International Journal of Radiation Oncology*Biology*Physics

AO_SCPLOWBSTRACTC_SCPLOWO_ST_ABSBackground and purposeC_ST_ABSMandibular osteoradionecrosis (ORN) is a severe side effect affecting patients undergoing radiation therapy for head and neck cancer. Variations in the bones vascularization and composition across the mandible may influence the susceptibility to ORN. Recently, deep learning-based models have been introduced for predicting mandibular ORN using radiation dose distribution maps to incorporate spatial information. These studies, however, only feature internal validation on a holdout subset of the data used for training. Materials and methodsThis study externally validated a 3D DenseNet-40 (DN40) ORN prediction model on an independent dataset. Model performance was evaluated in terms of discrimination and calibration, with Platt scaling applied for improved external calibration. The DN40 models discriminative ability on the external dataset was compared to a Random Forest model on corresponding dose-volume histogram (DVH) data. ResultsThe overall model performance was worse at external validation than at internal validation, with Platt scaling improving balance between recall and specificity but not significantly improving the overall calibration. Although the discrimination ability of the DN40 model was slightly lower at external validation (AUROC 0.63 vs. 0.69), this was statistically comparable to that of a DVH-based RF model for the same dataset (p-value 0.667). ConclusionsOur results suggest that, in addition to potential model overfitting issues, dosimetric data distribution differences between the two datasets could explain the low generalisability of the DN40 ORN prediction model. Future work will involve a larger and more diverse cohort.

Background and purposeStereotactic body radiotherapy (SBRT) has become a standard treatment option for localized prostate cancer, with low rates of clinically relevant late toxicity. However, the identification of robust dosimetric predictors of toxicity remains challenging due to the high dimensionality and collinearity of dose-volume histogram (DVH) metrics. This study aimed to explore whether principal component analysis (PCA) of DVHs can identify dose regions associated with late gastrointestinal and genitourinary toxicity after prostate SBRT. Materials and methodsWe analysed a single-institution cohort of patients treated with prostate SBRT. Rectum, rectal wall, bladder and bladder wall DVHs were extracted with a dose bin resolution of 0.5 Gy. PCA was applied separately to each structure to identify dominant patterns of dose-volume variability. PCA-derived dose metrics were subsequently evaluated using Spearman correlation analyses, receiver operating characteristic (ROC) curves, and exploratory logistic regression models. Late toxicity was scored according to CTCAE version 5.0, with grade [≥] 2 events at 12 months as the primary endpoint. ResultsPCA demonstrated that a limited number of components accounted for most DVH variability, with the largest contributions arising from intermediate-dose regions. For the whole rectum, intermediate-dose metrics showed the strongest association with late rectal toxicity. Rectal V18.1 Gy yielded the highest discriminative performance (AUC = 0.87), followed by V29 Gy (AUC = 0.83), whereas low-dose (V1.5 Gy) and high-dose (V42.5 Gy) metrics showed limited or no discrimination. Rectal wall metrics demonstrated weaker and less robust associations, and no clinically meaningful discriminative performance was observed for bladder or bladder wall DVH metrics. Exploratory regression analyses supported the association between intermediate rectal dose exposure and late rectal toxicity. ConclusionIn prostate SBRT, PCA of DVHs highlights intermediate rectal dose exposure as the primary dosimetric determinant of late rectal toxicity. Whole-rectum intermediate-dose metrics outperform both low- and high-dose parameters, as well as rectal wall and bladder-derived metrics. These findings support a parsimonious, data-driven focus on intermediate-dose rectal volumes for toxicity risk assessment and hypothesis generation in prostate SBRT planning.

