Faculty Bibliography

Physicians make critical time-constrained decisions every day. Clinical predictive models can help physicians and administrators make decisions by forecasting clinical and operational events. Existing structured data-based clinical predictive models have limited use in everyday practice owing to complexity in data processing, as well as model development and deployment^1-3. Here we show that unstructured clinical notes from the electronic health record can enable the training of clinical language models, which can be used as all-purpose clinical predictive engines with low-resistance development and deployment. Our approach leverages recent advances in natural language processing^4,5 to train a large language model for medical language (NYUTron) and subsequently fine-tune it across a wide range of clinical and operational predictive tasks. We evaluated our approach within our health system for five such tasks: 30-day all-cause readmission prediction, in-hospital mortality prediction, comorbidity index prediction, length of stay prediction, and insurance denial prediction. We show that NYUTron has an area under the curve (AUC) of 78.7-94.9%, with an improvement of 5.36-14.7% in the AUC compared with traditional models. We additionally demonstrate the benefits of pretraining with clinical text, the potential for increasing generalizability to different sites through fine-tuning and the full deployment of our system in a prospective, single-arm trial. These results show the potential for using clinical language models in medicine to read alongside physicians and provide guidance at the point of care.

BACKGROUND:The superficial temporal artery (STA)-to-middle cerebral artery (MCA) bypass requires precise preoperative planning, and 3-dimensional virtual reality (VR) models have recently been used to optimize planning of STA-MCA bypass. In the present report, we have described our experience with VR-based preoperative planning of STA-MCA bypass. METHODS:Patients from August 2020 to February 2022 were analyzed. For the VR group, using 3-dimensional models from the patients' preoperative computed tomography angiograms, VR was used to locate the donor vessels, potential recipient, and anastomosis sites and plan the craniotomy, which were referenced throughout surgery. Computed tomography angiograms or digital subtraction angiograms were used to plan the craniotomy for the control group. The procedure time, bypass patency, craniotomy size, and postoperative complication rates were assessed. RESULTS:The VR group included 17 patients (13 women; age, 49 ± 14 years) with Moyamoya disease (76.5%) and/or ischemic stroke (29.4%). The control group included 13 patients (8 women; age, 49 ± 12 years) with Moyamoya disease (92.3%) and/or ischemic stroke (7.3%). For all 30 patients, the preoperatively planned donor and recipient branches were effectively translated intraoperatively. No significant difference were found in the procedure time or craniotomy size between the 2 groups. Bypass patency was 94.1% for the VR group (16 of 17) and 84.6% for the control group (11 of 13). No permanent neurological deficits occurred in either group. CONCLUSIONS:Our early experience has shown that VR can serve as a useful, interactive preoperative planning tool by enhancing visualization of the spatial relationship between the STA and MCA without compromising the surgical results.

Background Knowledge regarding predictors of clinical and radiographic failures of middle meningeal artery (MMA) embolization (MMAE) treatment for chronic subdural hematoma (CSDH) is limited. Purpose To identify predictors of MMAE treatment failure for CSDH. Materials and Methods In this retrospective study, consecutive patients who underwent MMAE for CSDH from February 2018 to April 2022 at 13 U.S. centers were included. Clinical failure was defined as hematoma reaccumulation and/or neurologic deterioration requiring rescue surgery. Radiographic failure was defined as a maximal hematoma thickness reduction less than 50% at last imaging (minimum 2 weeks of head CT follow-up). Multivariable logistic regression models were constructed to identify independent failure predictors, controlling for age, sex, concurrent surgical evacuation, midline shift, hematoma thickness, and pretreatment baseline antiplatelet and anticoagulation therapy. Results Overall, 530 patients (mean age, 71.9 years ± 12.8 [SD]; 386 men; 106 with bilateral lesions) underwent 636 MMAE procedures. At presentation, the median CSDH thickness was 15 mm and 31.3% (166 of 530) and 21.7% (115 of 530) of patients were receiving antiplatelet and anticoagulation medications, respectively. Clinical failure occurred in 36 of 530 patients (6.8%, over a median follow-up of 4.1 months) and radiographic failure occurred in 26.3% (137 of 522) of procedures. At multivariable analysis, independent predictors of clinical failure were pretreatment anticoagulation therapy (odds ratio [OR], 3.23; P = .007) and an MMA diameter less than 1.5 mm (OR, 2.52; P = .027), while liquid embolic agents were associated with nonfailure (OR, 0.32; P = .011). For radiographic failure, female sex (OR, 0.36; P = .001), concurrent surgical evacuation (OR, 0.43; P = .009), and a longer imaging follow-up time were associated with nonfailure. Conversely, MMA diameter less than 1.5 mm (OR, 1.7; P = .044), midline shift (OR, 1.1; P = .02), and superselective MMA catheterization (without targeting the main MMA trunk) (OR, 2; P = .029) were associated with radiographic failure. Sensitivity analyses retained these associations. Conclusion Multiple independent predictors of failure of MMAE treatment for chronic subdural hematomas were identified, with small diameter (<1.5 mm) being the only factor independently associated with both clinical and radiographic failures. © RSNA, 2023 Supplemental material is available for this article. See also the editorial by Chaudhary and Gemmete in this issue.

