Platform for detection and analysis of brain tumor extracellular vesicles from tissue and biofluids

Abstract

A diagnostic platform for detecting and analyzing brain tumors includes a portable fiber Surface Enhanced Raman spectroscopy (SERS) module, a removable fiber probe for intraoperative use, disposable nanoplasmonic cartridges for individual sample loading, and an updatable spectral library collected from healthy and diseased samples. The platform is designed to analyze blood and tissue signatures and employ artificial intelligence (Al) models to classify SERS spectra based on an updatable proteomics and SERS spectral library encompassing distinct brain tumor types. The platform is further configured to utilize artificial intelligence (Al) algorithms for the analysis and classification of the recorded SERS spectra.

Description

PLATFORM FOR DETECTION AND ANALYSIS OF BRAIN TUMOR EXTRACELLULAR VESICLES FROM TISSUE AND BIOFLUIDSCROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 643,297 entitled "Glio-SERS Platform for Detection and Analysis of Brain Tumor EVs From Tissue and Biofluids" filed May 6, 2024, the contents of which are incorporated by reference herein in its entirety for all purposes.STATEMENT OF FEDERALLY SPONSORED RESEARCH OR DEVELOPMENT

[0002] Not applicable.TECHNICAL FIELD

[0003] The disclosure relates to the field of Surface-Enhanced Raman Spectroscopy (SERS) for biomedical applications, and more specifically, to a diagnostic platform utilizing SERS and artificial intelligence for the detection and analysis of brain tumors. The platform is designed for both liquid and solid biopsy, leveraging the principle of surface plasmon resonance to enhance the typically weak Raman signal, providing molecule-specific insights into the analyzed sample for both biofluid and tumor samples. The disclosure also involves the use of SERS, IR spectroscopy, and / or mass spectroscopy to identify disease biomarkers from body fluids and tissue samples.BACKGROUND

[0004] Glioblastoma ("GB" but which may in some portions of the descript and figures also be referred to as "GMB") is the most prevalent and lethal brain tumor in adults with less than 15-month survival. GB mostly manifests with neurological dysfunction, which can also be associated with other lesions such as abscess, cerebral toxoplasmosis, tumefactive demyelination, primary central nervous system lymphoma, or brain metastasis. Certain lesions require different treatment approaches, rendering it necessary to distinguish them from GBs.

[0005] However, current imaging techniques such as MRI are not able to make this distinction, and final diagnosis requires the histopathological and molecular genetic evaluation of surgical biopsy samples.SUMMARY

[0006] GB recurrence is the main cause of death and affects 90% of patients after treatment, making follow-up with MRI necessary. It is imperative to differentiate progression from treatment-related changes such as pseudo progression, pseudo response, and radiation necrosis that may mimic tumor recurrence / progression, which are indistinguishable from conventional imaging.

[0007] This dilemma poses a profound clinical challenge for subsequent decision making and survival assessment. Advanced radiological imaging techniques such as MRI, CT, and PET for the detection of glioma recurrence / progression could provide help to some extent but fail to predict or conclusively ascertain the occurrence of or true early progression and place another heavy financial burden on the family. Thus, developing a blood extracellular vesicle (EV)-based GB detection platform can be a complimentary noninvasive differential diagnostic tool to ensure an accurate preoperative assessment and influence the course of follow-up treatment.

[0008] GB diagnosis faces difficulties due to its molecular heterogeneity, which demands a combination of molecular pathology and histopathological analysis. The complexity of GB's molecular heterogeneity calls for additional molecular genetics tests as per 2021 World Health Organization Central Nervous System (WHO CNS) tumor classification guidelines. However, these tests are costly, time-consuming, and require trained expert staff, making them inaccessible to some patients.

[0009] Therefore, there remains a need for a rapid and cost-effective screening method to address this clinical challenge. However, no clinically applicable liquid biopsy assays are available today.

[0010] Furthermore, identifying tumor margins is important in terms of allowing safemaximum resection of the glioblastoma and reducing recurring tumors. Therefore, a moleculespecific non-invasive tool which can provide the information about the margin in between tumorous and healthy tissue would have significant value. Such a tool and platform are disclosed herein.

