Yu, Nathan Y.

Abstract Background While the radiation dose to left anterior descending (LAD) artery is associated with overall survival (OS) in NSCLC patients, manual segmentation is labor‐intensive. Several auto‐segmentation models were developed on average 4DCT with promising geometric agreement, yet their feasibility in clinical outcome analysis on lung CT remains unclear. Purpose This study evaluates whether automated LAD segmentation can reproduce, at the group level, the established dose‐survival association in a multi‐institutional clinical trial dataset. Methods Three deep learning‐based auto‐segmentation (DLAS) models for automatic segmentation of the LAD coronary artery were applied to 460 patients from the NRG Oncology/RTOG 0617 dataset. Kaplan–Meier curves and Cox proportional hazard ratio (HR) were calculated using each model's segmentation. The LAD V15 Gy was used to stratify patients into high‐and low‐risk groups, and the cut‐off values were determined by minimizing mean absolute error (MAE) between DLAS‐derived and manual OS curves. Results All three models demonstrated significantly different OS curves between the high risk and low risk groups ( p < 0.05). The HR was 1.40 (95% CI: 1.02–1.93; p = 0.0381; MAE = 7.1%), 1.32 (95% CI: 1.03–1.68; p = 0.0258; MAE = 9.6%), and 1.37 (95% CI: 1.03–1.84; p = 0.0331; MAE = 7.9%) for the three models using a common V15 Gy ≥ 10% cut‐off. After optimizing the cut‐off, one model aligned well with what was derived from manual segmentation (MAE = 4.3%). Conclusion Among the evaluated models, one DL model showed the strongest performance of survival groups separation and achieved group‐level overall survival stratification comparable to that previously reported using manual contours in the multi‐institutional cohort. Further validation with paired manual contours and pre‐specified cutoffs is warranted.

Abstract Background The left anterior descending (LAD) coronary artery is a critical organ‐at‐risk in thoracic radiotherapy. In thoracic stereotactic body radiation therapy (SBRT), sharp dose gradients and poor LAD contrast on 4DCT scans introduce uncertainty in LAD dose assessment. Methods We evaluated the performance of an AI‐generated LAD region structure as a surrogate for manual LAD contours in early‐stage NSCLC. We used deep inspiration breath hold (DIBH) scans from 49 patients with manually delineated LADs; planning risk volume contours were expanded by 3 mm. Then, 96 lung SBRT dose maps were registered by matching the left anterior portion of the heart contours. Dice similarity coefficient (DSC), inclusion ratios (IR), dose–volume histogram (DVH) metrics, linear regression, and linear mixed models were used to compare the structures. The correlation of the AI LAD region and LAD dose to mean heart dose (MHD) was assessed. Results The AI LAD region showed strong correlation with manual LAD contours for maximum dose (D1%, R 2 = 0.93) and volume receiving ≥10 Gy (V10Gy, R 2 = 0.81). The planning risk volume also demonstrated high correlation to the AI LAD region. MHD was a poor predictor of LAD dose ( R 2 ≈0.4). Additional dose metric correlations were assessed. Discussion To investigate dose volume effects to the LAD during radiotherapy for stage I NSCLC, a robust segmentation paradigm must be established. It must be easily implementable on typical imaging acquired during treatment, with an emphasis on clinical translatability. Our model has shown that the AI LAD region has a strong correlation with LAD dose under typical lung SBRT doses and supports its use for further investigation into LAD dose volume constraints during lung SBRT. Conclusion AI LAD region provides a clinically feasible and scalable alternative to manual LAD contouring for dose estimation in lung SBRT. Its strong correlation with key dose metrics validates its use in large‐scale retrospective and prospective studies.

