Feasibility of Left Anterior Descending Coronary Artery Sparing Radiation Therapy for Locally Advanced Lung Cancer.

Efforts to mitigate radiation therapy (RT)-associated cardiotoxicity have focused on constraining mean heart dose. However, recent studies have shown greater predictive power with cardiac substructure dose metrics, such as the left anterior descending (LAD) coronary artery volume (V) receiving 15 Gy (V15Gy) ≥10%. Herein, we investigated the feasibility of LAD radiation sparing in contemporary intensity modulated RT (IMRT)/volumetric modulated arc therapy (VMAT) lung cancer plans. Single institution retrospective analysis of 54 patients with locally advanced lung cancer treated with thoracic RT was conducted between February 2018 and August 2021. After excluding 33 (5 = non-IMRT/VMAT or intentionally LAD-optimized; 28 = LAD V15Gy <10%), 21 plans with LAD V15Gy ≥10% were identified for LAD reoptimization with intent to meet LAD V15Gy <10% while maintaining meeting organ at risk (OAR) metrics and target coverage with original plan parameters. Dosimetric variables were compared using paired t tests. Most patients (57.1%, 12/21) were treated with definitive RT, 8 of 21 patients (38.1%) with postoperative RT, and 1 with neoadjuvant RT. The median prescribed RT dose was 60 Gy (range, 50.4-66 Gy) in 30 fractions (range, 28-33 fractions). LAD reoptimized plans (vs original) led to significant reductions in mean LAD V15Gy (39.4% ± 13.9% vs 9.4% ± 13.0%; P < .001) and mean LAD dose (12.9 Gy ± 4.6 Gy vs 7.6 Gy ± 2.8 Gy; P < .001). Most (85.7%; 18/21) LAD reoptimized plans achieved LAD V15Gy <10%. There were no statistically significant differences in overall lung, esophageal, or spinal cord dose metrics. Only 1 reoptimization (1/21) exceeded an OAR constraint that was initially met in the original plan. To our knowledge, this is the first report describing the feasibility of LAD-optimized lung cancer RT planning using the newly identified LAD V15Gy constraint. We observed that LAD V15Gy <10% is achievable in more than 85% of plans initially exceeding this constraint, with minimal dosimetric tradeoffs. Our results support the feasibility of routine incorporation of the LAD as an OAR in modern thoracic IMRT/VMAT planning.

Feasibility and clinical implementation of MRI-guided surface brachytherapy.

Purpose: Best practices for high-dose-rate surface applicator brachytherapy treatment (SABT) have long relied on computed tomography (CT)-based imaging to visualize diseased sites for treatment planning. Compared with magnetic resonance (MR)-based imaging, CT provides insufficient soft tissue contrast. This work described the feasibility of clinical implementation of MR-based imaging in SABT planning to provide individualized treatment optimization. Material and methods: A 3D-printed phantom was used to fit Freiberg flap-style (Elekta, The Netherlands) applicator. Images were taken using an optimized pointwise encoding time reduction with radial acquisition (PETRA) MR sequence for catheter visualization, and a helical CT scan to generate parallel treatment plans. This clinical study included three patients undergoing SABT for Dupuytren's contracture/palmar fascial fibromatosis imaged with the same modalities.SABT planning was performed in Oncentra Brachy (Elekta Brachytherapy, The Netherlands) treatment planning software. A geometric analysis was conducted by comparing CT-based digitization with MR-based digitization. CT and MR dwell positions underwent a rigid registration, and average Euclidean distances between dwell positions were calculated. A dosimetric comparison was performed, including point-based dose difference calculations and volumetric segmentations with Dice similarity coefficient (DSC) calculations. Results: Euclidean distances between dwell positions from CT-based and MR-based plans were on average 0.68 ±0.05 mm and 1.35 ±0.17 mm for the phantom and patients, respectively. The point dose difference calculations were on average 0.92% for the phantom and 1.98% for the patients. The D95 and D90 DSC calculations were both 97.9% for the phantom, and on average 93.6% and 94.2%, respectively, for the patients. Conclusions: The sub-millimeter accuracy of dwell positions and high DSC's (> 0.95) of the phantom demonstrated that digitization was clinically acceptable, and accurate treatment plans were produced using MR-only imaging. This novel approach, MRI-guided SABT, will lead to individualized prescriptions for potentially improved patient outcomes.

Edge roughness quantifies impact of physician variation on training and performance of deep learning auto-segmentation models for the esophagus.