BACKGROUND:By 2030, nonacute subdural hematomas (NASHs) will likely be the most common cranial neurosurgery pathology. Treatment with surgical evacuation may be necessary, but the recurrence rate after surgery is as high as 30%. Minimally invasive middle meningeal artery embolization (MMAE) during the perioperative period has been posited as an adjunctive treatment to decrease the potential for recurrence after surgical evacuation. We evaluated the safety and efficacy of concurrent MMAE in a multi-institutional cohort. METHODS:Data from 145 patients (median age 73 years) with NASH who underwent surgical evacuation and MMAE in the perioperative period were retrospectively collected from 15 institutions. The primary outcome was the rate of recurrence requiring repeat surgical intervention. We collected clinical, treatment, and radiographic data at initial presentation, after evacuation, and at 90-day follow-up. Outcomes data were also collected. RESULTS:Preoperatively, the median hematoma width was 18 mm, and subdural membranes were present on imaging in 87.3% of patients. At 90-day follow-up, median NASH width was 6 mm, and 51.4% of patients had at least a 50% decrease of NASH size on imaging. Eight percent of treated NASHs had recurrence that required additional surgical intervention. Of patients with a modified Rankin Scale score at last follow-up, 87.2% had the same or improved mRS score. The total all-cause mortality was 6.0%. CONCLUSION/CONCLUSIONS:This study provides evidence from a multi-institutional cohort that performing MMAE in the perioperative period as an adjunct to surgical evacuation is a safe and effective means to reduce recurrence in patients with NASHs.

The transpalpebral "eyelid" approach is an innovative alternative to the traditional incisions for exposure of the anterior cranial fossa for neurosurgery. Yet, there is a paucity of data on such a surgical technique in the plastic surgery literature for accessing the anterior cranial fossa. A retrospective review was performed of patients who underwent supraorbital frontal craniotomy using an anterior skull base approach with transpalpebral exposure over eight years by a single plastic surgeon (D.A.S.). Surgical techniques, medical co-morbidities, intra-operative complications, and long-term complications were assessed. Twenty patients (mean age 52±12 years, 55% male, 45% female) underwent supraorbital frontal craniotomy using an anterior skull base approach with upper transpalpebral exposure. Operative indications included: 75% had anterior communicating aneurysms with a mean aneurysm size of 5.36±1.91 mm, 10% had meningiomas, 10% had dural fistulas, and 5% had an orbital hemangioma. Notably, 60% had a smoking history. No intra-operative complications were encountered, and no cases required conversion to a traditional open approach. Mean length of hospital stay was 3.2±1.5 days. Post-operative imaging revealed no residual or recurrent pathology. Mean follow up time was 62.2±30.6 months. No long-term neurological or ophthalmologic complications or infections occurred. No forehead paresthesias, and no brow ptosis or brow paralysis were noted. The transpalpebral technique is an excellent, minimally invasive alternative to approach lesions of the anterior cranial fossa. Successful application may require appropriate management of the frontal sinus and supraorbital nerve. As described, this approach does not limit neurosurgical access or results, and led to no neurosurgical complications.

OBJECTIVE:Morphological and angioarchitectural features of cerebral arteriovenous malformations (AVMs) have been widely described and associated with outcomes; however, few studies have conducted a quantitative analysis of AVM flow. The authors examined brain AVM flow and transit time on angiograms using direct visual analysis and a computer-based method and correlated these factors with the obliteration response after Gamma Knife radiosurgery. METHODS:A retrospective analysis was conducted at a single institution using a prospective registry of patients managed from January 2013 to December 2019: 71 patients were analyzed using a visual method of flow determination and 38 were analyzed using a computer-based method. After comparison and validation of the two methods, obliteration response was correlated to flow analysis, demographic, angioarchitectural, and dosimetric data. RESULTS:The mean AVM volume was 3.84 cm3 (range 0.64-19.8 cm3), 32 AVMs (45%) were in critical functional locations, and the mean margin radiosurgical dose was 18.8 Gy (range 16-22 Gy). Twenty-seven AVMs (38%) were classified as high flow, 37 (52%) as moderate flow, and 7 (10%) as low flow. Complete obliteration was achieved in 44 patients (62%) at the time of the study; the mean time to obliteration was 28 months for low-flow, 34 months for moderate-flow, and 47 months for high-flow AVMs. Univariate and multivariate analyses of factors predicting obliteration included AVM nidus volume, age, and flow. Adverse radiation effects were identified in 5 patients (7%), and 67 patients (94%) remained free of any functional deterioration during follow-up. CONCLUSIONS:AVM flow analysis and categorization in terms of transit time are useful predictors of the probability of and the time to obliteration. The authors believe that a more quantitative understanding of flow can help to guide stereotactic radiosurgery treatment and set accurate outcome expectations.

