BACKGROUND:Brain metastases (BMs) are manually contoured during stereotactic radiosurgery (SRS) treatment planning, which is both time-consuming and potentially inconsistent. To address these challenges, researchers have been actively developing deep learning-based approaches for the detection and segmentation of BMs. However, a comprehensive comparative analysis of deep learning models across different frameworks remains largely absent in the current literature. This study aimed to evaluate and compare deep learning models based on different frameworks for the detection and segmentation of BMs in T1-contrast MRI. MATERIALS AND METHODS:Eight deep learning models, based on CNN, Transformer, or Mamba architectures, were trained and validated for the task of detecting and segmenting brain metastatic lesions in T1-contrast MRI. A total of 934 patients were included, with 667 cases from publicly available datasets and 267 cases from our institution, designated for training and testing, respectively. Data were retrospectively collected and organized at our institution, and GTV defined as the total BM tumor volume delineated by the physician at the time of stereotactic radiosurgery (SRS). Additionally, labels in the publicly available dataset were modified under clinician guidance to create a BM GTV that met clinical criteria to improve ground-truth accuracy. A BM was considered detected if the ground-truth contour overlapped with a predicted structure. Sensitivity at both the patient-level (proportion of patients with at least one lesion detected) and lesion-level (proportion of ground-truth lesions detected) were used to evaluate BM detection. Segmentation performance was assessed using several metrics: dice similarity coefficient (DSC), positive predictive value (PPV), surface DSC (sDSC), and Hausdorff distance 95% (HD95). The performance across different BM diameters was also evaluated. RESULTS:Among the eight deep learning models, the U-Mamba (Bot) achieved a lesion-level sensitivity of 0.796 (95% CI: 0.779-0.812) for all sizes of BM, which was significantly higher than that of the other models, with a false positive rate of 2.46 ± 4.96 per patient. Further stratification by metastasis diameter, the sensitivity was 0.505 for BMs < 3 mm, 0.797 for BMs between 3 and 6 mm, and 0.885 for BMs between 6 and 9 mm. Moreover, U-Mamba (Enc) demonstrated significantly higher lesion-level segmentation performance, with DSC value of 0.632 ± 0.224. In terms of tumor boundary segmentation, nnU-Netv2 achieved the best performance, with Surface DSC and HD95 values of 0.877 ± 0.149 and 1.770 ± 1.458 mm. CONCLUSION:The nnU-Netv2 allows precise segmentation of lesion areas in T1-contrast MRI, while U-Mamba provide effective detection of brain metastasis, potentially aiding in treatment planning for SRS.

Abstract Purpose: Although radiotherapy (RT) is a predominant treatment for non-small cell lung cancer (NSCLC), radioresistance remains a major challenge and is strongly linked to dysregulated tumor immune microenvironments (TIME). Immunoradiotherapy has shown improved therapeutic activity by promoting antitumor immune responses and modifying the TIME. Recent studies positions STING activation as a potent catalyst of this immune reprogramming. Therefore, identifying upstream regulators capable of enhancing STING-mediated immune activation is critical for developing strategies to overcome radioresistance in NSCLC. Methods: Ubiquitin-specific protease3 (USP3) expression patterns, immune infiltration profiles, and pathway enrichment analyses were performed using public datasets and institutional sequencing data. Prognostic value was evaluated via TCGA datasets and validated in a clinical cohort of 105 NSCLC patients (Tianjin Medical University Cancer Institute & Hospital, 2013-2015). Subsequently, we investigated the interaction between USP3 and STING and the effects of USP3 on RT-induced STING/type I interferon (IFN-Is) signaling through co-immunoprecipitation and immunofluorescence. Finally, we evaluated the effects of USP3 on RT-induced T cell infiltration in tumor and sensitivity to RT and radioimmunotherapy. Results: USP3 expression was significantly downregulated in NSCLC specimens compared to normal tissues. High USP3 expression correlated with prolonged overall survival in public datasets and our independent clinical cohort. Functionally, USP3 overexpression enhanced the radiosensitivity of NSCLC cells and induced a pro-inflammatory phenotype. Mechanistically, USP3 physically interacted with STING, removed K48-linked polyubiquitin chains, and thereby prevented proteasomal degradation of STING. This stabilization amplified cGAS-STING signaling and downstream IFN-Is production. In vivo, USP3 overexpression remodeled the post-irradiation tumor immune microenvironment, as evidenced by increased infiltration of effector CD8+ T cells and enhanced systemic antitumor immunity. Conclusion: USP3 deubiquitinates STING and blocks its proteasomal degradation, thereby stabilizing the STING protein, potentiating the cGAS-STING axis and type I interferon signaling. These results highlight USP3 as a promising prognostic biomarker and a potential therapeutic target for overcoming radioresistance and improving radiotherapy-based treatment strategies. Citation Format: Zeyuan Cheng, Zhengkun Cai, Yihan Xu, jiazhuo yan, Zhiqiang Wu, Zhiyong Yuan. USP3 stabilizes STING to enhance radiosensitivity and antitumor immunity of NSCLC [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6605.