Manual segmentation of tumors and organs-at-risk (OAR) in 3D imaging for radiation-therapy planning is time-consuming and subject to variation between different observers. Artificial intelligence (AI) can assist with segmentation, but challenges exist in ensuring high-quality segmentation, especially for small, variable structures, such as the esophagus. We investigated the effect of variation in segmentation quality and style of physicians for training deep-learning models for esophagus segmentation and proposed a new metric, edge roughness, for evaluating/quantifying slice-to-slice inconsistency. This study includes a real-world cohort of 394 patients who each received radiation therapy (mainly for lung cancer). Segmentation of the esophagus was performed by 8 physicians as part of routine clinical care. We evaluated manual segmentation by comparing the length and edge roughness of segmentations among physicians to analyze inconsistencies. We trained eight multiple- and individual-physician segmentation models in total, based on U-Net architectures and residual backbones. We used the volumetric Dice coefficient to measure the performance for each model. We proposed a metric, edge roughness, to quantify the shift of segmentation among adjacent slices by calculating the curvature of edges of the 2D sagittal- and coronal-view projections. The auto-segmentation model trained on multiple physicians (MD1-7) achieved the highest mean Dice of 73.7 ± 14.8%. The individual-physician model (MD7) with the highest edge roughness (mean ± SD: 0.106 ± 0.016) demonstrated significantly lower volumetric Dice for test cases compared with other individual models (MD7: 58.5 ± 15.8%, MD6: 67.1 ± 16.8%, p < 0.001). A multiple-physician model trained after removing the MD7 data resulted in fewer outliers (e.g., Dice ≤ 40%: 4 cases for MD1-6, 7 cases for MD1-7, Ntotal = 394). While we initially detected this pattern in a single clinician, we validated the edge roughness metric across the entire dataset. The model trained with the lowest-quantile edge roughness (MDER-Q1, Ntrain = 62) achieved significantly higher Dice (Ntest = 270) than the model trained with the highest-quantile ones (MDER-Q4, Ntrain = 62) (MDER-Q1: 67.8 ± 14.8%, MDER-Q4: 62.8 ± 15.7%, p < 0.001). This study demonstrates that there is significant variation in style and quality in manual segmentations in clinical care, and that training AI auto-segmentation algorithms from real-world, clinical datasets may result in unexpectedly under-performing algorithms with the inclusion of outliers. Importantly, this study provides a novel evaluation metric, edge roughness, to quantify physician variation in segmentation which will allow developers to filter clinical training data to optimize model performance.

Observer preference of artificial intelligence-generated versus clinical prostate contours for ultrasound-based high dose rate brachytherapy.

BACKGROUND: For trans-rectal ultrasound (TRUS)-based high dose rate (HDR) prostate brachytherapy, prostate contouring can be challenging due to artifacts from implanted needles, bleeding, and calcifications. PURPOSE: To evaluate the geometric accuracy and observer preference of an artificial intelligence (AI) algorithm for generating prostate contours on TRUS images with implanted needles. METHODS: We conducted a retrospective study of 150 patients, who underwent HDR brachytherapy. These patients were randomly divided into training (104), validation (26) and testing (20) sets. An AI algorithm was trained/validated utilizing the TRUS image and reference (clinical) contours. The algorithm then provided contours for the test set. For evaluation, we calculated the Dice coefficient between AI and reference prostate contours. We then presented AI and reference contours to eight clinician observers, and asked observers to select their preference. Observers were blinded to the source of contours. We calculated the percentage of cases in which observers preferred AI contours. Lastly, we evaluate whether the presence of AI contours improved the geometric accuracy of prostate contours provided by five resident observers for a 10-patient subset. RESULTS: The median Dice coefficient between AI and reference contours was 0.92 (IQR: 0.90-0.94). Observers preferred AI contours for a median of 57.5% (IQR: 47.5, 65.0) of the test cases. For resident observers, the presence of AI contours was associated with a 0.107 (95% CI: 0.086, 0.128; p < 0.001) improvement in Dice coefficient for the 10-patient subset. CONCLUSION: The AI algorithm provided high-quality prostate contours on TRUS with implanted needles. Further prospective study is needed to better understand how to incorporate AI prostate contours into the TRUS-based HDR brachytherapy workflow.

Decoding biological age from face photographs using deep learning.

