Knowledge-based quality assurance of a comprehensive set of organ at risk contours for head and neck radiotherapy.

Introduction: Manual review of organ at risk (OAR) contours is crucial for creating safe radiotherapy plans but can be time-consuming and error prone. Statistical and deep learning models show the potential to automatically detect improper contours by identifying outliers using large sets of acceptable data (knowledge-based outlier detection) and may be able to assist human reviewers during review of OAR contours. Methods: This study developed an automated knowledge-based outlier detection method and assessed its ability to detect erroneous contours for all common head and neck (HN) OAR types used clinically at our institution. We utilized 490 accurate CT-based HN structure sets from unique patients, each with forty-two HN OAR contours when anatomically present. The structure sets were distributed as 80% for training, 10% for validation, and 10% for testing. In addition, 190 and 37 simulated contours containing errors were added to the validation and test sets, respectively. Single-contour features, including location, shape, orientation, volume, and CT number, were used to train three single-contour feature models (z-score, Mahalanobis distance [MD], and autoencoder [AE]). Additionally, a novel contour-to-contour relationship (CCR) model was trained using the minimum distance and volumetric overlap between pairs of OAR contours to quantify overlap and separation. Inferences from single-contour feature models were combined with the CCR model inferences and inferences evaluating the number of disconnected parts in a single contour and then compared. Results: In the test dataset, before combination with the CCR model, the area under the curve values were 0.922/0.939/0.939 for the z-score, MD, and AE models respectively for all contours. After combination with CCR model inferences, the z-score, MD, and AE had sensitivities of 0.838/0.892/0.865, specificities of 0.922/0.907/0.887, and balanced accuracies (BA) of 0.880/0.900/0.876 respectively. In the validation dataset, with similar overall performance and no signs of overfitting, model performance for individual OAR types was assessed. The combined AE model demonstrated minimum, median, and maximum BAs of 0.729, 0.908, and 0.980 across OAR types. Discussion: Our novel knowledge-based method combines models utilizing single-contour and CCR features to effectively detect erroneous OAR contours across a comprehensive set of 42 clinically used OAR types for HN radiotherapy.

Research Publication Rates Are Age- and Era-Dependent for Radiation Oncologists From Highly Ranked NCI-Designated Cancer Centers.

PURPOSE: Quantitative bibliometrics are increasingly used to evaluate faculty research productivity. This study benchmarks publication rates for radiation oncologists from highly ranked National Cancer Institute-designated cancer centers and reveals how productivity changes over the arc of a career and of the field over time. METHODS AND MATERIALS: Peer-reviewed articles from 1970 to 2022 were obtained using Scopus for the 348 radiation oncologists listed as faculty for the top 10 cancer hospitals ranked by US News and World Report in 2022. Bibliometrics were analyzed for authorships (A˙), authorships where the individual was first or last author (F˙L), the monograph equivalent of authorships (M˙E), h-index, and ha-index (an analog to h-index using M˙E in place of publications). Career start was defined as the year of first publication. Bibliometric inflation was explored by analyzing authorship and bibliometric changes between 1990 and 2022. RESULTS: Publication rates peak, with as much as a 500% increase, 20 to 25 years from the start of a career before declining until retirement. At career ages of 1, 10, 20, and 30 years, the median bibliometrics were A˙ = (1.5, 4.1, 6.5, 7.0) year-1, F˙L = (0.5, 0.9, 1.2, 0.6) year-1, M˙E= (0.2, 0.5, 0.7, 0.8) year-1, h-index = (1, 12, 22, 47), and ha-index = (0.4, 4.4, 6.9, 18.4). With regards to authorship patterns across eras, the median number of authors listed per paper increased by 240% between 1990 and 2022. Meanwhile, research productivity per individual as measured by F˙L and M˙E was unchanged. CONCLUSIONS: The research publication rates of the median radiation oncologist change substantially over the course of their career. Productivity improves steadily for more than 2 decades before peaking and declining. The culture of authorship has also changed between 1990 and 2022. The number of authors listed per paper has trended upwards, which has an inflationary effect on the number of authorships and h-index. Meanwhile, the rate of manuscripts published per faculty has not changed.

Sublobar Resection, Stereotactic Body Radiation Therapy, and Percutaneous Ablation Provide Comparable Outcomes for Lung Metastasis-Directed Therapy.

