Faculty Bibliography

INTRODUCTION/UNASSIGNED:The impact of multiple feeding arteries on clinical outcomes of cerebral arteriovenous malformations (AVMs) is not well understood. This study aims to compare outcomes between AVMs with multiple versus single feeding arteries. PATIENTS AND METHODS/UNASSIGNED:Data from the Multicenter International Study for Treatment of Brain AVMs (MISTA) consortium were analyzed. Propensity score matching (PSM) was used to balance cohorts. Subgroup analysis was conducted for ruptured and unruptured AVMs and different treatment options, and multivariable logistic regression was applied to assess the impact of feeding artery origin. RESULTS/UNASSIGNED: = 0.002). DISCUSSION AND CONCLUSION/UNASSIGNED:AVMs with a single feeding artery were more likely to present with rupture, but no significant differences in obliteration rates or complications were observed between the groups after PSM. These findings suggest that while the number of feeding arteries may influence the initial presentation, it does not appear to impact overall treatment success or patient prognosis. Further prospective studies are needed to confirm these findings.

BACKGROUND:Arteriovenous malformations (AVMs) are complex vascular anomalies with a high risk of hemorrhage and neurological deficits, especially when located in eloquent brain regions. The eloquence of an AVM location is a critical factor in the treatment planning, influencing both the risk of complications and long-term functional outcomes. This study aims to compare outcomes between eloquent and non-eloquent AVMs. METHODS:This multicenter, retrospective study utilized data from the Multicenter International Study for Treatment of Brain AVMs (MISTA) consortium. Patients with eloquent and non-eloquent AVMs were compared on baseline characteristics, angiographic outcomes, and functional outcomes using the modified Rankin Scale (mRS). Propensity score weighting (IPTW) was applied to adjust for confounding variables. RESULTS:The study included 1,013 patients, with 498 (49.2%) AVMs located in eloquent regions and 515 (50.8%) in non-eloquent regions. In unadjusted analysis, eloquent AVMs had lower complete obliteration rates (67.6% vs. 79.5%, OR: 0.53, 95% CI: 0.39-0.72, p < 0.001) and higher complication rates (24.5% vs. 19.0%, OR: 1.38, 95% CI: 1.02-1.86, p = 0.03) compared to non-eloquent AVMs. After IPTW adjustment, eloquent AVMs continued to show significantly higher odds of overall complications (OR: 1.68, 95% CI: 1.12-2.52, p = 0.01) and symptomatic complications (OR: 1.77, 95% CI: 1.12-2.80, p = 0.01). Secondary analysis within the eloquent group indicated that embolization was linked to an elevated risk of complications. Surgery and radiosurgery showed comparable functional outcomes at last follow-up and complications rates with higher complete obliteration rates in surgery. CONCLUSION/CONCLUSIONS:AVMs in eloquent brain areas present higher risks of complications and lower obliteration rates, emphasizing the need for cautious, individualized treatment planning. Within the eloquent group, embolization increased the risk of complications, while surgery and radiosurgery showed comparable functional outcomes at last follow-up and complication rates with higher complete obliteration rates in surgery. These findings highlight the importance of location in AVM management and support further research focusing on comparing treatment strategies for AVMs in eloquent brain areas.

The adoption of large language models (LLMs) in healthcare demands a careful analysis of their potential to spread false medical knowledge. Because LLMs ingest massive volumes of data from the open Internet during training, they are potentially exposed to unverified medical knowledge that may include deliberately planted misinformation. Here, we perform a threat assessment that simulates a data-poisoning attack against The Pile, a popular dataset used for LLM development. We find that replacement of just 0.001% of training tokens with medical misinformation results in harmful models more likely to propagate medical errors. Furthermore, we discover that corrupted models match the performance of their corruption-free counterparts on open-source benchmarks routinely used to evaluate medical LLMs. Using biomedical knowledge graphs to screen medical LLM outputs, we propose a harm mitigation strategy that captures 91.9% of harmful content (F1 = 85.7%). Our algorithm provides a unique method to validate stochastically generated LLM outputs against hard-coded relationships in knowledge graphs. In view of current calls for improved data provenance and transparent LLM development, we hope to raise awareness of emergent risks from LLMs trained indiscriminately on web-scraped data, particularly in healthcare where misinformation can potentially compromise patient safety.