BackgroundQuantitative mapping of the longitudinal relaxation rate (R1=1/T1) is a major building block for several multiparametric MRI protocols intended for adaptive radiation therapy planning. The implementation of these protocols is challenging in anatomical sites that experience large physiological motion. PurposeTo implement and validate a motion-resolved quantitative T1 mapping method on a 1.5T MR-Linac that combines non-Cartesian k-space sampling trajectories with compressed sensing (CS) reconstruction techniques. MethodsFour 3D non-Cartesian k-space trajectories were evaluated: radial and stack-of-stars sampling using half- and full-spoke coverage. A variable flip angle acquisition was performed using the spoiled gradient-echo sequence, and T1 mapping was validated using two standard phantoms. Gradient delay timing was optimized empirically to minimize trajectory-induced artifacts. Eight compressed sensing reconstruction strategies were tested using spatial and spatiotemporal regularization operators. Reconstructions were evaluated across multiple implementation parameters and ranked based on spatial resolution, bias, and variability. In vivo studies included one healthy volunteer and one patient undergoing radiotherapy to a target in the kidney. Motion-resolved imaging was performed using respiratory self-gating and phase-sorted reconstruction. ResultsAll non-Cartesian trajectories demonstrated high repeatability and low longitudinal bias in phantom studies, with coefficients of variation below 3.3%. Radial half-spoke sampling achieved the shortest scan times and highest agreement with Cartesian benchmarks. Reconstruction methods incorporating spatiotemporal regularization maintained spatial resolution and quantitative accuracy across undersampling factors up to 20-fold. In human subjects, non-Cartesian T1 mapping provided improved accuracy and reduced variability in mobile abdominal tissues compared to Cartesian acquisitions, particularly in the kidney cortex and medulla, where motion artifacts led to overestimation and higher variance in the reference method. ConclusionsT1 mapping using non-Cartesian trajectories and compressed sensing reconstruction is feasible on a 1.5T MR-Linac. The proposed approach enables accurate, motion-resolved quantitative imaging within clinically practical acquisition times. These results support integration of quantitative T1 mapping into adaptive MR-guided radiotherapy workflows and establish a foundation for future development of multiparametric imaging and response-adaptive treatment strategies.

PurposeEquivalent dose in 2 Gy fractions (EQD2), based on the original biological effective dose (BED) equation, is frequently used to guide treatment in the clinic. This work addresses the limitations of EQD2 in the context of voxelized dosimetry, clarifies potential sources of confusion, and provides an alternative formulation for improved precision. Methods and MaterialsThe EQD2 formula was evaluated by a simple insertion of the EQD2 dose into the BED equation. The mathematically exact form of EQD2, referred to here as equivalent physical dose (EPD), was provided by solving the linear-quadratic model BED equation for dose using the quadratic formula. The EPD derivation was compared in terms of absolute error to the EQD2 derivation, which separates the Relative Effect term from the BED equation. ResultsThe EQD2 expression implicitly assumed a homogenous dose, demonstrating that its use in voxelized dosimetry can mislead. As an alternative formulation, EPD was shown to adhere more closely to the first principles of radiobiological modeling. An error analysis identified absolute errors from EQD2 sometimes in excess of 10%. ConclusionsAssumptions in the standard EQD2 equation are inappropriate in the context of voxelized dosimetry, where voxels within a structure, such as a target volume, may receive a dose that differs from the prescribed dose. Using EPD (or BED) instead of EQD2 would address these areas of confusion. Optimizing therapy according to biological properties in this way could provide enhanced and more reliable radiobiological input to radiotherapy treatment planning.

PurposeThis study aims to evaluate the impact of varying Aperture Shape Controller (ASC)settings influence the optimization of VMAT plans for tongue carcinoma (Ca-Tongue), focusing on their role in modulating plan complexity, maintaining dosimetric integrity, and ensuring accurate treatment delivery. MethodsTwenty Ca-Tongue patients were retrospectively planned using four ASC settings: Off, Very Low, Moderate, and Very High, totaling 80 plans. Complexity metrics such as Modulation Complexity Score (MCSv), Small Aperture Score (SAS), and Monitor Units per cGy (MU/cGy) were computed using MATLAB from exported DICOM RT files. Each plan underwent portal dosimetry QA with gamma analysis (3%/3mm and 2%/2mm). Dosimetric quality was evaluated using Conformity Index (CI), Homogeneity Index (HI), and PTV D98%, along with doses to organs-at-risk (OARs). Statistical analysis included the Wilcoxon signed-rank test and linear regression. ResultsIncreasing ASC level significantly reduced plan complexity: MCSv increased from 0.32{+/-}0.02 (Off) to 0.38{+/-}0.03 (Very High), SAS decreased from 0.47{+/-}0.04 to 0.37{+/-}0.07, and MU/cGy dropped from 2.25{+/-}0.09 to 2.03{+/-}0.12 (p<0.05). However, higher ASC levels were associated with minor but consistent reductions in PTV coverage (D98%: 96.66% to 94.94%) and increases in OAR doses (e.g., spinal cord Dmax: 30.46Gy to 34.90Gy). CI and HI remained clinically acceptable across all settings. Gamma pass rates were uniformly high ([&ge;]98.85%), with no significant improvement across ASC levels. Weak or negligible correlations (R2 < 0.323) were found between complexity metrics and gamma outcomes. ConclusionThe ASC effectively reduces plan complexity in VMAT for Ca-Tongue without compromising delivery accuracy. While Very High ASC yields the greatest complexity reduction, it also introduces modest trade-offs in PTV coverage and OAR sparing. The Moderate ASC setting appears optimal, offering a balance between complexity control and dosimetric quality. Clinical implementation of ASC should be tailored to tumor site and anatomy, with Moderate ASC recommended for head and neck VMAT to ensure safety, efficiency, and robust QA performance.