[0011] Since utilizing the current imaging modalities (CT, MRI) only are insufficient in terms of brain tumor classification, the diagnosis of brain tumors requires more extensive analysis.Current diagnostic gold standard for brain tumor classification is histopathology and molecular pathology when needed, which requires invasive procedures such as biopsy or surgical resection. Moreover, current diagnostic imaging modalities are not sufficient to distinguish the tumor recurrence from pseudoprogression or radiation necrosis and may require biopsy or surgery for only diagnosis, when detected. There are currently no clinically applicable brain tumor biomarkers determined, as blood brain barrier selectively permits the passage of biomolecules to central blood circulation. On the other hand, extracellular vesicles are highly promising nanomolecules in cancer diagnostics with the ability to carry the parental information and to cross blood brain barrier due to the signaling molecules on their membrane.

[0012] First, this approach is innovative because it is the first time that GB EVs are studied using SERS-based sensors from plasma, and cancerous and healthy brain tissues simultaneously. GB-specific EVs can be captured by and detected through their moleculespecific SERS signals. Second, a SERS library of GB EVs can be created using tumor and plasma derived EVs to compare their SERS profiles to identify GB signatures. The platform can be tested on unique human-derived samples using EVs from non-cancerous and healthy samples for improved diagnostic precision and better treatment management of GB in clinics. Third, the development of a Glio-SERS-based detection tool for GB combines SERS and nanoplasmonic metasurface technologies into a single platform, providing molecule-specific unique signatures of EV cargo via non-invasive analysis of the samples. This method permits the development of a low-cost point-of-care biosensor (Glio-SERS) for the assessment of disease, based on the Raman signals of GB EVs. This capability tremendously benefits to empower the existing methods lacking from the toolset that is needed to support the improvement of precision medicine.

[0013] The three key innovative aspects can be leveraged to develop, test, and demonstrate the strength of the new platform for GB detection. First, as an EV-based approach to diagnose brain cancers does not exist currently, we here propose the innovative, paradigmshifting technique using of plasma EVs as markers for GB. For this, the use of rare samples of cancerous (GB), non-glial and brain-tumor-free (BTF) plasma and tissues from patients can be used to detect EVs on SERS surfaces. Using both plasma and tissue samples and their EV signatures can be uniquely informative in that this information can be longitudinally obtainedboth before and after surgery, alongside the removed brain tumor samples. Second, an extracellular vesicle-based GB detection technology platform is employed. For this, an extracellular vesicle-based SERS biosensor platform is created and a Raman spectral library of both plasma and tumor EVs can be constructed to compare their SERS profiles and identify GB signatures. The SERS technique can provide molecule-specific information that is up to 600-fold more sensitive than existing optical (e.g., colorimetric, SPR-based or fluorescence) sensors, rendering our approach as an especially powerful one to detect EV signatures. Overall, this integrated translational approach offers a low-cost biosensor for GB assessment with high sensitivity using Raman signals of GB EVs, making it a valuable addition to existing methods in precision medicine. Other biofluids including sweat, urine, serum, whole blood, CSF, saliva can be used to identify disease related EV and other biomarkers / biomolecules in addition to tissue samples.

[0014] Identifying the disease specific EVs and other disease relevant biomarkers from biofluids or tissue can allow rapid molecular analysis and will reduce invasive, costly, and / or time-consuming steps in the clinics, as well as intraoperatively assisting surgeons to distinguish the tumor margins based on SERS signatures via EVs or directly from tissue signatures.

[0015] The disclosed platform includes four main technical parts: a portable fiber Surface Enhanced Raman spectroscopy module, removable fiber probe for intraoperative utilization, disposable nanoplasmonic cartridges for individual sample loading, and an updateable spectral library collected from healthy and diseased samples. This technology has been developed to distinguish the molecule-specific Raman spectral signatures of diseased tissue or biofluid extracellular vesicles (EVs) and other biomarkers, isolated from patient biofluids using Al-based classification algorithms based on the spectral signatures for different tumors. The EV and non- EV biomarkers can be directly identified using SERS and / or targeted and captured by aptamers, antibodies, or other DNA / RNA sequences.