BACKGROUND: Prolonged survival of patients with metastatic disease has furthered interest in metastasis-directed therapy (MDT). RESEARCH QUESTION: There is a paucity of data comparing lung MDT modalities. Do outcomes among sublobar resection (SLR), stereotactic body radiation therapy (SBRT), and percutaneous ablation (PA) for lung metastases vary in terms of local control and survival? STUDY DESIGN AND METHODS: Medical records of patients undergoing lung MDT at a single cancer center between January 2015 and December 2020 were reviewed. Overall survival, local progression, and toxicity outcomes were collected. Patient and lesion characteristics were used to generate multivariable models with propensity weighted analysis. RESULTS: Lung MDT courses (644 total: 243 SLR, 274 SBRT, 127 PA) delivered to 511 patients were included with a median follow-up of 22 months. There were 47 local progression events in 45 patients, and 159 patients died. Two-year overall survival and local progression were 80.3% and 63.3%, 83.8% and 9.6%, and 4.1% and 11.7% for SLR, SBRT, and PA, respectively. Lesion size per 1 cm was associated with worse overall survival (hazard ratio, 1.24; P = .003) and LP (hazard ratio, 1.50; P < .001). There was no difference in overall survival by modality. Relative to SLR, there was no difference in risk of local progression with PA; however, SBRT was associated with a decreased risk (hazard ratio, 0.26; P = .023). Rates of severe toxicity were low (2.1%-2.6%) and not different among groups. INTERPRETATION: This study performs a propensity weighted analysis of SLR, SBRT, and PA and shows no impact of lung MDT modality on overall survival. Given excellent local control across MDT options, a multidisciplinary approach is beneficial for patient triage and longitudinal management.

Development and Dosimetric Characterization of a Customizable Shield for Subtotal Skin Electron Beam Therapy.

Purpose: Purpose: Subtotal skin electron beam therapy may be an option for patients with cutaneous lymphoma receiving radiation therapy to treat large areas of their skin but may benefit from sparing specific areas that may have had previous radiation therapy, are of specific cosmetic concern, and/or show no evidence of disease. We report here on the design, implementation, and dosimetric characteristics of a reusable and transparent customizable shield for use with the large fields used to deliver total skin electron beam therapy at extended distance with a conventional linear accelerator. Methods and Materials: A shield was designed and manufactured consisting of acrylic blocks that can be mounted on a steel frame to allow patient-specific shielding. The dosimetry of the device was measured using radiochromic film. Results: The shield is easy to use and well-tolerated for patient treatment, providing minimal electron transmission through the shield with a sharp penumbra at the field edge, with no increase in x-ray dose. We report on the dosimetry of a commercial device that has been used to treat more than 30 patients to date. Conclusions: The customizable shield is well suited to providing patient-specific shielding for subtotal skin electron beam therapy.

Validation of clinical acceptability of deep-learning-based automated segmentation of organs-at-risk for head-and-neck radiotherapy treatment planning.

Introduction: Organ-at-risk segmentation for head and neck cancer radiation therapy is a complex and time-consuming process (requiring up to 42 individual structure, and may delay start of treatment or even limit access to function-preserving care. Feasibility of using a deep learning (DL) based autosegmentation model to reduce contouring time without compromising contour accuracy is assessed through a blinded randomized trial of radiation oncologists (ROs) using retrospective, de-identified patient data. Methods: Two head and neck expert ROs used dedicated time to create gold standard (GS) contours on computed tomography (CT) images. 445 CTs were used to train a custom 3D U-Net DL model covering 42 organs-at-risk, with an additional 20 CTs were held out for the randomized trial. For each held-out patient dataset, one of the eight participant ROs was randomly allocated to review and revise the contours produced by the DL model, while another reviewed contours produced by a medical dosimetry assistant (MDA), both blinded to their origin. Time required for MDAs and ROs to contour was recorded, and the unrevised DL contours, as well as the RO-revised contours by the MDAs and DL model were compared to the GS for that patient. Results: Mean time for initial MDA contouring was 2.3 hours (range 1.6-3.8 hours) and RO-revision took 1.1 hours (range, 0.4-4.4 hours), compared to 0.7 hours (range 0.1-2.0 hours) for the RO-revisions to DL contours. Total time reduced by 76% (95%-Confidence Interval: 65%-88%) and RO-revision time reduced by 35% (95%-CI,-39%-91%). All geometric and dosimetric metrics computed, agreement with GS was equivalent or significantly greater (p<0.05) for RO-revised DL contours compared to the RO-revised MDA contours, including volumetric Dice similarity coefficient (VDSC), surface DSC, added path length, and the 95%-Hausdorff distance. 32 OARs (76%) had mean VDSC greater than 0.8 for the RO-revised DL contours, compared to 20 (48%) for RO-revised MDA contours, and 34 (81%) for the unrevised DL OARs. Conclusion: DL autosegmentation demonstrated significant time-savings for organ-at-risk contouring while improving agreement with the institutional GS, indicating comparable accuracy of DL model. Integration into the clinical practice with a prospective evaluation is currently underway.