Abstract Large Language Models (LLMs) demonstrate strong capabilities in healthcare applications. However, recruiting human evaluators to assess usability and effectiveness becomes challenging in the large-scale LLM-generated outcomes. Existing metrics are often outdated for evaluating LLM-based healthcare technologies. This paper proposes a Research through Evaluation (RtE) approach that can refine metrics and improve assessments of LLM-generated outputs. We deploy RtE in a retrospective comparative study of prostate cancer patient inquiries, comparing responses from human clinical care teams with those generated by a GPT-4-based in-basket bot. Through three rounds of co-evaluations with clinical professionals and an LLM-as-graders study with four LLMs, the RtE method shows that starting with small, iterative evaluations helps human graders adjust metrics to suit their daily practices better. Additionally, in the LLM-as-graders study, LLMs exhibit similar behavior, preferring the in-basket bot's responses over the clinical care team. Our methodology and findings demonstrate that the RtE method can enhance the transparency and robustness of evaluation metrics for LLM-generated content, with the potential to generalize to evaluations in other clinical practices.

Abstract Background The van Herk margin formula, derived for 3‐D conformal radiotherapy with a uniform 3.2 mm dose penumbra and no intrinsic dose buffer, remains widely used to design CTV‐to‐PTV expansions in contemporary image‐guided prostate radiotherapy. However, coplanar VMAT techniques often feature broader penumbrae and explicit PTV–isodose clearances, potentially violating the assumptions underlying the original formulation and leading to overly conservative margins. Purpose We investigated whether the widely adopted van Herk margin formula overestimates clinical target volume (CTV)‐to‐planning target volume (PTV) expansions for contemporary coplanar volumetric modulated arc therapy (VMAT) prostate treatments. Methods Fifty consecutive intact‐prostate VMAT radiotherapy plans (two coplanar arcs; clinical margins 3 mm, except for 2 mm posterior) were exported. Direction‐specific 90% isodose–to‐PTV gaps and penumbra widths were measured. Candidate anisotropic margins were tested by eroding the PTV to create CTV_eval. Monte‐Carlo simulations combined systematic (Σ) shifts with Gaussian random (σ′) blurring kernels of 0–2 mm were performed. Acceptability criteria of (i) CTV Dmin0.03 cc ≥ 90% Rx in ≥ 90% of simulated scenarios or (ii) population tumor‐control probability (TCP) loss < 1 % were used. Results VMAT plans exhibited intrinsic 90% isodose clearances of 3–5 mm laterally/anteriorly and 1–2 mm superior–inferiorly, while axial‐plane penumbras were up to fivefold broader than van Herk's assumption. With a [0, 0, 2] mm (LR/AP/SI) margin, ≥ 90% of patients maintained Dmin0.03 cc ≥ 90% Rx provided Σ lay within an ellipsoid of [2.0, 1.5, 1.8] mm and σ′ ≤ [1.5, 2.0, 1.5] mm. Under TCP criteria the safe Σ ellipsoid for high‐risk disease was [2.5, 1.9, 2.2] mm, while low‐intermediate risk was even less sensitive. Conclusions For image‐guided coplanar VMAT prostate radiotherapy, an anisotropic [0, 0, 2] mm CTV‐to‐PTV margin is sufficient for target coverage. A modified margin expression that subtracts the measured isodose–to‐PTV gap and uses reduced random‐error coefficients better reflects modern practice.

Abstract Purpose Lattice radiation therapy (LRT) is a form of spatially fractionated radiation therapy that allows increased total dose delivery aiming for improved treatment response without an increase in toxicities, commonly utilized for palliation of bulky tumors. The LRT treatment planning process is complex, while eligible patients often have an urgent need for expedited treatment start. In this study, we aimed to develop a simulation‐free workflow for volumetric modulated arc therapy (VMAT)‐based LRT planning via deep learning‐predicted synthetic CT (sCT) to expedite treatment initiation. Methods Two deep learning models were initially trained using 3D U‐Net architecture to generate sCT from diagnostic CTs (dCT) of the thoracic and abdomen regions using a training dataset of 50 patients. The models were then tested on an independent dataset of 15 patients using image similarity analysis assessing mean absolute error (MAE) and structural similarity index measure (SSIM) as metrics. VMAT‐based LRT plans were generated based on sCT and recalculated on the planning CT (pCT) for dosimetric accuracy comparison. Differences in dose volume histogram (DVH) metrics between pCT and sCT plans were assessed using the Wilcoxon signed‐rank test. Results The final sCT prediction model demonstrated high image similarity to pCT, with a MAE and SSIM of 38.93 ± 14.79 Hounsfield Units (HU) and 0.92 ± 0.05 for the thoracic region, and 73.60 ± 22.90 HU and 0.90 ± 0.03 for the abdominal region, respectively. There were no statistically significant differences between sCT and pCT plans in terms of organ‐at‐risk and target volume DVH parameters, including maximum dose (Dmax), mean dose (Dmean), dose delivered to 90% (D90%) and 50% (D50%) of target volume, except for minimum dose (Dmin) and (D10%). Conclusion With demonstrated high image similarity and adequate dose agreement between sCT and pCT, our study is a proof‐of‐concept for using deep learning predicted sCT for a simulation‐free treatment planning workflow for VMAT‐based LRT.

