Yongji Tian

Postoperative disturbances in cerebrospinal fluid (CSF) dynamics remain a major complication after pediatric medulloblastoma (MB) resection and may necessitate permanent ventriculoperitoneal shunting (VPS). Existing risk assessment tools have important methodological limitations: many are derived from heterogeneous posterior fossa tumor cohorts and rely predominantly on static preoperative anatomical variables, with limited consideration of tumor-specific invasion patterns or postoperative biological factors. To identify multidimensional factors associated with delayed postoperative ventriculoperitoneal shunt (VPS) requirement after pediatric medulloblastoma resection and to explore how preoperative, pathological, molecular, and postoperative features contribute to modeled risk patterns. We retrospectively analyzed 269 pediatric patients with medulloblastoma who underwent tumor resection between January 2019 and January 2024. The primary endpoint was delayed postoperative hydrocephalus requiring VPS more than 2 weeks after tumor resection for persistent intracranial hypertension and progressive ventricular enlargement. Firth penalized logistic regression was designated as the primary association model. Model performance was assessed using C-statistics, calibration curves, decision curve analysis, and internal bootstrap validation. A multi-algorithm AUC heatmap and ElasticNet-based SHAP analysis were used as secondary exploratory analyses, and the nomogram was derived from the Firth model. Among 269 children treated between 2019 and 2024, 18 (6.7%) developed delayed postoperative hydrocephalus requiring VPS more than 2 weeks after resection. In the primary Firth model, postoperative hemorrhage (OR 22.40, 95% CI 4.24–118.0, P < 0.001) and choroid plexus invasion (OR 18.00, 95% CI 3.34–96.9, P < 0.001) showed the strongest positive associations, although the wide confidence intervals indicate substantial uncertainty. Auxiliary SHAP analysis identified postoperative hemorrhage and age as prominent global contributors, while choroid plexus invasion showed large contributions in positive cases and Group 3 and Group 4 displayed contrasting higher- and lower-risk patterns. The findings support a multi-hit pattern in which limited baseline CSF compensatory reserve may be followed by postoperative hemorrhagic or inflammatory insults. The proposed model is exploratory and is not yet ready for individual clinical decision-making. Its performance and feature-contribution patterns should be tested in prospectively enrolled multicenter cohorts using prespecified variables and the same endpoint.

Objective The surgical management of sporadic optic pathway glioma (OPG) remains controversial due to its unpredictable visual outcomes and uncertain risk–benefit balance. This study aimed to identify key clinical and surgical determinants of long-term visual preservation following tumor debulking in children with sporadic OPG. Methods A total of 192 pediatric patients who underwent initial partial resection for sporadic OPG at Beijing Tiantan Hospital between 2011 and 2023 were retrospectively analyzed. Two complementary outcomes were assessed: follow-up visual acuity (VA) of the better-eye defined at baseline and individual visual change. Ordinal and binary logistic regression analyses were performed to identify predictors, supplemented by trend and visualization analyses. Results Baseline VA was the strongest predictor of follow-up VA of the baseline-defined better-eye (OR = 5.88, P < 0.001). Poorer outcomes were associated with younger age (< 3 years), delayed surgery, extent of resection ≥ 50%, and postoperative radiotherapy. Among patients with moderate baseline VA, a greater resection extent showed a trend toward worse outcomes. The anterior interhemispheric approach was linked to higher risk of visual deterioration, whereas transcallosal and frontotemporal approaches were protective. Early intervention, limited resection, and careful avoidance of high-risk surgical corridors were associated with better long-term visual preservation. Conclusions Preoperative baseline VA determines postoperative visual potential, while surgical timing, extent, and approach critically shape long-term outcomes. Patients with moderate baseline VA are the most intervention-sensitive subgroup. These findings support early, conservative, and function-oriented surgical strategies to optimize visual outcomes in pediatric sporadic OPG.

Precise diagnosis and treatment of glioblastoma (GBM) remain challenging. The overexpression level of carbonic anhydrase IX (CA IX), a key biomarker for GBM, is correlated with tumor malignancy. Herein, we reported a CA IX‐activated nanosensor (MPC@BDPCA NPs) for near‐infrared imaging and synergistic phototherapy of orthotopic GBM. The tailor‐made core agent, a BODIPY derivative (BDPCA) incorporates a benzenesulfonamide moiety that selectively binds to CA IX, inducing rotational restriction of the probe and resulting in fluorescence turn‐on response. Owing to this specific activating mechanism, BDPCA enables a high‐contrast fluorescence and photoacoustic dual‐modal imaging for quantitative CA IX sensing. Notably, an extended π‐conjugation combined with a donor–acceptor molecular design optimizes both radiative and nonradiative decay pathways, affording efficient photothermal conversion and reactive oxygen species generation under 660 nm irradiation. To enhance blood‐brain barrier (BBB) penetration in orthotopic GBM models, BDPCA was encapsulated within pH‐degradable polymer shell via in situ polymerization, employing 2‐methacryloyloxyethyl phosphorylcholine (MPC) as monomer to improve biocompatibility and facilitate receptor‐mediated BBB transport. The resulting MPC@BDPCA NPs exhibit selective CA IX sensing, multimodal imaging capability, efficient BBB penetration, and potent synergistic phototherapy. This work highlights a versatile nanoscale sensing and diagnostic platform for precision imaging and therapeutic intervention of GBM.