Background: Reliably predicting brain metastasis (BM) and brain metastasis‑free survival (BMFS) in patients with primary small cell carcinoma of the esophagus (PSCCE) remains a clinical challenge. Objectives: To develop and externally validate predictive nomograms for assessing BM probability and BMFS in patients with PSCCE. Design: A retrospective model development and external validation study. Methods: Using a training cohort, we constructed two separate nomograms. The first was developed using univariable and multivariable logistic regression to predict the probability of BM. The second was built using a Fine‑Gray competing-risk model to estimate BMFS, treating death without BM as a competing event. The performance of both nomograms was evaluated using the area under the receiver operating characteristic (ROC) curve/time-dependent ROC curve, calibration plots, and decision curve analysis (DCA). External validation was subsequently performed using an independent validation cohort. Results: The study sample included 492 patients in the training cohort (mean age at diagnosis 61.97 ± 8.86 years; 133 (27%) female) and 344 patients in the external validation cohort (mean age at diagnosis 62.97 ± 8.14 years; 108 (31%) female). Age, N stage, and M stage emerged as independent predictors and were included in a nomogram to estimate BM risk. Adding lesion length and initial treatments (radiotherapy, chemotherapy, and surgery) to these three factors allowed the model to predict BMFS. Both nomograms showed strong predictive performance in the training cohort, with area under the ROC curve values of 0.76 and 0.843, respectively, supported by calibration plots and DCA. These results were confirmed in the external validation cohort. The logistic regression‑based nomogram accurately predicted BM probability, while the Fine–Gray model provided reliable estimates of BMFS. Conclusion: We successfully developed and externally validated two complementary nomograms that reliably predict BM probability and BMFS in patients with PSCCE. These tools may aid in risk stratification and inform personalized surveillance strategies.

The role of cellular senescence as a tumor-associated hallmark and therapeutic target in influencing radiotherapy or radioimmunotherapy outcomes remains incompletely understood. To address this, we developed a senescence-associated gene signature to predict prognosis in non‑small cell lung cancer (NSCLC) patients treated with radiotherapy. A risk model based on nine genes—including COL4A1 and CSF1—effectively stratified patients into high‑ and low‑risk groups. High‑risk patients exhibited significantly poorer overall survival and a tumor microenvironment characterized by reduced immune infiltration and an immune‑excluded phenotype. CSF1 was identified as a pivotal gene within this signature. Mechanistically, radiotherapy induces the expansion of a CSF1‑high‑expressing exhausted T‑cell population, which exhibits characteristics of both exhausted and senescent T cells and forms a positive feedback loop with M2‑like macrophages, thereby reinforcing the immunosuppressive microenvironment. Preclinical studies demonstrated that combining a CSF1‑neutralizing antibody with radiotherapy and anti‑PD‑1 therapy effectively reduced exhausted T cells and M2 macrophages, leading to a significant enhancement in therapeutic efficacy. Taken together, our work establishes a nine‑gene signature for risk stratification in NSCLC and provides proof‑of‑concept that targeting CSF1 can potentiate radioimmunotherapy, offering a novel translational strategy for improving patient survival.

