Purpose: This study investigates the feasibility of personalizing radiotherapy prescription schemes (treatment margins and fractional doses) for glioblastoma (GBM) patients and their potential benefits using a proliferation and invasion (PI) glioma model on phantoms. Methods and Materials: We propose a strategy to personalize radiotherapy prescription schemes by simulating the proliferation and invasion of the tumor in 2D according to the PI glioma model. We demonstrate the strategy and its potential benefits by presenting virtual cases, where the standard and personalized prescriptions were applied to the tumor. Standard prescription was assumed to deliver 46 Gy in 23 fractions to the initial, gross tumor volume (GTV1) plus a 2 cm margin and an additional 14 Gy in 7 fractions to the boost GTV2 plus a 2 cm margin. The virtual cases include the tumors with a moving velocity of 0.029 (slow-move), 0.079 (average-move), and 0.13 (fast-move) mm/day for the gross tumor volume (GTV) with a radius of 1 (small) and 2 (large) cm. For each tumor size and velocity, the margin around GTV1 and GTV2 was varied between 0–6 cm and 1–3 cm, respectively. Equivalent uniform dose (EUD) to normal brain was constrained to the EUD value obtained by using the standard prescription. Various linear dose policies, where the fractional dose is linearly decreasing, constant, or increasing, were investigated to estimate the temporal effect of the radiation dose on tumor cell-kills. The goal was to find the combination of margins for GTV1 and GTV2 and a linear dose policy, which minimize the tumor cell-surviving fraction (SF) under a normal tissue constraint. The efficacy of a personalized prescription was evaluated by tumor EUD and the estimated survival time. Results: The personalized prescription for the slow-move tumors was to use 3.0–3.5 cm margins for GTV1, and a 1.5 cm margin for GTV2. For the average- and fast-move tumors, it was optimal to use a 6.0 cm margin for GTV1 and then 1.5–3.0 cm margins for GTV2, suggesting a course of whole brain therapy followed by a boost to a smaller volume. It was more effective to deliver the boost sequentially using a linearly decreasing fractional dose for all tumors. Personalized prescriptions led to surviving fractions of 0.001–0.465% compared to the standard prescription, and increased the tumor EUDs by 25.3–49.3% and estimated survival times by 7.6–22.2 months. Conclusions: Personalizing treatment margins based on the measured proliferative capacity of GBM tumor cells can potentially lead to significant improvements in tumor cell kill and related clinical outcomes.
Purpose To evaluate the impact of a digital whiteboard system integrated with data from the oncology information system (OIS) on the urgency of physics quality assurance (QA) tasks in the radiation oncology department. Methods Quality check list (QCL) items in the Mosaiq OIS corresponding to eight discrete, sequential steps in the treatment planning process were created. A whiteboard to graphically display active QCLs automatically and in real time was implemented in March 2020 using R shiny. QCL data with completion status were collected in two 12‐month time periods before and after whiteboard implementation: January 2019–December 2019 and July 2020–June 2021. For all plans requiring patient‐specific QA, we recorded when each plan was available for physics QA and which treatments started the following day. We further classified those plans into four categories (urgency levels 1–4 with 4 being the most urgent) depending on how much time was available to perform QA. We compared the proportion of these next‐day QAs in each category between time periods accounting for plan type, day of the week, and time of year. Results Overall QA numbers were similar between time periods with 797 and 765 QAs total. The total proportion of next‐day QA decreased by 27% and the proportions of urgency levels 1 and 4 both showed significant decreases after whiteboard implementation of 29.2% and 54.9%, respectively (p<0.05$p<0.05$). All plan types had reduced proportions of next‐day QAs, especially nonstereotactic body radiation therapy (non‐SBRT) (30.3% decrease, p<0.05$p < 0.05$). Fridays and the months of October–December had the highest proportion of next‐day QAs but showed significant reductions of 19.1% and 40.6% in the proportion of next‐day QAs, respectively (p<0.05$p<0.05$). Conclusions The integrated whiteboard system significantly reduced the proportion of last‐minute physics work, increasing patient safety. Advantages of the integrated whiteboard were low cost, low overhead with automatic interface to the OIS, and concurrent user support.
