Abstract Brain metastases affect up to 30% of patients with metastatic cancer and are a major cause for morbidity and mortality. Treatment approaches include neurosurgery, various approaches to radiotherapy and systemic pharmacotherapy. Encouraging response rates have been observed in patients with melanoma and non-small cell lung cancer (NSCLC) with asymptomatic or oligo-symptomatic brain metastases treated with novel systemic therapies, including immune checkpoint inhibitors and targeted therapy, challenging the need for immediate radiosurgery. Eligible patients for STRIKE must have newly diagnosed and untreated asymptomatic or oligo-symptomatic brain metastases from melanoma or NSCLC, with an indication for systemic therapy. The treatment regimen consists of standard systemic treatment with (Arm A) or without (Arm B) stereotactic radiosurgery. Systemic therapy follows the current standard of care according to the primary tumor. Primary endpoint is CNS-specific progression-free survival (PFS), locally assessed according to RANO criteria. Secondary endpoints include CNS-specific PFS per tumour cohort, objective CNS response rate, duration of CNS response, pattern of CNS-specific progression, extra-CNS progression, incidence of radionecrosis and pseudoprogression, overall survival, neurocognitive function, quality of life and functional independence, and toxicity. We assume that the addition of radiosurgery to systemic treatment will increase median CNS-specific PFS by 62% from a median of 4 and 8 months for melanoma and NSCLC, respectively, corresponding to an overall hazard ratio of 0.62 for time to CNS failure. According to the log-rank test, at a 5% one-sided significance level, 143 events provide 88% power for a sample size of 180 patients. The USZ-STRIKE trial is an academic study sponsored by ETOP IBCSG Partners Foundation, with substantial funding from the USZ Foundation. The trial is being conducted in 15 centres in Switzerland, Italy, the Netherlands, Spain and the United Kingdom. Current accrual is 57 patients. he study is registered on ClinicalTrials.gov: NCT05522660.

Non-invasive prediction of glioma molecular status from routine magnetic resonance imaging (MRI) has shown promising performance, but model generalization remains challenging given small-scale matched imaging-genomic datasets. Foundation models may address this bottleneck, but a comprehensive benchmark is needed to establish the impact of diverse architectures, pre-training domains, and objectives. Given the use case of isocitrate dehydrogenase (IDH) mutation prediction from FLAIR and post-contrast T1 MRIs, we compared four image-based foundation models, BrainIAC, MRI-CORE, BiomedCLIP, and BrainDINO, against radiomics-based TabPFN and logistic regression baselines. Prediction performance and calibration were assessed across four public adult glioma cohorts and an external post-treatment cohort. Within-cohort, TabPFN matched or outperformed all visual encoders, achieving 0.92 (0.03) AUROC and 0.74 (0.17) AUPRC (mean (SD) across all datasets). Among visual encoders, BiomedCLIP performed best (0.85 (0.08) AUROC), with BrainDINO competitive (0.82 (0.09) AUROC), while MRI-specific encoders (BrainIAC, MRI-CORE) consistently underperformed. Cross-cohort transfer showed moderate AUROC degradation but stronger AUPRC sensitivity to prevalence shifts. On the external cohort, BiomedCLIP achieved the highest AUROC (0.74 (0.07)), whereas TabPFN provided superior calibration (Expected Calibration Error 0.07 (0.01)). These results indicate that representation modality and evaluation context critically influence foundation-model performance in MRI-based molecular prediction. Tabular foundation models on radiomic features provide a strong, well-calibrated baseline, while image foundation models may offer complementary value under clinically distinct distribution shifts. Code available at https://github.com/nathanhollet/idh-status-prediction