Abstract Purpose: This study aimed to compare survival outcomes between surgery and stereotactic body radiation therapy (SBRT) for early-stage non-small cell lung cancer (ES-NSCLC) with interlobar pleural involvement (ILPI), and to further identify patients who benefit more from SBRT. Methods: This retrospective study analyzed 573 ES-NSCLC patients with ILPI, treated with surgery (n=379) or SBRT (n=194) across four centers. Propensity score matching (PSM) was employed to minimize confounding factors between treatment groups. Survival outcomes, including cancer-specific survival (CSS), disease-free survival (DFS) were estimated using the Kaplan-Meier method and compared with the log-rank test. Subgroup analyses utilized Cox proportional hazards models. We also conducted a comparative analysis of treatment-related adverse events (TRAEs) and assessed health-related quality of life (HRQoL) using the EORTC QLQ-C30. Results: In the entire cohort, 3-year CSS was comparable between surgery and SBRT (HR, 1.48, 95% CI, 1.0-2.3; P = 0.075). In the PSM-matched cohort, CSS and DFS showed no significant differences (CSS: HR, 1.02, 95% CI, 0.6-1.5; P = 0.822; DFS: HR, 1.38; 95% CI, 0.8-2.5; P = 0.374). Multivariate Cox analysis identified several independent prognostic factors for CSS, including IPI features (site: HR, 0.48, 95% CI, 0.3-0.8; P = 0.009; location: HR, 2.05, 95% CI, 1.2−3.6; P = 0.014; type: HR, 8.00, 95% CI, 1.8−35.1; P = 0.006) and patient characteristics (age: HR, 1.88, 95% CI, 1.1−3.3; P =0.027; smoking history: HR, 2.28 95% CI, 1.4−3.7; P = 0.001). Crucially, in the subgroup of patients over 70 years of age, SBRT was associated with significantly superior CSS compared to surgery (HR, 0.37, 95% CI, 0.2-0.8; P = 0.007). Specifically, CSS was improved in SBRT patients over 70 years with ILPI characterized by: Abutment-type ILPI (HR, 0.36, 95% CI, 0.1-1.0; P = 0.045), left lung oblique fissure (OF) involvement (HR, 0.21, 95% CI, 0.1-0.7; P = 0.011), or central ILPI location (HR, 0.10, 95% CI, 0.0-0.6; P = 0.012). Conversely, surgery showed better DFS in patients < 70 years and those with peripheral ILPI or right OF involvement. Regarding TRAEs, surgery was predominantly associated with pain (28.5%), while the SBRT cohort experienced fatigue (22.7%). The HRQoL showed SBRT was associated with significantly better preservation of emotional function and avoided the worsening of pain observed after surgery, which showed greater improvements in role and social function. Conclusions: SBRT offers survival outcomes comparable to surgery for ES-NSCLC with ILPI without increasing the burden of TRAEs or compromising HRQoL. Patient age and ILPI characteristics are critical factors in guiding treatment selection. SBRT is particularly beneficial for patients over 70 years, especially those presenting with abutment-type ILPI, left OF involvement, or central ILPI. Citation Format: Zhengkun Cai, Zhiyong Yuan, Yue Wang. Candidate identification for SBRT versus surgery in early-stage NSCLC with interlobar pleural involvement: A multicenter retrospective study [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 7364.

Abstract In this study, we aim to investigate the therapeutic potential of in vivo genetic therapy using AAV-mediated adenine base editor (ABE) delivery for non-small cell lung cancer (NSCLC) carrying the STK11 nonsense mutation STK11Q37* (c.109C>T), and further explore the molecular mechanisms underlying its impact on radiosensitivity. Radiotherapy is a standard treatment for locally advanced or inoperable NSCLC patients, radiation resistance driven by tumor-promoting somatic mutations severely limits clinical efficacy, underscoring the urgent need for targeted strategies such as precise gene correction to overcome this barrier. Through CRISPR-based in vivo mutation library screening combined with whole-exome sequencing, we identified STK11Q37* as a critical driver of radiation resistance. To restore STK11 function, we engineered a panel of ABEs with distinct PAM/TAM compatibilities and further developed a high-fidelity variant, A8E-N108Q-R26G, which enables precise correction of the pathogenic adenine without bystander editing. Using HEK293T reporter cells stably harboring the STK11Q37* locus, we found that spCas9-A8EQR (N108Q-R26G) achieved the highest overall correction efficiency, repairing up to 60% of mutant alleles with superior fidelity. The optimized editor was subsequently packaged into a dual-AAV system and delivered in vivo to humanized mice bearing subcutaneous patient-derived organoid xenografts. AAV-mediated base editing combined with radiotherapy produced synergistic antitumor effects and significantly suppressed tumor progression. Mechanistic studies revealed that precise correction of STK11Q37* restores endogenous LKB1 protein expression and kinase activity, reactivating downstream signaling required for maintaining redox homeostasis. Restored LKB1 directly stabilizes the transcriptional regulator BACH1 by limiting its ubiquitination and proteasomal degradation, thereby preserving BACH1-mediated repression of antioxidant gene programs. Concurrently, LKB1 re-expression dampens NRF2 nuclear accumulation and transcriptional activity, leading to reduced expression of NRF2-driven detoxification and antioxidant pathways. This coordinated regulation markedly elevates intracellular ROS levels following irradiation and reinstates radiation-induced cytotoxic stress, collectively enhancing tumor radiosensitivity. In summary, our study identifies STK11Q37* as a therapeutically actionable driver of radiation resistance and demonstrates that precise ABE-mediated correction provides a promising gene-editing-based strategy to improve treatment outcomes in NSCLC. Citation Format: Jiazhuo Yan, Yihan Xu, Qingxiao Fang, Jinpu Yu, Zhiyong Yuan. In vivo adenine base editing of STK11 Q37* reprograms tumor for radiosensitizing effect [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 278.