Objectives: Identifying patients’ advantageous radiotherapy modalities prior to CT simulation is challenging. This study aimed to develop a workflow using deep learning (DL)-predicted synthetic CT (sCT) for treatment modality comparison based solely on a diagnostic CT (dCT). Methods: A DL network, U-Net, was trained utilizing 46 thoracic cases from a public database to generate sCT images predicting planning CT (pCT) scans based on the latest dCT, and tested on 15 institutional patients. The sCT accuracy was evaluated against the corresponding pCT and a commercial algorithm deformed CT (MdCT) based on Mean Absolute Error (MAE) and Universal Quality Index (UQI). To determine advantageous treatment modality, clinical dose-volume histogram (DVH) metrics and Normal Tissue Complication Probability (NTCP) differences between proton and photon treatment plans were analyzed on the sCTs via concordance correlation coefficient (CCC). Results: The AI-generated sCTs closely resembled those of the commercial deformation algorithm in the tested cases. The differences in MAE and UQI values between the sCT-vs-pCT and MdCT-vs-pCT were 19.38 HU and 0.06, respectively. The mean absolute NTCP deviation between sCT and pCT was 1.54%, 0.21%, and 2.36% for esophagus perforation, lung pneumonitis, and heart pericarditis, respectively. The CCC between sCT and pCT was 0.90 for DVH metrics and 0.97 for NTCP, indicating moderate agreement for DVH metrics and substantial agreement. Conclusions: Radiation oncologists can potentially utilize this personalized sCT based approach as a clinical support tool to rapidly compare the treatment modality benefit during patient consultation and facilitate in-depth discussion on potential toxicities at a patient-specific level.

Background/Objectives: Pneumonitis caused by radiotherapy for lung cancer may be missed since it often occurs only several months later. In a previous trial including patients of any age, a scoring system was tested to facilitate the correct diagnosis of radiation pneumonitis. Since elderly lung cancer patients have a greater risk of developing this complication, a separate scoring system for this group appears useful. Our prospective multi-center trial (NCT06480734) investigates a specific tool for elderly patients irradiated for lung cancer. Methods: Patients aged ≥65 years with lung cancer will complete paper-based questionnaires and rate symptoms potentially caused by pneumonitis weekly during and up to 24 weeks following radiotherapy. The total score of this symptom-based scoring system ranging from 0 to 9 points is correlated to pneumonitis. The discriminative power of the scoring system is evaluated by calculating the area under the receiver operating characteristic curve. Optimality is defined as a cut-off score with sensitivity ≥90% and specificity ≥80%. Moreover, the Youden index will be applied. Fifty-nine patients are required for the full analysis set. Assuming 5% will not qualify for this set, 65 patients should be enrolled. Moreover, patient satisfaction with the scoring system is evaluated. If the dissatisfaction rate is >20%, the system needs modifications; if the dissatisfaction rate is >40%, it is considered not useful. An optimal cut-off score facilitating the diagnosis of pneumonitis and its discrimination from other lung diseases will contribute to a corresponding mobile application to be used by elderly lung cancer patients at home.