Because humans age at different rates, a person's physical appearance may yield insights into their biological age and physiological health more reliably than their chronological age. In medicine, however, appearance is incorporated into medical judgments in a subjective and non-standardized fashion. In this study, we developed and validated FaceAge, a deep learning system to estimate biological age from easily obtainable and low-cost face photographs. FaceAge was trained on data from 58,851 healthy individuals, and clinical utility was evaluated on data from 6,196 patients with cancer diagnoses from two institutions in the United States and The Netherlands. To assess the prognostic relevance of FaceAge estimation, we performed Kaplan Meier survival analysis. To test a relevant clinical application of FaceAge, we assessed the performance of FaceAge in end-of-life patients with metastatic cancer who received palliative treatment by incorporating FaceAge into clinical prediction models. We found that, on average, cancer patients look older than their chronological age, and looking older is correlated with worse overall survival. FaceAge demonstrated significant independent prognostic performance in a range of cancer types and stages. We found that FaceAge can improve physicians' survival predictions in incurable patients receiving palliative treatments, highlighting the clinical utility of the algorithm to support end-of-life decision-making. FaceAge was also significantly associated with molecular mechanisms of senescence through gene analysis, while age was not. These findings may extend to diseases beyond cancer, motivating using deep learning algorithms to translate a patient's visual appearance into objective, quantitative, and clinically useful measures.

Development and Validation of an Automated Image-Based Deep Learning Platform for Sarcopenia Assessment in Head and Neck Cancer.

Importance: Sarcopenia is an established prognostic factor in patients with head and neck squamous cell carcinoma (HNSCC); the quantification of sarcopenia assessed by imaging is typically achieved through the skeletal muscle index (SMI), which can be derived from cervical skeletal muscle segmentation and cross-sectional area. However, manual muscle segmentation is labor intensive, prone to interobserver variability, and impractical for large-scale clinical use. Objective: To develop and externally validate a fully automated image-based deep learning platform for cervical vertebral muscle segmentation and SMI calculation and evaluate associations with survival and treatment toxicity outcomes. Design, Setting, and Participants: For this prognostic study, a model development data set was curated from publicly available and deidentified data from patients with HNSCC treated at MD Anderson Cancer Center between January 1, 2003, and December 31, 2013. A total of 899 patients undergoing primary radiation for HNSCC with abdominal computed tomography scans and complete clinical information were selected. An external validation data set was retrospectively collected from patients undergoing primary radiation therapy between January 1, 1996, and December 31, 2013, at Brigham and Women's Hospital. The data analysis was performed between May 1, 2022, and March 31, 2023. Exposure: C3 vertebral skeletal muscle segmentation during radiation therapy for HNSCC. Main Outcomes and Measures: Overall survival and treatment toxicity outcomes of HNSCC. Results: The total patient cohort comprised 899 patients with HNSCC (median [range] age, 58 [24-90] years; 140 female [15.6%] and 755 male [84.0%]). Dice similarity coefficients for the validation set (n = 96) and internal test set (n = 48) were 0.90 (95% CI, 0.90-0.91) and 0.90 (95% CI, 0.89-0.91), respectively, with a mean 96.2% acceptable rate between 2 reviewers on external clinical testing (n = 377). Estimated cross-sectional area and SMI values were associated with manually annotated values (Pearson r = 0.99; P < .001) across data sets. On multivariable Cox proportional hazards regression, SMI-derived sarcopenia was associated with worse overall survival (hazard ratio, 2.05; 95% CI, 1.04-4.04; P = .04) and longer feeding tube duration (median [range], 162 [6-1477] vs 134 [15-1255] days; hazard ratio, 0.66; 95% CI, 0.48-0.89; P = .006) than no sarcopenia. Conclusions and Relevance: This prognostic study's findings show external validation of a fully automated deep learning pipeline to accurately measure sarcopenia in HNSCC and an association with important disease outcomes. The pipeline could enable the integration of sarcopenia assessment into clinical decision making for individuals with HNSCC.

Left Anterior Descending Coronary Artery Radiation Dose Association With All-Cause Mortality in NRG Oncology Trial RTOG 0617.

PURPOSE: A left anterior descending (LAD) coronary artery volume (V) receiving 15 Gy (V15 Gy) ≥10% has been recently observed to be an independent risk factor of major adverse cardiac events and all-cause mortality in patients with locally advanced non-small cell lung cancer treated with radiation therapy. However, this dose constraint has not been validated in independent or prospective data sets. METHODS AND MATERIALS: The NRG Oncology/Radiation Therapy Oncology Group (RTOG) 0617 data set from the National Clinical Trials Network was used. The LAD coronary artery was manually contoured. Multivariable Cox regression was performed, adjusting for known prognostic factors. Kaplan-Meier estimates of overall survival (OS) were calculated. For assessment of baseline cardiovascular risk, only age, sex, and smoking history were available. RESULTS: There were 449 patients with LAD dose-volume data and clinical outcomes available after 10 patients were excluded owing to unreliable LAD dose statistics. The median age was 64 years. The median LAD V15 Gy was 38% (interquartile range, 15%-62%), including 94 patients (21%) with LAD V15 Gy <10% and 355 (79%) with LAD V15 Gy ≥10%. Adjusting for prognostic factors, LAD V15 Gy ≥10% versus <10% was associated with an increased risk of all-cause mortality (hazard ratio [HR], 1.43; 95% confidence interval, 1.02-1.99; P = .037), whereas a mean heart dose ≥10 Gy versus <10 Gy was not (adjusted HR, 1.12; 95% confidence interval, 0.88-1.43; P = .36). The median OS for patients with LAD V15 Gy ≥10% versus <10% was 20.2 versus 25.1 months, respectively, with 2-year OS estimates of 47% versus 67% (P = .004), respectively. CONCLUSIONS: In a reanalysis of RTOG 0617, LAD V15 Gy ≥10% was associated with an increased risk of all-cause mortality. These findings underscore the need for improved cardiac risk stratification and aggressive risk mitigation strategies, including implementation of cardiac substructure dose constraints in national guidelines and clinical trials.