Cherenkov imaging provides real-time video of beam incidence upon the patient, for verification of safe and accurate radiotherapy delivery. However, the optical signal is inherently weak and is affected by non-optical radiation leakage and stray x-ray noise from the medical linear accelerator (Linac). This frequently leads to low signal-to-noise ratio (SNR) frames with background clutter, limiting video image clarity and beam visualization. To address this challenge, a wavelet-based deep video denoising method was proposed. The method was validated with three regular square fields and clinical data from two volumetric modulated arc therapy (VMAT) fractions administered to breast cancer patients. Image quality was assessed using the global gamma pass rate ( γ pass ) with 3%/3 mm criteria. The decision-making process of the network was visualized for interpretability. Results show that Cherenkov frames accumulated over five Linac pulses achieved γ pass of 96–97% for all square beams. For clinical VMAT cases, accumulated frames from selected control points reached γ pass exceeding 95%. We believe this to be the first demonstration of a deep video denoising framework sufficiently fast for real-time Cherenkov imaging.

Adding a Planning Target Volume (PTV) margin remains a straightforward and effective strategy to ensure adequate target coverage under various uncertainties. Appropriately reduced margins can minimize treatment-related toxicity without compromising tumor control. However, respiratory motion introduces complex interactions with Internal Target Volume (ITV) management, making conventional PTV recipes less reliable. This study aims to refine the definition of PTV margin for lung stereotactic body radiotherapy (SBRT) using a moving target dose (MTD) model to account for respiratory motion uncertainty under free breathing, and to compare it with alternative margin calculation methods. Data from 31 patients with non-small-cell lung cancer (15 in the upper lobe, 16 in the lower lobe) were retrospectively analyzed. All patients underwent Four-Dimensional (4D) CT simulation and tumor motion tracking during treatment, represented by three-dimensional coordinates extracted from log files generated by Synchrony Respiratory Tracking System, which were used to characterize displacement along the superior-inferior (S–I), anterior-posterior (A–P), and left-right (L–R) axis. The MTD model is proposed to assess the accumulative dose of the moving target considering both motion and dose fall-off outside of the PTV from VMAT treatment, then margins were calculated to ensure Gross Tumor Volume (GTV) receives a minimum of 95

OBJECTIVE:We aimed to develop clinical decision support tools for patients with treatment-naïve advanced esophageal squamous cell carcinoma (ESCC). METHODS:Patients who received first-line immunochemotherapy with or without radiotherapy between 2018 and 2023 across 9 centers were included in the current study. Candidate predictors were routine clinical variables such as patient demographics, tumor characteristics, and treatment details. Feature selection was conducted using the least absolute shrinkage and selection operator (LASSO), Boruta, and stepwise feature selection method. The models were fitted using 4 machine learning algorithms and a Cox proportional hazards model was set as a benchmark, comparing model predictive performance by computing the time-dependent area under the receiver operator curve (tAUC). Concordance index (C-index), Brier score, calibration plots, precision-recall (PR) curves, decision curve analysis (DCA), and held-out testing and external validation were utilized for evaluating model robustness, transportability, and clinical utility. SHapley Additive exPlanations (SHAP) analysis was used to rank feature importance and explain the models. RESULTS:A total of 1,048 patients were included, comprising the training/test (n = 828) and validation cohorts (n = 220). The random survival forest model that was constructed following feature selection using Boruta [Boruta-random survival forest (RSF)] showed the highest predictive performance. The tAUCs for 6-, 12-, and 18-month OS were 0.886, 0.775, and 0.772, respectively, in the training set, which significantly outperformed all other models (P-adjust < 0.05). This superiority persisted in the external validation cohort (all tAUCs > 0.750). The Boruta-RSF model exhibited excellent calibration and the highest clinical net benefit on DCA. A publicly accessible web calculator (https://escc.shinyapps.io/ESCC/) was implemented, enabling individualized overall survival prediction under varying treatment scenarios. CONCLUSIONS:The Boruta-RSF model, leveraging routinely available clinical variables with the corresponding web calculator, facilitates personalized prognostic assessment and optimization of treatment strategies.