Abstract Background Inter‐fractional and intra‐fractional positioning and motion monitoring based on implanted fiducials through kilovoltage (kV) imaging is a cost‐effective approach to enhance intact prostate radiotherapy treatment accuracy. However, comprehensive studies for three‐dimensional (3D) position extraction and the corrective action threshold are lacking. Purpose To develop and verify a fiducial motion monitoring system with 3D position information based on triggered kV images taken during treatment delivery, and to comprehensively evaluate the recommended thresholds for corrective action during treatment. Methods An in‐house fiducial triangulation algorithm was developed to monitor fiducial positions using subsequent kV triggered images. The precision of fiducial triangulation and motion detection was validated on a pelvis phantom with four fiducial inserts. A retrospective study was conducted on prostate cancer patients who received either moderately hypofractionated radiotherapy (MHRT, n = 16) or stereotactic body radiotherapy (SBRT, n = 12), categorized by endo‐rectal balloon (ERB) use. Intra‐fractional positions relative to isocenter were computed, and these positions were analyzed in comparison to the original marker locations of the patient's treatment plan. A linear regression fit was used per fraction to determine correlation coefficients between motions in the left–right (LR), superior–inferior (SI), and anterior–posterior (AP) directions. Average fraction time (AFT) of treatment was reported based on the observed average time spent on steps in the workflow and the percentage of fiducials caught outside of the current two‐dimensional (2D) tolerance threshold of 5 mm. The observed 3D offsets were used to estimate AFT for various 3D tolerance thresholds. Results Overall, our triangulation method proved to be very precise for static cases, where phantom measurements revealed a fiducial position triangulation precision of mm for stationary targets, but had a spread of mm for targets with 1 mm of motion. Employing triangulation, true motion exceeding 5 mm was detected above 4.0 mm in magnitude 90.9% of the time, a noticeable improvement in comparison to the 28%–65% successful detection rate reported by 2D projection methods. Most detection errors were attributed to depth disparity. Similar to other reports, correlation coefficients for intra‐fractional motion generally indicated no LR/SI, no LR/AP, and weak positive SI/AP correlations for MHRT and SBRT patients. Offsets beyond a 3D tolerance threshold of 5 mm were observed at a rate of 4.25%–5.25%, while the 5 mm 2D out‐of‐tolerance catch‐rate was 1.6%. The AFT was 8.1–8.2 min using the 2D tolerance threshold of 5 mm. In comparison, the estimated AFT for the proposed 3D tolerance monitoring of offsets beyond 2–5 mm was slightly higher at 8.2–10.9 min due to the higher amount of out‐of‐tolerance instances for the higher precision intra‐fractional motion management. Conclusions Our study showcases a promising subsequent kV‐based triangulation method for intra‐fractional prostate motion monitoring. Acquiring 3D motion information results in higher out‐of‐tolerance catch‐rates particularly in the depth dimension of the kV images, which is perpendicular to the treatment beam. Failure to properly observe and catch these offsets would result in sub‐optimal conformality and accuracy of the dose delivery.