Prospective deployment of an automated implementation solution for artificial intelligence translation to clinical radiation oncology.

Introduction: Artificial intelligence (AI)-based technologies embody countless solutions in radiation oncology, yet translation of AI-assisted software tools to actual clinical environments remains unrealized. We present the Deep Learning On-Demand Assistant (DL-ODA), a fully automated, end-to-end clinical platform that enables AI interventions for any disease site featuring an automated model-training pipeline, auto-segmentations, and QA reporting. Materials and methods: We developed, tested, and prospectively deployed the DL-ODA system at a large university affiliated hospital center. Medical professionals activate the DL-ODA via two pathways (1): On-Demand, used for immediate AI decision support for a patient-specific treatment plan, and (2) Ambient, in which QA is provided for all daily radiotherapy (RT) plans by comparing DL segmentations with manual delineations and calculating the dosimetric impact. To demonstrate the implementation of a new anatomy segmentation, we used the model-training pipeline to generate a breast segmentation model based on a large clinical dataset. Additionally, the contour QA functionality of existing models was assessed using a retrospective cohort of 3,399 lung and 885 spine RT cases. Ambient QA was performed for various disease sites including spine RT and heart for dosimetric sparing. Results: Successful training of the breast model was completed in less than a day and resulted in clinically viable whole breast contours. For the retrospective analysis, we evaluated manual-versus-AI similarity for the ten most common structures. The DL-ODA detected high similarities in heart, lung, liver, and kidney delineations but lower for esophagus, trachea, stomach, and small bowel due largely to incomplete manual contouring. The deployed Ambient QAs for heart and spine sites have prospectively processed over 2,500 cases and 230 cases over 9 months and 5 months, respectively, automatically alerting the RT personnel. Discussion: The DL-ODA capabilities in providing universal AI interventions were demonstrated for On-Demand contour QA, DL segmentations, and automated model training, and confirmed successful integration of the system into a large academic radiotherapy department. The novelty of deploying the DL-ODA as a multi-modal, fully automated end-to-end AI clinical implementation solution marks a significant step towards a generalizable framework that leverages AI to improve the efficiency and reliability of RT systems.

Image quality comparisons of coil setups in 3T MRI for brain and head and neck radiotherapy simulations.

PURPOSE: MRI is increasingly used for brain and head and neck radiotherapy treatment planning due to its superior soft tissue contrast. Flexible array coils can be arranged to encompass treatment immobilization devices, which do not fit in diagnostic head/neck coils. Selecting a flexible coil arrangement to replace a diagnostic coil should rely on image quality characteristics and patient comfort. We compared image quality obtained with a custom UltraFlexLarge18 (UFL18) coil setup against a commercial FlexLarge4 (FL4) coil arrangement, relative to a diagnostic Head/Neck20 (HN20) coil at 3T. METHODS: The large American College of Radiology (ACR) MRI phantom was scanned monthly in the UFL18, FL4, and HN20 coil setup over 2 years, using the ACR series and three clinical sequences. High-contrast spatial resolution (HCSR), image intensity uniformity (IIU), percent-signal ghosting (PSG), low-contrast object detectability (LCOD), signal-to-noise ratio (SNR), and geometric accuracy were calculated according to ACR recommendations for each series and coil arrangement. Five healthy volunteers were scanned with the clinical sequences in all three coil setups. SNR, contrast-to-noise ratio (CNR) and artifact size were extracted from regions-of-interest along the head for each sequence and coil setup. For both experiments, ratios of image quality parameters obtained with UFL18 or FL4 over those from HN20 were formed for each coil setup, grouping the ACR and clinical sequences. RESULTS: Wilcoxon rank-sum tests revealed significantly higher (p < 0.001) LCOD, IIU and SNR, and lower PSG ratios with UFL18 than FL4 on the phantom for the clinical sequences, with opposite PSG and SNR trends for the ACR series. Similar statistical tests on volunteer data corroborated that SNR ratios with UFL18 (0.58 ± 0.19) were significantly higher (p < 0.001) than with FL4 (0.51 ± 0.18) relative to HN20. CONCLUSIONS: The custom UFL18 coil setup was selected for clinical application in MR simulations due to the superior image quality demonstrated on a phantom and volunteers for clinical sequences and increased volunteer comfort.