Abstract Background: Radiotherapy remains one of the principal therapeutic modalities for non-small cell lung cancer (NSCLC), yet its therapeutic efficacy is frequently compromised by tumor recurrence, metastasis, and the development of radioresistance. Accumulating evidence identifies cancer stem cells (CSCs) as a critical driver of radioresistance. Therefore, elucidating the underlying molecular mechanisms and developing therapeutic targets to enhance radiosensitivity are of critical importance. Methods: To identify key mediators of adaptive radioresistance in NSCLC, we established radioresistant NSCLC cell lines through repeated cycles of irradiation and subsequently performed RNA-seq analysis comparing them with their parental counterparts. The effects of SLFN5 on radiosensitivity were evaluated by flow cytometry and colony formation assays, and further validated in vivo using a nude mouse xenograft tumor model. Cancer stem-like properties were assessed by RT-PCR, Western blot, flow cytometry, sphere formation assay, and in vivo limiting dilution tumorigenesis assays. Subsequently, we employed mass spectrometry, co-immunoprecipitation, and proximity ligation assay to identify the interaction between SLFN5 and NOTCH1. Furthermore, phase separation assay, fluorescence recovery after photobleaching, CUT&Tag-seq, and ATAC-seq were used to investigate the role of NOTCH1 phase separation in promoting stemness and radioresistance, and the regulatory effect of SLFN5 on this process. Finally, the methylation status of the SLFN5 promoter was analyzed using bisulfite sequencing PCR. Results: Our results demonstrate that SLFN5 was significantly downregulated in radioresistant NSCLC cell lines. Overexpressing SLFN5 effectively suppressed cancer stemness and epithelial-mesenchymal transition, thereby enhancing radiosensitivity both in vitro and in vivo. Conversely, knockdown of SLFN5 had significantly opposite effects. Mechanistically, liquid-liquid phase separation property is critical for NOTCH1 mediated stemness and radioresistance. Notably, SLFN5 interacts with NOTCH1 and induces the liquid-to-solid phase transition of NOTCH1, thereby impairing NOTCH1 signaling and the subsequent stemness and radioresistance. Finally, we identified DNMT3A and DNMT3B as the epigenetic regulators responsible for promoter hypermethylation and consequent silencing of SLFN5. Conclusion: Our study reveals that DNA methylation downregulated SLFN5 resulting in radioresistance of NSCLC via liberating NOTCH1 from gel-like phase to liquid droplet to potentiating stemness of cancer cells. This research provides a potential therapeutic strategy to overcome radioresistance in NSCLC. Citation Format: Mi Tang, Lu Zhang, Jiaxin Zhao, Hongji Dai, Zhiyong Yuan, Zeyun Mi, Zhiqiang Wu. SLFN5 mediates liquid-to-solid phase transition of NOTCH1 to suppress stemness and radioresistance of non-small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 2192.

