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

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.

Glioma subtype classification from histopathological images using in-domain and out-of-domain transfer learning: An experimental study

We provide in this paper a comprehensive comparison of various transfer learning strategies and deep learning architectures for computer-aided classification of adult-type diffuse gliomas. We evaluate the generalizability of out-of-domain ImageNet representations for a target domain of histopathological images, and study the impact of in-domain adaptation using self-supervised and multi-task learning approaches for pretraining the models using the medium-to-large scale datasets of histopathological images. A semi-supervised learning approach is furthermore proposed, where the fine-tuned models are utilized to predict the labels of unannotated regions of the whole slide images (WSI). The models are subsequently retrained using the ground-truth labels and weak labels determined in the previous step, providing superior performance in comparison to standard in-domain transfer learning with balanced accuracy of 96.91% and F1-score 97.07%, and minimizing the pathologist's efforts for annotation. Finally, we provide a visualization tool working at WSI level which generates heatmaps that highlight tumor areas; thus, providing insights to pathologists concerning the most informative parts of the WSI.

Impact of Formalin- and Cryofixation on Raman Spectra of Human Tissues and Strategies for Tumor Bank Inclusion

Reliable training of Raman spectra-based tumor classifiers relies on a substantial sample pool. This study explores the impact of cryofixation (CF) and formalin fixation (FF) on Raman spectra using samples from surgery sites and a tumor bank. A robotic Raman spectrometer scans samples prior to the neuropathological analysis. CF samples showed no significant spectral deviations, appearance, or disappearance of peaks, but an intensity reduction during freezing and subsequent recovery during the thawing process. In contrast, FF induces sustained spectral alterations depending on molecular composition, albeit with good signal-to-noise ratio preservation. These observations are also reflected in the varying dual-class classifier performance, initially trained on native, unfixed samples: The Matthews correlation coefficient is 81.0% for CF and 58.6% for FF meningioma and dura mater. Training on spectral differences between original FF and pure formalin spectra substantially improves FF samples' classifier performance (74.2%). CF is suitable for training global multiclass classifiers due to its consistent spectrum shape despite intensity reduction. FF introduces changes in peak relationships while preserving the signal-to-noise ratio, making it more suitable for dual-class classification, such as distinguishing between healthy and malignant tissues. Pure formalin spectrum subtraction represents a possible method for mathematical elimination of the FF influence. These findings enable retrospective analysis of processed samples, enhancing pathological work and expanding machine learning techniques.

Computational Assessment of Spectral Heterogeneity within Fresh Glioblastoma Tissue Using Raman Spectroscopy and Machine Learning Algorithms

Understanding and classifying inherent tumor heterogeneity is a multimodal approach, which can be undertaken at the genetic, biochemical, or morphological level, among others. Optical spectral methods such as Raman spectroscopy aim at rapid and non-destructive tissue analysis, where each spectrum generated reflects the individual molecular composition of an examined spot within a (heterogenous) tissue sample. Using a combination of supervised and unsupervised machine learning methods as well as a solid database of Raman spectra of native glioblastoma samples, we succeed not only in distinguishing explicit tumor areas-vital tumor tissue and necrotic tumor tissue can correctly be predicted with an accuracy of 76%-but also in determining and classifying different spectral entities within the histomorphologically distinct class of vital tumor tissue. Measurements of non-pathological, autoptic brain tissue hereby serve as a healthy control since their respective spectroscopic properties form an individual and reproducible cluster within the spectral heterogeneity of a vital tumor sample. The demonstrated decipherment of a spectral glioblastoma heterogeneity will be valuable, especially in the field of spectroscopically guided surgery to delineate tumor margins and to assist resection control.

Unusual Case of Splenic Metastasis in Adenosquamous Carcinoma of the Cervix Uteri: Diagnosis and Treatment Considerations

BACKGROUND Due to several factors such as its specific cellular and biochemical microenvironment, the spleen is not a predestined organ of frequent metastatic colonization in the case of primary solid carcinoma. Hence, the mode of diagnosis and the preferred treatment of a lesion highly suspicious of splenic metastasis must be decided on a case-by-case basis, considering not only the biological tumor entity but also the stage of the primary disease. CASE REPORT In the present case, we demonstrate the clinical course of a 37-year-old female patient who initially presented to our clinic with irregular vaginal bleeding. A consecutive gynecological examination revealed a 3×3-cm large mass of the cervix uteri, and the subsequent histomorphological workup led to the diagnosis of an adenosquamous carcinoma of the cervix uteri. Therapeutically, the patient received multimodal treatment, namely radical hysterectomy with adjuvant radio-chemotherapy. After 1.5 years, the patient presented to our Emergency Department with intermittent left-sided abdominal pain. Subsequent abdominal imaging (computed tomography scan, magnetic resonance imaging, positron emission tomography) determined a metabolically active splenic lesion with a central necrosis - signs of malignancy in line with a splenic metastasis. Presentation and discussion of the case within our interdisciplinary tumor board led to the decision of splenectomy followed by chemotherapy, a procedure that could be considered as therapeutic treatment in such exceptional cases. CONCLUSIONS The collection and reporting of atypical clinical courses remains a key factor in precision medicine to enable the most evidence-based decision making in such cases.

