IntroductionPituitary adenomas (PAs) alter the anatomy of the pituitary gland (PG) and internal carotid arteries (ICAs), requiring accurate segmentation for surgical planning and risk assessment. Multi-sequence magnetic resonance imaging (MRI)-based automatic segmentation methods have been employed but remain limited in routine clinical practice due to protocol inconsistencies. This study aimed to evaluate the feasibility and clinical relevance of Swin-Unet and nnU-Net for automated segmentation of the PA, PG, and ICA using only contrast-enhanced T1-weighted (T1CE) coronal MRI.MethodsA retrospective cohort of 255 patients with surgically confirmed PAs was analyzed. T1CE coronal MRI, acquired from heterogeneous scanners and protocols, was used for model development. Manual annotations of the PA, PG, and ICA by expert neurosurgeons served as ground truth. Two deep learning architectures were evaluated: 1) Swin-Unet, a transformer-based U-Net variant, and 2) nnU-Net, a self-configuring CNN framework. Models were trained using standard preprocessing and five-fold cross-validation, with ensemble predictions generated for final evaluation. Model performance was assessed using the Dice similarity coefficient (DSC), 95th percentile Hausdorff distance (HD95), true-positive rate (TPR), and false-positive rate (FPR). Additionally, slice-level presence detection and per-patient processing time were evaluated.ResultsBoth models demonstrated comparable voxel-wise performance, with mean DSC = 0.70 (95% CI: 0.66-0.74) for Swin-Unet and 0.69 (95% CI: 0.65-0.73) for nnU-Net. Segmentation accuracy was lower for the PG compared to the PA and ICA. Boundary-based evaluation showed a mean HD95 of 5.21 mm (95% CI: 4.40-6.03) for Swin-Unet and 5.30 mm (95% CI: 4.44-6.16) for nnU-Net. Slice-level recognition yielded a mean F1-score of 0.89 ± 0.09 for Swin-Unet and 0.87 ± 0.08 for nnU-Net across the PA, PG, and ICA, respectively. The mean computation time was 78.4 seconds per patient (95% CI: 61.5–95.5), and the segmentation outputs facilitated approximate 3D reconstructions for qualitative assessment.DiscussionOur findings demonstrate the feasibility of automated segmentation of the PA, PG, and ICA using single-sequence T1CE coronal MRI. While voxel-level accuracy was modest, the combination of reliable slice-level recognition, rapid processing, and 3D visualization suggests potential utility as an adjunct tool for radiologic assessment and surgical planning.

INTRODUCTION:The tumor control rate after stereotactic radiosurgery (SRS) for neurofibromatosis type 2-associated vestibular schwannomas (NF2-VSs) compared to sporadic vestibular schwannomas (S-VSs) remains unclear. This nationwide, multicenter, retrospective study (KGKRS-21-001) aimed to clarify this issue. METHODS:A total of 4718 patients treated with SRS for vestibular schwannomas were analyzed from 13 nationwide institutions in Korea. NF2-VS cases were propensity score-matched with S-VS cases at a ratio of 1:1, based on age, tumor volume, and marginal dose, resulting in 122 cases in each group. RESULTS:No significant differences in age, tumor volume, or marginal dose were observed between the matched cohorts. The overall tumor control rates at 1, 3, and 10 years after SRS were 93.3%, 87.7%, and 80.7%, respectively, with no significant difference between NF2-VS and S-VS groups (p = 0.63). Subgroup analysis showed that age ≤ 19 years was a significant negative prognostic factor for tumor control in NF2-VS patients (p < 0.001), whereas no such correlation was found in the S-VS cohort (p = 0.78). CONCLUSIONS:SRS provides comparable tumor control for NF2-VSs and S-VSs. However, among NF2-VS patients, younger age (≤ 19 years) was associated with poorer tumor control, suggesting that age may be a critical factor in treatment decisions.

