Faculty Bibliography

BACKGROUND AND OBJECTIVES/OBJECTIVE:Multiple-choice questions are the primary assessment format for neurosurgical board certification. Creating high-quality examination questions requires significant expert time and resources. The goal of this study was to develop an automated system to generate board-style neurosurgical multiple-choice questions using state-of-the-art vision-language models and compare their quality with authentic self-assessment questions. METHODS:articles. We generated 89 587 synthetic questions: 45 689 with GPT-4o and 43 898 with Claude. Each question was associated with a single image extracted from the articles' figures. We evaluated the quality of synthetic questions through 5 surveys comparing 20 synthetic questions (10 from each model) with 10 authentic questions from the Self-Assessment for Neurological Surgeons (SANS) question bank. Each survey was completed by a neurosurgery resident and an attending who guessed the source [human vs artificial intelligence (AI)-generated] and rated suitability for board examination use. We also evaluated the question-answering performance of the generalist GPT-4o and the specialized CNS-Obsidian. RESULTS:). CONCLUSION/CONCLUSIONS:Although quality gaps exist between AI-generated and human-created neurosurgical board examination questions, our approach demonstrates the potential of vision-language models to augment assessment development in specialized medical fields, reducing the burden on examination boards and credentialing organizations.

BACKGROUND AND OBJECTIVES/OBJECTIVE:Normal pressure hydrocephalus (NPH) affects approximately 0.2% to 2.9% of adults older than 65 years and is characterized by progressive gait impairment, cognitive decline, and urinary incontinence. Cerebrospinal fluid shunting is the mainstay of treatment but carries a risk of overdrainage, leading to subdural collections. Evidence guiding the management of these collections in patients with shunted NPH remains limited. We aimed to characterize treatment strategies and outcomes for subdural collections in patients with shunted NPH and to propose a practical management framework for nonemergent presentations. METHODS:We performed a retrospective cohort study of patients with shunted NPH who developed subdural collections requiring treatment between January 2014 and September 2025. Interventions included valve pressure adjustment, middle meningeal artery (MMA) embolization, and surgical evacuation. Clinical, radiographic, and treatment data were extracted from the electronic health record. Time-to-resolution and recurrence were assessed using Kaplan-Meier analysis. Receiver operating characteristic analysis determined the predictive value of baseline collection volume for failure of valve adjustment. RESULTS:Sixty patients with 96 total collections underwent 91 valve pressure increases, 8 craniotomies for 8 collections, 23 MMA embolization procedures for 32 collections, and 15 subdural evacuating port system procedures for 16 collections. The median time to resolution for collections that underwent valve adjustment only, MMA embolization without evacuation, evacuation without MMA embolization, and both MMA embolization and evacuation was 1.91, 3.99, 1.77, and 5.17 months, respectively. Using pretreatment volume to predict valve adjustment failure demonstrated an area under the curve of 0.75 with an optimal cutoff of 40.05 cm3, corresponding to a sensitivity of 75.0% and a specificity of 71.4%. CONCLUSION/CONCLUSIONS:Baseline hematoma volume may help identify which shunted NPH patients with subdural hematomas are at higher risk of failing valve adjustment alone. We propose a volume-informed management framework in which larger hematomas (approximately >40 cm3) prompt earlier consideration of treatment escalation.

BACKGROUND AND OBJECTIVES/OBJECTIVE:Cerebrospinal fluid shunting is a common treatment for hydrocephalus. Several techniques for ventricular catheter placement exist, yet comparative safety data remain limited. This study evaluated whether stereotactic navigation-assisted ventricular catheter placement is associated with reduced 30-day postoperative intracranial hemorrhage (ICH) rates. METHODS:This was a retrospective analysis of adults who underwent frontal or occipital shunt placement for hydrocephalus between 2014 and 2023. The primary exposure variable was the use of stereotactic navigation for ventricular catheter placement. The primary outcomes were hospital length of stay (LOS) and 30-day postoperative ICH. RESULTS:A total of 541 shunt placements were included (median age 72.5 years; 51.6% male). Normal pressure hydrocephalus represented the most common underlying diagnosis (60.3%). Stereotactic navigation was used in 221 cases (40.9%) and was more frequently employed in younger patients (66.0 vs 73.9 years; P < .001), more recent years (median 2021 vs 2019; P < .001), and cases involving non-normal pressure hydrocephalus diagnoses (P < .001). Overall, 12 patients (2.2%) developed ICH within 30 days of surgery. The 30-day ICH rate was significantly lower with stereotactic navigation (0.5% vs 3.4%; odds ratio 0.13, 95% CI 0.02-1.00; P = .033), corresponding to an absolute risk reduction of 2.99% and a number needed to treat of 34. Kaplan-Meier analysis demonstrated superior hemorrhage-free survival with navigation-assisted placement (log-rank P = .021), with most events occurring within the first 14 days postoperatively. Propensity-matched analysis demonstrated a similar effect size but fell short of statistical significance. Navigation use was not independently associated with LOS or any other complications. CONCLUSION/CONCLUSIONS:Stereotactic navigation-assisted ventricular catheter placement was associated with lower rates of 30-day postoperative ICH compared with non-navigation-assisted placement, with a number needed to treat of 34. Given the absence of adverse effects on LOS or overall complication rates, these findings support routine consideration of neuronavigation during shunt placement when available.