Clinical validation of deep learning algorithms for radiotherapy targeting of non-small-cell lung cancer: an observational study.

BACKGROUND: Artificial intelligence (AI) and deep learning have shown great potential in streamlining clinical tasks. However, most studies remain confined to in silico validation in small internal cohorts, without external validation or data on real-world clinical utility. We developed a strategy for the clinical validation of deep learning models for segmenting primary non-small-cell lung cancer (NSCLC) tumours and involved lymph nodes in CT images, which is a time-intensive step in radiation treatment planning, with large variability among experts. METHODS: In this observational study, CT images and segmentations were collected from eight internal and external sources from the USA, the Netherlands, Canada, and China, with patients from the Maastro and Harvard-RT1 datasets used for model discovery (segmented by a single expert). Validation consisted of interobserver and intraobserver benchmarking, primary validation, functional validation, and end-user testing on the following datasets: multi-delineation, Harvard-RT1, Harvard-RT2, RTOG-0617, NSCLC-radiogenomics, Lung-PET-CT-Dx, RIDER, and thorax phantom. Primary validation consisted of stepwise testing on increasingly external datasets using measures of overlap including volumetric dice (VD) and surface dice (SD). Functional validation explored dosimetric effect, model failure modes, test-retest stability, and accuracy. End-user testing with eight experts assessed automated segmentations in a simulated clinical setting. FINDINGS: We included 2208 patients imaged between 2001 and 2015, with 787 patients used for model discovery and 1421 for model validation, including 28 patients for end-user testing. Models showed an improvement over the interobserver benchmark (multi-delineation dataset; VD 0·91 [IQR 0·83-0·92], p=0·0062; SD 0·86 [0·71-0·91], p=0·0005), and were within the intraobserver benchmark. For primary validation, AI performance on internal Harvard-RT1 data (segmented by the same expert who segmented the discovery data) was VD 0·83 (IQR 0·76-0·88) and SD 0·79 (0·68-0·88), within the interobserver benchmark. Performance on internal Harvard-RT2 data segmented by other experts was VD 0·70 (0·56-0·80) and SD 0·50 (0·34-0·71). Performance on RTOG-0617 clinical trial data was VD 0·71 (0·60-0·81) and SD 0·47 (0·35-0·59), with similar results on diagnostic radiology datasets NSCLC-radiogenomics and Lung-PET-CT-Dx. Despite these geometric overlap results, models yielded target volumes with equivalent radiation dose coverage to those of experts. We also found non-significant differences between de novo expert and AI-assisted segmentations. AI assistance led to a 65% reduction in segmentation time (5·4 min; p<0·0001) and a 32% reduction in interobserver variability (SD; p=0·013). INTERPRETATION: We present a clinical validation strategy for AI models. We found that in silico geometric segmentation metrics might not correlate with clinical utility of the models. Experts' segmentation style and preference might affect model performance. FUNDING: US National Institutes of Health and EU European Research Council.

Elevated Coronary Artery Calcium Quantified by a Validated Deep Learning Model From Lung Cancer Radiotherapy Planning Scans Predicts Mortality.

PURPOSE: Coronary artery calcium (CAC) quantified on computed tomography (CT) scans is a robust predictor of atherosclerotic coronary disease; however, the feasibility and relevance of quantitating CAC from lung cancer radiotherapy planning CT scans is unknown. We used a previously validated deep learning (DL) model to assess whether CAC is a predictor of all-cause mortality and major adverse cardiac events (MACEs). METHODS: Retrospective analysis of non-contrast-enhanced radiotherapy planning CT scans from 428 patients with locally advanced lung cancer is performed. The DL-CAC algorithm was previously trained on 1,636 cardiac-gated CT scans and tested on four clinical trial cohorts. Plaques ≥ 1 cubic millimeter were measured to generate an Agatston-like DL-CAC score and grouped as DL-CAC = 0 (very low risk) and DL-CAC ≥ 1 (elevated risk). Cox and Fine and Gray regressions were adjusted for lung cancer and cardiovascular factors. RESULTS: The median follow-up was 18.1 months. The majority (61.4%) had a DL-CAC ≥ 1. There was an increased risk of all-cause mortality with DL-CAC ≥ 1 versus DL-CAC = 0 (adjusted hazard ratio, 1.51; 95% CI, 1.01 to 2.26; P = .04), with 2-year estimates of 56.2% versus 45.4%, respectively. There was a trend toward increased risk of major adverse cardiac events with DL-CAC ≥ 1 versus DL-CAC = 0 (hazard ratio, 1.80; 95% CI, 0.87 to 3.74; P = .11), with 2-year estimates of 7.3% versus 1.2%, respectively. CONCLUSION: In this proof-of-concept study, CAC was effectively measured from routinely acquired radiotherapy planning CT scans using an automated model. Elevated CAC, as predicted by the DL model, was associated with an increased risk of mortality, suggesting a potential benefit for automated cardiac risk screening before cancer therapy begins.

