Answers are grounded in this paper and cite their sources.
Background Predictive modeling using artificial intelligence has expanded rapidly across neurosurgical subspecialties, but a focused synthesis of recent evidence is needed to assess performance and methodological quality. Such models may support neurosurgical practice by improving risk stratification, outcome prediction, surgical planning, and individualized decision-making in clinically complex patients Objectives To identify, classify, and critically synthesize predictive modeling studies in neurosurgery published from March 1, 2021, through March 31, 2026, and to evaluate subspecialty distribution, model types, discrimination, external validation, calibration, and methodological quality. Methods A PRISMA 2020-compliant systematic review was conducted using PubMed/MEDLINE and Web of Science. Risk of bias was assessed with PROBAST, and reporting quality was evaluated using TRIPOD. Results Twenty-nine eligible publications were included: 23 primary predictive model studies, 2 large-language-model benchmark studies, and 4 systematic review or quality appraisal studies. Primary studies covered neuro-oncology, cerebrovascular neurosurgery, spine surgery, traumatic brain injury, epilepsy or functional neurosurgery, and pediatric hydrocephalus. Machine learning was the dominant model type. Among the 23 primary predictive-model studies, all reported an extractable discrimination metric, ranging from 0.672 to 0.940, with a median of 0.827 and a mean of 0.818 (SD 0.092). These values represent reported AUC (C-statistic) measures of model discrimination. External validation was reported in 52.2% of primary studies, whereas formal calibration was reported in 26.1% (6/23). Calibration refers to the agreement between predicted probabilities and observed clinical outcomes. Conclusions Predictive AI in neurosurgery shows promising performance across subspecialties, but inadequate calibration reporting and inconsistent external validation remain major barriers to responsible clinical translation. Clinical trial number Not applicable.
No takes yet. Share an insight, caveat, or question.
Alizada et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: