INTELLIGENT INTRAOPERATIVE NEUROPATHOLOGY IN BRAIN TUMOR SURGERY: STIMULATED RAMAN HISTOLOGY, RAPID NANOPORE EPIGENOMICS, AND AI-GUIDED MARGIN ASSESSMENT

Аннотация

Abstract
Background. Intraoperative neuropathology is moving from morphology-only frozen-section consultation toward integrated optical, molecular, and artificial-intelligence workflows capable of informing surgery before closure. Stimulated Raman histology provides label-free microscopic images from  fresh  tissue,  whereas  rapid  nanopore  sequencing,  microfluidic  polymerase-chain-reaction systems, and foundation models can estimate tumor class, molecular subtype, and infiltration within minutes. 
Materials  and  methods. Prospective  multicenter  studies,  translational  investigations, diagnostic-accuracy  cohorts,  and  platform-validation  studies  published  through  July  2026  were synthesized. Evidence was organized according to specimen acquisition, optical histology, machine- learning interpretation, rapid epigenomic or genetic profiling, margin assessment, turnaround time, diagnostic confidence, and clinical decision impact. 
Results. Stimulated Raman histology combined with deep learning achieves near-real-time tumor  classification  and  can  detect  diffuse-glioma  infiltration  more  accurately  than  several conventional intraoperative adjuncts. FastGlioma produced infiltration scores within seconds in an international prospective cohort. Rapid-CNS2 generated methylation classification and copy-number information  within  a  thirty-minute  intraoperative  window,  while  Sturgeon  and  MethyLYZR demonstrated  classification  from  sparse  nanopore  methylation  data.  Hetairos  extended  routine hematoxylin-and-eosin analysis toward prediction of 102 methylation-associated central-nervous- system tumor subtypes. Nevertheless, low tumor purity, sampling error, rare entities, domain shift, overconfident predictions, and uncertain action thresholds remain limitations. 
Conclusion. Intelligent intraoperative neuropathology should augment rather than replace neuropathology.  The  proposed  SMART-PATH  framework  integrates  sampling,  multimodal acquisition,  algorithmic  confidence,  rapid  molecular  testing,  topographic  mapping,  pathology verification, actionable interpretation, and human oversight to support maximal safe resection and adequate biopsy.
Keywords: stimulated Raman histology; artificial intelligence; brain tumor surgery; nanopore sequencing; DNA methylation; glioma infiltration; intraoperative diagnosis; digital neuropathology; molecular classification; surgical margin.