% IMPORTANT: The following is UTF-8 encoded. This means that in the presence
% of non-ASCII characters, it will not work with BibTeX 0.99 or older.
% Instead, you should use an up-to-date BibTeX implementation like “bibtex8” or
% “biber”.
@ARTICLE{Scheffler:307512,
author = {P. Scheffler and N. Neidert$^*$ and J. Straehle and D. Erny
and M. Prinz and U. Hubbe and R. Rölz and D. H. Heiland and
J. Beck and A. El Rahal},
title = {{A}rtificial intelligence-based analysis and diagnosis of
intradural extramedullary spinal tumors by stimulated
{R}aman histology.},
journal = {Neuro-oncology advances},
volume = {7},
number = {1},
issn = {2632-2498},
address = {Oxford},
publisher = {Oxford University Press},
reportid = {DKFZ-2026-00017},
pages = {vdaf211},
year = {2025},
abstract = {Intraoperative Stimulated Raman Histology (SRH) has been
reported to be fast and accurate in the assessment of
neuro-oncological lesions. However, its application to
spinal tumors, especially intradural extramedullary tumors
(IDEM), remains underexplored. IDEM primarily include
meningiomas and schwannomas, as well as less common entities
such as metastases or ependymomas. Given that surgical
resection is the primary treatment modality, rapid,
artificial intelligence (AI)-driven intraoperative tumor
classification based on SRH could enhance surgical
decision-making and subsequent management.We acquired 232
SRH images from patients with IDEM using the NIO Laser
Imaging System (Invenio Imaging Inc.). We categorized images
into three diagnostic classes: 'Meningioma,' 'Schwannoma,'
and 'Other.' Images were divided into 224 × 224 pixel
patches and used to train and test AI-based image
classifiers employing CTransPath, ResNet, and Vision
Transformer architectures.Our best-performing model,
utilizing the CTransPath architecture, achieved a
classification accuracy of $94.3\%$ on the test dataset.
Vision Transformer-based models also performed well,
exceeding $90\%$ accuracy, while ResNet models attained
slightly lower accuracies $(79.6\%-88.8\%).$ Qualitative
analysis indicates that the top-performing model primarily
relies on cellular morphology for classification.Our
findings confirm the feasibility and effectiveness of
AI-assisted SRH analysis for distinguishing IDEM tumor
types. This approach may complement conventional
intraoperative neuropathology by providing rapid, reliable,
and clinically actionable diagnostic information.},
keywords = {artificial intelligence (Other) / intradural extramedullary
tumors (Other) / stimulated raman histology (Other)},
cin = {FR01},
ddc = {610},
cid = {I:(DE-He78)FR01-20160331},
pnm = {899 - ohne Topic (POF4-899)},
pid = {G:(DE-HGF)POF4-899},
typ = {PUB:(DE-HGF)16},
pubmed = {pmid:41473747},
pmc = {pmc:PMC12746601},
doi = {10.1093/noajnl/vdaf211},
url = {https://inrepo02.dkfz.de/record/307512},
}