Single and Multifraction Spine Stereotactic Body Radiation Therapy and the Risk of Radiation Induced Myelopathy.

Purpose: This study reports on the risk of radiation-induced myelitis (RM) of the spinal cord from a large single-institutional experience with 1 to 5 fraction stereotactic body radiation therapy (SBRT) to the spine. Methods and Materials: A retrospective review of patients who received spine SBRT to a radiation naïve level at or above the conus medullaris between 2007 and 2019 was performed. Local failure determination was based on SPIne response assessment in Neuro-Oncology criteria. RM was defined as neurologic symptoms consistent with the segment of cord irradiated in the absence of neoplastic disease recurrence and graded by Common Toxicity Criteria for Adverse Events, version 4.0. Rates of adverse events were estimated and dose-volume statistics from delivered treatment plans were extracted for the planning target volumes and spinal cord. Results: A total of 353 lesions in 277 patients were identified that met the specified criteria, for which 270, 70, and 13 lesions received 1-, 3-, and 5-fraction treatments, respectively, with a median follow-up of 46 months (95% confidence interval [CI], 41-52 months) for all surviving patients. The median overall survival was 33.0 months (95% CI, 29-43). The median D0.03cc to the spinal cord was 11.7 Gy (interquartile range [IQR], 10.5-12.4), 16.7 Gy (IQR, 12.8-20.6), and 26.0 Gy (IQR, 24.1-28.1), for 1-, 3-, 5-fractions. Using an a/b = 2Gy for the spinal cord, the median single-fraction equivalent-dose (SFED2) was 11.7 Gy (IQR, 10.2-12.5 Gy) and the normalized biological equivalent dose (nBED2/2) was 19.9 Gy (IQR, 15.4-22.8 Gy). One patient experienced grade 2 RM after a single-fraction treatment. The cumulative probability of RM was 0.3% (95% CI, 0%-2%). Conclusions: Spine SBRT is safe while limiting the spinal cord (as defined on treatment planning magnetic resonance imaging or computed tomography myelogram) D0.03cc to less than 14 Gy, 21.9 Gy, and 30 Gy, for 1, 3, and 5-fractions, consistent with standard guidelines.

Scalable radiotherapy data curation infrastructure for deep-learning based autosegmentation of organs-at-risk: A case study in head and neck cancer.

In this era of patient-centered, outcomes-driven and adaptive radiotherapy, deep learning is now being successfully applied to tackle imaging-related workflow bottlenecks such as autosegmentation and dose planning. These applications typically require supervised learning approaches enabled by relatively large, curated radiotherapy datasets which are highly reflective of the contemporary standard of care. However, little has been previously published describing technical infrastructure, recommendations, methods or standards for radiotherapy dataset curation in a holistic fashion. Our radiation oncology department has recently embarked on a large-scale project in partnership with an external partner to develop deep-learning-based tools to assist with our radiotherapy workflow, beginning with autosegmentation of organs-at-risk. This project will require thousands of carefully curated radiotherapy datasets comprising all body sites we routinely treat with radiotherapy. Given such a large project scope, we have approached the need for dataset curation rigorously, with an aim towards building infrastructure that is compatible with efficiency, automation and scalability. Focusing on our first use-case pertaining to head and neck cancer, we describe our developed infrastructure and novel methods applied to radiotherapy dataset curation, inclusive of personnel and workflow organization, dataset selection, expert organ-at-risk segmentation, quality assurance, patient de-identification, data archival and transfer. Over the course of approximately 13 months, our expert multidisciplinary team generated 490 curated head and neck radiotherapy datasets. This task required approximately 6000 human-expert hours in total (not including planning and infrastructure development time). This infrastructure continues to evolve and will support ongoing and future project efforts.