Uncommon Coexistence of Pleomorphic Adenoma and Warthin's Tumor in a Painfully Swollen Left Parotid Gland: A Surgical Case Report

BACKGROUND Benign pleomorphic adenoma is the most common primary tumor of the salivary glands and mainly arises in the parotid gland. Warthin's tumor, or papillary cystadenoma lymphomatosum, represents <30% of benign parotid tumors. The simultaneous occurrence of multiple parotid tumors is rarely described - depending on the corresponding histology (different/identical), the time of their occurrence (synchronous/metachronous), as well as their location (unilateral/bilateral), multiple parotid tumors can be further sub-classified. CASE REPORT We describe the case of a 54-year-old female patient with progressive and painful swelling of the left parotid gland for the last 6 months. During extra-oral examination, a bulging, displaceable mass of approximately 3 cm was determined. A subsequent MRI (magnetic resonance imaging) examination revealed a multifocal lesion but failed to provide a decisive clue as to the tumor entity of the lesion, and a lateral (superficial) parotidectomy was performed. Postoperative histomorphological interpretation allowed the final pathological diagnosis of synchronous, unilateral occurrence of a pleomorphic adenoma as well as a Warthin's tumor. CONCLUSIONS This report presents a rare case of synchronous unilateral parotid tumors and supports that benign pleomorphic adenoma and Warthin's tumor are the most common associations. Since clinical examination, MRI imaging, and even cytological assessment could be misleading in the detection of synchronous ipsilateral multiple parotid gland tumors, our report also highlights the importance of timely and accurate diagnosis with histopathology to plan surgery and to exclude malignant transformation, which is a rare but important association with both types of primary salivary gland tumor.

The Therapy of Vulvar Carcinoma-Evaluation of Surgical Options in a Retrospective Monocentric Study

(1) Background: Surgical-oncological treatment methods are continuously put to the test in times of evidence-based medicine-notably, a constant reevaluation remains key, especially for tumor entities with increasing incidence such as vulvar carcinoma. (2) Methods: In order to determine the postoperative clinical course of different methods of vulvar excision (vulvectomy, hemivulvectomy) as well as inguinal lymph node removal (lymphadenectomy, sentinel lymph node biopsy) with regard to postoperative wound-healingprocess, perioperative hemorrhage, and re-resection rates, we retrospectively analyzed surgical, morphological and laboratory data of 76 patients with a pathological diagnosed vulvar cancer. (3) Results: Analysis of our data from a single center revealed a comparable perioperative clinical course regardless of the chosen method of vulvar excision and inguinal lymph node removal. (4) Conclusions: Thus, our results emphasize the current multimodality in surgical therapy of vulvar carcinoma, in which consideration of known prognostic factors together with the individual patient's clinical situation allow guideline-based therapy aimed at maximizing surgical safety.

Accuracy of Breast Ultrasonography and Mammography in Comparison with Postoperative Histopathology in Breast Cancer Patients after Neoadjuvant Chemotherapy

INTRODUCTION

Nowadays chemotherapy in breast cancer patients is optionally applied neoadjuvant, which allows for testing of tumor response to the chemotherapeutical treatment in vivo, as well as allowing a greater number of patients to benefit from a subsequent breast-conserving surgery.

MATERIAL AND METHODS

We compared breast ultrasonography, mammography, and clinical examination (palpation) results with postoperative histopathological findings after neoadjuvant chemotherapy, aiming to determine the most accurate prediction of complete remission and tumor-free resection margins. To this end, clinical and imaging data of 184 patients (193 tumors) with confirmed diagnosis of breast cancer and neoadjuvant therapy were analyzed.

RESULTS

After chemotherapy, tumors could be assessed by palpation in 91.7%, by sonography in 99.5%, and by mammography in 84.5% (chi-square p < 0.0001) of cases. Although mammography proved more accurate in estimating the exact neoadjuvant tumor size than breast sonography in total numbers (136/163 (83.44%) vs. 142/192 (73.96%), n.s.), 29 tumors could be assessed solely by means of breast sonography. A sonographic measurement was feasible in 192 cases (99.48%) post-chemotherapy and in all cases prior to chemotherapy.