In a phase III randomized trial, adding a radiation boost to visible tumor(s) on MRI improved prostate cancer disease-free and metastasis-free survival without additional toxicity. However, radiation oncologists ability to identify prostate tumors is critical and represents a major barrier to widely adopting intraprostatic tumor radiotherapy boost for patients. We previously developed a quantitative diffusion MRI biomarker for prostate cancer, called the Restriction Spectrum Imaging restriction score (RSIrs), that has been shown to improve radiologist identification of clinically significant prostate cancer. 42 radiation oncologists (participants) from multiple, international institutions contoured prostate tumors on 40 patient cases using standard MRI with or without RSIrs map, producing 1646 target volumes. Use of RSIrs maps significantly improved all evaluated accuracy metrics, including participants percent overlap with consensus expert target volume (73% vs. 42%, p<0.001). A mixed effects model confirmed that RSIrs maps were the main variable driving the improvement in all metrics. System Usability Scores indicated RSIrs maps significantly improved the contouring experience (72 vs. 58, p<0.002). The expert-defined tumor was completely missed 158 times on standard MRI alone and only 19 times with RSIrs maps. RSIrs maps improve the accuracy of target delineation for prostate tumor boost. Patient SummaryAdding an extra boost of radiation to tumor(s) visible on MRI has been shown to prevent cancer recurrence and cancer spread beyond the prostate without adding additional side effects; however, drawing the prostate tumor on MRI is difficult, and most radiation oncologists have not been trained to do this. We have developed an advanced MRI technique (RSIrs maps) that increases tumor visibility. We found that RSIrs maps improve radiation oncologists accuracy in targeting prostate tumors.

PurposeIn order to improve segmentation accuracy in head and neck cancer (HNC) radiotherapy treatment planning for the 1.5T MR-Linac, 3D fat-suppressed T2-weighted MRI sequences were developed and optimized. MethodsAfter initial testing of fat suppression techniques, SPectral Attenuated Inversion Recovery (SPAIR) was chosen as the fat suppression technique. Five candidate SPAIR sequences and a non-suppressed T2-weighted sequence were acquired on five HNC patients on the Unity MR-Linac. The primary tumor, metastatic lymph nodes, parotid glands, and pterygoid muscles were delineated by five segmentors. A robust image quality analysis platform was developed to objectively score the SPAIR sequences based on a combination of qualitative and quantitative metrics. ResultsSequences were analyzed for signal-to-noise (SNR), contrast-to-noise (CNR) compared to fat and muscle, conspicuity, pairwise distance metrics, segmentor assessment, and MR physicist assessment. From this analysis, the non-suppressed sequence was inferior to each of the SPAIR sequences for the primary tumor, lymph nodes, and parotid glands, but was superior for the pterygoid muscles. Two SPAIR sequences consistently received the highest scores among the analysis categories and are recommended for use to Unity MR-Linac users for HNC radiotherapy treatment planning. ConclusionsTwo deliverables resulted from this study. First, an optimized 3D fat-suppressed T2-weighted sequence was developed that can be disseminated to Unity MR-Linac users. Second, a robust image quality analysis process pathway, used to objectively score the various SPAIR sequences, was developed and can be customized and generalized to any image quality optimization. Improved segmentation accuracy with the proposed SPAIR sequence can potentially lead to improved treatment outcomes and reduced toxicity by maximizing target coverage and minimizing organ-at-risk exposure.