BACKGROUND AND PURPOSE:Chemoradiotherapy (CRT) followed by surgery is a treatment option for esophageal cancer (EC). However, concerns persist regarding cardiopulmonary toxicity and inconsistent daily target coverage due to anatomical changes. To address these challenges, we implemented a CBCT-based online adaptive radiotherapy (oART) workflow for EC. This study provides updated dosimetric results and evaluates toxicity and treatment response using daily oART on the Ethos™ platform for EC in the neoadjuvant setting. MATERIALS AND METHODS:We analyzed 26 EC patients treated with oART and concurrent chemotherapy, comparing dosimetric data from scheduled and adapted treatment plans. Esophagectomy was performed 8-12 weeks post-CRT, and treatment response was evaluated using Becker tumor regression grading (Grades 1a and 1b indicating pathologic major response, pMR). We documented Grade ≥3 toxicities (CTCAE v5) and analyzed associations between target volume parameters, pMR, and toxicity. RESULTS:Adapted plans improved PTVD99 % and CTVD99 % by 20.6 % (absolute: 15.6 %, p < 0.001) and 6.1 % (absolute: 5.4 %, p < 0.001), respectively, compared to scheduled plans. Mean heart dose decreased by 3.8 % (0.8 Gy, p = 0.038), and mean lung V20Gy reduced by 3.3 % (0.7 %, p = 0.049). Acute Grade ≥3 CRT-related toxicities occurred in 7 (27 %) patients, and Grade ≥3 post-operative complications in 17 (65.4 %). A pMR was achieved in 18 (69.2 %) patients. No significant correlation was found between target dose parameters and toxicity or pMR. CONCLUSION:This study confirms dosimetric improvements and high pMR rates with manageable toxicity following neoadjuvant CBCT-based oART in EC, suggesting oART could enhance treatment outcomes and support non-operative strategies for selected patients.

Purpose: Radiation therapy (RT) plays a key role in the management of esophageal cancer (EC). However, toxicities caused by proximity of organs at risk (OAR) and daily target coverage caused by interfractional anatomic changes are of concern. Daily online adaptive RT (oART) addresses these concerns and has the potential to increase OAR sparing and improve target coverage. We present the first clinical experience and dosimetric investigations of cone beam CT-based oART in EC using the ETHOS platform. Methods and Materials: Treatment fractions of the first 10 EC patients undergoing cone beam CT-based oART at our institution were retrospectively analyzed. The prescription dose was 50.4 Gy in 28 fractions. The same clinical target volume (CTV) and planning target volume (PTV) margins as for nonadaptive treatments were used. For all sessions, the timestamp of each oART workflow step, PTV size, target volume doses, mean heart dose, and lung V20Gy of both the scheduled and the adapted treatment plan were analyzed. Results: Following automatic propagation, the CTV was adapted by the physician in 164 (59%) fractions. The adapted treatment plan was selected in 276 (99%) sessions. The median time needed for an oART session was 28 minutes (range, 14.8-43.3). Compared to the scheduled plans, a significant relative reduction of 9.5% in mean heart dose (absolute, 1.6 Gy; P = .006) and 16.9% reduction in mean lung V20Gy (absolute, 2.3%; P < .001) was achieved with the adapted treatment plans. Simultaneously, we observed a significant relative improvement in D99%PTV and D99%CTV by 15.3% (P < .001) and 5.0% (P = .008), respectively, along with a significant increase in D95%PTV by 5.1% (P = .003). Conclusions: Although being resource-intensive, oART for EC is feasible in a reasonable timeframe and results in increased OAR sparing and improved target coverage, even without a reduction of margins. Further studies are planned to evaluate the potential clinical benefits.

PURPOSE:Integrating auto-contouring in radiotherapy workflows is shifting the role of radiation oncologists from manual delineation to reviewing and correcting automatically generated contours. However, we postulate that this process is hindered by significant inter-evaluator variability in assessing the dosimetric impact of contour variations. This study investigates how radiation oncologists and medical physicists evaluate the impact of glioblastoma target volume (TV) variations on the dose to organs at risk (OARs), focusing on understanding inter-evaluator variability and decision-making patterns. METHODS:A qualitative survey was conducted involving four radiation oncologists and three medical physicists. Participants classified 54 glioblastoma TV contour variations using up to four changes each across 14 patients as "better," "no change," or "worse" regarding their expected impact on the dose to OARs. The corresponding ground truth labels were derived from standardized treatment plans. Inter-evaluator variability was analyzed using Cohen's Kappa. RESULTS:Substantial variability was observed, with Cohen's Kappa values ranging from weak to moderate agreement (0.33-0.74). Evaluators frequently overestimated the negative impact of contour variations, misclassifying 46% of "no change" variations as "Worse." No evaluator judged contour variations as resulting in "better" doses to OARs, despite this being the case for 4 variations. CONCLUSION:Significant variability in estimating the dosimetric impact of contour variations underscores the critical need for standardized guidelines to reduce inconsistencies and allow for the assessment of automatically generated contours based on clinically meaningful factors. Evaluators frequently overestimated the negative impact of contour variations, potentially leading to inefficiencies and unnecessary contour corrections in clinical practice.