[0016] The database includes a spectral library composed of tumor tissue and / or corresponding plasma or biofluid samples obtained pre, intra, and / or post operatively, non- tumorous tissue, and / or plasma samples from tumor free healthy volunteers. Biofluid derived EVs or biomarkers are isolated and deposited on the disposable SERS cartridges and be installed into portable fiber SERS reader for signal acquisition. Once the EV signal acquisition iscompleted, the embedded Al algorithms can analyze the Raman spectral signatures of the sample and classify the data according to the spectral library.

[0017] The EVs or other disease biomarkers (i.e., DNA, RNA, protein) can be used to assess molecular alterations (i.e., mutations, copy number variations, expression levels, chromosome gain / loss) of the tumor and can be used to support physicians in tumor diagnosis.

[0018] The product can be used by hospitals, diagnostic and research laboratories to distinguish the Surface Enhanced Raman Spectroscopy (SERS) signatures of disease related extracellular vesicles (EVs) or other biomarkers derived from patient tumor or biofluids (such as plasma, tissue fluid, cerebrospinal fluid, serum, urine, sweat, saliva) from the healthy sample derived EVs using the updateable spectral library as a diagnostic platform. In addition, this technology has applications beyond brain cancers such as neurodegenerative diseases.

[0019] According to one aspect, a diagnostic platform for detecting and analyzing brain tumors is disclosed. The platform includes a portable fiber surface enhanced Raman spectroscopy (SERS) module, a removable fiber probe for intraoperative use, disposable nanoplasmonic cartridges for individual or multiple sample loading, and an updatable spectral library collected from healthy and diseased samples. The platform is configured to analyze blood and tissue signatures and employ artificial intelligence (Al) models to classify SERS spectra based on an updatable proteomics and SERS spectral library encompassing distinct brain tumor types.

[0020] In some forms of the platform, the portable fiber surface enhanced Raman spectroscopy (SERS) module may be configured to amplify Raman scattering signals from molecules adsorbed on or in close proximity to engineered surfaces.

[0021] In some forms of the platform, the removable fiber probe may be integrated into a Cavitron Ultrasonic Surgical Aspirator (CUSA) instrument with an embedded camera for intraoperative utilization.

[0022] In some forms, the disposable nanoplasmonic cartridges may be designed for the isolation and deposition of extracellular vesicles (EVs) from patient biofluids and tissues.

[0023] In some forms, the updatable spectral library may include Raman spectral signatures of diseased tissue, blood derivatives, or biofluids, and / or EVs isolated from patient biofluids and tissues.

[0024] In some forms, the platform may be further configured to utilize artificial intelligence (Al) algorithms for the analysis and classification of the recorded SERS spectra.

[0025] In some forms, the platform may be further configured to update the spectral library with new spectral data collected from additional healthy and diseased samples.

[0026] According to another aspect, a method for early cancer detection is provided. The method includes performing surface enhanced Raman spectroscopy on a sample containing extracellular vesicles to obtain a Raman spectra for the sample and analyzing the Raman spectra for the sample against a proteomics and SERS spectral library and providing a classification for the sample.

[0027] In some forms, the method may further include obtaining the sample containing extracellular vesicles via liquid biopsy.

[0028] In some forms, analyzing the Raman spectra for the sample against the proteomics and SERS spectral library and providing the classification for the sample may involve employing artificial intelligence models to perform the classification.

[0029] In some forms, providing a classification for the sample may encompass providing a distinct brain tumor type based on comparison of the Raman spectra for the sample against the proteomics and SERS spectral library.

[0030] In some forms, the proteomics and SERS spectral library may be updatable. In such case, the method may further involve adding the Raman spectra for the sample to the proteomics and SERS spectral library. Adding the Raman spectra for the sample to the proteomics and SERS spectral library may involve providing a classification (for example, a separate clinical classification) for the sample which clinical classification information is used to further develop an artificial intelligence model to perform classification. In this way, the accuracy of the models used may be improved.

[0031] In some forms, providing a classification for the sample may entail providing a classification relating to a presence of cancer based on comparison of the Raman spectra for the sample against the proteomics and SERS spectral library. In some forms, the classification relating to a presence of cancer may further include information relating to a type of cancer.

[0032] In some forms of the method, the method may further involve isolating extracelluar vesicles from a liquid biopsy sample before the step of performing surface enhanced Raman spectroscopy on a sample containing extracellular vesicles.