PURPOSE/OBJECTIVE:Pediatric low-grade gliomas (pLGGs) typically have excellent long-term outcomes; intratumoral hemorrhage is a rare, potentially dangerous complication. Hemorrhage risk in the context of molecular alterations and targeted therapies remains poorly characterized. We analyzed the incidence, timing, and independent risk factors for hemorrhage in a large contemporary pLGG cohort. METHODS:We conducted a retrospective cohort study of 236 children with pLGG treated at a single center (2011-2025). Clinical, radiographic, and molecular variables were abstracted. The primary endpoint was spontaneous tumoral hemorrhage. Time-to-event analyses utilized Kaplan-Meier methods and Cox proportional hazards modeling; penalized regression mitigated overfitting given the event rarity. RESULTS:Twelve patients (5.1%) experienced hemorrhage over 2,234 person-years (incidence: 0.54/100 person-years). Hemorrhage typically occurred years after initial tumor diagnosis (median 6.4 years). The presence of a KIAA1549::BRAF fusion in the tumor had the strongest association with hemorrhage, persisting across multivariable models (approximately sixfold increased risk), although estimates were limited by low event number. MAPK inhibitor exposure (specifically binimetinib and tovorafenib) was associated with hemorrhage in univariate analysis but partially confounded by fusion status. CSF diversion independently increased risk at brainstem, optic pathway, and hypothalamic locations. No hemorrhages occurred among patients with underlying genetic syndromes, including neurofibromatosis type 1 and tuberous sclerosis. CONCLUSION/CONCLUSIONS:Hemorrhage in pLGG is an infrequent late complication associated with tumor biology. KIAA1549::BRAF fusion may identify a higher-risk subgroup, with MAPK inhibitor exposure and CSF diversion further modifying risk. These findings support biology-informed surveillance and personalized management strategies for at-risk children.

INTRODUCTION/BACKGROUND:Subway-related trauma is an understudied category of urban injury. Prior work has focused on high-acuity train-contact events, with less attention to more common mechanisms and the roles of psychiatric illness and substance use. We analyzed a large contemporary cohort to define epidemiology, injury patterns, and outcome predictors across the full spectrum of subway trauma. METHODS:We performed a retrospective cohort study of adults presenting with subway-related injury to a Level I trauma center (2018-2024). Registry data and manual chart review captured demographics, mechanisms, comorbidities, intoxication, injury severity, and outcomes. Injuries were categorized using validated natural language processing. Multivariable logistic regression identified predictors of train-contact mechanism, major trauma (ISS ≥ 15), TBI, and assault. RESULTS:Among 809 patients, median age was 50 years and 81.2% were male. Falls were most common (57.2%), followed by train contact (16.8%) and assault (16.7%). Median ISS was 9, with 25.0% sustaining major trauma. TBI occurred in 40.5% and did not differ by mechanism. Train contact was the most severe mechanism and the only independent predictor of major trauma (aOR 5.08). Psychiatric diagnosis (aOR 1.59) and acute intoxication (aOR 1.69) independently predicted train contact, while intoxication predicted TBI (aOR 1.66). Psychiatric illness and substance use disorder were associated with longer hospitalization independent of injury severity. Assault exposure varied by race/ethnicity. CONCLUSION/CONCLUSIONS:Most subway injuries result from falls rather than intentional mechanisms. Train contact drives severity, while intoxication increases TBI risk. Psychiatric and substance use comorbidities prolong hospitalization, supporting routine behavioral health screening and targeted prevention strategies.

