Meta-analysis of intraoperative neurophysiological monitoring in neuro-oncology using Artificial Intelligence

<h3>Introduction</h3> Gliomas, particularly Glioblastoma, are among the most aggressive brain tumors, often infiltrating regions crucial for vital functions. Intraoperative neurophysiological monitoring (ION) is integral during resections within eloquent brain areas to ensure maximal tumor removal while preserving functionality. This study appraises Artificial Intelligence (AI)/Machine Learning (ML) methodologies enhancing ION during surgery. <h3>Methods</h3> Utilizing a PRISMA-guided review and a DerSimonian and Laird random effects model, we evaluated AI/ML-augmented ION’s sensitivity, specificity, and predictive values in functional assessment. Data were extracted from patients undergoing surgery with ION assistance, and statistical significance was assessed at p&lt;0.05. <h3>Results</h3> Analysis of 3117 patients revealed an AI/ML-based ION sensitivity effect size of 75.7%, specificity of 87.2%, an Area Under the Receiver Operator Characteristic Curve (AUC) of 84.9%, and a positive predictive value (PPV) of 67.1%. Notably, the negative predictive value (NPV) reached 93.9% without heterogeneity, indicating a high efficacy in predicting non-functional alterations. <h3>Conclusion</h3> AI/ML-enhanced ION provides reliable delineation of functional zones during glioma surgery, with robust sensitivity, specificity, and AUC, alongside a substantial NPV. These findings suggest that AI/ML methods are particularly effective in predicting the absence of functional change, a critical factor in surgical decision-making. The development of AI-based ION diagnostic biomarkers could be furthered by employing a meta-analytical Youden index. The accuracy of these techniques, vital for future research, holds promise for advancing ION’s precision in brain tumor surgery.