Artificial intelligence in preoperative brain mapping for functional neurosurgery: A narrative review of current evidence and clinical limitations
Advanced Neurology · 2026-06-12 · 5 (4) · 025480120
DOI: 10.36922/an025480120 ↗Abstract
Preoperative brain mapping localizes eloquent regions before tumor surgery. The goal is to maximize resection while avoiding permanent deficits. Conventional techniques—navigated transcranial magnetic stimulation (nTMS), functional magnetic resonance imaging (fMRI), and magnetoencephalography (MEG)—have well-known limitations: patient cooperation requirements, acquisition variability, and interpretation challenges. nTMS combined with tractography predicts motor deficits better than fMRI in most comparative studies, because fMRI is prone to false signals from neurovascular uncoupling and co-activation of non-essential areas. MEG offers good spatial accuracy but is expensive and technically complex. Artificial intelligence (AI) can automate segmentation, integrate multimodal data, and improve functional localization. Early studies report increased mapping precision. Real-time AI-assisted neuronavigation, robotics integration, augmented/virtual reality, and explainable AI models are under development. But challenges remain: data heterogeneity, lack of external validation, and unresolved ethical and legal issues. Large-scale multicenter studies and prospective AI-driven trials are missing. Most published evidence is retrospective and single-center. This review synthesizes current AI applications in preoperative brain mapping, identifies what works and what does not, and argues that rigorous validation—not hype—will determine whether AI truly advances precision neurosurgery.
EUVIMED