To decrease the recurrence rate after complete resection of a brain metastasis, removal of a surgical safety margin is advocated. This is not always feasible when resecting a metastasis in an eloquent location. We aimed to assess the recurrence rate after resection of metastases in an eloquent location followed by postoperative stereotactic radiotherapy to the resection cavity. We retrospectively included patients with 1–3 brain metastases undergoing gross total resection and postoperative stereotactic radiotherapy between 2010 and 2022. Primary endpoint was local recurrence free survival (LRFS). Secondary endpoints were overall survival and distant brain failure free survival. Patients were grouped according to the location of their metastasis into eloquent and non-eloquent. Eloquent localization was considered a surrogate for resection without a surgical safety margin according to our institutional practice. We included 193 patients with 201 resected metastases. Ninety-five metastases (47.3

BACKGROUND AND PURPOSE:Manual delineation of target volumes in glioblastoma (GBM) radiotherapy (RT) is time-consuming and variable. This study evaluates the clinical applicability of a preliminary deep learning model (Neosoma Glioma) for automating gross tumor volume (GTV) segmentation in postoperative GBM per ESTRO-EANO guidelines. MATERIALS AND METHODS:We retrospectively analyzed 100 GBM cases treated at Inselspital University Hospital, Bern (2016-2020) with standardized multi-modal MRI. Auto-segmented GTVs were compared to expert-defined contours using geometric metrics. Radiation oncologists reviewed and adjusted the best-performing configuration. Time savings, geometric similarity, and dosimetric impact were assessed. RESULTS:Optimal auto-segmentation (resection cavity plus enhancing tumor with 1 mm margin) achieved a mean Dice similarity coefficient of 0.79 (SD = 0.14) vs. ground truth. Manual adjustment took 5.9 (SD = 4.6) minutes vs. 12.3 (SD = 6.8) minutes for manual contouring (>50 % time reduction). The mean Dice between auto-segmented and adjusted GTVs was 0.84 (SD = 0.18). Dosimetric evaluation showed plans from adjusted auto-segmentations were equivalent to those based on consensus contours, with no clinically relevant differences in target coverage or organ-at-risk sparing. CONCLUSION:The Neosoma Glioma model generates clinically useful postoperative GTV segmentations, with geometric performance comparable to expert variability and dosimetric equivalence to consensus contours. It reduces contouring time by over 50%, enabling faster RT workflows. Its consistency across diverse GBM presentations supports its practical value. AI-based segmentation can help standardize GBM target definition when integrated into RT planning with proper quality assurance.

BACKGROUND:Manual contouring of organs at risk in radiotherapy is time-consuming, taking 1-4 hours per case. Automatic segmentation using deep learning has emerged as a promising solution, with many commercial options now available. However, these methods require rigorous validation before clinical use, and current evaluation approaches lack consistency and comprehensive assessment across publications. METHODS:We developed the Comprehensive Multifaceted Technical Evaluation framework, which integrates four key assessment components: quantitative geometric measures, qualitative expert evaluation, time efficiency analysis, and dosimetric evaluation. We demonstrated this framework using an in-house automatic segmentation model for brain organs at risk, trained on 100 cases and evaluated by 8 radiation oncology experts from 4 institutions. The evaluation included geometric accuracy measurements, expert ratings of clinical acceptability, time-saving assessments, and dosimetric impact analysis comparing treatment plans. RESULTS:Here we show that our automatic segmentation model achieved an overall geometric accuracy of 0.78 and outperformed manual inter-rater variability. Expert evaluation revealed that 88% of automatically segmented structures were clinically acceptable with only minor adjustments needed. The evaluation and adjustment process averaged 22 minutes compared to 69 minutes for manual contouring. Dosimetric analysis showed minimal impact on treatment plans, with average dose differences of 0.30 Gray for mean dose and 0.23 Gray for maximum dose. CONCLUSIONS:The framework provides a robust method for validating automatic segmentation models in radiotherapy. However, establishing standardized benchmarks and consensus guidelines within the radiotherapy community remains essential for proper clinical implementation and comparison of different segmentation tools.