[0033] According to yet another aspect, a nanoplasmonic cartridge is provided for processing a liquid biopsy sample. The nanoplasmonic cartridge is capable of collecting and isolating extracelluar vesicles from the liquid biopsy sample for surface enhanced Raman spectroscopy (SERS) spectral collection. Accordingly, the nanoplasmonic cartridge may employ an extracellular vesicle isolation process and then present the processed extracelluar vesicles is a state for surface enhanced Raman spectroscopy (SERS) spectral collection.

[0034] These and still other advantages of the invention will be apparent from the detailed description and drawings. What follows is merely a description of some preferred embodiments of the present invention. To assess the full scope of the invention, the claims should be looked to as these preferred embodiments are not intended to be the only embodiments within the scope of the claims.BRIEF DESCRIPTION OF THE FIGURES

[0035] FIG. 1 is a schematic illustration of a study including, in various steps, sample collection, plasma separation, extracellular vesicle isolation and characterization from Glioblastoma (GB) and meningioma (MNG) patients and healthy individuals, SERS spectra collection, data analysis, and spectral classification. In step (a), blood is withdrawal from brain tumor patients and healthy individuals. In step (b), plasma is separated and extracellular vesicles are isolated from plasma samples via ExoTIC and characterized according to MISEV2023 guidelines; in step (c), Raman spectral measurement of plasma extracellular vesicles on a gold nanopillar patterned surface are obtained; in step (d) pre-processing steps including cosmic ray removal, denoising, and baseline correction were performed; in step (e), a pre-processed spectra dataset was created for Al classification; and in step (f) multiple Al models were trained and tested with plasma EV Raman spectra from GB, MNG and HC samples, and EV SERS spectra were classified.

[0036] FIGS. 2A through 2D illustrates characterization of extracellular vesicles. In FIG. 2ARoom Temperature Transmission Electron Microscopy (TEM) images are shown in the left-side column while Cryo- Transmission Electron Microscopy (Cryo-TEM) are shown in the right-sidecolumn. FIG. 2B shows Nanoparticle Tracking Analysis (NTA) results in the leftmost column and Interferometric Scattering Microscopy (iSCAT) size distribution and image in the center and rightmost columns, respectively. FIG. 2C shows Western Blot results. FIG. 2D shows Bead- Captured Flow Cytometry (BC-FC) analysis of extracellular vesicles.

[0037] FIGS. 3A through 3F show Surface-Enhanced Raman Spectral analysis of plasma extracellular vesicles from brain tumor patients and healthy individuals. FIG. 3A shows an overall SERS Spectra of Glioblastoma (GB), Meningioma (MNG), and Healthy Control (HC) samples. FIG. 3B shows PC loadings of overall SERS spectra. FIG. 3C shows 3D PCA distribution of SERS spectra. FIGS. 3D, 3E, and 3F show violin plots for the peaks corresponding to the distinguished peaks attributed to tryptophane (1069 cmx), myosin (1304 cm-1) and nucleid acids:DNA / RNA (1422 cm1), respectively.

[0038] FIG. 4A through 4E show Surface-Enhanced Raman Spectroscopy (SERS) and Al- based analysis of extracellular vesicles obtained from participants with Glioblastoma (GB), Meningioma (MNG), and Healthy Controls (HC). FIG. 4A illustrates an overall pipeline of the Al- based analysis of the SERS spectra. FIG. 4B provides a schematic explanation of artificial intelligence-based SERS data analysis structure in which the SERS data obtained from small extracellular vesicles (sEV) from participants with GB, MNG, and HC. Two different approaches were used to analyze SERS data: the data split by patient (left) and split by spectra (right). FIGS. 4C and 4D provide a confusion matrix for the classification of brain tumor (BT) patients with GB and MNG and from brain tumor free (BTF) individuals and for multi-class classification, respectively. FIG. 4E provides an ROC curve of extracellular vesicle SERS spectra derived from GB and MNG patients and HC individuals.

[0039] FIGS. 5A through 5G shows EGFR and relevant amino acid analysis GB, MNG, and HC sample derived extracellular vesicles. FIGS, 5A and 5C show regression analysis and FIGS. 5B and 5D show RMSE predictions of Tryptophane and Cysteine SERS peak in GB, MNG, and HC sample derived extracellular vesicles. FIGS. 5E and 5F show predicted presence of Tryptophane and Cysteine in clinical sample derived extracellular vesicle samples. FIG. 5G shows the spectra differences of EGRF versus EGFRvI 11.

