Machine Learning-Assisted Classification of Paraffin-Embedded Brain Tumors with Raman Spectroscopy

Gilbert Georg Klamminger,Laurent Mombaerts,Françoise Kemp,Finn Jelke,Karoline Klein,Rédouane Slimani,Giulia Mirizzi,Andreas Husch,Frank Hertel,Michel Mittelbronn,Felix B. Kleine Borgmann
DOI: https://doi.org/10.3390/brainsci14040301
IF: 3.333
2024-03-24
Brain Sciences
Abstract:Raman spectroscopy (RS) has demonstrated its utility in neurooncological diagnostics, spanning from intraoperative tumor detection to the analysis of tissue samples peri- and postoperatively. In this study, we employed Raman spectroscopy (RS) to monitor alterations in the molecular vibrational characteristics of a broad range of formalin-fixed, paraffin-embedded (FFPE) intracranial neoplasms (including primary brain tumors and meningiomas, as well as brain metastases) and considered specific challenges when employing RS on FFPE tissue during the routine neuropathological workflow. We spectroscopically measured 82 intracranial neoplasms on CaF2 slides (in total, 679 individual measurements) and set up a machine learning framework to classify spectral characteristics by splitting our data into training cohorts and external validation cohorts. The effectiveness of our machine learning algorithms was assessed by using common performance metrics such as AUROC and AUPR values. With our trained random forest algorithms, we distinguished among various types of gliomas and identified the primary origin in cases of brain metastases. Moreover, we spectroscopically diagnosed tumor types by using biopsy fragments of pure necrotic tissue, a task unattainable through conventional light microscopy. In order to address misclassifications and enhance the assessment of our models, we sought out significant Raman bands suitable for tumor identification. Through the validation phase, we affirmed a considerable complexity within the spectroscopic data, potentially arising not only from the biological tissue subjected to a rigorous chemical procedure but also from residual components of the fixation and paraffin-embedding process. The present study demonstrates not only the potential applications but also the constraints of RS as a diagnostic tool in neuropathology, considering the challenges associated with conducting vibrational spectroscopic analysis on formalin-fixed, paraffin-embedded (FFPE) tissue.
neurosciences
What problem does this paper attempt to address?
This paper aims to solve several key problems in neuropathological diagnosis, as follows: 1. **Classifying formalin - fixed paraffin - embedded (FFPE) brain tumors using Raman Spectroscopy (RS)**: - Researchers used Raman Spectroscopy to monitor the changes in molecular vibration characteristics of various FFPE intracranial tumors (including primary brain tumors, meningiomas, and brain metastases). - Through machine - learning algorithms, researchers attempted to extract features from spectral data to distinguish different types of brain tumors. 2. **Overcoming the challenges of FFPE tissues in Raman spectral analysis**: - Some interfering factors may be introduced during the processing of FFPE tissues, such as nucleic acid degradation, protein cross - linking, and strong Raman background of paraffin signals. - Researchers reduced the influence of these interfering factors and improved the quality of spectral data by dewaxing treatment and selecting appropriate substrate materials (such as CaF₂ slides). 3. **Improving classification accuracy through machine - learning algorithms**: - Researchers established a random forest (RF) classification model for training and validating different types of brain tumors. - By evaluating the performance indicators of the model (such as AUROC and AUPR values), researchers verified the effectiveness of the model in distinguishing different tumor types. 4. **Exploring the spectral characteristics of specific tumor regions**: - Researchers not only focused on different types of tumors but also specifically studied different tissue regions within the tumors, such as the viable tumor area and the necrotic area. - Through spectral analysis, researchers were able to identify pure necrotic tissue, which is difficult to achieve with traditional optical microscopes. 5. **Classification tasks in practical applications**: - Researchers developed a classifier for quickly identifying the primary sites of brain metastases (such as the lung, breast, and colorectum). - Another classifier was used to distinguish different types of gliomas and predict their genetic characteristics (such as IDH mutation status). In conclusion, by combining Raman Spectroscopy and machine - learning methods, this paper aims to improve the diagnostic accuracy of brain tumors. Especially when dealing with FFPE tissue samples, it provides a label - free, fast, and accurate diagnostic tool.