Dual-layer multileaf collimator (MLC) accelerators, such as Halcyon, are increasingly used in clinical settings; however, the ability of treatment planning systems (TPS) to effectively optimize treatment plans for such accelerators remains an important consideration. This study evaluates the feasibility of using uTPS to optimize treatment plans for dual-layer MLC accelerators, with Eclipse as a clinical reference. Standard radiation data from the Halcyon 2.0 accelerator were used to model the beam in the uTPS (Shanghai United Imaging Healthcare Co., Ltd., version R001). Twenty cases were selected, with five each for hypopharyngeal, esophageal, breast, and cervical cancers. Volumetric modulated arc therapy plans were generated using the uTPS stochastic platform optimizer and the Eclipse photon optimizer (PO). All plans were calculated using the Acuros external beam (AXB) algorithm for dosimetric comparison. Key indicators, such as the conformity index (CI), homogeneity index (HI), and organ-at-risk (OAR) doses, were analyzed and compared. The feasibility of delivering uTPS plans on the Halcyon accelerator was validated. The dosimetric accuracy was confirmed using ArcCheck, and Wilcoxon signed-rank tests with Benjamini–Hochberg false discovery rate correction were performed. uTPS and Eclipse plans exhibited similar dosimetric qualities for esophageal and cervical cancer cases, with no significant differences in CI, HI, and OAR doses. However, uTPS demonstrated superior OAR sparing in specific cases: in hypopharyngeal cancer, uTPS significantly reduced the brainstem D_0.1cc by 34.48

PURPOSE:Evidence supports stereotactic radiosurgery (SRS) or fractionated stereotactic radiosurgery (fSRS) for brainstem metastases (BSMs). The optimal dose-fractionation schedule remains undefined. We evaluated tumor control probability (TCP), overall survival (OS), and treatment-related adverse events after SRS and fSRS. METHODS AND MATERIALS:We conducted a comprehensive review of studies from the PubMed, Embase, and Cochrane databases and from our institutional cohort. Logistic dose-response models compared TCP and OS using biological effective dose (BED) calculated using the linear-quadratic model and equivalent doses for 1-5 fractions. The α/β ratio was estimated by fitting TCP data using maximum likelihood estimation across three representative radiobiological models. RESULTS:A total of 2,237 patients (2,423 lesions) from 28 articles and our institutional cohort were included in the analysis. The median tumor volume was 0.4 cm3 (range, 0.04-4.2; interquartile range [IQR], 0.19-0.995), and the median follow-up duration was 10 months (range, 3.2-37.7; IQR, 5.8-14.15). Fitting the clinical TCP data from SRS and fSRS consistently yielded α/β ratios of approximately 20 Gy across all three radiobiological models. SRS and fSRS achieved an estimated 1-year TCP of 90% at a BED20 ≈ 36.8 Gy (≈ 18.9 Gy/1 fx, 23.3 Gy/2 fx, 25.7 Gy/3 fx, 27.4 Gy/4 fx, and 28.6 Gy/5 fx) and a 2-year TCP of 90% at a BED20 ≈ 41.4 Gy (≈ 20.5 Gy/1 fx, 25.3 Gy/2 fx, 28.2 Gy/3 fx, 30.1 Gy/4 fx, and 31.5 Gy/5 fx). Estimated 1- and 2-year TCPs of 80%, 85%, and 90% were achieved with single-fraction doses of 15.8, 17.2, and 18.9 Gy as well as 17.4, 18.8, and 20.5 Gy, respectively. A trend toward significance was observed for BED20 and equivalent dose in relation to 1- and 2-year OS following SRS and fSRS. Grade ≥ 3 adverse events were infrequent (3.1%), with only one patient experiencing grade 5 hemorrhage (0.04%). CONCLUSIONS:For carefully selected patients with BSMs, SRS or fSRS should be delivered at a BED20 of at least 41.4 Gy, corresponding to 20.5-31.5 Gy in 1-5 fractions, yielding favorable 2-year LC with acceptable incidences of grade ≥ 3 adverse events. These findings are warranting validation through ongoing and planned prospective clinical trials.

BackgroundStereotactic Radiation Therapy (SRT) has proven effective for various stages of Hepatocellular Carcinoma (HCC), however, its role in managing intra- or extrahepatic recurrence after liver transplantation remains underexplored.ObjectivesThis study evaluates the safety and efficacy of SRT delivered using the CyberKnife® system for recurrent HCC after liver transplantation and introduces a novel nomogram for predicting survival to guide individualized management.MethodsIn a single-center retrospective study conducted between 2007 and 2020, 79 patients with recurrent HCC after transplantation presenting 133 intra- or extrahepatic lesions were treated with SRT. Treatment response, survival outcomes, and local control rates were evaluated. Multivariate analysis identified significant prognostic factors, which were incorporated into a predictive nomogram for survival.ResultsWith a median follow-up of 11.4 months, median overall survival (OS) was 15.1 months, and the local control rate was 90.5%. The OS rates at 6 months, 1 year, and 2 years were 78.9%, 57.1%, and 38.9%, respectively. Key factors associated with improved survival included fewer than three lesions, AFP <500 ng/ml, KPS ≥70, and total gross tumor volume (GTV) <40 mL. The nomogram demonstrated good predictive accuracy with a validated C-index of 0.760. Treatment was well tolerated, with no severe treatment-related toxicities observed.ConclusionSRT provides effective local control for both intra- and extrahepatic recurrences of HCC after liver transplantation. The proposed nomogram offers a valuable tool for personalized surveillance and treatment planning.

