Artificial intelligence-enhanced intraoperative neurosurgical workflow: current knowledge and future perspectives.

Area of Science:

• Neurosurgery
• Artificial Intelligence
• Machine Learning

Background:

• AI and ML enhance surgical decision-making and productivity across perioperative and intraoperative phases.
• These technologies support surgeons in diagnosis, preoperative planning, and real-time surgical assistance.
• This review focuses on AI applications within neurosurgical workflows across various subspecialties.

Purpose of the Study:

• To identify and describe current AI platforms used in neurosurgical perioperative and intraoperative settings.
• To synthesize the role and performance of AI and ML models in neurosurgical applications.
• To categorize AI applications by neurosurgical subspecialty and identified prediction tasks.

Main Methods:

• Systematic literature review following PRISMA guidelines (PubMed, EMBASE, Scopus, up to Dec 31, 2020).
• Inclusion criteria: AI platforms in perioperative/intraoperative settings with reported ML performance metrics.
• Qualitative synthesis due to application heterogeneity; risk of bias and applicability assessed using PROBAST tool.

Main Results:

• Forty-one articles included, all evaluating supervised learning algorithms.
• Most frequent ML models: neural networks (15) and tree-based models (13).
• Medium-high risk of bias noted, but positive applicability across studies; AI applications categorized into neuro-oncology, spine, functional, and other neurosurgery subspecialties.

Conclusions:

• AI shows potential for augmenting neurosurgical workflows by reducing errors and enabling personalized surgical plans.
• Current ML models can enhance surgical team performance.
• Further high-quality studies are essential to validate and advance AI integration in neurosurgery.