Artikel
Differentiation between meningioma, dura mater and overlapping-areas by robotised, macroscopic Raman spectroscopy (RS)
Unterscheidung zwischen Meningeom, Dura mater und Überlappungsbereichen durch robotisierte, makroskopische Raman Spektroskopie (RS)
Suche in Medline nach
Autoren
-
Giulia Mirizzi
- Universität des Saarlandes, Homburg, Deutschland
-
Finn Jelke
- Universität des Saarlandes, Homburg, Deutschland
-
Felix Kleine Borgmann
- Universität des Saarlandes, Homburg, Deutschland; University of Luxembourg, Medical Clinic IV Endocrinology, Esch an der Alzette, Luxembourg; National Health Laboratory Luxembourg, Düdelingen, Luxembourg; Luxembourg Institute of Health, Luxemburg, Luxembourg
-
Andreas Husch
- University of Luxembourg, Medical Clinic IV Endocrinology, Esch an der Alzette, Luxembourg; Hospital Centre of Luxembourg, National Department of Neurosurgery, Luxemburg, Luxembourg
-
Rédouane Slimani
- Luxembourg Institute of Health, Luxemburg, Luxembourg
-
Michel Mittelbronn
- University of Luxembourg, Medical Clinic IV Endocrinology, Esch an der Alzette, Luxembourg; National Health Laboratory Luxembourg, Düdelingen, Luxembourg; Luxembourg Institute of Health, Luxemburg, Luxembourg
-
Frank Hertel
- Universität des Saarlandes, Homburg, Deutschland; University of Luxembourg, Medical Clinic IV Endocrinology, Esch an der Alzette, Luxembourg; Hospital Centre of Luxembourg, National Department of Neurosurgery, Luxemburg, Luxembourg
| Veröffentlicht: | 26. Juni 2020 |
|---|
Gliederung
Text
Objective: Incomplete intraoperative excision of meningiomas from the dura mater is often the cause for tumor-recurrence. This problem requires a quick and precise intra- or perioperative diagnostic tool, which is able to accurately determine the tumor-infiltration zone. We use RS as a non-destructive procedure to trace precisely the tumor-borders directly inside the operating room.
Methods: We employed a fully robotized system equipped with a visible light imaging camera as well as a 785nm laser, for the analysis on native samples immediately after surgical excision. The samples are afterwards subjected to a full pathological diagnostic. The combined data are analyzed by custom MATLABTM tools. We performed exploratory data analysis after dimensionality reduction by unsupervised learning. Supervised learning based on support vector machines was employed to train a classifier to discriminate the different tissue types. Additionally, we compared different peaks in areas characteristic for collagen.
Results: We were able to distinguish between healthy dura mater and meningioma with a sensitivity of 99.7% and a specificity of 95.9%. We were not able to distinguish sufficiently between different meningioma subtypes. Cluster-visualization by dimension reduction (TSNE) of dura mater and meningioma spectra forms distinct clusters, the measuring points that are scattered in between originated from infiltration zones. This could be used as basis for mapping of tumor-dura overlapping areas and was confirmed by retrospective pathological examination. Next to classification, the spectra can also indicate specific biochemical differences between tissue types. The disappearance of the most prominent peaks in the dura sample, after treatment with collagenase, suggests that these originate from collagen.
Conclusion: RS is a promising, complimentary and easily-usable diagnostic tool to distinguish perioperatively between healthy dura mater and meningioma without additional fixation and microscopy. In the near future, RS might be used to diagnose directly the tumor-borders as well as the tumor-origin of the sample. This could potentially reduce the number of left-over tumor-cells after surgical excision and, as a conclusion, decrease the number of recurrences.