[0040] FIG. 6 shows a schematic of plasma and tissue collection from a patient, a portableSERS instrument, and a reader smart device to analyze spectral data (i.e., SERS spectra, mass spectra, IR spectra.).

[0041] FIG. 7 provides a schematic depicting a removable fiber probe from portable reader instrument for intraoperative use for the assessment of brain tumor margins, identifying healthy and tumorous region boundaries using optical and proteomics techniques.

[0042] FIGS. 8A through 8F provide proteomics analysis of plasma EVs derived from glioblastoma (GB), meningioma (MNG), and healthy control (HC) samples. FIG. 8A shows a Venn diagram of identified proteins detected in at least one injection per sample across GB, MNG, and GB groups. FIGS. 8B and 8C provide volcano plots that show the differential protein abundances between the MNG vs GB and the HC vs brain tumor (GB and MNG) groups, respectively. The x-axis represents the fold change (log2scale) of protein abundance, while the y-axis shows the statistical significance as -logio(p-value). Proteins with a significant increase in abundance in one group are shown in red (positive fold change, generally in the upper right areas), while proteins significantly more abundant in the other group are displayed in blue (negative fold change, generally in the upper left areas). Non-significant proteins are shown in black (generally centering at 0 at the bottom of the plot). FIG. 8D is a heatmap of differentially abundant proteins (P < 0.05) between MNG and GB samples. The heatmap displays proteins with a significant difference in abundance (P < 0.05) when comparing MNG and GB groups. Rows represent individual proteins, while columns represent samples within each group. Z- scores indicate relative abundance levels, with red representing higher abundance and blue having lower abundance. The proteins are ranked based on their Z-score differences between groups and the statistical significance is represented by the asterisks (* for p < 0.05, ** for p < 0.01, etc.). FIG. 8E is a proteomics heatmap. FIG. 8F is a box plot representation of significant Raman shifts attributed to proteins.DETAILED DESCRIPTION

[0043] Surface-Enhanced Raman Spectroscopy (SERS) is a powerful analytical technique that amplifies Raman scattering signals from molecules adsorbed on or in close proximity to engineered surfaces. This technique leverages the principle of surface plasmon resonance to enhance the typically weak Raman signal, providing molecule-specific insights into the analyzedsample. SERS has been widely used in various fields, including biomedical applications, for its ability to provide detailed molecular information.

[0044] In the field of cancer diagnostics, the detection and analysis of brain tumors present a considerable challenge. Traditional methods often involve invasive procedures such as surgical biopsies, which can be risky and uncomfortable for patients. Furthermore, these methods often require histopathological and molecular genetic evaluation of the biopsy samples, which can be costly and time-consuming.

[0045] Liquid biopsy technologies are rapidly transforming clinical diagnostics, providing non-invasive insights into disease progression, treatment response, patient stratification across diverse medical fields, including oncology, cardiovascular, and neurological diseases. Traditional diagnostic methods such as histopathology, molecular testing, and imaging often involve invasive procedures, specialized expertise, high costs, and delayed results, highlighting the immense need for rapid, minimally invasive, and broadly applicable diagnostic alternatives.

[0046] Circulating biomolecules, including micro RNA (miRNA), cell-free DNA (cfDNA), circulating tumor DNA (ctDNA), natural killer (NK) cells, cancer stem cells (CSCs), extracellular vesicles (EVs), proteins, and metabolites have emerged as promising targets, reflecting the pathological state of their parental cells and tissues. Extracellular vesicles are broadly defined as a small, lipid-bound particle secreted by cells into the extracellular space. These vesicles play a role in intercellular communication. EVs can be further categorized into subtypes such as exosomes, microvesicles, and apoptotic bodies based on their size, origin, and function.

[0047] Among these targets, small extracellular vesicles (sEVs), nanoscale vesicles (<220 nm), are particularly attractive as universal biomarkers due to their abundance, stability, and capacity to reflect disease-specific biological processes. Circulating EVs provide unique molecular insights into pathological conditions, carrying immunogenic, tumorigenic, and angiogenic information that mirrors the complexity of their microenvironment.