BACKGROUND:Stereotactic radiosurgery (SRS) is an established treatment for vestibular schwannoma (VS), primarily for tumors smaller than 3 cm in diameter. Its role in larger tumors remains controversial and is generally limited to carefully selected patients. This study evaluated long-term outcomes of primary single-session SRS for VS, stratified by tumor size. METHODS:We retrospectively analyzed 160 patients with unilateral sporadic VS treated with SRS between July 2009 and September 2020. SRS was primarily indicated for tumors <3 cm in maximum diameter, while larger tumors were selectively treated in patients without significant brainstem compression or severe neurological deficits. Patients were stratified by Koos grade: Group A (I), Group B (II-III), and Group C (IV). Median follow-up was 96 months. Primary outcomes included progression-free survival (PFS), hearing preservation, and treatment-related complications. RESULTS:The overall tumor control rate was 98.1%. The 10-year PFS rates were 100%, 97.4%, and 96.6% for Groups A, B, and C, respectively, with no significant difference (p=0.739). Among patients with serviceable hearing, 5-year preservation rates were 91.7%, 36.0%, and 53.6% (p=0.002). New-onset trigeminal neuralgia occurred in 0%, 0.9%, and 6.1% of patients in Groups A, B, and C, respectively. Post-SRS hydrocephalus requiring ventriculoperitoneal shunting was observed in 0%, 5.6%, and 27.3%, respectively. CONCLUSION:Within appropriately selected patients, SRS provides excellent tumor control across tumor sizes. However, the risks of trigeminal neuralgia and hydrocephalus requiring shunting increased with tumor size, underscoring the need for individualized treatment strategies.

Objective : The aim of this study was to report the early clinical experience with stereotactic biopsy using the Medtronic Stealth Autoguide Cranial Robotic Guidance Platform, representing the first relevant clinical study in the Republic of Korea (ROK). We evaluated potential advantages and limitations related to workflow integration, instrument setup, procedural efficiency, accuracy, and safety. Methods : A retrospective case series was conducted across three centers in the ROK. From April to July 2025, 17 consecutive patients underwent frameless stereotactic biopsy using the Stealth Autoguide system. Data on demographics, operative time, and complications were collected, and descriptive analyses were performed. Results : Biopsy using the Stealth Autoguide System was successfully completed in 16 of 17 patients. The median operative time was 48 minutes (interquartile range, 33-64), and the median target alignment error was 0.4 mm (interquartile range, 0.3-0.6). One case required conversion to the conventional manual method due to targeting inaccuracy. Three cases of asymptomatic intracerebral hemorrhage occurred, including one at the tissue-harvest site and two with cortical hemorrhage, all of which were managed conservatively. Conclusion : The Stealth Autoguide system facilitates a more efficient frameless stereotactic biopsy than conventional methods by reducing operative time, enhancing accuracy, and allowing for a smaller, muscle-sparing incision. However, some concerns remain, and further clinical validation is needed to define its optimal indications and safety profile.

Glioblastoma is an aggressive brain cancer with limited treatment options and poor patient survival, driven in part by cellular diversity within tumors. While individual cell types have been catalogued, how malignant, vascular, and immune cells are spatially organized inside human tumors remains incompletely understood. Here we show a spatially resolved, multi-modal atlas of human glioblastoma that integrates gene expression profiling across tissue sections with matched single-cell and protein measurements at subcellular resolution. Using a targeted 348 gene panel enriched for vascular and stromal markers, we identify less well-characterized endothelial, perivascular, and fibroblast-like cell states and define their spatial associations with malignant and immune compartments. We further identify a distinct oligodendrocyte population restricted to tumor core and perivascular regions that exhibits gene expression patterns associated with tumor recurrence and poor clinical outcome. This publicly accessible atlas provides a high-resolution framework for studying the spatial organization of glioblastoma and highlights region-specific cellular interactions that may represent therapeutically actionable vulnerabilities.