Graphical Abstract For image description, please refer to the figure legend and surrounding text.

Preoperative prognosis of Ependymoma is critical for treatment planning but challenging due to the lack of semantic insights in MRI compared to post-operative surgical reports. Existing multimodal methods fail to leverage this privileged text data when it is unavailable during inference. To bridge this gap, we propose HyperPriv-EPN, a hypergraph-based Learning Using Privileged Information (LUPI) framework. We introduce a Severed Graph Strategy, utilizing a shared encoder to process both a Teacher graph (enriched with privileged post-surgery information) and a Student graph (restricted to pre-operation data). Through dual-stream distillation, the Student learns to hallucinate semantic community structures from visual features alone. Validated on a multi-center cohort of 311 patients, HyperPriv-EPN achieves state-of-the-art diagnostic accuracy and survival stratification. This effectively transfers expert knowledge to the preoperative setting, unlocking the value of historical post-operative data to guide the diagnosis of new patients without requiring text at inference.

Background: Approximately 20% of medulloblastoma (MB) patients are diagnosed with metastatic disease and typically exhibit extremely poor clinical outcomes. This study aimed to investigate potential prognostic factors affecting the survival of patients with metastatic MB. Methods: Patients with initial diagnosis of metastatic MB (M2/3) at Beijing Tiantan Hospital were included. Radiological characteristics were discerned through a retrospective review. Overall survival (OS) and event-free survival (EFS) were calculated using the Kaplan-Meier analysis. Multivariable Cox proportional hazards model was employed to identify potential prognostic factors. Results: This study included 115 patients with M2/3 MB. Group 4 MBs accounted for 59.1% of cases (68/115). Kaplan-Meier analysis indicated that the 5-year EFS and OS rates were 56.1% and 68.7%, respectively. Patients with metastatic Group 3 MB and spinal metastases exhibited dismal outcomes. Postoperative “sandwich” strategy significantly prolonged 5-year EFS and OS rates. Multivariate COX regression models demonstrated that molecular subgrouping, metastatic patterns, and “sandwich” strategy were independent prognostic predictors for both EFS and OS in patients with M2/3 MB. Conclusion: Our study provides a novel risk stratification model for metastatic medulloblastoma that could potentially facilitate the development of individualized therapeutic strategies for the very high-risk patient population.

To develop and validate a Pediatric Anatomo-Endocrine (PAE)-based nomogram for predicting severe postoperative endocrine dysfunction in children undergoing craniopharyngioma (CP) resection, integrating anatomical classification, surgical factors, and baseline endocrine status. This retrospective study included 370 pediatric CP patients (≤ 18 years) who underwent primary resection at a single center between 2017 and 2024. Patients were randomly divided into a primary (n = 300) and a validation (n = 70) cohort. Endocrine dysfunction was assessed based on the number and severity of hypothalamic–pituitary axis abnormalities. A novel three-tier PAE classification was developed from preoperative MRI. Logistic regression was used to identify predictors of severe dysfunction. A nomogram was constructed and validated via ROC, calibration, and decision curve analyses. In the primary cohort, 83.4% of patients had worsened endocrine dysfunction postoperatively, with severe dysfunction increasing from 20.3% to 79.3%. Independent predictors included PAE classification, baseline endocrine status, and pituitary stalk sacrifice. The nomogram demonstrated strong performance (AUC 0.84 in primary; 0.82 in validation). Among PAE subtypes, the Global subtype posed the highest risk, with an odds ratio of 16.30 versus the Sella–Infundibulum subtype. The PAE classification stratifies hypothalamic–pituitary involvement and serves as an independent predictor of endocrine outcomes. Although validated internally in a large single-center cohort, its long-term benefit remains to be confirmed. The proposed nomogram may provide a practical, non-invasive tool for individualized risk prediction and surgical planning, but requires external multicenter validation to establish generalizability.

Purpose To develop a deep learning model for predicting molecular subgroups of medulloblastoma (MB) using preoperative brain MRI. Materials and methods This study included a cohort of 350 patients with MB for model development. Preoperative multiparametric brain MRIs were acquired, and molecular classification data for tumor samples were analyzed. A dual-task deep learning model, composed of a 3D Swin Transformer backbone and a Transformer-based mask decoder, was developed for the prediction of MB molecular subgroups. The model was jointly optimized with a parallel task of tumor and cerebellum segmentation. Ablation analysis was conducted to verify the effectiveness of the dual-task model design. An independent test cohort of 126 patients with MB was established to validate the predictive performance of the dual-task model. Results Our dual-task deep learning model demonstrated superior performance for MB molecular subgroup prediction, achieving an AUC of 0.877, accuracy of 88.9%, sensitivity of 71.6%, and specificity of 91.9%. The performance remained robust across both adult and pediatric age populations, with AUCs of 0.915 and 0.871, respectively. Furthermore, our approach exhibited effective generalization to the independent test cohort, yielding an AUC of 0.853, accuracy of 89.7%, sensitivity of 73.5%, and specificity of 92.1%. Ablation analysis demonstrated a significant improvement in AUC of 0.169 (95% CI 0.097–0.244) when using the dual-task model design. In comparison with the radiomics-based model, our deep learning model achieved a higher AUC by 0.156 (95% CI 0.079–0.233). Conclusion Our proposed dual-task deep learning model enables automated and accurate prediction of MB molecular subgroups.