Purpose: During the treatment for cervical cancer, target volumes vary due to the tumour’s response and changes in bladder and rectal filling. Online adaptive radiotherapy, oART, uses a CBCT image of the patient before each fraction to calculate a new treatment plan based on the patient’s anatomy-of-the day. The goal is to determine dosimetric benefits of oART for cervical cancer patients and which patients benefit the most from this technique.

 

Methods: Ten patients with locally advanced cervical cancer, previously treated at our institution, were identified. Three CBCTs were used to simulate an adaptive session on EthosTM (Varian Medical Systems, California, USA) for each patient. We compared EthosTM reference plans against the treated Eclipse plan, as well as EthosTM scheduled plans vs adapted plans.

 

Results: When comparing the EthosTM reference plan to the treated VMAT plan, the IMRT plans showed improved target coverage and a decrease in maximum dose delivered to the OARs. For PTV V95% and V93%, the dose goals for the scheduled plan did not satisfy our institutional requirements for 16/30 fractions and 22/30 fractions, respectively. OARs showed a statistically significant decrease in dose for the adaptive plan. The patients that showed varying bladder and rectum volumes benefited the most from adaptive radiotherapy.

 

Conclusion: In simulation, adaptive radiotherapy provided treatment plans that had superior coverage of the target volumes while limiting the dose to the OARs when compared to non-adaptive radiotherapy. More work will be done to determine, prior to treatment, which patients would benefit the most from oART.

Purpose: To investigate advanced TG-186 model-based dose calculations using a clinically implemented Monte Carlo algorithm in the largest multi-institution cohort of patients treated with permanent implant prostate brachytherapy (PIPB).

 

Methods: The analysis included 1500 patients who received I-125 PIPB across four institutions. Dose calculations were performed using eb_gui (egs_brachy) software under TG-186 (full tissue, with interseed attenuation (ISA)) and TG-43 (all water, no ISA) assumptions. Dose-volume metrics for the target and organs at risk were compared between the two approaches. Additionally, the effect of calcification volume on dose calculations was examined. Statistical analysis was performed for the entire cohort and by institution.

 

Results and Discussions: TG-186 patient-specific dose calculations generally show lower dose coverage for the prostate, rectum, bladder, and urethra compared to TG-43, with differences of up to 17%. This discrepancy is attributed to TG-43's failure to account for tissue heterogeneities and interseed effects. Larger dosimetric discrepancies were observed in the 540 patients with intraprostatic calcifications. TG-43 overestimated the dose in the prostate by up to 26% and 6% for patients with and without calcifications, respectively.

 

Conclusions: This study demonstrates that TG-43 dose evaluations result in notable inaccuracies compared to patient-specific TG-186 calculations. The presence of calcifications is crucial for accurate dose distribution, highlighting the need for revised target prescription and organs-at-risk tolerance levels.

Purpose: Tattoos are used in breast radiotherapy (RT) to position patients during treatment, but these permanent skin markings can have lasting emotional impacts on cancer survivors. This study investigates the operational impact and cost of surface guided positioning compared to the current standard of tattoo-based setup.

 

Methods: Setup time and positional accuracy prior to radiographic localization are compared between a linac equipped with AlignRT surface guidance technology and one without for patients receiving RT for breast cancer. An unequal variance one-tailed t-test is used to determine statistically significant differences in setup time and positional shifts. A cost analysis comparing the two positioning modalities is presented.

 

Results: The two positioning modalities were equivalent for IMRT, but AlignRT was faster for VMAT by an average of 62 seconds. Breathhold setup times were equivalent for both positioning modalities, but for free breathing setups, AlignRT was faster by an average of 20 seconds. Excluding first fractions did not impact the results. There were statistically significant differences in the magnitude of shifts in the lateral and rotational directions, but not in the vertical and longitudinal directions. On average, setup with AlignRT resulted in smaller shifts by 0.2 cm laterally and 0.16° rotationally. The item and staffing costs of both modalities yielded near equivalent costs over a 10-year lifespan of an AlignRT tool, with tattooing costs including a conservative estimate of tattoo removal.

 

Conclusions: Surface guided patient positioning for breast radiotherapy is non-inferior to permanent tattoos in terms of setup time, accuracy, and cost.

Purpose: Since early 2023, our institution has integrated AI automation into routine clinical ultrasound-guided prostate brachytherapy. This study evaluates the clinical impact, efficiency gains, and limitations of AI automation over this two-and-a-half-year period.

 

Methods: We retrospectively analyzed 120 clinical cases treated using AI automation. Key metrics included contouring accuracy (% of contoured imaging slices retained) for prostate, rectum, and urethra; catheter reconstruction accuracy; and time savings in contouring and catheter reconstruction steps.