[0048] Raman spectroscopy, a label-free analytical technique based on molecular vibrational modes, provides detailed molecular fingerprints reflecting the biochemical composition of biological samples. Surface-Enhanced Raman Spectroscopy (SERS), an ultrasensitive variant of Raman spectroscopy, combined with artificial intelligence (Al)-d riven data analytics offers a powerful approach for rapid, accurate, and high-throughput detection ofbiomolecular signatures. Although SERS-based EV analyses have demonstrated potential using various biofluids including blood, urine, saliva, cell-culture, tumor, and CSF, clinical adoption has been hindered by analytical complexity, scalability challenges, and insufficient validation across diverse disease contexts.

[0049] Herein, we present an integrated liquid biopsy platform that combines rapid and standardized EV isolation from plasma using ExoTIC (see, e.g., U.S. patent no. 11,073,511 which is incorporated herein by reference in its entirety for a description of the ExoTIC platform), molecular fingerprinting using SERS, and Al-driven deep learning classification. To illustrate the versatility, robustness, and translational potential of the EV signature detection approach, 73 clinical plasma samples from patients diagnosed with glioblastoma (GB, n=20), meningioma (MNG, n=23), and healthy controls (HC, n=30). Proteomic analyses complemented spectral data were analyzed, further validating identified molecular signatures. This platform provides a scalable, minimally invasive, and generalizable diagnostic framework with significant potential to advance precision medicine across a wide spectrum of diseases.

[0050] Accordingly, one object of the present disclosure is to provide a diagnostic platform for the detection and analysis of brain tumors. This platform is designed to analyze blood and tissue signatures and employ artificial intelligence (Al) models to classify Surface-Enhanced Raman Spectroscopy (SERS) spectra. The platform comprises a portable fiber SERS module, a removable fiber probe for intraoperative use, disposable nanoplasmonic cartridges for individual sample loading, and an updatable spectral library collected from healthy and diseased samples.

[0051] The present disclosure relates, at least in part, to a diagnostic platform for detecting and analyzing brain tumors. More specifically, the platform may include a portable fiber Surface Enhanced Raman spectroscopy (SERS) module, a removable fiber probe for intraoperative use, disposable nanoplasmonic cartridges for individual sample loading, and an updatable spectral library collected from healthy and diseased samples. The platform is designed to analyze blood and tissue signatures and employ artificial intelligence (Al) models to classify SERS spectra based on an updatable proteomics and SERS spectral library encompassing distinct brain tumor types.

[0052] In some aspects, the portable fiber Surface Enhanced Raman spectroscopy (SERS) module may amplify Raman scattering signals from molecules adsorbed on or in close proximity to engineered surfaces. The removable fiber probe may be integrated into a Cavitron Ultrasonic Surgical Aspirator (CUSA) instrument with an embedded camera for intraoperative utilization. The disposable nanoplasmonic cartridges are designed for the isolation and deposition of extracellular vesicles (EVs) from patient biofluids and tissues. The updatable spectral library may comprise Raman spectral signatures of diseased tissue, blood derivatives, or biofluids, and / or EVs isolated from patient biofluids and tissues.

[0053] The platform may utilize artificial intelligence (Al) algorithms for the analysis and classification of the recorded SERS spectra. The spectral library can be updated with new spectral data collected from additional healthy and diseased samples. This platform provides a rapid, non-invasive, and analytical liquid and solid biopsy technique for the detection and monitoring of brain tumors. It also offers an intraoperative tool for accurate assessment of tumor boundaries, which can guide neurosurgeons during surgical procedures.

[0054] The diagnostic platform may offer several technical advantages. For instance, it may provide a more effective and efficient method for detecting and monitoring brain tumors compared to existing techniques. The platform's ability to analyze blood and tissue signatures and classify SERS spectra based on an updatable proteomics and SERS spectral library may enhance the accuracy of brain tumor detection and analysis. Furthermore, the platform's intraoperative tool may assist in accurately identifying tumor boundaries during surgical procedures, potentially improving surgical outcomes and patient prognosis.

[0055] In some embodiments, the diagnostic platform may include a portable fiber Surface Enhanced Raman spectroscopy (SERS), and / or mass spectroscopy, and / or IR spectroscopy module. This module or these modules may be configured to amplify Raman scattering / IR spectra and / or mass spectra signals from molecules adsorbed on or in close proximity to engineered surfaces. This amplification of Raman scattering signals may enhance the detection sensitivity of the platform, enabling the detection of minute concentrations of tumor biomarkers in the analyzed samples.