BACKGROUND:Spetzler-Martin (SM) Grade III brain arteriovenous malformations (BAVMs) represent a transitional risk zone between low- and high-grade BAVMs, characterized by diverse angioarchitecture. The primary treatment options are endovascular embolization, microsurgical resection (MS), and stereotactic radiosurgery (SRS). This study compares the efficacy and outcomes of MS and SRS. METHODS:We conducted a multicenter, retrospective study involving patients from the MISTA database with SM Grade III BAVMs treated with MS or SRS between 2010 and 2023. Propensity matching was based on age, favorable modified Rankin Score (mRS) at presentation, nidus size, rupture status, location depth, and eloquence. RESULTS:, p = 0.6) were similar. MS showed higher obliteration rates (93.3 %) compared to SRS (46.7 %) at the last follow-up (p < 0.001). The median time to obliteration post-SRS was 31.5 months (IQR: 15.3-60.0). SRS obliteration rates were 19 %, 29 %, and 59 % at 24, 36, and 60 months, respectively. Overall complication rates (MS: 30 % vs. SRS: 20 %, p = 0.4) and permanent complications (MS: 10 % vs. SRS: 13.3 %, p > 0.9) were similar. Hemorrhage occurred once in the MS group and none in the SRS (p > 0.9). Favorable outcomes (mRS 0-2) were higher with SRS than MS (93.3 % vs 80.0 %, p = 0.3), with one AVM-related mortality in the MS group. CONCLUSION/CONCLUSIONS:MS and SRS are viable treatments for SM Grade III BAVMs. Treatment choice should be individualized by a multidisciplinary team, considering patient goals.

BACKGROUND:Arteriovenous malformations (AVMs) are uncommon cerebral lesions that can cause significant neurological complications. Surgical resection is the gold standard for treatment, but endovascular embolization and stereotactic radiosurgery (SRS) are viable alternatives. OBJECTIVE:To compare the outcomes of endovascular embolization versus SRS in the treatment of AVMs with Spetzler-Martin grades I-III. METHODS:This study combined retrospective data from 10 academic institutions in North America and Europe. Patients aged 1 to 90 years who underwent endovascular embolization or SRS for AVMs with Spetzler-Martin grades I-III between January 2010 and December 2023 were included. RESULTS:The study included 244 patients, including 84 who had endovascular embolization and 160 who had SRS. Before propensity score matching (PSM), complete obliteration at the last follow-up was achieved in 74.5% of the SRS group compared with 57.8% of the embolization group (OR=0.47; 95% CI 0.26 to 0.48; P=0.01). After propensity score matching, SRS still achieved significantly higher occlusion rates at last follow-up (78.9% vs 55.3%; OR=0.32; 95% CI 0.12 to 0.90; P=0.03).Hemorrhagic complications were higher in the embolization group than in the SRS group, although this difference did not reach statistical significance after PSM (13.2% vs 2.6%; OR=5.6; 95% CI 0.62 to 50.47; P=0.12). Similarly, re-treatment rate was higher in the embolization group (10.5% vs 5.3%; OR=2.11; 95% CI 0.36 to 12.31; P=0.40) compared with the SRS group. CONCLUSION/CONCLUSIONS:Our findings indicate that SRS has a significantly higher obliteration rate at last follow-up compared with endovascular embolization. Also, SRS has a higher tendency for fewer hemorrhagic complications and lower re-treatment rate. Further prospective studies are needed.

BACKGROUND AND OBJECTIVES/OBJECTIVE:Scholarly output is accelerating in medical domains, making it challenging to keep up with the latest neurosurgical literature. The emergence of large language models (LLMs) has facilitated rapid, high-quality text summarization. However, LLMs cannot autonomously conduct literature reviews and are prone to hallucinating source material. We devised a novel strategy that combines Reference Publication Year Spectroscopy-a bibliometric technique for identifying foundational articles within a corpus-with LLMs to automatically summarize and cite salient details from articles. We demonstrate our approach for four common spinal conditions in a proof of concept. METHODS:Reference Publication Year Spectroscopy identified seminal articles from the corpora of literature for cervical myelopathy, lumbar radiculopathy, lumbar stenosis, and adjacent segment disease. The article text was split into 1024-token chunks. Queries from three knowledge domains (surgical management, pathophysiology, and natural history) were constructed. The most relevant article chunks for each query were retrieved from a vector database using chain-of-thought prompting. LLMs automatically summarized the literature into a comprehensive narrative with fully referenced facts and statistics. Information was verified through manual review, and spine surgery faculty were surveyed for qualitative feedback. RESULTS:Our tandem approach cost less than $1 for each condition and ran within 5 minutes. Generative Pre-trained Transformer-4 was the best-performing model, with a near-perfect 97.5% citation accuracy. Surveys of spine faculty helped refine the prompting scheme to improve the cohesion and accessibility summaries. The final artificial intelligence-generated text provided high-fidelity summaries of each pathology's most clinically relevant information. CONCLUSION/CONCLUSIONS:We demonstrate the rapid, automated summarization of seminal articles for four common spinal pathologies, with a generalizable workflow implemented using consumer-grade hardware. Our tandem strategy fuses bibliometrics and artificial intelligence to bridge the gap toward fully automated knowledge distillation, obviating the need for manual literature review and article selection.

