PRECISION-GUIDED MAXIMAL SAFE RESECTION OF DIFFUSE GLIOMAS IN THE MOLECULAR ERA: INTEGRATION OF FUNCTIONAL MAPPING, 5-ALA, INTRAOPERATIVE IMAGING, RAPID OPTICAL HISTOLOGY, AND ARTIFICIAL INTELLIGENCE

Аннотация

Background. Surgical  management  of  diffuse  gliomas  is  undergoing  a  transition  from anatomically defined tumor removal toward biologically and functionally individualized resection. The conventional objective of gross-total resection based on contrast-enhanced or T2/FLAIR magnetic resonance  imaging  incompletely  reflects  the  infiltrative  architecture  of  gliomas  and  does  not incorporate  interindividual  differences  in  molecular  subtype,  eloquent-network  organization,  or microscopic tumor burden. Recent advances in awake cortical–subcortical mapping, 5-aminolevulinic acid fluorescence, intraoperative ultrasound and magnetic resonance imaging, stimulated Raman histology,  confocal  laser  endomicroscopy,  nanopore-based  molecular  profiling,  and  artificial intelligence provide complementary sources of intraoperative information.
Materials  and  Methods. A  structured  narrative  review  of  clinical  trials,  prospective multicenter studies, molecularly annotated surgical cohorts, and translational investigations published through August 2026 was performed. Evidence was synthesized around extent of resection, molecular subtype, functional preservation, intraoperative tumor detection, optical histology, and rapid molecular classification.
Results. Increasing  extent  of  resection  remains  associated  with  improved  oncological outcomes, but the magnitude and safe limit of cytoreduction vary according to patient age, tumor biology,  residual  contrast-enhancing  and  non-enhancing  volume,  and  functional  anatomy. Supramaximal resection has shown particularly favorable associations in IDH-mutant grade 2 gliomas when functionally feasible. 5-ALA improves detection of metabolically active high-grade tumor, while intraoperative ultrasound and MRI compensate for brain shift. FastGlioma, combining stimulated Raman histology with a foundation model, detected the degree of glioma infiltration with a mean AUROC of 92.1% in a prospective international cohort of 220 patients. DeepGlioma and rapid nanopore methylation classifiers demonstrate that molecular information can increasingly be generated within an intraoperative time frame.
Conclusion. The contemporary surgical  endpoint should not  be defined by radiographic completeness  alone.  A  precision-guided  resection  strategy  should  integrate  biological  tumor probability, functional-network boundaries, updated intraoperative anatomy, molecular subtype, and anticipated postoperative neurological cost. The optimal boundary of resection is therefore patient- specific rather than purely MRI-defined.
Keywords: diffuse glioma; glioblastoma; maximal safe resection; supramaximal resection; awake mapping; 5-ALA; intraoperative ultrasound; stimulated Raman histology; artificial intelligence; molecular neurosurgery.