Dosimetric Planning Tradeoffs to Reduce Heart Dose Using Machine Learning-Guided Decision Support Software in Patients with Lung Cancer.

PURPOSE: Cardiac toxicity is a well-recognized risk after radiation therapy (RT) in patients with non-small cell lung cancer (NSCLC). However, the extent to which treatment planning optimization can reduce mean heart dose (MHD) without untoward increases in lung dose is unknown. METHODS AND MATERIALS: Retrospective analysis of RT plans from 353 consecutive patients with locally advanced NSCLC treated with intensity modulated RT (IMRT) or 3-dimensional conformal RT. Commercially available machine learning-guided clinical decision support software was used to match RT plans. A leave-one-out predictive model was used to examine lung dosimetric tradeoffs necessary to achieve a MHD reduction. RESULTS: Of all 232 patients, 91 patients (39%) had RT plan matches showing potential MHD reductions of >4 to 8 Gy without violating the upper limit of lung dose constraints (lung volume [V] receiving 20 Gy (V20 Gy) <37%, V5 Gy <70%, and mean lung dose [MLD] <20 Gy). When switching to IMRT, 75 of 103 patients (72.8%) had plan matches demonstrating improved MHD (average 2.0 Gy reduction, P < .0001) without violating lung constraints. Examining specific lung dose tradeoffs, a mean ≥3.7 Gy MHD reduction was achieved with corresponding absolute increases in lung V20 Gy, V5 Gy, and MLD of 3.3%, 5.0%, and 1.0 Gy, respectively. CONCLUSIONS: Nearly 40% of RT plans overall, and 73% when switched to IMRT, were predicted to have reductions in MHD >4 Gy with potentially clinically acceptable tradeoffs in lung dose. These observations demonstrate that decision support software for optimizing heart-lung dosimetric tradeoffs is feasible and may identify patients who might benefit most from more advanced RT technologies.

Deep-learning system to improve the quality and efficiency of volumetric heart segmentation for breast cancer.

Although artificial intelligence algorithms are often developed and applied for narrow tasks, their implementation in other medical settings could help to improve patient care. Here we assess whether a deep-learning system for volumetric heart segmentation on computed tomography (CT) scans developed in cardiovascular radiology can optimize treatment planning in radiation oncology. The system was trained using multi-center data (n = 858) with manual heart segmentations provided by cardiovascular radiologists. Validation of the system was performed in an independent real-world dataset of 5677 breast cancer patients treated with radiation therapy at the Dana-Farber/Brigham and Women's Cancer Center between 2008-2018. In a subset of 20 patients, the performance of the system was compared to eight radiation oncology experts by assessing segmentation time, agreement between experts, and accuracy with and without deep-learning assistance. To compare the performance to segmentations used in the clinic, concordance and failures (defined as Dice < 0.85) of the system were evaluated in the entire dataset. The system was successfully applied without retraining. With deep-learning assistance, segmentation time significantly decreased (4.0 min [IQR 3.1-5.0] vs. 2.0 min [IQR 1.3-3.5]; p < 0.001), and agreement increased (Dice 0.95 [IQR = 0.02]; vs. 0.97 [IQR = 0.02], p < 0.001). Expert accuracy was similar with and without deep-learning assistance (Dice 0.92 [IQR = 0.02] vs. 0.92 [IQR = 0.02]; p = 0.48), and not significantly different from deep-learning-only segmentations (Dice 0.92 [IQR = 0.02]; p ≥ 0.1). In comparison to real-world data, the system showed high concordance (Dice 0.89 [IQR = 0.06]) across 5677 patients and a significantly lower failure rate (p < 0.001). These results suggest that deep-learning algorithms can successfully be applied across medical specialties and improve clinical care beyond the original field of interest.

Knowledge-based inverse treatment planning for low-dose-rate prostate brachytherapy.