[0056] In other embodiments, the diagnostic platform may include a removable fiber probe for intraoperative use. This probe may be integrated into a Cavitron Ultrasonic SurgicalAspirator (CUSA) instrument with an embedded camera. This integration may allow for realtime imaging and spectroscopic analysis during surgical procedures, potentially aiding in the accurate identification of tumor boundaries and improving surgical outcomes. To perform volumetric imaging of the tissue, photonic crystals and wavefront shaping options are provided for intraoperative deep tissue imaging and analysis. The fiber probe may be designed to be removable, allowing for easy replacement or cleaning, further enhancing the practicality and usability of the platform in various settings.

[0057] In yet other embodiments, the diagnostic platform may include disposable nanoplasmonic cartridges for individual sample loading. These cartridges may be designed for the isolation and deposition of extracellular vesicles (EVs) or biospecimens from patient biofluids and tissues. This feature may facilitate the efficient and effective isolation of EVs, which serve as non-invasive biomarkers for brain cancers, from patient samples.

[0058] In further embodiments, the diagnostic platform may include an updatable spectral library collected from healthy and diseased samples. This library may comprise spectral signatures of diseased tissue, blood derivatives, or biofluids, and / or EVs isolated from patient biofluids and tissues. The updatable nature of the spectral library may allow for the continuous enrichment of the library with new spectral data, potentially improving the accuracy and reliability of the platform's diagnostic capabilities.

[0059] The diagnostic platform may provide several technical advantages. For instance, the integration of a portable fiber SERS module, a removable fiber probe, disposable nanoplasmonic cartridges, and an updatable spectral library may enable the platform to provide rapid, non-invasive, and accurate detection and analysis of brain tumors. Furthermore, the platform's ability to analyze blood and tissue signatures and classify SERS spectra based on an updatable proteomics, IR and SERS spectral library may enhance the accuracy of brain tumor detection and analysis. Additionally, the platform's intraoperative tool may assist in accurately identifying tumor boundaries during surgical procedures, potentially improving surgical outcomes and patient prognosis with or without using targeting molecules including aptamers, antibodies, DNA / RNA fragments, and proteins.

[0060] In some cases, the portable fiber Surface Enhanced Raman spectroscopy (SERS) module may be configured to amplify Raman scattering signals from molecules adsorbed on orin close proximity to engineered surfaces. This amplification of Raman scattering signals may enhance the detection sensitivity of the platform, enabling the detection of minute concentrations of tumor biomarkers in the analyzed samples. The SERS module may utilize the principle of surface plasmon resonance to enhance the typically weak Raman signal, providing molecule-specific insights into the analyzed sample. This may allow for a more detailed and accurate analysis of the sample, potentially improving the diagnostic accuracy of the platform.

[0061] In other embodiments, the SERS module may be designed to be portable, allowing for easy transportation and use in various settings, such as hospitals, clinics, and research laboratories. This portability may increase the accessibility and convenience of the platform, potentially facilitating its widespread adoption in the field of brain tumor diagnostics. The SERS module may also be designed to be user-friendly, requiring little to no specialized training to operate, further enhancing its usability and practicality in various settings.

[0062] The use of the portable fiber SERS module in the diagnostic platform may provide several technical advantages. For instance, the amplification of Raman scattering signals may enhance the detection sensitivity of the platform, enabling the detection of minute concentrations of tumor biomarkers in the analyzed samples. This may allow for early detection and monitoring of brain tumors, potentially improving patient prognosis. Furthermore, the portability and user-friendliness of the SERS module may increase the accessibility and convenience of the platform, potentially facilitating its widespread adoption in the field of brain tumor diagnostics.

[0063] In other embodiments, the diagnostic platform may include an integrated compartment for easy in-cavity application to assist neurosurgeons to guide the surgery. This integrated compartment may be designed to accommodate the removable fiber probe, facilitating its easy insertion and removal during surgical procedures. This feature may allow for the precise positioning of the fiber probe within the surgical cavity, potentially improving the accuracy of tumor boundary identification and enhancing surgical outcomes.

