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: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.

OBJECTIVES/OBJECTIVE:To review the current evidence regarding thyroid artery embolization (TAE) for the management of benign and select malignant thyroid diseases and to present expert consensus recommendations from the North American Society for Interventional Thyroidology (NASIT) on patient selection, indications, technical considerations, safety, clinical outcomes, and the role of TAE in contemporary thyroid disease management. METHODS:A multidisciplinary consensus was developed through literature review and expert input addressing indications, technique, outcomes, and controversies, including comparisons with surgery, thermal ablation and radioactive iodine therapy. RESULTS:TAE is effective for large-volume disease (>30 mL) and substernal extension, achieving 32-73% volume reduction at 6 months with improvement in symptoms and cosmesis. In hyperthyroidism (Graves' disease and toxic multinodular goiter), euthyroid rates exceed 70% with durability up to 50 months. Preoperative use reduces operative time and blood loss. Complications are typically mild and self-limited; serious events are rare. Advances include pressure-enabled embolization and collateral flow management. CONCLUSIONS:TAE is a safe, effective, and durable option for selected patients with large symptomatic benign nodules and goiters, with additional roles in hyperthyroidism and as a surgical adjunct. Further studies are needed to standardize technique and define its role relative to other therapies.

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.

Deep venous drainage (DVD) is considered a negative prognostic factor in AVM surgery, yet its effect on postoperative functional decline remains incompletely defined. This study evaluates whether DVD predicts worsened functional status after surgical resection of Spetzler-Martin Grade II-III AVMs. This retrospective multicenter study analyzed 129 patients with Spetzler-Martin Grade II-III AVMs across nine centers in North America and Europe who underwent primary surgical resection. We excluded cases with prior endovascular or stereotactic interventions. The primary outcome measured was poor functional status, defined as modified Rankin Scale (mRS) score 3-6 at last follow up. Among 129 patients with Spetzler-Martin Grade II-III AVMs, 38 (29.5%) exhibited deep venous drainage (DVD). Poor functional outcome (mRS ≥ 3) at last follow-up occurred in 14 patients (10.9%). This occurred in 6 of 38 patients with DVD (15.8%) compared with 8 of 91 without DVD (8.8%; Fisher's exact p = 0.244). On univariate Firth-penalized logistic regression, DVD was not significantly associated with poor outcome (OR 1.96, 95% CI 0.65-5.89; p = 0.228). In the primary reduced Firth model adjusting for age and pre-existing functional disability, DVD was independently associated with poor outcome (OR 6.87, 95% CI 1.07-44.20; p = 0.042). Increasing age (OR 1.08 per year, 95% CI 1.02-1.13; p = 0.004) and pre-existing functional disability (OR 6.53, 95% CI 1.63-26.22; p = 0.008) were also independently associated with poor outcome. DVD is associated with functional decline following surgical resection of Spetzler-Martin Grade II-III AVMs after adjustment for age and pre-existing functional disability.

BACKGROUND AND OBJECTIVES/OBJECTIVE:Traditional medical board examinations present clinical information in static vignettes with multiple-choices (MC), fundamentally different from how physicians gather and integrate data in practice. Recent advances in large language models (LLMs) offer promising approaches to creating more realistic clinical interactive conversations. However, these approaches are limited in neurosurgery, where patient communication capacity varies significantly and diagnosis heavily relies on objective data such as imaging and neurological examinations. We aimed to develop and evaluate a multi-artificial intelligence (AI) agent conversation framework for neurosurgical case assessment that enables realistic clinical interactions through simulated patients and structured access to objective clinical data. METHODS:We developed a framework to convert 608 Self-Assessment in Neurological Surgery first-order diagnosis questions into conversation sessions using 3 specialized AI agents: patient AI for subjective information, system AI for objective data, and clinical AI for diagnostic reasoning. We evaluated generative pretrained transformer 4o's (GPT-4o's) diagnostic accuracy across traditional vignettes, patient-only conversations, and patient + system AI interactions, with human benchmark testing from 10 neurosurgery residents. RESULTS:= .0030) using fewer interactions and reported high educational value of the interactive format. CONCLUSION/CONCLUSIONS:This multi-AI agent framework provides both a more challenging evaluation method for LLMs and an engaging educational tool for neurosurgical training. The significant performance drops in conversational formats suggest that traditional MC testing may overestimate LLMs' clinical reasoning capabilities, while the framework's interactive nature offers promising applications for enhancing medical education.

OBJECTIVE:The cause of perimesencephalic subarachnoid hemorrhage (pmSAH) is unclear but has historically been attributed to a venous source. The authors hypothesized that high-resolution cone-beam CT (CBCT) during angiography could better identify pmSAH etiology. METHODS:All patients with pmSAH treated at the authors' institution between January 2023 and December 2024 were retrospectively analyzed. Patients were excluded if CBCT was not performed as part of the digital subtraction angiography (DSA), if CBCT source data were not available for review, or if the images were deemed to be low quality. All images were reviewed by 2 neuroangiographers with extensive neurovascular imaging experience and discussed until consensus agreement. Data were recorded as counts and percentages. RESULTS:Among 152 patients who presented with spontaneous SAH in 2023-2024, 22 had a pmSAH defined according to the Rinkel criteria. These 22 patients had a catheter angiogram performed on 1 of 2 biplane machines. Thirteen of those patients had high-quality CBCT data available for review, 8 (61%) of whom were found to harbor a basilar perforator focal outpouching consistent with a site of rupture. All patients with pmSAH, including the 8 found to have a basilar perforator aneurysm, achieved an excellent neurological recovery with resolution of the basilar perforator finding on follow-up DSA with CBCT and without experiencing a re-rupture event or clinically significant vasospasm. CONCLUSIONS:In the setting of pmSAH, high-resolution CBCT acquired as part of catheter angiography frequently identifies a basilar perforator pseudoaneurysm. Conservative management was associated with excellent outcomes in this series. The authors propose that in the setting of pmSAH, a high suspicion of an arterial etiology should be considered until proven otherwise.