Bulky solid tumors present significant therapeutic challenges. Spatially fractionated radiotherapy (SFRT), a technique delivering alternating high- and low-dose subvolumes, alters the tumor microenvironment while minimizing toxicity. This phase II trial assesses the efficacy and safety of SFRT combined with immune checkpoint inhibitors (ICIs) and anti-angiogenic agents in advanced malignancies. This prospective phase II trial enrolled 34 patients with bulky solid tumors between October 2024 and July 2025. All patients underwent SFRT using GRID, LATTICE, or Stereotactic central/core ablative radiation therapy techniques. Multimodal therapy, incorporating pre-radiotherapy administration of granulocyte-macrophage colony-stimulating factor and thymalfasin, as well as concurrent ICIs and anti-angiogenic agents during SFRT, was administered according to clinical recommendations and patient preferences. The endpoints were treatment-related adverse events and the objective response rate (ORR). In addition, a prognostic analysis was performed to identify factors associated with clinical outcomes. Among the 37 treatment courses in 34 patients, 4 patients did not complete the planned therapy, and 1 was lost to follow-up. The median follow-up duration was 6.0 months. Of the 32 evaluable lesions from patients who completed the study, the ORR was 65.63

High-resolution computed tomography (HRCT), with its precise fine structure imaging character, is of growing importance in modern clinical practice. Emerging deep-learning-based CT super-resolution (SR) approaches, particularly those arbitrary-scale SR networks employing implicit neural representation (INR), are demonstrated with promising results. Nevertheless, existing INR algorithms are still hindered by the inferior ability to capture CT fine structural detail since the multilayer perceptron module (MLP) in INR is biased to learn high-frequency components. In this paper, we propose to incorporate a pixel-wise hybrid high-dimension mapping (HHM) module into INR to alleviate the above issues. The HHM module applies sinusoidal function simultaneously on both latent features and spatial coordinates before MLP, projecting the incorporated spatial and image features into a higher dimensional space, to force the INR network to learn more high-frequency details. The fine structural detail in arbitrary-scale SR CT is thus enhanced. We train the proposed network using real low- and high- resolution clinical CT images rather than using down-sampling images, which is more practical in the clinic. We validate the proposed method in detail using qualitative and quantitative evaluations on the thoracic and pelvic dataset, and the results indicate that our method is accurate and robust, outperforming the other deep learning methods.

Cherenkov imaging is widely recognized as a promising technique for large-area, non-contact dose monitoring in radiation oncology. However, its quantitative reliability is fundamentally limited by the characteristic hybrid noise pattern, particularly under limited exposure conditions inherent to video-based acquisition. As such, a learning-based Cherenkov video denoising framework is proposed, allowing for single-frame integration time down to similar to 17.5 mu s. To preserve model generalizability and training practicability, a non-blind self-supervised paradigm is developed through prior-guided cycle-consistent degradation regression. The causal effects of the denoising mechanism were systematically visualized and interpreted. Quantitative evaluation using gamma analysis (3%/3 mm criterion) was performed to assess simultaneous dose and positioning verification. Phantom studies demonstrate that the proposed method improves the gamma passing rate (gamma(pass)) by an average of 35.90% compared with the best-performing self-supervised baseline across three regular radiation fields. In clinical validation across patients with various tumor sites, the average gamma(pass) reaches 90.09%, representing a 93.48% improvement. In addition, the gamma-based sensitivity to translational, yaw, and roll motions increases to 150.75%, 63.84%, and 80.18%, respectively. Overall, the results demonstrate a quantitatively robust and interpretable Cherenkov imaging framework, highlighting its broader potential for general video denoising under extremely low signal-to-noise conditions.