Stereotactic body radiotherapy (SBRT) has demonstrated clinical benefit for patients with metastatic and/or unresectable cancer. Technical considerations of treatment delivery and sensitive organs at risk (OARs) limit the use of SBRT in large tumors or those in unfavorable locations. Spatially fractionated radiotherapy (SFRT) delivers high-dose radiation to discrete sub-volume vertices within a tumor target while restricting the remainder of the target to low dose. SFRT has been utilized for treatment of large tumors with reported dramatic tumor response and minimal side effects. Lattice is a modern approach to SFRT that can be delivered with arc-based therapy, which allows for the rapid dose fall-off required for high quality SBRT. In order to overcome the limitations of SBRT for large tumors, we developed Lattice SBRT. Here we report the results of a dosimetry and quality assurance (QA) feasibility study of Lattice SBRT in 11 patients with 12 tumor targets, each [&ge;] 10 cm in an axial dimension. Prior CT simulation scans were used to generate volumetric-modulated arc therapy (VMAT) Lattice SBRT plans that were then delivered on clinically available Linacs. QA testing included external portal imaging device (EPID) and ion chamber (IC) analysis. All generated plans were able to meet the standard SBRT dose constraints, such as those from AAPM Task Group 101. Additionally, we provide a step-by-step approach for generating and delivering Lattice SBRT plans using commercially available treatment technology. Lattice SBRT is currently being tested in a prospective trial for patients with metastatic cancer needing palliation of large tumors (NCTXXXX).

PurposeDesign and evaluate a workflow using commercially available artificial intelligence tools for automated hippocampal segmentation and treatment planning to efficiently generate clinically acceptable hippocampal-avoidant whole brain (HA-WBRT) radiotherapy plans. Methods and MaterialsWe retrospectively identified 100 consecutive adult patients treated for brain metastases outside the hippocampal region. Each patients T1 post-contrast brain MRI was processed using FDA-approved software that provides segmentations of brain structures in 5-7 minutes. Automated hippocampal segmentations were reviewed for accuracy and edited manually if necessary, then converted to files compatible with a commercial treatment planning system, where hippocampal avoidance regions and planning target volumes (PTV) were generated. Other organs-at-risk (OARs) were previously contoured per clinical routine. A RapidPlan knowledge-based planning routine was applied for a prescription of 30 Gy in 10 fractions using volumetric modulated arc therapy (VMAT) delivery. Plans were evaluated based on NRG CC001 dose-volume objectives. ResultsOf the 100 cases, 99 (99%) had acceptable automated hippocampi segmentations without manual intervention. Knowledge-based planning was applied to all cases; the median processing time was 9 minutes 59 seconds (range 6:53 - 13:31). All plans met per-protocol dose-volume objectives for PTV per the NRG CC001 protocol. For comparison, only 66.0% of plans on NRG CC001 met PTV goals per protocol, with 26.3% within acceptable variation. In this study, 43 plans (43%) met OAR constraints, and the remaining 57 (57%) were within acceptable variation, compared to 42.9% and 48.6% on NRG CC001, respectively. No plans in this study had unacceptable dose to OARs, compared to 0.8% of manually generated plans from NRG CC001. ConclusionAn automated pipeline harnessing the efficiency of commercially available artificial intelligence tools can generate clinically acceptable VMAT HA-WBRT plans with minimal manual intervention. This process could improve clinical efficiency for a treatment established to improve patient outcomes over standard WBRT.

IntroductionProstate cancer is one of the most commonly diagnosed cancers among men worldwide. Advances in radiotherapy, particularly Stereotactic Body Radiotherapy (SBRT), have enabled ultra-hypofractionated treatment schedules that enhance tumor control while reducing treatment time. This study focuses on evaluating the dosimetric accuracy and plan quality of prostate SBRT using RapidArc technology on a Varian Millennium Multileaf Collimator (MLC) system. Materials and MethodsA total of 24 patients with localized prostate adenocarcinoma received SBRT with a prescribed dose of 36.25 Gy in five fractions. Treatment planning was performed using Eclipse v15.6 with Acuros XB algorithm, utilizing three 6 MV flattening filter-free (FFF) arcs. Planning Target Volume (PTV) coverage, OAR doses, Paddick Conformity Index (PCI), Gradient Index (GI), and Monitor Units (MU) were analyzed. Multileaf collimator motion was evaluated through log files, including leaf speed and position dynamics. Quality assurance was performed using electronic portal imaging devices (EPID) and gamma pass rate evaluation. ResultsThe mean PTV D95 was 35.80 {+/-} 0.46 Gy, with mean Dmax and Dmean being 39.80 {+/-} 1.05 Gy and 37.10 {+/-} 0.40 Gy, respectively. V95% averaged 99.13 {+/-} 1.14%, confirming adequate coverage. Slight violations of rectum and bladder Dmax constraints were observed but remained clinically acceptable. The mean PCI and GI were 0.905 {+/-} 0.18 and 3.08 {+/-} 0.16, respectively. Gamma pass rates exceeded 99.6% for 2%/2 mm criteria. MLC leaf speed remained below the 2.5 cm/s threshold, ensuring mechanical safety and dose delivery accuracy. ConclusionProstate SBRT delivered via RapidArc on a Varian Millennium MLC system demonstrated high plan conformity, efficient delivery, and acceptable OAR sparing. The integration of intra-fraction CBCT improved setup accuracy, while MLC analysis confirmed mechanical precision. This approach supports the clinical adoption of RapidArc-based SBRT as a feasible and effective option for localized prostate cancer treatment.