BACKGROUND:Lesions of the internal auditory canal presenting with partial hearing loss are almost always vestibular schwannomas (VSs). Intracanalicular anterior inferior cerebellar artery (AICA) aneurysms are extremely rare but can mimic VS based on symptoms and imaging. The authors report the case of a flow-related intracanalicular AICA aneurysm from a pial brainstem arteriovenous malformation (AVM) masquerading as VS. OBSERVATIONS/METHODS:A 57-year-old male with partial left-sided hearing loss and an intracanalicular enhancing lesion was initially diagnosed with VS and managed conservatively at an outside institution with surveillance imaging over 3 years. When he was referred for VS follow-up, new imaging raised radiological suspicion for vascular pathology. Cerebral angiography revealed a small pial AVM located at the trigeminal root entry zone with an associated flow-related intracanalicular AICA aneurysm. The AVM was obliterated with open surgery, during which intraoperative angiography confirmed no AVM filling, preservation of the AICA, and no further aneurysm filling. LESSONS/CONCLUSIONS:Intracanalicular AICA aneurysms and other lesions, including cavernous malformations, can mimic radiographic features of VS and present with hearing loss or facial weakness. Modern vascular neurosurgical techniques such as endovascular intervention and open surgery in a hybrid operating room allowed definitive management of both lesions without untoward morbidity.

BACKGROUND:Middle meningeal artery (MMA) embolization is an apparently efficacious minimally invasive treatment for nonacute subdural hematomas (NASHs), but how different embolisates affect outcomes remains unclear. Our objective was to compare radiographic and clinical outcomes after particle or liquid MMA embolization. METHODS:Patients who had MMA embolization for NASH were retrospectively identified from a multi-institution database. The primary radiographic and clinical outcomes-50% NASH thickness reduction and need for surgical retreatment within 90 days, respectively-were compared for liquid and particle embolizations in patients treated 1) without surgical intervention (upfront), 2) after recurrence, or 3) with concomitant surgery (prophylactic). RESULTS:The upfront, recurrent, and prophylactic subgroups included 133, 59, and 16 patients, respectively. The primary radiographic outcome was observed in 61.8%, 61%, and 72.7% of particle-embolized patients and 61.3%, 55.6%, and 20% of liquid-embolized patients, respectively (p = 0.457, 0.819, 0.755). Hazard ratios comparing time to reach radiographic outcome in the particle and liquid groups or upfront, recurrent, andprophylactic timing were 1.31 (95% CI 0.78-2.18; p = 0.310), 1.09 (95% CI 0.52-2.27; p = 0.822), and 1.5 (95% CI 0.14-16.54; p = 0.74), respectively. The primary clinical outcome occurred in 8.0%, 2.4%, and 0% of patients who underwent particle embolization in the upfront, recurrent, and prophylactic groups, respectively, compared with 0%, 5.6%, and 0% who underwent liquid embolization (p = 0.197, 0.521, 1.00). CONCLUSIONS:MMA embolization with particle and liquid embolisates appears to be equally effective in treatment of NASHs as determined by the percentage who reach, and the time to reach, 50% NASH thickness reduction and the incidence of surgical reintervention within 90 days.

OBJECTIVE:Carotid stenosis is currently treated by carotid endarterectomy (CEA), carotid artery stenting (CAS), or transcarotid artery revascularization (TCAR). This study sought to add to the literature by providing real-world data comparing the safety and effectiveness associated with the performance of these carotid revascularization techniques by dual-trained neurosurgeons. METHODS:The authors performed a retrospective review of carotid stenosis databases at two US centers. Patients treated by CEA, transfemoral CAS, or TCAR for atherosclerotic carotid artery disease were included. Clinical outcomes were compared at 30 days after the procedure. RESULTS:Seven hundred eighty patients were included (583 with CAS, 165 with CEA, and 32 with TCAR). Overall, 486 patients (62.3%) were men, and 393 (50.4%) had left-sided carotid stenosis. Most patients (n = 617, 79.1%) had symptomatic disease. Among the three treatment groups, there were no statistically significant differences with respect to 30-day ischemic events (CAS 3.8%, CEA 1.8%, TCAR 6.3%; p = 0.267) or 30-day mortality rates (CAS 3.6%, CEA 2.4%, TCAR 3.1%; p = 0.857). Male sex had significantly lower odds of 30-day transient ischemic attack (TIA) or stroke in both univariable (p = 0.024) and multivariable (p = 0.023) regression models. Increasing age had significantly higher odds of 30-day mortality on univariable (p = 0.006) and multivariable (p = 0.003) regression. Patients with the occurrence of 30-day TIA or stroke also had significantly higher odds of 30-day mortality on univariable (p < 0.001) and multivariable (p < 0.001) regression. CONCLUSIONS:This real-world experience reflects the current practice of hybrid neurosurgery at two high-volume tertiary care centers and suggests that all three treatment modalities have comparable safety and effectiveness if patients are properly selected.