MULTIMODAL ARTIFICIAL INTELLIGENCE FOR PRECISION GLIOMA SURGERY: INTEGRATING RADIOMICS, INTRAOPERATIVE ULTRASOUND, BRAIN-SHIFT CORRECTION, AND EXPLAINABLE DECISION SUPPORT

Аннотация

Background. In diffuse adult-type gliomas, including glioblastoma and IDH-mutant astrocytoma, neurosurgical decisions remain difficult because preoperative neuronavigation loses spatial validity after dural opening, while conventional image interpretation cannot continuously integrate molecular phenotype, functional anatomy, residual-tumor probability, and evolving brain deformation.

Materials and methods. A structured narrative review is proposed using clinical trials, prospective cohorts, diagnostic-accuracy studies, systematic reviews, and relevant technical investigations. Evidence is organized by biological plausibility, technical performance, patient selection, safety, clinical effectiveness, reproducibility, and implementation. A prospective comparative study is specified for adults undergoing first or repeat resection of supratentorial diffuse glioma, with standard neuronavigation with conventional visual interpretation of preoperative MRI and intraoperative ultrasound as the reference strategy.

Results. The synthesis addresses five domains: preoperative molecular prediction, intraoperative lesion localization, dynamic compensation for brain shift, transparent risk estimation, workflow integration and human factors. The proposed primary endpoint is extent of resection normalized to molecularly defined tumor burden and the rate of new persistent neurological deficit at 90 days; secondary endpoints include brain-shift registration error, residual tumor volume, progression-free survival, operative time, calibration error, subgroup fairness, and surgeon cognitive workload. Originality derives from linking technical accuracy to patient-centered outcomes, explicitly modeling uncertainty, and requiring external validation.

Conclusion. Standardized endpoints support reliable clinical comparison. Independent adjudication strengthens causal interpretation. Longitudinal follow-up clarifies therapeutic durability. Equity analysis supports responsible implementation. Transparent reporting improves scientific reproducibility. Rigorous independent prospective multicenter transparent external validation remains essential.