Development of transparent eye shields for total skin electron beam radiotherapy.

PURPOSE: For total skin electron (TSE) beam radiation therapy, the anterior eye and conjunctiva can be protected with eye shields to prevent keratitis, xerophthalmia, and cataractogenesis. Conventional metal eye shields can reduce patient balance by obscuring vision and thus increasing the risk for falls. We report on the design, fabrication, and clinical use of transparent acrylic eye shields for TSE. METHODS: The primary design goals were a seven-fold reduction in the dose to the anterior eye and conjunctiva to meet published dose-recommendations, preservation of vision for the wearer, and biocompatibility for external use. Resembling thick swim goggles, the design features 23 mm thick acrylic lenses that are mounted in a 3-D printed support structure that conforms to the eye socket and can be worn with a strap. Dose measurements were performed in a simulated Stanford-technique treatment with an anthropomorphic phantom using Gafchromic EBT film RESULTS: The transparent eye shields were manufactured using a 3D-printer and CNC-machine. Based on measurements from the simulated treatments for each of the eye shields, the eye shields provided a 12-fold reduction in dose to the lens. After use in more than 200 fractions, the shields were well tolerated by patients, and there were no reports of any incidents or adverse events. CONCLUSION: Transparent TSE eye shields are able to reduce the dose to the eyes while maintaining vision during treatment at a reasonable cost.

Ablative radiotherapy for ultracentral lung cancers: Dosimetric, geometric, and volumetric predictors of outcomes and toxicity.

BACKGROUND: Ultracentral lung cancers arise near the proximal bronchial tree (PBT), trachea, or esophagus, and have been associated with worse outcomes and increased toxicity after radiotherapy. We sought to associate dosimetric and anatomic factors with oncologic outcomes and toxicities. METHODS: One-hundred ten patients treated with ablative, curative-intent radiotherapy for ultracentral, node-negative, non-small cell lung cancer were included. Dosimetric and geometric data obtained using custom software that calculated volumes of target structures and organs-at-risk and measured the shortest vector length between these volumes were associated with outcomes and toxicity. RESULTS: Common dose/fractionation schemes included 50 Gy in 5 fractions (57%), 60 Gy in 8 fractions (15%), and 48 Gy in 4 fractions (13%). Overall survival at 1, 2, and 5 years was 78%, 57%, and 32%, respectively. Factors significantly associated with death included endobronchial tumor, gross tumor volume (GTV) or planning target volume (PTV) contacting PBT, shorter distance from GTV to PBT or esophagus, volume of PBT receiving prescription dose, higher pericardium max dose, lung V20Gy, and mean lung dose. Local progression at 1, 2, and 5 years was 4%, 16%, and 21%. Factors associated with local progression were lower GTV minimum dose and higher GTV/PTV volume ratio. Acute and late grade 2 + toxicity was seen in 18% and 27%, respectively. Four patients (4%) had fatal toxicity. CONCLUSIONS: Lower GTV minimum dose and smaller volumetric PTV expansions may increase risk of local progression, and should be balanced against normal tissue doses including pericardium maximum dose, lung V20Gy, and mean lung dose.

Monte Carlo calculation of dose distributions in oligometastatic patients planned for spine stereotactic ablative radiotherapy.

Dosimetric consequences of plans optimized using the analytical anisotropic algorithm (AAA) implemented in the Varian Eclipse treatment planning system for spine stereotactic body radiotherapy were evaluated by re-calculating with BEAMnrc/DOSXYZnrc Monte Carlo. Six patients with spinal vertebral metastases were planned using volumetric modulated arc therapy. The planning goal was to cover at least 80% of the planning target volume with a prescribed dose of 35 Gy in five fractions. Tissue heterogeneity-corrected AAA dose distributions for the planning target volume and spinal canal planning organ-at-risk volume were compared against those obtained from Monte Carlo. The results showed that the AAA overestimated planning target volume coverage with the prescribed dose by up to 13.5% (mean 8.3% +/- 3.2%) when compared to Monte Carlo simulations. Maximum dose to spinal canal planning organ-at-risk volume calculated with Monte Carlo was consistently smaller than calculated with the treatment planning system and remained under spinal cord dose tolerance. Differences in dose distribution appear to be related to the dosimetric effects of accounting for body composition in Monte Carlo simulations. In contrast, the treatment planning system assumes that all tissues are water-equivalent in their composition and only differ in their electron density.