PurposeAccurate dose prediction is essential for automating radiotherapy planning. In spot scanning proton therapy (SSPT), dose evaluation is required at both the plan and field level. Evaluating individual treatment fields is critical to ensuring optimal beam angles are chosen to ensure target coverage and maximum organ-at-risk (OAR) sparing. Currently, however, no knowledge-based tools exist for predicting field-level doses for head and neck cancer (HNC) treated with SSPT. In this work, we aim to develop the first deep learning-based dose prediction model capable of field-level dose prediction for HNC treated with SSPT. MethodsA cohort of 62 HNC patients treated with SSPT was compiled for model development and evaluation. Collected patient data included treatment planning CTs, OAR masks, signed distance maps (SDMs), generated beam masks, and dose distributions. An encoder-decoder architecture enhanced with a cross-attention transformer bottleneck was used as the field prediction model. Comparison and ablation studies evaluated the models performance and determined the benefits of individual model components. Evaluation imaging metrics included mean absolute error, structural similarity index measure, and peak signal-to-noise ratio. Clinical performance was evaluated using dose-volume histogram metrics. ResultsThe best performing model from the ablation study was the full model using OAR masks, SDMs, generated beam masks and four-field dose prediction. The model outperformed the Distance Guided Dose Prediction (DGDP) and DeepLabV3 comparison models. The DGDP and DeepLabV3 comparison models had a mean validation set MAE performance of 1.268 Gy and 1.325 Gy, respectively, compared to our models mean validation set MAE performance of 0.949 Gy. The models final mean test set performance was MAE 1.024 Gy, SSIM 0.913, and PSNR 28.495 dB. ConclusionsWe developed a cross-attention transformer-enhanced deep learning model that accurately predicts per-field dose for HNC treated with SSPT, demonstrating superior performance over state-of-the-art models limited to plan-level dose prediction.

PurposeConformal dose distributions in proton radiotherapy promise to reduce normal tissue toxicity such as radiation-induced pneumonitis, but this has not been fully realized in clinical trials. To further investigate dose and toxicity, we employ voxel-based normal tissue evaluation techniques such as ventilation maps throughout treatment. We hypothesize that ventilation change after 1 week of treatment (WK1) predicts for ventilation change at the end of treatment (EOT). MethodsFor 48 photon and 23 proton lung cancer patients, 4DCT-based ventilation maps were generated using stress-based methods at planning, WK1, and EOT. Voxel-wise ventilation change from planning to WK1 and EOT was calculated and binned by planned dose, and median ventilation change at WK1 and EOT was calculated across all patients in each dose bin. Patients were stratified into 6 groups based on modality and increased, decreased, or stable ventilation at WK1. Mann-Whitney U tests were performed to determine if median ventilation change at WK1 and EOT in each dose bin was significantly different from zero. Univariate analysis was performed to correlate ventilation change at EOT with change at WK1 and other clinical factors. A linear regression model was developed to predict ventilation at EOT using a variety of input features including ventilation at planning, ventilation at WK1, tumor response information, and tumor location. Accuracy of the model was assessed through R2. ResultsFor patients that decreased in ventilation at WK1, 90% of photon patients and 92% of proton patients were stratified similarly at EOT. Patients that were stratified as increased ventilation at WK1 were stratified similarly (72% for photon, 80% for proton) at EOT. These patients were more likely to develop Grade 2+ pneumonitis though the difference was not significant when computing a Fishers exact test. Univariate analysis indicated that only ventilation change at WK1 was correlated with ventilation change at EOT. The linear regression model achieved R2 of 0.65. ConclusionVentilation changes at EOT can be predicted using ventilation information from planning and WK1. Patients that increased in ventilation at WK1 were more likely to develop pneumonitis. Further work is needed to characterize the relationship between ventilation change with pneumonitis development.