 

Results: AI-generated contours for rectum and urethra were used without modifications in 69.2% and 76.9% of cases, respectively. AI-generated prostate contours were almost always extensively modified before planning, with only 7.7% of cases having more than 50% of AI-contoured slices retained. AI automation reduced contouring time by 29% on average (6.6min), with a 41% reduction (9.3min) observed when more than 2/3 of contoured slices were used (21% of cases). Catheter reconstruction was fully adequate in 21% of cases, with 1 or 2 catheter modifications required in 58% of cases. AI automation led to 40% reduction in mean catheter reconstruction time (14.1min), and 50% reduction (17.4 minutes) when >90% of catheters were correctly reconstructed (41% of cases).

 

Conclusion: AI automation in prostate HDR brachytherapy has demonstrated significant time savings and workflow efficiency improvements at our center. While AI-generated contours and reconstructions often require modifications, the technology has proven to be valuable in reducing manual workload and enhancing clinical throughput. Further refinements in AI algorithms are warranted to improve accuracy and reduce the need for manual adjustments.

Purpose: Creating a safe and effective radiation treatment plan is a time-consuming process. The purpose of this study is to implement a fully automated treatment planning pipeline for lung stereotactic ablative radiotherapy (SABR) delivered via volumetric modulated arc therapy.

 

Methods: Treatment plans of patients who received single target lung SABR at a tertiary institution were collected and divided into 80% for training and 20% for testing. Three common dose prediction models (a Unet, a generative adversarial network, and a hierarchically densely connected Unet) were trained for dose prediction. After training, the three models were used to generate dose distributions for the testing set plans. The predicted doses were used as optimization targets in inverse planning, carried out in a custom version of the open source software matRad.

 

Results: In total, 244 treatment plans were collected. The predicted dose distributions of all three models were accurate, with average mean absolute differences (MAD) of <1Gy, and gamma pass fractions (GPF) of >90% across all organs at risk (OARs), with the HDUnet performing significantly better for some OARs. After inverse planning, the optimized doses from all three models were similar with MAD and GPF remaining <1Gy and >90% across all OARs. 90% of both predicted and optimized doses passed every hard constraint except the R100.

 

Conclusions: A two-stage approach of dose prediction followed by inverse optimization is a feasible approach for fully automated lung SABR treatment planning. Further improving the quality of the dose prediction will lead to higher quality optimized plans.

Purpose: Tilted multi-leaf collimators (MLC) exhibit different transmission properties than stepped tongue and groove MLC. The purpose of this research is to provide the community with optimized RayStation beam modeling parameters for various beam qualities on the Elekta Agility MLC.

 

Methods: Saez & Hernandez (2017) proposed using a farmer chamber and water phantom to measure transmission from synchronous and asynchronous sliding gap dynamic apertures. These measurements are used in the RayPhysics beam modelling software to generate transmission modeling parameters. This methodology was followed for four beam qualities (6MVFFF, 6MV, 10MVFFF, 10MV) on 10 beam-matched Elekta linear accelerators with Agility MLC. An additional methodology was created to automatically choose the optimized parameters based on measurements. Beam models with optimized MLC transmission parameters were used to assess delivery accuracy with patient specific quality assurance (PSQA) measurements.

 

Results: Identical fields delivered on all 10 linear accelerators agreed within 0.5%. Optimizing transmission parameters for flat field beam qualities found the parameters published by Hussein et al. (2024) were appropriate within less than 5% of the measured transmission. However, the FFF beam qualities both required 10% reductions in the published Tongue & Groove parameter to agree with measurement within 5%. PSQA measurements with film, ion chambers and diode arrays all passed within current clinical tolerances (3%/2mm).

 

Conclusion: This research validated beam model parameters for commonly used beam qualities on 10 beam-matched Elekta linear accelerators. Tilted MLC transmission measurements are influenced by FFF beam qualities and should be separately characterized in the beam modelling process.

Purpose: Proton therapy (PT) has demonstrated greater precision compared to photon radiotherapy for various sites; however, PT systems are expensive, which limits global availability. The dielectric wall accelerator (DWA) is proposed as a low-cost, compact PT system. In this work, we compare several clinical characteristics of our DWA model with conventional PT.

 

Methods: A linear beam optics model of the DWA was developed using TRANSOPTR. Proton bunches (20 to 230 MeV) were simulated through a DWA beamline and nozzle components in TOPAS. For comparison, an IBA proton beam was simulated in TOPAS, with and without range shifters. Spot sizes, surface doses, and energy ranges for twelve pediatric central nervous system (CNS) patients in RayStation were recorded.

 

Results: Compared to IBA, DWA protons exhibited a 67.0% average reduction in lateral spot size across 7–21 cm ranges (100–175 MeV) and a 76.8% reduction for ranges below 7 cm, averaged over both range shifters. The longitudinal spot size was reduced by 21.5%. Surface doses from DWA increased from 14.5% to 49.7% between 100 and 20 MeV, though they remained lower than those from IBA. This is beneficial, as in RayStation, energies between 44 and 151 MeV were selected during optimization for twelve pediatric CNS patients.

 

Conclusion: DWA technology may lead to improved spot sizes and wider energy ranges compared to conventional PT, providing opportunity for more conformal treatment plans.