Faculty Bibliography

Object/UNASSIGNED:The purpose of the current study is to determine the sensitivity and specificity of an eye tracking method as a classifier for identifying concussion. Methods/UNASSIGNED:Brain injured and control subjects prospectively underwent both eye tracking and Sport Concussion Assessment Tool 3. The results of eye tracking biomarker based classifier models were then validated against a dataset of individuals not used in building a model. The area under the curve (AUC) of receiver operating characteristics was examined. Results/UNASSIGNED:An optimal classifier based on best subset had an AUC of 0.878, and a cross-validated AUC of 0.852 in CT- subjects and an AUC of 0.831 in a validation dataset. The optimal misclassification rate in an external dataset (n = 254) was 13%. Conclusion/UNASSIGNED:If one defines concussion based on history, examination, radiographic and Sport Concussion Assessment Tool 3 criteria, it is possible to generate an eye tracking based biomarker that enables detection of concussion with reasonably high sensitivity and specificity.

Predictions of patient outcomes after a given therapy are fundamental to medical practice. We employ a machine learning approach towards predicting the outcomes after stereotactic radiosurgery for cerebral arteriovenous malformations (AVMs). Using three prospective databases, a machine learning approach of feature engineering and model optimization was implemented to create the most accurate predictor of AVM outcomes. Existing prognostic systems were scored for purposes of comparison. The final predictor was secondarily validated on an independent site's dataset not utilized for initial construction. Out of 1,810 patients, 1,674 to 1,291 patients depending upon time threshold, with 23 features were included for analysis and divided into training and validation sets. The best predictor had an average area under the curve (AUC) of 0.71 compared to existing clinical systems of 0.63 across all time points. On the heldout dataset, the predictor had an accuracy of around 0.74 at across all time thresholds with a specificity and sensitivity of 62% and 85% respectively. This machine learning approach was able to provide the best possible predictions of AVM radiosurgery outcomes of any method to date, identify a novel radiobiological feature (3D surface dose), and demonstrate a paradigm for further development of prognostic tools in medical care.

OBJECT Randomized-controlled trials (RCTs) are advocated to provide high-level medical evidence. However, in neurosurgery, there are barriers to conducting RCTs. The authors of this study sought to analyze the quality of neurosurgical RCTs since 2000 to determine the adequacy of their design and reporting. METHODS A search of the MEDLINE and EMBASE databases (2000-2014) was conducted. The medical subject heading (MeSH) terms used in the search included: "neurosurgery" OR "neurosurgical procedure," "brain neoplasms," "infarction" and "decompression," "carotid stenosis," "cerebral hemorrhage," and "spinal fusion." These studies were limited to RCTs, in humans, and in the English language. The Consolidated Standards for Reporting of Trials (CONSORT) and Jadad scales were used to assess the quality of RCT design and reporting. The standardized median times cited (median citations divided by years since publication) were used to assess impact. A pragmatic-explanatory continuum indicator summary-based scale was used to assess the design of the studies as primarily pragmatic or explanatory. RESULTS Sixty-one articles were identified, and the following subspecialties were the most common: vascular (23, 37%), followed by functional neurosurgery and neurooncology (both 13, 21%). The following nations were the primary leaders in RCTs: US (25 studies, 41%), Germany (8 studies, 13%), and the United Kingdom (7 studies, 11%). Median sample size was 100 (interquartile range [IQR] 41.5-279). The majority of the studies (40, 66%) had pragmatic objectives. The median number of times cited overall was 69 (IQR 20.5-193). The combined median CONSORT score was 36 (IQR 27.5-39). Blinding was most deficiently reported. Other areas with a relatively low quality of reporting were sample size calculation (34.2% of surgical, 38.5% of drug, and 20% of device studies), allocation concealment (28.9% of surgical, 23.1% of drug, and 50% of device studies), and protocol implementation (18.4% of surgical, 23% of drug, and 20% of device studies). The quality of reporting did not correlate with the study impact. All studies had a median Jadad score