OBJECTIVE:Operating-room exposure in neurosurgical training is unevenly distributed. We developed a patient-specific virtual reality (VR) neurosurgical training platform of 20 patient-derived cases spanning aneurysm, meningioma, and trauma, running standalone on a commercial head-mounted display with cloud-based logging, and report its first independent evaluation. METHODS:A summative usability test of one case-a convexity meningioma-was conducted under IEC 62366-1 by an institutional medical-device testing center. Fifteen neurosurgeons (2 specialists, 13 residents) were recruited and compensated by the center against pre-specified criteria. Independent assessors classified performance on 20 predefined tasks as Correct Use, Use Error, Close Call, or Use Difficulty, and participants completed two 5-point Likert questionnaires. Groups were compared with Fisher's exact and Mann-Whitney U tests. RESULTS:All 15 participants completed the protocol without dropout or deviation. Subjective ratings were high (pooled median 4.0/5.0 on both instruments), highest for display refresh, realism, usefulness, and interface consistency. Specialists had fewer flagged tasks than residents on all three metrics-Use Error 2/20 versus 6/20 (p = 0.235), Close Call 1/20 versus 7/20 (p = 0.044), Use Difficulty 0/20 versus 3/20 (p = 0.231)-but no metric survived correction, and the task-level denominator does not scale with group size. The 'Control' domain was lowest and most variable (trainee median 4.0, range 1-5); errors concentrated in instrument selection and bimanual manipulation. CONCLUSIONS:A patient-specific neurosurgical VR platform running standalone on commercial hardware is feasible and was rated favorably by independently recruited evaluators. This pilot characterizes usability and defines engineering targets-instrument manipulation and user control-rather than testing efficacy.

Training for complex cranial neurosurgery requires safe, objective methods beyond traditional models. Virtual reality (VR) simulation offers a powerful solution, and to address this need, we developed and validated a novel VR platform for advanced surgical training. Our metaverse-based platform enables trainees to practice entire surgical procedures using high-fidelity 3D models derived from real patient clinical data, processed via optimized segmentation algorithms. The platform features a library of 21 cases, including meningiomas, and for this validation study, neurosurgeons performed tasks across three representative surgical scenarios. We assessed performance by tracking use errors and collecting subjective feedback on user satisfaction and educational value. The platform was met with high satisfaction; its intuitive interface led to successful task completion with minimal errors by both expert and trainee surgeons. Participants confirmed the platform’s significant educational value, praising its effectiveness for understanding complex surgical workflows and anatomy in a risk-free environment. Key suggestions for enhancement included greater user autonomy and graphical realism, underscoring user engagement. Validated by formal usability testing and endorsed by practicing neurosurgeons, our platform represents a significant advancement in surgical training. Its use of real patient data and a diverse case library, combined with proven user satisfaction, confirms its efficacy as a safe, scalable, and realistic environment for mastering critical skills. We believe this platform is poised to become an indispensable tool for enhancing neurosurgical proficiency and advancing patient safety in neurosurgery.

Vestibular schwannoma (VS) is the most common lesion of the cerebellopontine angle, accounting for approximately 8–10% of all intracranial tumors. Current treatment options include microsurgery, stereotactic radiosurgery (SRS), and conservative observation. SRS has been established as a primary treatment option and a viable alternative to microsurgery, offering high tumor control rates with low morbidity. This review summarizes current discussions regarding SRS for VS, based on recent studies and international guidelines. It also provides evidence-based treatment strategies according to tumor size to inform clinical decision-making.

This study aimed to determine whether trigeminal neuralgia can be diagnosed using convolutional neural networks (CNNs) based on plain X-ray skull images. A labeled dataset of 166 skull images from patients aged over 16 years with trigeminal neuralgia was compiled, alongside a control dataset of 498 images from patients with unruptured intracranial aneurysms. The images were randomly partitioned into training, validation, and test datasets in a 6:2:2 ratio. Classifier performance was assessed using accuracy and the area under the receiver operating characteristic (AUROC) curve. Gradient-weighted class activation mapping was applied to identify regions of interest. External validation was conducted using a dataset obtained from another institution. The CNN achieved an overall accuracy of 87.2%, with sensitivity and specificity of 0.72 and 0.91, respectively, and an AUROC of 0.90 on the test dataset. In most cases, the sphenoid body and clivus were identified as key areas for predicting trigeminal neuralgia. Validation on the external dataset yielded an accuracy of 71.0%, highlighting the potential of deep learning-based models in distinguishing X-ray skull images of patients with trigeminal neuralgia from those of control individuals. Our preliminary results suggest that plain x-ray can be potentially used as an adjunct to conventional MRI, ideally with CISS sequences, to aid in the clinical diagnosis of TN. Further refinement could establish this approach as a valuable screening tool.