BACKGROUND AND OBJECTIVE/OBJECTIVE:This investigation evaluates the safety and efficacy of stereotactic radiosurgery (SRS) vs observation for Koos grade I vestibular schwannomas (VS). METHODS:In a multicenter study, we retrospectively analyzed data of patients with Koos grade I VS who underwent SRS (SRS group) or were observed (observation group). Propensity score matching was used to equilibrate demographics, tumor size, and audiometric data across groups. The outcome analyzed included tumor control, preservation of serviceable hearing, and neurological function. RESULTS:The study matched 142 patients, providing a median follow-up period of 36 months. SRS significantly enhanced tumor control compared with observation, with a 100% control rate at both 5- and 8-year marks in the SRS group vs 48.6% and 29.5% in the observation group at the same time intervals, respectively ( P < .001). Preservation of serviceable hearing outcomes between groups showed no significant difference at 5 and 8 years, ensuring a comparable quality of auditory function (SRS 70.1% vs observation 53.4% at 5 years; P = .33). Furthermore, SRS was associated with a reduced likelihood of tinnitus (odds ratio [OR] = 0.46, P = .04), vestibular dysfunction (OR = 0.17, P = .002), and overall cranial nerve dysfunction (OR = 0.49, P = .03) at last follow-up. CONCLUSION/CONCLUSIONS:SRS management of patients with Koos grade I VS was associated with superior tumor control and reduced odds for cranial nerve dysfunction, while not compromising hearing preservation compared with observation. These findings support the safety and efficacy of SRS as a primary care approach for this patient population.

BACKGROUND AND OBJECTIVES/OBJECTIVE:This study assessed the efficacy and safety of stereotactic radiosurgery (SRS) in comparison with watchful waiting for managing Koos grade II vestibular schwannomas (VS). METHODS:A retrospective, multicentric analysis was conducted, focusing on patients with Koos grade II VS who either received SRS (SRS group) or were observed (observation group). To ensure comparability between groups, propensity score matching was used, including factors such as demographic characteristics, tumor dimensions, and hearing assessments. The primary end points examined were tumor control, maintenance of serviceable hearing, and neurological outcomes. RESULTS:A total of 92 patients were equally matched across both cohorts, with a median follow-up of 37 months for the SRS group and 27.5 months for those observed. The SRS cohort exhibited superior tumor control over observation across 3, 5, and 8 years, achieving a 100% control rate vs 47.9%, 40.1%, and 34.3% for the observation group at these time intervals, respectively ( P < .001). Serviceable hearing preservation rates were comparable between the 2 groups throughout 3, 5, and 7 years (72.9% for SRS vs 65.4% for observation at 3 years; P = .86). Moreover, SRS management correlated with a lower incidence of vestibular symptoms (odds ratio = 0.11, P = .002), with no significant disparity in the deterioration of cranial nerve (CN) V or CN VII functions. Notably, the likelihood of experiencing any CN impairment was significantly diminished in the SRS cohort (odds ratio = 0.47, P = .04). CONCLUSION/CONCLUSIONS:For patients with Koos grade II VS, SRS offers superior tumor control rate and a lower risk of CN dysfunction without sacrificing hearing preservation.

PURPOSE/OBJECTIVE:The purpose of this study was to investigate an extended self-adapting nnU-Net framework for detecting and segmenting brain metastases (BM) on magnetic resonance imaging (MRI). METHODS AND MATERIALS/METHODS:Six different nnU-Net systems with adaptive data sampling, adaptive Dice loss, or different patch/batch sizes were trained and tested for detecting and segmenting intraparenchymal BM with a size ≥2 mm on 3 Dimensional (3D) post-Gd T1-weighted MRI volumes using 2092 patients from 7 institutions (1712, 195, and 185 patients for training, validation, and testing, respectively). Gross tumor volumes of BM delineated by physicians for stereotactic radiosurgery were collected retrospectively and curated at each institute. Additional centralized data curation was carried out to create gross tumor volumes of uncontoured BM by 2 radiologists to improve the accuracy of ground truth. The training data set was augmented with synthetic BMs of 1025 MRI volumes using a 3D generative pipeline. BM detection was evaluated by lesion-level sensitivity and false-positive (FP) rate. BM segmentation was assessed by lesion-level Dice similarity coefficient, 95-percentile Hausdorff distance, and average Hausdorff distance (HD). The performances were assessed across different BM sizes. Additional testing was performed using a second data set of 206 patients. RESULTS:. Mean values of Dice similarity coefficient, 95-percentile Hausdorff distance, and average HD of all detected BMs were 0.758, 1.45, and 0.23 mm, respectively. Performances on the second testing data set achieved a sensitivity of 0.907 at an FP rate of 0.57 ± 0.85 for all BM sizes, and an average HD of 0.33 mm for all detected BM. CONCLUSIONS:Our proposed extension of the self-configuring nnU-Net framework substantially improved small BM detection sensitivity while maintaining a controlled FP rate. Clinical utility of the extended nnU-Net model for assisting early BM detection and stereotactic radiosurgery planning will be investigated.