Twenty Years of Advancements in a Radiotherapy Facility: Clinical Protocols, Technology, and Management.

BACKGROUND: Hypo-fractionation can be an effective strategy to lower costs and save time, increasing patient access to advanced radiation therapy. To demonstrate this potential in practice within the context of temporal evolution, a twenty-year analysis of a representative radiation therapy facility from 2003 to 2022 was conducted. This analysis utilized comprehensive data to quantitatively evaluate the connections between advanced clinical protocols and technological improvements. The findings provide valuable insights to the management team, helping them ensure the delivery of high-quality treatments in a sustainable manner. METHODS: Several parameters related to treatment technique, patient positioning, dose prescription, fractionation, equipment technology content, machine workload and throughput, therapy times and patients access counts were extracted from departmental database and analyzed on a yearly basis by means of linear regression. RESULTS: Patients increased by 121 ± 6 new per year (NPY). Since 2010, the incidence of hypo-fractionation protocols grew thanks to increasing Linac technology. In seven years, both the average number of fractions and daily machine workload decreased by -0.84 ± 0.12 fractions/year and -1.61 ± 0.35 patients/year, respectively. The implementation of advanced dose delivery techniques, image guidance and high dose rate beams for high fraction doses, currently systematically used, has increased the complexity and reduced daily treatment throughput since 2010 from 40 to 32 patients per 8 h work shift (WS8). Thanks to hypo-fractionation, such an efficiency drop did not affect NPY, estimating 693 ± 28 NPY/WS8, regardless of the evaluation time. Each newly installed machine was shown to add 540 NPY, while absorbing 0.78 ± 0.04 WS8. The COVID-19 pandemic brought an overall reduction of 3.7% of patients and a reduction of 0.8 fractions/patient, to mitigate patient crowding in the department. CONCLUSIONS: The evolution of therapy protocols towards hypo-fractionation was supported by the use of proper technology. The characteristics of this process were quantified considering time progression and organizational aspects. This strategy optimized resources while enabling broader access to advanced radiation therapy. To truly value the benefit of hypo-fractionation, a reimbursement policy should focus on the patient rather than individual treatment fractionation.

Multicentric evaluation of a machine learning model to streamline the radiotherapy patient specific quality assurance process.

PURPOSE: Patient-specific quality assurance (PSQA) is performed to ensure that modulated treatment plans can be delivered as intended, but constitutes a substantial workload that could slow down the radiotherapy process and delay the start of clinical treatments. In this study, we investigated a machine learning (ML) tree-based ensemble model to predict the gamma passing rate (GPR) for volumetric modulated arc therapy (VMAT) plans. MATERIALS AND METHODS: 5622 VMAT plans from multiple treatment sites were selected from a database of Institution 1 and the ML model trained using 19 metrics. PSQA analyses were performed automatically using criteria 3%/1 mm (global normalization, absolute dose, 10% threshold) and 95% action limit. Model's performance was evaluated on an out-of-sample test set of Institution 1 and on two independent sets of measurements collected at Institution 2 and Institution 3. Mean absolute error (MAE), as well as the model's sensitivity and specificity, were computed. RESULTS: The model obtained a MAE of 2.33%, 2.54% and 3.91% for the three Institutions, with a specificity of 0.90, 0.90 and 0.68, and a sensitivity of 0.61, 0.25, and 0.55, respectively. Small positive median values of the residuals (i.e., the difference between measurements and predictions) were observed for each Institution (0.95%, 1.66%, and 3.42%). Thus, the model's predictions were, on average, close to the real values and provided a conservative estimation of the GPR. CONCLUSIONS: ML models can be integrated into clinical practice to streamline the radiotherapy workflow, but they should be center-specific or thoroughly verified within centers before clinical use.

Knowledge-based DVH estimation and optimization for breast VMAT plans with and without avoidance sectors.

BACKGROUND: To analyze RapidPlan knowledge-based models for DVH estimation of organs at risk from breast cancer VMAT plans presenting arc sectors en-face to the breast with zero dose rate, feature imposed during the optimization phase (avoidance sectors AS). METHODS: CT datasets of twenty left breast patients in deep-inspiration breath-hold were selected. Two VMAT plans, PartArc and AvoidArc, were manually generated with double arcs from ~ 300 to ~ 160°, with the second having an AS en-face to the breast to avoid contralateral breast and lung direct irradiation. Two RapidPlan models were generated from the two plan sets. The two models were evaluated in a closed loop to assess the model performance on plans where the AS were selected or not in the optimization. RESULTS: The PartArc plans model estimated DVHs comparable with the original plans. The AvoidArc plans model estimated a DVH pattern with two steps for the contralateral structures when the plan does not contain the AS selected in the optimization phase. This feature produced mean doses of the contralateral breast, averaged over all patients, of 0.4 ± 0.1 Gy, 0.6 ± 0.2 Gy, and 1.1 ± 0.2 Gy for the AvoidArc plan, AvoidArc model estimation, RapidPlan generated plan, respectively. The same figures for the contralateral lung were 0.3 ± 0.1 Gy, 1.6 ± 0.6 Gy, and 1.2 ± 0.5 Gy. The reason was found in the possible incorrect information extracted from the model training plans due to the lack of knowledge about the AS. Conversely, in the case of plans with AS set in the optimization generated with the same AvoidArc model, the estimated and resulting DVHs were comparable. Whenever the AvoidArc model was used to generate DVH estimation for a plan with AS, while the optimization was made on the plan without the AS, the optimizer evidentiated the limitation of a minimum dose rate of 0.2 MU/°, resulting in an increased dose to the contralateral structures respect to the estimation. CONCLUSIONS: The RapidPlan models for breast planning with VMAT can properly estimate organ at risk DVH. Attention has to be paid to the plan selection and usage for model training in the presence of avoidance sectors.