Carotid webs are increasingly recognized as an underdiagnosed etiology of ischemic stroke, especially in young, otherwise healthy patients. These fibrous intimal protrusions create regions of flow stasis within the internal carotid artery, predisposing to thromboembolism. Diagnosis remains challenging due to their subtle radiographic appearance and underappreciation in clinical practice. While antiplatelet therapy or anticoagulation used to be the cornerstone of management, medical therapy alone has been found to be insufficient for stroke prevention in symptomatic patients. Definitive intervention includes carotid artery stenting or carotid endarterectomy; both have demonstrated excellent safety and efficacy. Risk stratification for symptomatic and asymptomatic carotid webs remains an area of active research, with emerging evidence suggesting that specific anatomic features, termed the carotid web angioarchitecture, may help predict stroke risk. Further studies are needed to determine the role of preventative intervention. A deeper understanding of carotid web pathogenesis, natural history, and hemodynamic impact is critical for guiding clinical decision-making.

BACKGROUND AND OBJECTIVES/OBJECTIVE:Arteriovenous malformations (AVMs) with perinidal aneurysms and single draining vein are associated with an elevated risk of rupture and increased procedural complexity. The role of preoperative embolization in this high-risk anatomical subset remains unclear. This study aimed to evaluate the safety and efficacy of microsurgery with preoperative embolization, compared with microsurgery alone in patients with such AVMs. METHODS:We conducted a multicenter retrospective analysis of an AVM registry from the MISTA (Multicenter International Study for Treatment of Brain AVMs) consortium and included AVMs with perinidal aneurysms and a single draining vein. Baseline characteristics, angiographic outcomes, functional outcomes, and complication rates were compared. Propensity score weighting (PSW) using the covariate balancing method was applied to adjust for baseline differences. RESULTS:Out of a total of 1919 patients, 65 met the inclusion criteria; 45 patients underwent preoperative embolization followed by microsurgery, and 20 underwent microsurgery alone. After adjustment, complete obliteration rates were similar between groups (OR 0.87, 95% CI 0.04 to 16.33, P=0.92), as were rates of functional independence at discharge and follow-up. Overall complication, symptomatic complication, and mortality rates did not differ significantly between groups. However, permanent complications were significantly lower in patients with preoperative embolization (OR 0.06, 95% CI 0.004 to 0.84, P=0.03). DISCUSSION/CONCLUSIONS:In patients with AVMs featuring perinidal aneurysms and single draining vein, preoperative embolization followed by microsurgery was associated with fewer permanent complications and no increase in adverse outcomes compared with microsurgery alone. However, given the small number of events, this finding should be interpreted cautiously.

OBJECTIVE:To validate the carotid web (CW) risk stratification assessment described in previous works within a larger cohort of patients with symptomatic and incidentally found asymptomatic CWs. METHODS:A retrospective analysis of our institution's electronic medical records identified all patients with a diagnosis of CW from 2017 to 2024. We included symptomatic patients and those with asymptomatic CWs, that is, incidentally found webs without history of stroke or transient ischemic attack. Patient charts were reviewed for demographics, imaging, comorbidities, and a diagnosis of stroke after diagnosis of asymptomatic CW. All angles were measured as described in previous work on a sagittal reconstruction of neck CT angiography in which the common carotid artery (CCA), external carotid artery, and internal carotid artery (ICA) were well visualized, together with the CW itself. Principal component analysis and logistic regression were performed to evaluate the association between high-risk angles and stroke risk.  RESULTS: Twenty-six symptomatic and 26 asymptomatic patients were identified. Of note, the number of patients with hypertension, hyperlipidemia, and smoking history was 17 (65.0%), 16 (62.0%), and 8 (31.0%) for symptomatic patients and 18 (69.0%), 17 (65.0%), and 15 (58.0%) for asymptomatic patients. All angular measurements showed statistically significant associations with stroke status. The CCA-web-pouch angle showed the strongest association (p=2.07×10⁻⁴), followed by the CCA-pouch-tip angle (p=3.23×10⁻⁴), ICA-web-pouch angle (p=0.004), and ICA-pouch-tip angle (p=0.005). Each additional high-risk angle increased the odds of stroke by 9.47-fold (p<0.0001). The associated probability of stroke increased from 6.3% with no high-risk angles to 39.1% with one high-risk angle and further to 85.9% with two high-risk angles. The model demonstrated high sensitivity, correctly identifying 84.6% of positive cases, and high specificity, correctly identifying 88.5% of negative cases. The F1 score was 0.863, indicating good overall model performance.  CONCLUSION: Given this successful stratification of CWs into high- and low-risk groups, the utilization of geometric CW parameters may play a role in improving patient selection for intervention in the setting of incidentally diagnosed CW. .