BackgroundAccurate delineation of orodental structures on radiotherapy CT images is essential for dosimetric assessments and dental decisions. We propose a deep-learning auto-segmentation framework for individual teeth and mandible/maxilla sub-volumes aligned with the ClinRad ORN staging system. MethodsMandible and maxilla sub-volumes were manually defined, differentiating between alveolar and basal regions, and teeth were labelled individually. For each task, a DL segmentation model was independently trained. A Swin UNETR-based model was used for the mandible sub-volumes. For the smaller structures (e.g., teeth and maxilla sub-volumes) a two-stage segmentation model first used the ResUNet to segment the entire teeth and maxilla regions as a single ROI that was then used to crop the image input of the Swin UNETR. In addition to segmentation accuracy and geometric precision, a dosimetric comparison was made between manual and model-predicted segmentations. ResultsSegmentation performance varied across sub-volumes - mean Dice values of 0.85 (mandible basal), 0.82 (mandible alveolar), 0.78 (maxilla alveolar), 0.80 (upper central teeth), 0.69 (upper premolars), 0.76 (upper molars), 0.76 (lower central teeth), 0.70 (lower premolars), 0.71 (lower molars) - and exhibited limited applicability in segmenting teeth and sub-volumes often absent in the data. Only the maxilla alveolar central sub-volume showed a statistically significant dosimetric difference (Bonferroni-adjusted p-value = 0.02). ConclusionWe present a novel DL-based auto-segmentation framework of orodental structures, enabling spatial localization of dose-related differences in the jaw. This tool enhances image-based bone injury detection, including ORN, and improves clinical decision-making in radiation oncology and dental care for head and neck cancer patients.

Accurate delineation of orodental structures on computed tomography (CT) is critical for image-guided assessments of radiation-associated bone injury. This dataset comprises curated CT imaging and expert-defined segmentation masks for 60 patients with head and neck cancer treated with radiotherapy (RT), including delineations of mandibular and maxillary sub-volumes and individual teeth. Segmentation guidelines were informed by anatomical differences across sub-regions and aligned with the ClinRad osteoradionecrosis (ORN) staging system. The dataset includes converted NIfTI files of simulation CT images, RT dose distributions, and delineated structures. All segmentations were performed manually using a standardized protocol in a commercial treatment planning system and converted to research-ready formats using open-source tools. This dataset may facilitate the development and validation of automated segmentation tools, dose mapping applications, and image-based ORN detection pipelines in head and neck cancer survivors.

PurposeRadiation therapy planning for locally-advanced non-small cell lung cancer (NSCLC) is challenging due to the balancing of target coverage and organs-at-risk (OAR) sparing. Using the Varian Ethos Treatment Planning System (TPS), we developed a methodology to automatically generate efficient, high-quality treatment plans for locally-advanced lung cancer patients. Methods and MaterialsFifty patients previously treated with Eclipse-generated plans for inoperable Stage IIIA-IIIC NSCLC were included in this Institutional Review Board (IRB)-approved retrospective study. Fifteen patients were used to iteratively optimize an Ethos TPS planning template, and the remaining thirty-five patients had plans automatically generated without manual intervention using the created template. Ethos and Eclipse plan quality was then assessed using 1) standard dose volume histogram (DVH) metrics, 2) adherence to clinical trial objectives, and 3) radiation oncologist qualitative review. ResultsEthos-generated plans showed improved primary and nodal planning target volume (PTVp and PTVn, respectively) V100% and V95% coverage (p<0.001) and reduced PTVp Dmax values (p=0.023). Furthermore, the Ethos template-generated plans had lower spinal cord Dmax, lungs V5Gy, and heart V25Gy, V30Gy, and V45Gy values (p[&le;]0.021). However, Ethos esophagus metrics (mean, V35Gy, V50Gy, Dmax) and brachial plexus metrics (Dmax) were greater than Eclipse (p[&le;]0.008), but were still clinically acceptable. A large majority (80%) of automatically generated plans had entirely "per protocol" or "variation acceptable" metrics. Three radiation oncologists qualitatively scored the Ethos plans; 78% of plans were scored as clinically acceptable during physician evaluation, with zero plans receiving scores requiring major changes. ConclusionsA standard Ethos template generated lung cancer radiotherapy plans with greater target coverage, increased spinal cord, heart, and lung V5Gy sparing, but increased esophagus and brachial plexus dose, compared to manually generated Eclipse plans. This template elucidates an efficient approach for generating automated, high quality lung radiation therapy treatment plans.