INTRODUCTION/BACKGROUND:Hemispherotomy is an effective treatment for children with drug-resistant epilepsy (DRE). While hemispherotomy techniques and indications have evolved, access remains predominantly constrained to high-resource settings. METHODS:We performed a retrospective analysis of children who underwent hemispherotomy from 2011 to 2023 by a hybrid team, including local Panamanian and US neurologists, neurosurgeons, and EEG technicians and analyzed surgical, epilepsy, and quality of life (QoL) parameters. Follow-up data were collected according to the International Consortium for Health Outcomes Measurement (ICHOM) guidelines for children with epilepsy. RESULTS:Twenty-three children underwent hemispherotomy. The median age at surgery was 10 years (range 2-20). The median follow-up time was 6 years (range 1-13). The etiology of DRE included malformations of cortical development in 14 children (60.8%), including 8 (34.8%) with schizencephaly, and secondary causes in 9 children (39.1%). Seizure frequency improved for all 23 children (100%): Engel I was achieved in 15 children (65.2%), Engel II (26.1%) in six children, and Engel III (8.7%) in two children. Patients with seizure freedom had significantly fewer preoperative seizures per day than patients with seizure recurrence. Complications occurred in six children (26.1%): 2 wound infections, 2 meningitis, 1 femoral vein thrombosis, and 1 wound hematoma with return to OR. There were no perioperative mortality and no postoperative hydrocephalus or CSF diversion. QoL-related outcomes were available for 16 children: 16/16 (100%) reported that the surgery was a worthwhile and repeatable choice, 14 (87.5%) reported improved cognitive function, the median QOLCE-16 score was 62.5 ± 21. CONCLUSION/CONCLUSIONS:Hemispherotomy for DRE in selected children is a safe and effective surgery in a public children's hospital in a low-resource setting. At last follow-up, the majority of children were seizure-free, and all children had decreased seizure frequency. Families reported improved cognitive function, improved QoL and high satisfaction with their decision to pursue this surgery.

BACKGROUND AND OBJECTIVES/OBJECTIVE:General purpose vision-language models (VLMs) demonstrate impressive capabilities, but their opaque training on uncurated internet data poses critical limitations for high-stakes decision making, such as in neurosurgery. We present CNS-Obsidian, a neurosurgical VLM trained on peer-reviewed neurosurgical literature, and demonstrate its clinical utility compared with GPT-4o in a real-world setting. METHODS:We compiled 23 984 articles from Neurosurgery Publications journals, yielding 78 853 figures and captions. Using GPT-4o and Claude Sonnet-3.5, we converted these image-text pairs into 263 064 training samples across 3 formats: instruction fine-tuning, multiple-choice questions, and differential diagnosis. We trained CNS-Obsidian, a fine-tune of the 34-billion parameter Large Language and Visual Assistant-Next model. In a blinded, randomized deployment trial at NYU Langone Health (August 30-November 30, 2024), neurosurgeons were assigned to use either CNS-Obsidian or a Health Insurance Portability and Accountability Act-compliant GPT-4o end point as a diagnostic copilot after patient consultations. Primary outcomes were diagnostic helpfulness and accuracy, assessed through user ratings and presence of the correct diagnosis within the VLM-provided differential, respectively. RESULTS:CNS-Obsidian matched GPT-4o on synthetic questions (76.13% vs 77.54%, P = .235), but only achieved 46.81% accuracy on human-generated questions vs GPT-4o's 65.70% (P < 10-15). In the randomized trial, 70 consultations were evaluated (32 CNS-Obsidian, 38 GPT-4o) from 959 total consults (7.3% utilization). CNS-Obsidian received positive ratings in 40.62% of cases vs 57.89% for GPT-4o (P = .230). Both models included correct diagnosis in approximately 60% of cases (59.38% vs 65.79%, P = .626). CONCLUSION/CONCLUSIONS:Domain-specific VLMs trained on curated scientific literature can approach frontier model performance in specialized medical domains despite being orders of magnitude smaller and less expensive to train. This establishes a transparent framework for scientific communities to build specialized artificial intelligence models. However, low clinical utilization suggests chatbot interfaces may not align with specialist workflows, indicating need for alternative artificial intelligence integration strategies.