Despite the use of T1CE coronal MRI in PA diagnosis, the effectiveness of automatic segmentation remains suboptimal, with research predominantly focused on convolutional neural networks. We aim To develop a transformer-based model for the automatic segmentation of pituitary adenoma (PA), pituitary gland (PG), and internal carotid artery (ICA) in contrast-enhanced T1-weighted (T1CE) coronal MR images. Patients with PA were retrospectively selected for inclusion in the training and test datasets using electronic medical and radiological records. Pretreatment T1CE coronal MR images were collected, annotated, and used as ground truth. A Swin-Unet-based transformer model was developed for segmentation, incorporating preprocessing techniques and loss functions to address class imbalance. To evaluate performance, the nnU-Net V2 model was trained using the same training dataset and compared against the Swin-Unet model. A total of 255 patients with PA were analyzed for model training and evaluation. The Swin-Unet model demonstrated superior performance compared to raw images, with Dice similarity coefficients (DSCs) of 0.825, 0.565, and 0.716 for PA, PG, and ICA, respectively (p=1.000, <0.01, and 0.425). Combo loss revealed the best segmentation performance, with the highest DSC of 0.702, a 95th percentile Hausdorff distance (95HD) of 4.846 mm, a true-positive rate (TPR) of 0.706, and a false-positive rate (FPR) of 0.002 for the mean of three classes. The nnU-Net V2 model exhibited comparable performance to Swin-Unet, with DSC of 0.691, 95HD of 3.686 mm, TPR of 0.775 (p < 0.001), and FPR of 0.002 for the mean of three classes. A novel transformer-based segmentation model using Swin-Unet was introduced for the automated detection of PA, PG, and ICA in T1CE coronal MRI.

The perivascular niche plays a crucial role in glioblastoma (GBM) progression and therapy resistance. However, recapitulating the dynamic interaction between GBM cells and human blood vessels remains a challenge in preclinical models. In this study, we utilized patient-derived GBM tissues and engineered vascular spheroids to establish an in vitro platform for studying tumor–vascular interactions. Vascular spheroids were generated from primary human endothelial cells and co-cultured with patient-derived GBM tissues or dissociated tumor organoids. Morphological changes, vessel infiltration, and spatial associations were analyzed using immunofluorescence and 3D confocal microscopy. Expression of angiogenic and stemness markers (PDGFRβ, CD31, Nestin, OLIG2) was assessed via qPCR and immunostaining. GBM tissues exhibited directional migration toward vascular spheroids and induced vessel sprouting and reorganization. Confocal analysis revealed close association of GBM cells with endothelial networks, mimicking perivascular niche characteristics. Molecular profiling indicated upregulation of angiogenesis-associated markers in the co-culture model compared to monoculture controls. Our in vitro co-culture platform effectively models the vascular–GBM interface, enabling mechanistic studies on perivascular invasion and vessel co-option. This system holds potential for screening anti-angiogenic and anti-invasive therapeutics in a patient-specific context.

Objective : Various skull base reconstruction techniques have been proposed for the endoscopic trans-sphenoidal approach (eTSA). These methods often involve multilayer closure using autologous fat grafts and fascia lata, frequently combined with lumbar drainage (LD). This study suggests a modification of graded reconstruction methods for low-grade (grade 0-2) cerebrospinal fluid (CSF) leaks, particularly grade 2 leaks, following eTSAfor pituitary neuroendocrine tumor (PitNET). Methods : A total of 401 consecutive patients who underwent eTSA for PitNET between 2017 and 2022 at a single institution were retrospectively reviewed, and ultimately 338 patients with intraoperative low-grade CSF leaks were included. Skull base reconstruction was performed according to intraoperative CSF leak grades. Grade 2 leaks were further subcategorized into 2a and 2b based on the size of the diaphragmatic defect. Grades 0, 1, and 2a leaks were reconstructed using fibrin sealant patches, hydrogel sealant, and reinforcement with nasoseptal flap where required. Grade 2b cases underwent rigid reconstruction with hydroxyapatite. Autologous fat grafts, fascia lata, and perioperative LD were not used. Postoperative CSF leaks and other complications were analyzed. Results : Among the 338 cases, there were 235 grade 0, 55 grade 1, 40 grade 2a and eight grade 2b intraoperative CSF leaks. No postoperative CSF leaks occurred. Two cases of postoperative meningitis were observed in grade 2a, all of whom were treated successfully with antibiotics. Conclusion : Our graded reconstruction approach is an effective and efficient method to prevent CSF leaks after eTSA in PitNET patients with low-grade intraoperative CSF leaks. Further subdividing grade 2 leaks based on defect size enables more precise and reliable skull base reconstruction. Moreover, autologous fat grafts, fascia lata, and perioperative LD appear unnecessary in low-grade cases.