PURPOSE: Permanent low-dose-rate brachytherapy is a widely used treatment modality for managing prostate cancer. In such interventions, treatment planning can be a challenging task and requires experience and skills of the planner. We developed a novel knowledge-based (KB) optimization method based on previous treatment plans. The purpose of this method was to generate clinically acceptable plans that do not require extensive manual adjustments in clinical scenarios. METHODS: Objective functions used in current inverse planning methods are preferably based on spatial invariant dose objectives rather than spatial dose distributions. Therefore, they are prone to return suboptimal plans resulting in time consuming plan adjustments. To overcome this limitation, a KB approach is introduced. The KB model uses the dose distributions of previous clinical plans projected onto a standardized geometry. From those standardized distributions a template plan is generated. The treatment plans were optimized with an in-house developed planning system by solving a constraint inverse optimization problem that mimics the projected template dose plan constraint to DVH metrics. The method is benchmarked under an IRB-approved retrospective study by comparing optimization time, dosimetric performance, and clinical acceptability against current clinical practice. The quality of the KB model is evaluated with a Turing test. RESULTS: The KB model consists of five high-quality treatment plans. Those plans were selected by one of our experts and showed all desired dosimetric features. After generating the model treatment plans were created with one run of the optimizer for the remaining 20 patients. The optimization time including needle optimization ranged from 6 to 29 s. Based on a Wilcoxon signed rank test the new plans are dosimetrically equivalent to current clinical practice. The Turing test showed that the proposed method generates plans that are equivalent to current clinical practice and that the dose prediction drives the optimization to achieve high-quality treatment plans. CONCLUSIONS: This study demonstrated that the proposed KB model was able to capture user-specific features in isodose lines which can be used to generate acceptable treatment plans with a single run of the optimization engine in under a minute. This could potentially reduce the time in the operating room and the time a patient is under anesthesia.

Artificial intelligence in radiation oncology.

Artificial intelligence (AI) has the potential to fundamentally alter the way medicine is practised. AI platforms excel in recognizing complex patterns in medical data and provide a quantitative, rather than purely qualitative, assessment of clinical conditions. Accordingly, AI could have particularly transformative applications in radiation oncology given the multifaceted and highly technical nature of this field of medicine with a heavy reliance on digital data processing and computer software. Indeed, AI has the potential to improve the accuracy, precision, efficiency and overall quality of radiation therapy for patients with cancer. In this Perspective, we first provide a general description of AI methods, followed by a high-level overview of the radiation therapy workflow with discussion of the implications that AI is likely to have on each step of this process. Finally, we describe the challenges associated with the clinical development and implementation of AI platforms in radiation oncology and provide our perspective on how these platforms might change the roles of radiotherapy medical professionals.

Development and clinical implementation of semi-automated treatment planning including 3D printable applicator holders in complex skin brachytherapy.

PURPOSE: High-dose-rate brachytherapy (HDR-BT) is a treatment option for malignant skin diseases compared to external beam radiation therapy, HDR-BT provides improved target coverage, better organ sparing, and has comparable treatment times. This is especially true for large clinical targets with complex topologies. To standardize and improve the quality and efficacy of the treatments, a novel streamlined treatment approach in complex skin HDR-BT was developed and implemented. This approach consists of auto generated treatment plans and a 3D printable applicator holder (3D-AH). MATERIALS AND METHODS: The in-house developed planning system automatically segments computed tomography simulation images (a), optimizes a treatment plan (b), and generates a model of the 3D-AH (c). The 3D-AH is used as an immobilization device for the flexible Freiburg flap applicator used to deliver treatment. The developed, automated planning is compared against the standard clinical plan generation process for a flat 10 × 10 cm2 field, curved fields with radii of 4, 6, and 8 cm, and a representative clinical case. The quality of the 3D print is verified via an additional CT of the flap applicator latched into the holder, followed by an automated rigid registration with the original planning CT. Finally, the methodology is implemented and tested clinically under an IRB approval. RESULTS: All automatically generated plans were reviewed and accepted for clinical use. For the clinical workflow, the coverage achieved at a prescription depth for the flat 4, 6, and 8 cm applicator was (100.0 ± 4.9)%, (100.0 ± 4.9)%, (96.0 ± 0.3)%, and (98.4 ± 0.3)%, respectively. For auto planning, the coverage was (99.9 ± 0.3)%, (100.0 ± 0.2)%, (100.0 ± 0.3)%, and (100.1 ± 0.2)%. For the clinical test case, the D90 for the clinical workflow and auto planning was found to be 93.5% and 100.29% of the prescribed dose, respectively. Processing of the patient's CT to generate trajectories and positions as well as the 3D model of the applicator took <5 min. CONCLUSION: This workflow automates time intensive catheter digitizing and treatment planning. Compared to printing full applicators, the use of 3D-AH reduces the complexity of the 3D prints, the amount of the material to be used, the time of 3D printing, and amount of quality assurance required. The proposed methodology improves the overall treatment plan quality in complex HDR-BT and impact patient treatment outcomes potentially.