PurposeTemporally feathered radiation therapy (TFRT) for head-and-neck cancer (HNC) radiotherapy combines variable-dose daily subplans to increase the rest time of organs-at-risk (OARs) as sought in intensity modulated radiation therapy (IMRT). While the standard TFRT recommends uniform rest time for each OAR, improved toxicity outcomes may be achieved through variable rest time for OARs by incorporating the OARs variable radiosensitivity profiles. Methods and MaterialsA decision-making model was constructed to maximize the combined recovery of OARs by determining OARs optimal rest times. Two main components were incorporated: the cumulative biologically effective dose based on the linear-quadratic model; and a dynamical model capturing the adjusted recovery of OARs as a function of delivered dose. Further, variable radiosensitivity profiles were allowed across the OARs to capture their variable recovery time. Individual recoveries of each OAR under IMRT and the standard TFRT (sTFRT) was compared against optimized TFRT (oTFRT). ResultsFive OARs (larynx, esophagus, parotid, spinal cord, brainstem) were considered. When the cumulative dose delivered under TFRT and IMRT remains the same, three OARs exhibited higher recovery under oTFRT compared to the second-best approach (larynx (81.8% vs. 74.1%), esophagus (95.9% vs. 93.9%), parotid (85.6% vs. 83.5%), while the recovery of spinal cord (90.5% vs. 90.8%) and brainstem (96.2% vs. 96.6%) remained comparable under TFRT and IMRT approaches. With different cumulative dose under TFRT and IMRT, oTFRT achieved significantly higher recovery for larynx (95.5% vs. 81.8%) and parotid (92.9% vs. 85.6%), while it is slightly outperformed by IMRT for esophagus (93.4% vs. 95.9%), spinal cord (87.1% vs. 90.5%), and brainstem (90.2% vs. 96.6%). When considering the minimum end-of-treatment recovery, oTFRT always achieved higher recovery among the other two approaches. ConclusionsBy considering non-identical radiosensitivity profiles of OARs in HNC radiotherapy, TFRT can optimize their rest time to enhance recovery at the end of treatment, potentially reducing patient toxicities.

IntroductionTreatments encompassing multiple courses of radiation are becoming increasingly common in the management of oligometastatic disease, offering opportunities to extend progression-free and overall survival. However, a major challenge in clinical practice is the lack of standardized methods to assess and mitigate toxicity risks associated with successive radiation treatments. Furthermore, normal tissue recovery post-radiation remains poorly characterized, and the absence of standardized documentation for radiotherapy history complicates large-scale research efforts. To address these limitations, we propose the development of a novel DICOM-compatible object for integration into patient medical records. Materials and MethodsWe generated software designs and bundle mathematics that demonstrate the utility of this DICOM object and how various dose forgiveness algorithms can be applied to the data. We include simple linear, exponential, logarithmic, and Gaussian recovery algorithms as well as complex non-linear algorithms based on the literature currently available. ResultsWe applied the tool to an anonymized patient dataset, demonstrating the mathematical analysis applied to the data found in the new DICOM object. Noting ease and efficacy, we demonstrated that, in contrast to the current practice of gathering and structuring information distributed across electronic medical records, ready access to prior radiation courses accomplished two goals. (1) Facilitate data collection and analysis by streamlining access to comprehensive radiotherapy history, enabling researchers to conduct large-scale studies, and ultimately improve our understanding of tissue recovery. (2) Enhance clinical decision-making by enabling clinicians or software tools to leverage this data to personalize treatment plans, support clinical decision making to minimize toxicity risks during re-irradiation. For the anonymized patient, our analysis demonstrates safer delivery of re-radiation plans when viewed in the lens of dose forgiveness. ConclusionsA novel DICOM object which keeps track of radiation treatments enables clinicians to factor tissue recovery and response into planning safer multiple radiation therapy courses and facilitates cross-institution research on re-irradiation and dose forgiveness.