745 Background: The combination of EV-P therapy has recently been approved as first-line (1L) therapy for patients with LA/mUC based on the results of the EV-302 trial. Herein, we report the updated results from a cohort of patients with LA/mUC who received EV-P in real world clinical setting. Methods: This retrospective study included patients with LA/mUC (07/2022-08/2024) at the Mayo Clinic who completed at least one cycle of EV-P. The best overall response (BOR) was evaluated using radiographic imaging and was adjudicated as complete response (CR), partial response (PR), stable disease (SD), mixed response (MR), or progressive disease (PD). The median progression-free survival (PFS), defined as the time from treatment initiation to disease progression or death, was estimated using the Kaplan-Meier method. Patients who had not experienced disease progression by their last follow-up were censored for the PFS analysis. Results: A total of 120 patients were included; 79 (65.8%) were males, 41 (34.1%) females, 109 (91%) White, and 119 (99%) were non-Hispanic/Latino. The median age was 72 years (IQR: 65-77). Of these, only 22 (18.3%) had locally advanced disease while the remaining had mUC; 62 (51.6%) had lower tract urothelial carcinoma (LUTC), 36 (30%) had upper tract urothelial carcinoma (UTUC). The median follow-up time was 7.1 months (IQR: 4.93-9.26). In terms of BOR, 56 patients (46.6%) experienced PR, 34 (28.3%) CR, 12 (10%) PD, 7 (5.8%) MR, and 8 (6.6%) experienced SD. 3 (2.5%) patients were not evaluable. The BOR across subgroups, LA and mUC, including LUTC and UTUC, is summarized in the table. Among the responders, the median number of treatment cycles to first response was 3 (IQR: 2.25–4). Within the follow-up, 28 patients (23.3%) experienced eventual radiographic disease progression; death was observed in 25 patients (20.8%). The median PFS was 12.7 months (95% CI: 9.8-NE), and OS was 25.1 months (16.3-NE). Conclusions: In real world setting, the combination of EV-P in LA/mUC demonstrated clinical efficacy compatible with the findings from the EV-302 trial. Our results showed comparable efficacy with EV-P in both UTUC and LUTC, highlighting it's broad therapeutic potential. However, clinical data at longer follow-up is required to fully evaluate overall survival and the durability of response. mUC LA Overall LUTC UTUC Total patients 98 62 36 22 BOR CR 23 (23.4%) 13 (20.9%) 10 (28.5%) 11 (50%) PR 49 (52.6%) 31 (50%) 18 (50%) 7 (31.8%) SD 6 (6.4%) 2 (3.2%) 4 (11.4%) 2 (9.0%) PD 10 (10.7%) 9 (15.5%) 1 (2.8%) 2 (9.0%) MR 7 (7.5%) 5 (8.6%) 2 (5.7%) 0 (0%)

136 Background: The ECOG performance status (PS) is a critical measure for evaluating functional capacity and determining clinical trial eligibility in patients with advanced prostate cancer. We aim to assess the capability of large language models (LLMs) to accurately infer ECOG PS from unstructured oncology notes. Methods: This retrospective study included patients with metastatic castration-resistant prostate cancer (mCRPC) receiving Lutetium-177–PSMA-617 ( 177 Lu) at the Mayo Clinic (2022-2023). Baseline ECOG PS scores were manually curated by a trained clinician from clinical notes recorded within four weeks prior to 177 Lu initiation. Structured zero-shot prompts were then used to prompt GPT-4 to infer ECOG PS scores from unstructured oncology notes. Total dataset was randomly sampled to prompt development (20%) and held-out test set (80%). Three experimental setups were employed to assess model performance: (i) utilizing unmasked clinical notes, (ii) clinical notes with only numerical ECOG PS scores masked while retaining narrative descriptive/assessment by oncologist of ECOG score, and (iii) clinical notes with both numerical scores and narrative ECOG descriptors masked. Final performance was assessed using evaluation metrics (precision, recall, F1 score) on the held-out test set. Results: A total of 240 patients (n: 50 prompt development; n: 190 test set) were included in the analysis. The median age was 70 (IQR: 66-76); a majority of patients were White (n: 229; 95%) and non-Hispanic (n: 233; 97%). The most prevalent ECOG PS score was 0 (n: 140; 58%), followed by 1 (n: 88; 37%), and 2 (n: 12; 5%). The performance was numerically higher when using unmasked notes (weighted average precision: 0.89; recall: 0.89; F1: 0.89), compared to using notes with masked scores only (weighted average precision: 0.76; recall: 0.79; F1: 0.75) and notes with both scores and narrative masked (weighted average precision: 0.57; recall: 0.53; F1: 0.48). Metrics by each ECOG PS score is outlined (Table). Conclusions: Large language models demonstrate significant potential in precisely extracting ECOG performance status from unstructured oncology notes, offering valuable support for treatment decision-making and clinical trial enrollment. However, model performance declines as reliance on implicit reporting within clinical notes increases. Future efforts focusing on prediction of ECOG using a series of notes over time will be the next step. ECOG Unmasked oncology notes ECOG Score masked only ECOG Score + ECOG narrative marked Precision Recall F1 Score Precision Recall F1 Score Precision Recall F1 Score 0 0.89 0.92 0.91 0.75 0.93 0.83 0.59 0.74 0.66 1 0.88 0.86 0.87 0.76 0.74 0.75 0.19 0.47 0.27 2 1.00 0.82 0.90 0.78 0.33 0.47 0.78 0.16 0.26