gms | German Medical Science

73. Jahrestagung der Deutschen Gesellschaft für Neurochirurgie (DGNC)
Joint Meeting mit der Griechischen Gesellschaft für Neurochirurgie

Deutsche Gesellschaft für Neurochirurgie (DGNC) e. V.

29.05. - 01.06.2022, Köln

Integrated intraoperative imaging of glioblastoma using stimulated raman histology and 5-ALA mediated fluorescence

Integrierte intraoperative Bildgebung von Glioblastomgewebe mittels Stimulierter Raman Histologie und 5-ALA vermittelter Fluoreszenz

Meeting Abstract

  • presenting/speaker Jakob Straehle - Universitätsklinikum Freiburg, Klinik für Neurochirurgie, Freiburg i. Br., Deutschland
  • Nicolas Neidert - Universitätsklinikum Freiburg, Klinik für Neurochirurgie, Freiburg i. Br., Deutschland
  • Vlad Sacalean - Universitätsklinikum Freiburg, Klinik für Neurochirurgie, Freiburg i. Br., Deutschland
  • Daniel Erny - Universitätsklinikum Freiburg, Institut für Neuropathologie, Freiburg i. Br., Deutschland
  • Peter Christoph Reinacher - Universitätsklinikum Freiburg, Klinik für Neurochirurgie, Department of Stereotactic and Functional Neurosurgery, Freiburg i. Br., Deutschland; Fraunhofer Institut für Lasertechnik (ILT), Aachen, Deutschland
  • Volker Arnd Coenen - Universitätsklinikum Freiburg, Klinik für Neurochirurgie, Department of Stereotactic and Functional Neurosurgery, Freiburg i. Br., Deutschland
  • Boris Mizaikoff - Universität Ulm, Institut für Analytische und Bioanalytische Chemie, Ulm, Deutschland; Hahn-Schickard Institute for Microanalysis Systems, Ulm, Deutschland
  • Marco Prinz - Universitätsklinikum Freiburg, Institut für Neuropathologie, Freiburg i. Br., Deutschland
  • Oliver Schnell - Universitätsklinikum Freiburg, Klinik für Neurochirurgie, Freiburg i. Br., Deutschland
  • Jürgen Beck - Universitätsklinikum Freiburg, Klinik für Neurochirurgie, Freiburg i. Br., Deutschland
  • Dieter Henrik Heiland - Universitätsklinikum Freiburg, Klinik für Neurochirurgie, Freiburg i. Br., Deutschland

Deutsche Gesellschaft für Neurochirurgie. 73. Jahrestagung der Deutschen Gesellschaft für Neurochirurgie (DGNC), Joint Meeting mit der Griechischen Gesellschaft für Neurochirurgie. Köln, 29.05.-01.06.2022. Düsseldorf: German Medical Science GMS Publishing House; 2022. DocV269

doi: 10.3205/22dgnc261, urn:nbn:de:0183-22dgnc2618

Published: May 25, 2022

© 2022 Straehle et al.
This is an Open Access article distributed under the terms of the Creative Commons Attribution 4.0 License. See license information at http://creativecommons.org/licenses/by/4.0/.


Outline

Text

Objective: Intraoperative histological evaluations of CNS malignancies is challenging and limited to labour- and time-intensive neuropathological assessments. Integrated Stimulated Raman Histology (SRH) and 5-ALA mediated fluorescence imaging enables novel perspectives of interactive histo- and molecular feedback during brain tumor surgeries. Our investigations aimed to discover the heterogeneity of SRH-histomorphological pattern in glioblastoma correlated to 5-ALA mediated fluorescence signals to enable precise intraoperative interpretation of SRH.

Methods: 5-ALA mediated fluorescence and stimulated Raman scattering microscopy was performed on 50 glioblastoma specimens from 18 surgical resections and 5 stereotactic biopsies. All images with an input shape of [9000,9000,3] were randomly segregated into 1000 patches with the shape [160,160,3], resulting in a total of 50k patches. The data was then split 1:1 into training and validation data. A generative deep learning network architecture (variational autoencoder) was used to simplify the image information by learning to reconstruct the image based on a reduced information content (latent space). The reduced information was further clustered to uncover unique histomorphological patterns.

Results: Image reconstruction with Deep Learning showed the highest accuracy (F1 score 96.2 p<0.021) using a convolutional neural network embedded in a variational autoencoder (CNN-VAE). In a preliminary analysis, the clustering of the reduced image information yielded 6 clusters, with one cluster representing images with artifacts (5.43%). The other clusters reflected vascular proliferations (12.91%), high nuclear density (16.23%), large saturated fibers (19.43%), and small saturated fibers (20.36%) and nonsaturated fibers (25.64%). Integration of 5-ALA fluorescence revealed significantly enriched signal only in the clusters of saturated fibers. Samples obtained from the tumor periphery showed significant enrichment of the cluster of large saturated fibers (p=2.4e10-6).

Conclusion: Our analysis provides new insights into intraoperative analysis of combined SRS and 5-ALA fluorescence. We identified spatial heterogeneity of 5-ALA enrichment across different histopathological patterns in glioblastoma, suggesting that SRH may improve intraoperative diagnostics. The underlying pathophysiology of individual clusters, particularly saturated fiber bundles, a morphology absent in classical H&E staining, requires further investigation.