BackgroundRadiotherapy (RT) dose optimization is often labor-intensive, requiring repeated manual adjustments to achieve clinically acceptable plans. PurposeIn this work, we introduce nnDoseNet, a deep learning framework designed to automate and streamline RT dose prediction. MethodsBuilding on the nnU-Net segmentation engine, nnDoseNet adapts this architecture for dose regression by incorporating specialized loss functions (including dose-volume histogram terms) and multi-channel input (CT, targets, organs-at-risk, and body mask). It also supports clinically relevant evaluation metrics (e.g., gamma analysis and D95). ResultsWe evaluated nnDoseNet on the OpenKBP challenge dataset comprising 340 head-and- neck cancer cases (240 for training and 100 for testing). Multiple hyperparameters (U-Net depth, patch size, batch size, and loss function) were tested. The best-performing configuration achieved a dose score of 2.579 and a DVH score of 1.540 on the test set--competitive with top submissions in the original challenge. Additional validation on an institutional cohort of 80 prostate cancer patients (45 training, 35 testing) demonstrated good agreement with clinical dose distributions (mean-squared error 0.817) and improved target coverage compared to clinical plans. ConclusionBy offering automated data preprocessing, systematic model training, and robust dose evaluation--all within a single framework--nnDoseNet reduces the complexity of building and testing dose prediction models. It accommodates diverse prescription doses, organ-at-risk definitions, and hardware configurations, making it a suitable benchmark for multi-institutional research. With its balance of simplicity, flexibility, and performance, nnDoseNet aims to accelerate the development, comparison, and clinical integration of advanced AI-driven dose prediction methods in radiotherapy.

1.PurposeRadiotherapy treatment planning is a resource-intensive process characterized by multiple manual steps and clinical hand-offs that contribute to treatment delays and inter-observer variability. The Radiation Planning Assistant (RPA) is a web-based platform designed to deliver automated contouring and planning approaches tailored to low-resource settings. This work expands the RPA to develop and clinically validate end-to-end, AI-driven workflows for prostate and cervical cancers, designed to improve efficiency, consistency, and accessibility in low- and middle-income countries (LMICs). MethodsWe developed deep learning-based auto-contouring models using nnU-Net and integrated them with knowledge-based planning (KBP) models trained on curated datasets from over 1,000 prostate and 110 cervical cancer treatment plans. For prostate cancer, models were developed to accommodate prostate directed, prostate bed, and nodal treatment scenarios. Cervical cancer planning followed EMBRACE II guidelines and included pelvic and para-aortic nodal volumes. These tools were integrated into the RPA. Clinical acceptability of the auto-contours and plans was assessed retrospectively by radiation oncologists using a five-point Likert scale. ResultsIn total, 50 test patients (40 prostate, 10 cervical) were evaluated. For prostate cancer, 70% of target auto-contours and 73% of treatment plans were clinically acceptable without edits; for cervical cancer, these rates were 80% and 80%, respectively. For prostate cancer planning, 77% of target and 98% of organ-at-risk structures met all per-protocol compliance criteria. For cervical cancer planning, all EMBRACE II protocol hard constraint criteria were met. Bowel and vaginal contours demonstrated lower performance, but these did not compromise plan quality. ConclusionWe present validated, end-to-end radiotherapy planning workflows for prostate and cervical cancers that leverage the RPAs infrastructure to streamline treatment planning in a globally accessible platform and demonstrate high clinical acceptability. By reducing reliance on specialist input, this work addresses key barriers to equitable radiotherapy access in resource-limited settings and responds to global calls from the IAEA and WHO to expand radiotherapy capacity. FundingNational Institute of Health, National Science Foundation, Rising Tide Foundation, University of Texas MD Anderson Cancer Center

PurposeTo develop a new patient consult program, where patients are invited to meet directly with a clinical medical physicist to learn and ask questions about the technical aspects of their care. MethodsPatients are invited to meet voluntarily with a clinical medical physicist directly after the treatment planning CT appointment, and then again after treatment starts. Each consult starts with an overview of the clinical medical physicists role in patient care. This is followed by a detailed explanation of the treatment planning CT, treatment planning, and treatment delivery processes. Data are collected after each patient encounter, including: age, gender, treatment intent, treatment site, consult duration, discussion points, overall impression, and a summary of the questions asked. Qualitative data analysis focused on understanding the number and types of questions asked during the physics consults. Additional analyses focused on evaluating the encounter notes for interesting insights regarding meeting tone, number of meeting attendees, and other non-clinical discussion points. ResultsSixty three patients were seen between August 2016 and December 2017, accounting for 29% of the total department patient load. The average physics consult duration was 24 minutes. When evaluating the patient encounter notes for overall tone, 55 patients (87%) had positive descriptors such as "pleasant conversation". Thirty three patients (52%) brought at least one other person into the consult, and 27 patients (43%) contributed personal stories or professional background information to the conversation. When the collection of patient questions was grouped into question types, the data show that the majority of the consult discussion addresses questions related to treatment delivery, treatment planning, and other technical questions. ConclusionsIncorporation of a medical physics patient consult program into clinical practice requires modest time commitment, and has the benefits of increasing medical physics engagement with patient care and improving patient satisfaction through better education.