Background: Delayed postoperative hyponatremia (DPH) is the most common cause of readmission after pituitary surgery. In this study, we aimed to evaluate the cutoff values of serum copeptin and determine the optimal timing for copeptin measurement for the prediction of the occurrence of DPH in patients who undergo endoscopic transsphenoidal approach (eTSA) surgery and tumor resection.Methods: This was a prospective observational study of 73 patients who underwent eTSA surgery for pituitary or stalk lesions. Copeptin levels were measured before surgery, 1 hour after extubation, and on postoperative days 1, 2, 7, and 90.Results: Among 73 patients, 23 patients (31.5%) developed DPH. The baseline ratio of copeptin to serum sodium level showed the highest predictive performance (area under the curve [AUROC], 0.699), and its optimal cutoff to maximize Youden’s index was 2.5×10–11, with a sensitivity of 91.3% and negative predictive value of 92.0%. No significant predictors were identified for patients with transient arginine vasopressin (AVP) deficiency. However, for patients without transient AVP deficiency, the copeptin-to-urine osmolarity ratio at baseline demonstrated the highest predictive performance (AUROC, 0.725). An optimal cutoff of 6.5×10–12 maximized Youden’s index, with a sensitivity of 92.9% and a negative predictive value of 94.1%.Conclusion: The occurrence of DPH can be predicted using baseline copeptin and its ratio with serum sodium or urine osmolarity only in patients without transient AVP deficiency after pituitary surgery.

PURPOSEPrimary CNS tumors (PCNSTs) are tumors originating from the brain and surrounding tissues. These tumors account for a significant proportion of cancer deaths and morbidity globally. Accurate epidemiologic data are essential for shaping clinical practices, research priorities, and health care policies. This study presents the latest 2020 national data on PCNSTs from the Republic of Korea (ROK) and explores the trends in incidence and their societal implications in the context of an aging population.METHODSThis is a cross-sectional, observational study conducted using data sourced from the Korea National Cancer Incidence Database by the Korea Central Cancer Registry. The study analyzed national data on PCNSTs in the ROK for the years 2010, 2013, 2016, and 2020.RESULTSIn 2020, 15,568 new PCNST cases were diagnosed in the ROK. The overall crude rate was 30.32, and the age-standardized rate was 19.37 per 100,000 persons. A decade-long trend analysis revealed an increasing trend in newly diagnosed glioblastoma and lymphoma, and a decreasing trend in embryonal tumors, in relation to the aging population of the ROK.CONCLUSIONThis study shows the significant impact of demographic shifts on the epidemiologic patterns of PCNSTs in the ROK. Our findings emphasize the need for collaborative efforts to address the rising challenges posed by the changing incidence of PCNSTs related to an aging population.

Giant cell tumors (GCTs) are locally aggressive primary bone tumors of osteoclast-like cells. Most GCTs occur within the long bones, and primary GCTs involving the clivus are extremely rare. We present the case of an 18-year-old boy with binocular horizontal diplopia with an insidious onset who was found to have a hypointense enhancing mass involving the clivus and left side dorsum sellae on magnetic resonance images. The tumor was completely resected via an endoscopic endonasal transclival approach, and histopathologic examination via immunohistochemistry indicated a GCT. The patient's left abducens nerve palsy improved slightly after surgery. Because of the rarity of GCTs, there is no consensus about the definitive treatment protocol. However, we suggest that gross total resection is the treatment of choice, and denosumab plays a critical role in patients with subtotal resection.

