Truace
Subscribe

AI helps confirm brain tumor tissue during biopsy surgery, but struggles with some non-tumor lesions

Study title: Artificial intelligence–based tumor detection using stimulated Raman histology in patients undergoing neurosurgical biopsies

Researchers retrospectively reviewed consecutive neurosurgical biopsy patients at a single center who had intraoperative stimulated Raman histology imaging between December 2020 and October 2025. A CE-certified algorithm, Glioma Reveal, scored each scan for tumor cells, and results were compared against routine histopathology.

Read the study abstract

OBJECTIVE Intraoperative confirmation of diagnostic tumor tissue is critical in neurosurgical biopsies, particularly in stereotactic procedures, in which sampling error remains a major limitation. Stimulated Raman histology (SRH) enables rapid, label-free tissue imaging directly in the operating room. This study aimed to evaluate the performance of intraoperative SRH combined with artificial intelligence (AI)–based tumor detection using a CE (Conformité Européenne)–certified algorithm under routine clinical cond…

By the Truvace desk2 min read1 sourceEvidence: Moderate · 56/100 — Observational study · n = 89 · 1 independent sourcePeer-reviewedTRV-2026-2298· cite & verify
Artificial intelligence–based tumor detection using stimulated Raman histology in patients undergoing neurosurgical biopsies
CT scan Rasmussen's encephalitis by Mikhail Kalinin. CC BY-SA 3.0 · https://creativecommons.org/licenses/by-sa/3.0
Trace impact readingPositive state
P 51The 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 59The 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 sampling error is a major limitation in biopsies, especially stereotactic ones, and fast confirmation of representative tissue could help. Still, the strongest accuracy figures depended on neuropathologist review of discordant cases, and the tool performed poorly on nonneoplastic lesions, so its standalone reliability remains uncertain.

Main points

  1. The study covered 204 scans from 89 patients, analyzed at the scan level for specific regions of interest.
  2. Initial overall accuracy was 80.9%, increasing to 94.1% after a neuropathologist reviewed misclassified cases.
  3. For gliomas and lymphomas (70 patients, 168 scans), sensitivity was 78.8% and specificity 100.0%.
  4. Expert review found 80.6% (25 scans) of false negatives were morphologically correct within the specific region despite whole-specimen positivity.
  5. Accuracy for nonneoplastic lesions stayed at 62.5% because of reactive mimics.

The gain

In 204 scans from 89 biopsy patients, the Glioma Reveal algorithm detected tumor tissue with 80.9% initial accuracy, rising to 94.1% after expert review, giving rapid intraoperative confirmation of tumor tissue.

The problem

The algorithm reached only 62.5% accuracy on nonneoplastic lesions because of reactive mimics, and sensitivity for gliomas and lymphomas was 78.8%, with false negatives driven by tissue heterogeneity.

The rundown

Patients were included if they had neurosurgical biopsy with intraoperative stimulated Raman histology, a rapid, label-free imaging method used in the operating room. Each tissue sample then received routine histopathological analysis, which served as the reference.

Cases were grouped into common tumor entities, rare tumor entities and nonneoplastic lesions. Within gliomas and lymphomas, initial accuracy was 81.5% and rose to 96.4% after correcting for false negatives that the expert judged morphologically correct.

In rare tumors (8 patients, 20 scans), accuracy rose from 90.0% to 100.0% after expert review. The authors conclude that performance is robust but that intraspecimen heterogeneity and spatial sampling differences significantly affect diagnostic metrics.

What this doesn’t fix

This was a retrospective single-center study, and the higher accuracy figures relied on expert re-review of discordant cases rather than the algorithm alone; heterogeneity and sampling differences affected metrics.

Sources

  1. Peer-reviewedNeurosurgical Focus2026-10-28

The debate