 BBs, the maximum and mean 2D differences in expected and detected BB positions are 2.00 mm and 1.11 mm, respectively. There are no clear trends as function of BB distance from isocentre. The maximum and mean radiation field area discrepancies are 4.46 cm2 and 2.73 cm2, respectively. Refinements to methods for calculating radiation field area discrepancy are under investigation. 

 

Conclusion: The proposed QA system defines plan-specific machine isocentricity for single-isocentre multi-target SRS according to a chosen tolerance level. The results of this QA test could contribute to the choice of PTV margin.

Purpose:Purpose: This work describes the implementation of an automated, quantitative cone beam computed tomography (CBCT) quality assurance (QA) program for linac and gamma knife mounted CBCT at our clinic.
Methods:The open-source Pylinac library was used as a starting point to establish a quantitative CBCT QA program at our clinic using the CatPhan®503, CatPhan®504, and CatPhan®604 phantoms. Several modifications and additions to the library were required to accurately localize of the phantom modules, accurate localization of the line-pairs within the high-resolution module, and automatic identification of the CatPhan® version. Historical data consisting of 182 CBCT acquisitions performed over 18 months were used to evaluate the performance of the automated method before clinical deployment and statistical process control was used to define tolerances.  

Results: Without modification, the Pylinac library failed to accurately localize the CatPhan® modules in 3% (n=5) cases and failed to localize the line-pairs of the high-resolution module in 7% (n=13) cases. Our changes resulted in all cases successfully localizing the phantom modules and line-pairs. Our custom routine to automatically identify the CatPhan® version (503, 504, or 604) was functional in all cases and facilitated the use of a single automated pipeline for CBCT QA in our clinic. The automated method is now used clinically at our center and has replaced operator-dependent subjective evaluation of image quality with algorithmic methods. 

Conclusions: This work builds upon the open-source Pylinac library, improving accuracy and robustness, in the development of an automated QA program for linac and gamma knife mounted CBCT imaging.