Introduction The mechanism of hearing loss following stereotactic radiosurgery (SRS) for vestibular schwannomas (VSs) remains unclear. There is conflicting evidence regarding cochlear nerve damage by transient volume expansion of VSs after radiosurgery and radiation-induced cochlear damage. This study aimed to investigate whether there is a specific patient population that can achieve definite hearing preservation after SRS for VSs. Methods A total of 37 consecutive patients with sporadic unilateral intracanalicular VSs and serviceable hearing (Gardner-Roberson [G-R] class I or II) were treated with SRS from 2009 to 2023. This is a retrospective study. Survival analysis with Cox regression for hearing deterioration was performed. Results The median age was 55 years old. The median tumor volume was 0.089 cm3, and the median marginal dose was 12.0 Gy. Nonserviceable hearing deterioration occurred in 9 patients (24.3%), with a median onset of 11.9 months after SRS. The actuarial rates of serviceable hearing preservation were 86%, 82%, and 70% at 1, 2, and 3 years after SRS, respectively. In a multivariate analysis, only baseline pure tone average > 30 dB increased the risk of nonserviceable hearing deterioration with significant hazard ratio. There were 13 patients with petit VSs whose tumor volume was smaller than 0.05 cm(3), and 11 of them were treated by a 4-mm single shot with a marginal dose of 12 Gy. None of the 13 patients had nonserviceable hearing deterioration. Conclusions Petit VSs that can be treated with 4-mm single or double shots with a marginal dose of 12 Gy may achieve hearing preservation after SRS.

Abstract The natural growth pattern of vestibular schwannoma (VS) associated with the rare disease neurofibromatosis type 2 (NF2) has not been sufficiently elucidated, and the rationale for proactive treatment for VS in NF2 is not firmly established. This study aimed to uncover the growth pattern of VS associated with NF2 and identify risk factors for rapid growth. A total of 68 VSs from 41 NF2 patients who were diagnosed in Seoul National University Bundang Hospital from July 2006 to July 2023 were retrospectively reviewed. NF2 was clinically diagnosed according to the Manchester criteria. The mean age was 34±16 years (mean ± standard deviation) at diagnosis. The initial tumor volume was a median of 0.263 cm³ (interquartile range [IQR], 0.053-0.754 cm³), and 51 VSs exhibited serviceable hearing classified as Gardner-Robertson grade 1 or 2. Patients were followed for a median of 45 months (IQR, 30-80 months), during which the median volume increased by 0.145 cm³ (IQR, 0.020-0.783 cm³), representing an annual growth rate of 0.044 cm³/yr (IQR, 0.003-0.202 cm³) or 12.4%/yr (IQR, 2.2-27.5%). Patients younger than 25 years (15 patients with 28 VSs) had significantly smaller initial tumor sizes (0.170 cm³ [IQR, 0.031-0.339 cm³] vs. 0.510 cm³ [0.099-1.399 cm³]) but faster tumor growth rates (annual growth rate; 27.3%/yr [IQR, 11.3-81.4%/yr] vs. 4.7%/yr [IQR, 0.0-16.1%/yr]) compared to those 25 years or older (26 patients with 40 VSs). Among the 51 VSs with serviceable hearing at diagnosis, 40 VSs were followed up for more than 12 months. Among them, 13 (32.5%) experienced unserviceable hearing deterioration at a median of 58 months. The risk of hearing loss significantly increased with faster tumor growth rates. The growth rate of NF2-related VS was higher in patients under the age of 25, which may be helpful in determining indications for proactive treatments like radiosurgery for NF2-related VS.

Abstract Glioblastoma (GBM) is the most common malignant brain tumor of the central nervous system and has a significantly low survival rate of approximately 15 months. Current preclinical GBM models are limited by the lack of a normal human microenvironment and the inability of many tumor cell lines to accurately reproduce GBM biology. To address these limitations, we describe a detailed procedure to generate brain-vessel- GBM assembloids, which are fusions of glioblastoma organoids (GBOs) from resected patient tumor tissue and cerebral and vessel organoids derived from human pluripotent stem cells. The proposed methodology will be used to detect and treat brain tumors, which will enable personalized treatment for brain tumor diseases. By universalizing the methodology we present, we aim to encourage other scientists to further develop existing models for studying these lethal tumors.