Truace
Subscribe

AI tools can read brain tumor type during surgery, but no trials show patient benefit

Study title: Intraoperative artificial intelligence for brain tumor surgery: from tissue diagnosis to real-time surgical guidance

This review surveys AI models that can be used in the operating room during brain tumor resections. Molecular markers such as IDH status, 1p/19q codeletion and methylation class now define glioma diagnosis, but results normally arrive days to weeks after surgery. The models extract this information from tissue and imaging data available in real time.

Read the study abstract

The 2021 WHO CNS tumor classification marked a fundamental shift in neuro-oncology: molecular markers rather than histological appearance now define glioma diagnosis and guide treatment. Yet, in clinical practice the molecular information that matters most to a surgeon—IDH status, 1p/19q codeletion, and methylation class—typically arrives days to weeks after the operation it was meant to inform. Artificial intelligence (AI) is beginning to close that gap. Over the past several years, a growing body of work has d…

By the Truvace desk2 min read1 sourceEvidence: Emerging · 30/100 — Commentary, opinion, or narrative review · 1 independent sourcePeer-reviewedTRV-2026-2299· cite & verify
Intraoperative artificial intelligence for brain tumor surgery: from tissue diagnosis to real-time surgical guidance
CT-scan-4th-and-3rd-ventricles by Aceofhearts1968. CC BY 3.0 · https://creativecommons.org/licenses/by/3.0
Trace impact readingContested
P 45The P score combines the specificity and measured human impact of the grounded problem claim with the strength of this Trace’s cited sources.

Both sides are scored from claims and sources, not community votes.

G 45The G score combines the specificity and measured human impact of the grounded gain claim with the strength of this Trace’s cited sources.

In brief

The work matters because earlier molecular answers could shape decisions while the surgeon is still operating. However, the review states that the field lacks randomized trials showing that patients actually do better, so the clinical value of these tools remains unproven, and the reported accuracy figures come from varied studies and settings.

Main points

  1. The cryosection model CHARM reached areas under the curve of 0.88–0.93 across glioma subtypes.
  2. Hyperspectral imaging with machine learning classified tumor type, grade and IDH mutation from the 5-ALA signal with up to 93% accuracy.
  3. The ROBIN nanopore ensemble achieved 90% concordance with the final diagnosis prospectively in under 2 hours.
  4. Stimulated Raman histology models exceeded 90% glioma subtype accuracy in multi-institutional cohorts.

The gain

AI models can give surgeons molecular tumor information during the operation, for example ROBIN nanopore classification reached 90% concordance with the final diagnosis in under 2 hours, rather than days to weeks later.

The problem

Widespread clinical adoption of intraoperative AI is held back because no randomized clinical trials yet show that these tools benefit patients.

The rundown

The review covers AI added to established intraoperative techniques, including cryosection and fluorescence-guided surgery. LGNet was reported to discriminate primary CNS lymphoma, while CHARM performed molecular subtyping of glioma.

It also covers newer technologies such as nanopore sequencing and stimulated Raman histology. In SRH, foundation models for infiltration detection outperformed FLAIR, contrast enhancement and 5-ALA fluorescence for identifying tumor margins.

At the single-cell level, morphometric profiling of SRH images showed nuclear features that distinguish IDH-mutant from IDH-wildtype tumors, offering a route to interpretability. The authors point to multimodal systems that combine imaging, pathology and molecular data as a next step.

What this doesn’t fix

The review says adoption is limited by the absence of randomized clinical trials showing patient-level benefit, so improved patient outcomes are not established.

Sources

  1. Peer-reviewedNeurosurgical Focus2026-10-28

The debate