Fluence map optimization for intensity-modulated radiation therapy planning can be formulated as a large-scale inverse problem with competing objectives and constraints associated with the tumors and organs at risk. Unfortunately, clinically relevant dose–volume constraints are nonconvex, so standard algorithms for convex problems cannot be directly applied. Although prior work focuses on convex approximations for these constraints, we propose a novel relaxation approach to handle nonconvex dose–volume constraints. We develop efficient, provably convergent algorithms based on partial minimization, and show how to adapt them to handle maximum-dose constraints and infeasible problems. We demonstrate our approach using the CORT data set and show that it is easily adaptable to radiation treatment planning with dose–volume constraints for multiple tumors and organs at risk. Summary of Contribution: This paper proposes a novel approach to deal with dose–volume constraints in radiation treatment planning optimization, which is inherently nonconvex, mixed-integer programming. The authors tackle this NP-hard problem using auxiliary variables and continuous optimization while preserving the problem’s nonconvexity. Algorithms to efficiently solve the nonconvex optimization problem presented in this paper yield computation speeds suitable for a busy clinical setting.
Background: Clinical medical physics duties include routine tasks, special procedures, and development projects. It can be challenging to distribute the effort equitably across all team members, especially in large clinics or systems where physicists cover multiple sites. The purpose of this work is to study an equitable workload distribution system in radiotherapy physics that addresses the complex and dynamic nature of effort assignment. Methods: We formed a working group that defined all relevant clinical tasks and estimated the total time spent per task. Estimates used data from the oncology information system, a survey of physicists, and group consensus. We introduced a quantitative workload unit, "equivalent workday" (eWD), as a common unit for effort. The sum of all eWD values adjusted for each physicist's clinical full-time equivalent yields a "normalized total effort" (nTE) metric for each physicist, that is, the fraction of the total effort assigned to that physicist. We implemented this system in clinical operation. During a trial period of 9 months, we made adjustments to include tasks previously unaccounted for and refined the system. The workload distribution of eight physicists over 12 months was compared before and after implementation of the nTE system. Results: Prior to implementation, differences in workload of up to 50% existed between individual physicists (nTE range of 10.0%-15.0%). During the trial period, additional categories were added to account for leave and clinical projects that had previously been assigned informally. In the 1-year period after implementation, the individual workload differences were within 5% (nTE range of 12.3%-12.8%). Conclusion: We developed a system to equitably distribute workload and demonstrated improvements in the equity of workload. A quantitative approach to workload distribution improves both transparency and accountability. While the system was motivated by the complexities within an academic medical center, it may be generally applicable for other clinics.
Purpose: Patient-specific quality assurance (PSQA) failures in radiotherapy can cause a delay in patient care and increase the workload and stress of staff. We developed a tabular transformer model based directly on the multi-leaf collimator (MLC) leaf positions (without any feature engineering) to predict IMRT PSQA failure in advance. This neural model provides an end-to-end differentiable map from MLC leaf positions to the probability of PSQA plan failure, which could be useful for regularizing gradient-based leaf sequencing optimization algorithms and generating a plan that is more likely to pass PSQA.
Method: We retrospectively collected DICOM RT PLAN files of 968 patient plans treated with volumetric arc therapy. We constructed a beam-level tabular dataset with 1873 beams as samples and MLC leaf positions as features. We trained an attention-based neural network FT-Transformer to predict the ArcCheck-based PSQA gamma pass rates. In addition to the regression task, we evaluated the model in the binary classification context predicting the pass or fail of PSQA. The performance was compared to the results of the two leading tree ensemble methods (CatBoost and XGBoost) and a non-learned method based on mean-MLC-gap.
Results: The FT-Transformer model achieves 1.44% Mean Absolute Error (MAE) in the regression task of the gamma pass rate prediction and performs on par with XGBoost (1.53% MAE) and CatBoost (1.40% MAE). In the binary classification task of PSQA failure prediction, FT-Transformer achieves 0.85 ROC AUC (compared to the mean-MLC-gap complexity metric achieving 0.72 ROC AUC). Moreover, FT-Transformer, CatBoost, and XGBoost all achieve 80% true positive rate while keeping the false positive rate under 20%.
Conclusions: We demonstrated that reliable PSQA failure predictors can be successfully developed based solely on MLC leaf positions. FT-Transformer offers an unprecedented benefit of providing an end-to-end differentiable map from MLC leaf positions to the probability of PSQA failure.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.