CONCLUSIONS

We determined a superiority of mammography and breast sonography over clinical palpation in predicting neoadjuvant tumor size. However, neither examination method can predict either pCR or tumor margins with high confidence.

PATH-01. EVALUATION OF RAMAN SPECTROSCOPY AS A DIAGNOSTIC TOOL IN NEUROPATHOLOGY FOR TUMOR CLASSIFICATION

BACKGROUND

Raman spectroscopy (RS) has shown its applicability in neurooncological diagnostics ranging from intraoperative tumor identification to peri- and postoperative tissue analyses. In the present study, we applied RS to track changes in the molecular vibrational status of a broad spectrum of formalin fixed paraffin-embedded (FFPE) intracranial neoplasms (primary brain tumors, meningiomas, brain metastases) and evaluated its potential as an additional method in the neuropathology toolbox, considering specific challenges when employing RS on FFPE tissue. Material and

METHODS

We examined 82 cases of intracranial neoplasms (679 individual measurements) by RS and applied a machine learning pipeline for recognition of spectral properties. The discrimination potential of the machine learning algorithms was evaluated using standard performance metrics such as AUROC and AUPR values, macro and weighted average of accuracy, precision, recall, and f1 scores. To address occurring misclassifications and further evaluate our models we searched for important Raman bands usable for tumor identification.

RESULTS

Using our trained machine learning model, we differentiated between different types of gliomas and determined the primary origin in case of a brain metastasis. We further spectroscopically diagnosed tumor types solely based on biopsy fragments of necrosis, something not possible by means of light microscopy. During the validation process we confirmed a high complexity within the spectroscopic data, possibly resulting not only from biological tissue which has undergone a rough chemical procedure but also from residual components of the fixation/paraffination process.

CONCLUSIONS

Our study demonstrates possibilities and limits of RS as a potential diagnostic tool in neuropathology, considering accompanying difficulties in the vibrational spectroscopic examination of FFPE tissue.

From Research to Diagnostic Application of Raman Spectroscopy in Neurosciences: Past and Perspectives

In recent years, Raman spectroscopy has been more and more frequently applied to address research questions in neuroscience. As a non-destructive technique based on inelastic scattering of photons, it can be used for a wide spectrum of applications including neurooncological tumor diagnostics or analysis of misfolded protein aggregates involved in neurodegenerative diseases. Progress in the technical development of this method allows for an increasingly detailed analysis of biological samples and may therefore open new fields of applications. The goal of our review is to provide an introduction into Raman scattering, its practical usage and also commonly associated pitfalls. Furthermore, intraoperative assessment of tumor recurrence using Raman based histology images as well as the search for non-invasive ways of diagnosis in neurodegenerative diseases are discussed. Some of the applications mentioned here may serve as a basis and possibly set the course for a future use of the technique in clinical practice. Covering a broad range of content, this overview can serve not only as a quick and accessible reference tool but also provide more in-depth information on a specific subtopic of interest.

PATH-44. RAMAN SPECTROSCOPY AS A DIAGNOSTIC TOOL IN NEUROPATHOLOGY

BACKGROUND

Although microscopic assessment is still the diagnostic gold standard in pathology, non-light microscopic methods such as new imaging methods and molecular pathology have considerably contributed to more precise diagnostics. As an upcoming method, Raman spectroscopy (RS) offers a "molecular fingerprint" which could be used to differentiate tissue heterogeneity or diagnostic entities. RS has so far been successfully applied on fresh and frozen tissue, however more aggressively, chemically treated tissue such as formalin-fixed, paraffin-embedded (FFPE) samples are challenging for RS.

METHODS

To address this issue, we examined FFPE samples of a broad range of intracranial tumors (e.g. glioblastoma and primary CNS lymphoma) and also different areas of morphologically highly heterogeneous glioblastoma tumor tissue. The latter in order to classify not only the tumor entity but also histologically defined GBM areas according to their spectral properties. We applied linear and nonlinear machine learning algorithms (Logistic Regression, Random Forest, Support Vector Machine) on our spectroscopic data and compared statistical performance of resulting classifiers.

RESULTS

We found that Random Forest classification distinguished between glioblastoma and primary CNS lymphoma with a balanced accuracy of 94%, only using Raman measurements on FFPE tissue. Furthermore, our established support vector machine-based classifier identified distinct histological areas in glioblastoma such as tumor core and necroses with an overall accuracy of 70.5% and showed a clear separation between the areas of necrosis and peritumoral zone.