BACKGROUND:The role and optimal timing of SRT for patients with advanced NSCLC and asymptomatic brain metastases treated with immune checkpoint inhibitors (ICI) are controversial. METHODS:Efficacy and safety outcomes of patients with newly diagnosed non oncogene- addicted NSCLC with asymptomatic brain metastases (1-10 lesions, max. diameter of lesions 3 cm) treated with a first-line ICI-containing regimen at 11 Swiss cancer centers were retrospectively analyzed. RESULTS:A total of 128 patients in two cohorts (58 patients with upfront SRT and 69 patients without upfront SRT) were included in this analysis. The median intracranial progression-free survival (PFS) was significantly longer in patients with upfront SRT (12.6 vs. 8.2 months, Hazard ratio (HR) 0.62 [95 % CI 0.41 vs. 0.95], p = 0.026). This benefit remained significant after correcting for number and size of lesions and programmed cell death 1 ligand (PD-L1) status. The proportion of patients with symptomatic progression of brain metastases and of patients receiving further local treatment to the brain was similar between cohorts (3 % vs. 12 %, p = 0.11 and 33 % vs. 42 %, p = 0.3). No significant difference in median overall survival (OS) was observed between the cohorts (22.8 vs. 21.7 months, p = 0.4). Only two patients with upfront SRT experienced a clinically significant Central Nervous System (CNS) adverse event (AE). CONCLUSION:In this multicentric retrospective analysis of patients with asymptomatic brain metastases upfront SRT was associated with an improved intracranial PFS and was well tolerated but median OS and the rate of patients developing symptomatic brain progression were similar to patients without upfront SRT.

Accurate MRI-based detection of brain metastases (BM) is essential for planning stereotactic radiosurgery (SRS). Although spin-echo (SE) sequences such as T1-SPACE have shown superior lesion detectability compared with gradient-recalled echo (GRE)–based T1-MPRAGE, direct dosimetric comparisons and evaluations of clinical impact are lacking. This study aimed to quantitatively and qualitatively compare T1-SPACE and T1-MPRAGE sequences for SRS planning, focusing on lesion detectability, target volume delineation, dosimetric effects, and oncological outcomes. Quantitative, qualitative, and dosimetric analyses were performed in 51 patients who underwent MRI with T1-SPACE and T1-MPRAGE sequences prior to SRS (SPACE group). An experienced neuroradiologist identified BM on both sequences as the reference standard. For outcome evaluation, distant brain metastasis-free survival (DBMFS) and overall survival (OS) were compared between the SPACE group and a matched control group (n = 51) planned exclusively on the T1-MPRAGE sequence. A senior resident identified significantly more BM on T1-SPACE (94.7

Background and Purpose Despite widespread adaptation of automatic segmentation (AS), manual review and adjustment of generated contours are still essential. This process is time-consuming and identifying clinically relevant corrections remains challenging. Inter-observer variability and the risk of overlooking significant errors further complicate the workflow. A dedicated quality assurance tool is highly relevant to assure quality and speed up the manual review task. The primary aim of this work is to identify critical segmentation errors while reducing unnecessary manual review, enabling efficient integration of AS into routine radiotherapy. Materials and Methods We developed an evaluation assistant that assesses contour quality through the geometric measures Dice similarity coefficient and the Hausdorff distance. This was combined with a dose prediction model to determine the clinical relevance. The system was validated on 30 glioblastoma cases with ground truth and manually modified organ at risk (OAR) contours. A traffic light decision matrix classified contours based on geometric and dose parameters, flagging structures for human review. Results Out of 507 analyzed OARs, 180 were classified as critical. Our approach identified 173 of these critical structures (sensitivity: 0.96, specificity: 0.55). The system flagged 317 organs (61%) as critical, effectively ruling out 39% as non-critical with only 7 false negatives comprising structures. Conclusions Our dual-layer QA approach effectively identifies critical OAR segmentations with high sensitivity and acceptable specificity, potentially reducing manual review requirements significantly. By focusing on clinically relevant dose/volume metric endpoints, this method assures the quality of brain AS results in clinical radiotherapy practice.