BACKGROUND AND OBJECTIVES/OBJECTIVE:Normal pressure hydrocephalus (NPH) is characterized by the classic triad of cognitive decline, gait instability, and urinary incontinence in the setting of ventriculomegaly with normal intracranial pressure. Cerebrospinal fluid diversion is the current standard treatment, yet it carries a risk of overdrainage, resulting in subdural hematoma or hygroma. Different valves have been developed to mitigate this risk, yet consensus remains unclear regarding optimal valve for NPH. METHODS:We performed a retrospective cohort study on all patients with NPH who underwent cerebrospinal fluid shunting or revision between January 2014 and September 2025 at our institution. Demographic, clinical, and radiological data were collected from the electronic health record. Kaplan-Meier survival analysis, univariate logistic regression, and multivariate modeling were used to identify predictors of subdural collections and the need for surgical treatment. RESULTS:Since our change in practice from the Integra NPH Low Flow Valve (Low Flow OSV) to other valves in 2022, we observed a rise in symptomatic subdural collections. Programmable valves were associated with a markedly increased 1-year risk of both subdural collection formation and need for surgical intervention compared with the Low Flow OSV. Overall, Certas and Strata valves demonstrated higher rates of subdural collections requiring surgery than the Low Flow OSV (14.6% vs 2.1%, P < .001; 10.5% vs 2.1%, P = .005, respectively). On multivariate analysis, both the Strata and Certas valves were independently associated with increased odds of developing any subdural collection and necessitating surgery. Vascular disease and dual antiplatelet therapy also increased risk. CONCLUSION/CONCLUSIONS:In this large single-center cohort study, programmable valves, specifically the Certas and Strata, were associated with an increased rate and severity of subdural collections compared with the Low Flow OSV. The use of low-flow designs may mitigate complications for the NPH population, and the use of lower programmable valve settings should be carefully considered.

In children undergoing craniotomy, the impact of postoperative pain on recovery is receiving growing recognition. While opioids are often the primary treatment, their administration requires a delicate balance between achieving sufficient analgesia and mitigating side effects like sedation, nausea, vomiting, and respiratory depression. We review the emerging adjunct treatment modality regional scalp block (RSB) infiltration for post-craniotomy pain. Postoperative pain after pediatric craniotomy can be challenging to manage and may contribute to unnecessary suffering as well as the development of long-term neurocognitive and psychological sequelae. Pain during the PICU stay is also a major risk factor for post-PICU syndrome, which involves persistent impairments in children's physical, cognitive, or mental health persisting beyond acute hospitalization. Despite increasing awareness and treatment strategies for post-craniotomy pain in adults, significant gaps remain in understanding its assessment and management in children. This review examines the current literature surrounding post-craniotomy pain management in children with a special emphasis on RSB, a treatment option increasingly used in adults and children. RSB has been shown in randomized trials to reduce postoperative pain and opioid use. However, while pediatric perioperative trials support its safety and feasibility, robust clinical evidence supporting RSB's efficacy for post-craniotomy pain in children remains limited, hindering wider translation into clinical standard. RSB infiltration is an emerging and promising technique for pediatric post-craniotomy pain management. Early evidence suggests it is both safe and effective, with potential to enhance postoperative recovery and to be integrated into clinical practice. Further research is critical to validate initial findings and better define the benefits across diverse pediatric populations.

PURPOSE/OBJECTIVE:To identify clinical variables associated with ventricular shunt infection and shunt failure in pediatric hydrocephalus. METHODS:Patients ≤ 18 years treated with ventricular shunts between 2013 and 2024 were identified from one institution's electronic medical record. Children with a confirmed diagnosis of hydrocephalus and ≥ 6 months of postoperative follow-up were included. Primary and revision shunt surgeries were included. Records were manually reviewed for clinical variables. Statistical analyses were performed using R (version 4.2.3). RESULTS:The dataset included 474 surgeries, 146 primary and 328 revisions, undergone by 226 patients. Infection necessitating removal of a previously placed shunt occurred following 3.59% (17/474) of cases. Discharge in ≤ 4 days had a 75% lower relative risk for infection compared to stays > 4 days (1.5% vs. 6% 100-day infection risk; p = 0.011). Patients who underwent revision surgeries for shunt infections were more likely to experience subsequent infections in the first 100 days postoperatively than those revised for other causes (2.42% vs. 21.05%; p < 0.0001). Patient characteristics associated with shunt failure during the 10-year study included younger age (median age: 2.23 years in those with failure vs. 6.62 years in those without; p < 0.0002) and lower weight (median weight: 11.8 kg vs. 20.3 kg; p < 0.0002) at the time of admission. Congenital hydrocephalus (OR = 1.86; p = 0.0045) and aqueductal stenosis (OR = 1.75; p = 0.025) were also associated with shunt failure. CONCLUSIONS:Length of stay > 4 days and previous shunt infection are associated with an increased risk of infection after shunt surgery. These findings are important to consider when counseling pediatric patients and during postoperative monitoring.