Combined stereotactic biopsy and stepping-source interstitial irradiation of glioblastoma multiforme.

Patients diagnosed with glioblastoma multiforme receiving stereotactic biopsy only either due to tumor localization or impaired clinical status face a devastating prognosis with very short survival times. One strategy to provide an initial cytoreductive and palliative therapy at the time of the stereotactic biopsy is interstitial irradiation through the pre-defined trajectory of the biopsy channel. We designed a novel treatment planning system and evaluated the treatment potential of a fixed-source and a stepping-source algorithm for interstitial radiosurgery on non-spherical glioblastoma in direct adjacency to risk structures. Using both setups, we show that radiation doses delivered to 100% of the gross tumor volume shifts from sub-therapeutic (10-12 Gy) to sterilizing single doses (25-30 Gy) when using the stepping source algorithm due to improved sparing of organs-at-risk. Specifically, the maximum doses at the brain stem were 100% of the PTV dose when a fixed central source and 38% when a stepping-source algorithm was used. We also demonstrated precision of intracranial target points and stability of superficial and deep trajectories using both a phantom and a body donor study. Our setup now for the first time provides a basis for a clinical proof-of-concept trial and may widen palliation options for patients with limited life expectancy that should not undergo time-consuming therapies.

A fast inverse treatment planning strategy facilitating optimized catheter selection in image-guided high-dose-rate interstitial gynecologic brachytherapy.

PURPOSE: Interstitial high-dose rate (HDR) brachytherapy is an important therapeutic strategy for the treatment of locally advanced gynecologic (GYN) cancers. The outcome of this therapy is determined by the quality of dose distribution achieved. This paper focuses on a novel yet simple heuristic for catheter selection for GYN HDR brachytherapy and their comparison against state of the art optimization strategies. The proposed technique is intended to act as a decision-supporting tool to select a favorable needle configuration. MATERIALS: The presented heuristic for catheter optimization is based on a shrinkage-type algorithm (SACO). It is compared against state of the art planning in a retrospective study of 20 patients who previously received image-guided interstitial HDR brachytherapy using a Syed Neblett template. From those plans, template orientation and position are estimated via a rigid registration of the template with the actual catheter trajectories. All potential straight trajectories intersecting the contoured clinical target volume (CTV) are considered for catheter optimization. Retrospectively generated plans and clinical plans are compared with respect to dosimetric performance and optimization time. RESULTS: All plans were generated with one single run of the optimizer lasting 0.6-97.4 s. Compared to manual optimization, SACO yields a statistically significant (P ≤ 0.05) improved target coverage while at the same time fulfilling all dosimetric constraints for organs at risk (OARs). Comparing inverse planning strategies, dosimetric evaluation for SACO and "hybrid inverse planning and optimization" (HIPO), as gold standard, shows no statistically significant difference (P > 0.05). However, SACO provides the potential to reduce the number of used catheters without compromising plan quality. CONCLUSION: The proposed heuristic for needle selection provides fast catheter selection with optimization times suited for intraoperative treatment planning. Compared to manual optimization, the proposed methodology results in fewer catheters without a clinically significant loss in plan quality. The proposed approach can be used as a decision support tool that guides the user to find the ideal number and configuration of catheters.

A fast multitarget inverse treatment planning strategy optimizing dosimetric measures for high-dose-rate (HDR) brachytherapy.

PURPOSE: In this study, we introduce a novel, fast, inverse treatment planning strategy for interstitial high-dose-rate (HDR) brachytherapy with multiple regions of interest solely based on dose-volume-histogram-related dosimetric measures (DMs). METHODS: We present a new problem formulation of the objective function that approximates the indicator variables of the standard DM optimization problem with a smooth logistic function. This problem is optimized by standard gradient-based methods. The proposed approach is then compared against state-of-the-art optimization strategies. RESULTS: All generated plans fulfilled prescribed DMs for all organs at risk. Compared to clinical practice, a statistically significant improvement (p=0.01) in coverage of target structures was achieved. Simultaneously, DMs representing high-dose regions were significantly reduced (p=0.01). The novel optimization strategies run-time was (0.8 ± 0.3) s and thus outperformed the best competing strategies of the state of the art. In addition, the novel DM-based approach was associated with a statistically significant (p=0.01) increase in the number of active dwell positions and a decrease in the maximum dwell time. CONCLUSIONS: The generated plans showed a clinically significant increase in target coverage with fewer hot spots, with an optimization time approximately three orders of magnitude shorter than manual optimization currently used in clinical practice. As optimization is solely based on DMs, intuitive, interactive, real-time treatment planning, which motivated the adoption of manual optimization in our clinic, is possible.