CONCLUSIONS

This relatively cheap and easy-to-apply tool may serve useful to complement histopathological and molecular diagnostics. It provides an unbiased approach to tumor diagnostics with very little requirements (e.g. histopathological feature completeness of the tumor entity) to the sample. As a conclusion, we propose RS as a potential future additional method in the (neuro)-pathological toolbox for tumor diagnostics.

PATH-43. RAMAN SPECTROSCOPY AS A TOOL IN NEUROSURGERY

BACKGROUND

Raman Spectra have been shown to be sufficiently characteristic to their samples of origin that they can be used in a wide range of applications including distinction of intracranial tumors. While not replacing pathological analysis, the advantage of non-destructive sample analysis and extremely fast feedback make this technique an interesting tool for surgical use.

METHODS

We sampled intractanial tumors from more than 300 patients at the Centre Hospitalier Luxembourg over a period of three years and compared the spectra of different tumor entities, different tumor subregions and healthy surrounding tissue. We created machine-learning based classifiers that include tissue identification as well as diagnostics.

RESULTS

To this end, we solved several classes in the intracranial tumor classification, and developed classifiers to distinguish primary central nervous system lymphoma from glioblastoma, which is an important differential diagnosis, as well as meningioma from the surrounding healthy dura mater for identification of tumor tissue. Within glioblastoma, we resolve necrotic, vital tumor tissue and peritumoral infiltration zone.We are currently developing a multi-class classifier incorporating all tissue types measured.

CONCLUSIONS

Raman Spectroscopy has the potential to aid the surgeon in the surgery theater by providing a quick assessment of the tissue analyzed with regards to both tumor identity and tumor margin identification. Once a reliable classifier based on sufficient patient samples is developed, this may even be integrated into a surgical microscope or a neuronavigation system.

Application of Raman spectroscopy for detection of histologically distinct areas in formalin-fixed paraffin-embedded glioblastoma

Background

Although microscopic assessment is still the diagnostic gold standard in pathology, non-light microscopic methods such as new imaging methods and molecular pathology have considerably contributed to more precise diagnostics. As an upcoming method, Raman spectroscopy (RS) offers a “molecular fingerprint” that could be used to differentiate tissue heterogeneity or diagnostic entities. RS has been successfully applied on fresh and frozen tissue, however more aggressively, chemically treated tissue such as formalin-fixed, paraffin-embedded (FFPE) samples are challenging for RS.

Methods

To address this issue, we examined FFPE samples of morphologically highly heterogeneous glioblastoma (GBM) using RS in order to classify histologically defined GBM areas according to RS spectral properties. We have set up an SVM (support vector machine)-based classifier in a training cohort and corroborated our findings in a validation cohort.

Results

Our trained classifier identified distinct histological areas such as tumor core and necroses in GBM with an overall accuracy of 70.5% based on the spectral properties of RS. With an absolute misclassification of 21 out of 471 Raman measurements, our classifier has the property of precisely distinguishing between normal-appearing brain tissue and necrosis. When verifying the suitability of our classifier system in a second independent dataset, very little overlap between necrosis and normal-appearing brain tissue can be detected.

Conclusion

These findings show that histologically highly variable samples such as GBM can be reliably recognized by their spectral properties using RS. As conclusion, we propose that RS may serve useful as a future method in the pathological toolbox.

Differentiation of primary CNS lymphoma and glioblastoma using Raman spectroscopy and machine learning algorithms

Objective and Methods: Timely discrimination between primary CNS lymphoma (PCNSL) and glioblastoma is crucial for diagnosis and therapy, but also determines the intraoperative surgical course. Advanced radiological methods allow for their distinction to a certain extent but ultimately, biopsies are still necessary for final diagnosis. As an upcoming method that enables tissue analysis by tracking changes in the vibrational state of molecules via inelastic scattered photons, we used Raman Spectroscopy (RS) as a label free method to examine specimens of both tumor entities intraoperatively, as well as postoperatively in formalin fixed paraffin embedded (FFPE) samples. Results: We applied and compared statistical performance of linear and nonlinear machine learning algorithms (Logistic Regression, Random Forest and XGBoost), and found that Random Forest classification distinguished the two tumor entities with a balanced accuracy of 82.4% in intraoperative tissue condition and with 94% using measurements of distinct tumor areas on FFPE tissue. Taking a deeper insight into the spectral properties of the tumor entities, we describe different tumor-specific Raman shifts of interest for classification. Conclusions: Due to our findings, we propose RS as an additional tool for fast and non-destructive tumor tissue discrimination, which may help to choose the proper treatment option. RS may further serve as a useful additional tool for neuropathological diagnostics with little requirements for tissue integrity.