Abstract Background: Radiotherapy is is a pivotal treatment for locally advanced or inoperable non-small cell lung cancer (NSCLC), however its efficacy is limited by radioresistance. Approximately 44%-46% of patients develop recurrence or metastasis after chemoradiotherapy, with stage III patients showing a 30%-40% local recurrence rate within five years. Identification of effective therapeutic targets to enhance radiosensitivity remains a critical challenge in NSCLC management. This study aims to elucidate the molecular basis of radioresistance and develop novel approaches to enhance tumor radiosensitivity. Methods: Through RNA transcriptome sequencing of treatment-naive NSCLC patients (n=87) stratified by RECIST into radiotherapy-sensitive (n=55) and -resistant (n=32) cohorts, and based on histopathological and clinical diagnostic data from Tianjin Medical University Cancer Hospital (2015-2020), we analyzed the association between CYLD expression and radiotherapy outcomes through RNA sequencing and tissue microarrays. The impact of CYLD on radiosensitivity was assessed through flow cytometry and colony formation assays, and further verified in nude mouse xenograft models. To investigate CYLD-related pathways, we performed next-generation sequencing and examined its role in ferroptosis through Western blotting, qPCR, glutathione assays, malondialdehyde measurements, lipid peroxidation assessment, and transmission electron microscopy. Potential CYLD-interacting proteins were identified via mass spectrometry, immunofluorescence, and co-immunoprecipitation assays, while CYLD-mediated deubiquitination of ALOX12B was characterized using ubiquitination immunoprecipitation.Furthermore, we combined database predictions with Western blot, qPCR, chromatin immunoprecipitation, and luciferase reporter assays to establish EGR1 as a transcriptional regulator CYLD expression. Result: Our findings identify CYLD as a positive regulator of radiosensitivity in NSCLC, We demonstrate that CYLD enhances radiation sensitivity in NSCLC by stabilizing ALOX12B. Mechanistically, CYLD enhances ALOX12B protein stability by deubiquitinating K63- and K48-linked ubiquitin chains, thereby suppressing its proteasomal degradation. This stabilization potentiates cellular ferroptosis and consequently increases radiosensitivity. Furthermore, we identified EGR1 as an upstream transcriptional activator of CYLD that promotes ferroptosis in a CYLD-dependent manner. Conclusion: Our study demonstrates that CYLD, driven by its transcriptional activator EGR1, promotes ferroptosis and enhances radiosensitivity in NSCLC by mediating the deubiquitination and stabilization of ALOX12B. This study identifies CYLD as both a valuable prognostic biomarker and a potential molecular target for radiosensitization in NSCLC. Citation Format: Yihan Xu, Jiazhuo Yan, Huiwen Yu, Lu Zhang, Jinpu Yu, Zhiyong Yuan, . CYLD induces ferroptosis through ALOX12B to enhance radiosensitivity in non-small cell lung cancer [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2026; Part 1 (Regular Abstracts); 2026 Apr 17-22; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2026;86(7 Suppl):Abstract nr 6620.

Widespread clinical implementation of rapidly evolving auto-segmentation tools remains constrained by a scarcity of high-quality prospective evidence. Here we show the results of a prospective, multicenter, observational trial (NCT05787522) evaluating the clinical performance of a deep learning model (iCurveE) for artificial intelligence (AI)-assisted delineation of organs at risk (OARs) in thoracic and breast cancer radiotherapy. Computed tomography images from 500 patients across five centers are annotated by 37 physicians using manual, AI-generated, and AI-assisted methods. Eleven thoracic OARs are evaluated based on the primary endpoints of volumetric Dice similarity coefficient (vDSC) and contouring time, alongside secondary metrics including 95% Hausdorff Distance (HD95). We prospectively annotate 2,483 OAR sets (27,043 OARs): 993 manual, 497 AI-generated, and 993 AI-assisted. AI-assisted delineation achieves significantly better vDSC (mean, 0.902) and HD95 (mean, 5.20 mm) than manual delineation (mean vDSC, 0.857; mean HD95, 8.01 mm; p < 0.0001) while improving time efficiency by 81.63% (median: 10.0 vs. 55.0 min; p < 0.0001). AI-assisted delineation reduces performance variability across centers and physicians with varying expertise. This study validates the clinical applicability of AI-assisted delineation in improving delineation performance and promoting healthcare equity.