[0064] The use of the removable fiber probe and the integrated compartment in the diagnostic platform may provide several technical advantages. For instance, the integration of the fiber probe into a CUSA instrument with an embedded camera may allow for real-time imaging and spectroscopic analysis during surgical procedures, potentially aiding in the accurateidentification of tumor boundaries and improving surgical outcomes. Furthermore, the inclusion of an integrated compartment for easy in-cavity application may facilitate the precise positioning of the fiber probe within the surgical cavity, potentially improving the accuracy of tumor boundary identification and enhancing surgical outcomes. Additionally, the design of the fiber probe to be removable may allow for easy replacement or cleaning, further enhancing the practicality and usability of the platform in various settings.

[0065] In some embodiments, the diagnostic platform may include disposable nanoplasmonic cartridges for individual sample loading. These cartridges may be designed for the isolation and deposition of extracellular vesicles (EVs) from patient biofluids and tissues. The EVs, which may include extracellular vesicles, microvesicles, small and large EVs, oncosomes, endosomes, exomeres, among others, serve as non-invasive biomarkers for brain cancers. The cartridges may be designed to allow individual or multiple simultaneous analysis of samples, facilitating efficient and effective isolation and analysis of EVs from patient samples.

[0066] In other embodiments, the diagnostic platform may be applied for cerebrospinal fluid (CSF), interstitial tissue fluid, or solid organ tumor in addition to the blood and blood derivatives such as serum, plasma, platelets etc. This versatility in sample types may enhance the applicability of the platform in various clinical and research settings, potentially improving the detection and monitoring of brain tumors.

[0067] The use of disposable nanoplasmonic cartridges in the diagnostic platform may provide several technical advantages. For instance, the cartridges may facilitate the efficient and effective isolation and deposition of EVs from patient biofluids and tissues, potentially enhancing the sensitivity and specificity of the platform in detecting and analyzing brain tumors. Furthermore, the ability of the cartridges to allow individual or multiple simultaneous analysis of samples may increase the throughput of the platform, potentially facilitating its use in large-scale clinical and research studies. Additionally, the versatility of the platform in analyzing various types of samples may enhance its applicability in various clinical and research settings, potentially improving the detection and monitoring of brain tumors.

[0068] In some embodiments, the diagnostic platform may include an updatable spectral library collected from healthy and diseased samples. This library may comprise Raman spectral signatures of diseased tissue, blood derivatives, or biofluids, and / or EVs isolated from patientbiofluids and tissues. The spectral library may be designed to be updatable, allowing for the continuous enrichment of the library with new spectral data collected from additional healthy and diseased samples. This feature may facilitate the efficient and effective analysis of the recorded SERS spectra, potentially enhancing the diagnostic accuracy of the platform.

[0069] In other embodiments, the spectral library may include spectral data collected from tumor tissue and / or corresponding blood and / or blood derivatives or biofluid samples obtained pre, intra, and / or post operatively, non-tumorous tissue, from tumor free and healthy volunteers. This comprehensive collection of spectral data may provide a rich source of information for the classification of SERS spectra, potentially improving the specificity and sensitivity of the platform in detecting and analyzing brain tumors.

[0070] The use of an updatable spectral library in the diagnostic platform may provide several technical advantages. For instance, the continuous enrichment of the spectral library with new spectral data may enhance the accuracy and reliability of the platform's diagnostic capabilities. Furthermore, the comprehensive collection of spectral data from various types of samples may enhance the specificity and sensitivity of the platform in detecting and analyzing brain tumors. Additionally, the updatable nature of the spectral library may allow for the adaptation of the platform to evolving diagnostic challenges, potentially enhancing its longterm usability and effectiveness in the field of brain tumor diagnostics.

[0071] In some embodiments, the diagnostic platform may utilize artificial intelligence (Al) algorithms for the analysis and classification of the recorded SERS spectra. These Al algorithms may include machine learning models, deep learning models, or other types of Al models suitable for spectral data analysis. The Al algorithms may be designed to analyze the Raman spectral signatures of the sample and classify the data according to the spectral library. This feature may facilitate the efficient and effective analysis of the recorded SERS spectra, potentially enhancing the diagnostic accuracy of the platform.

[0072] The use of Al algorithms in the diagnostic platform may provide several technical advantages. For instance, the Al algorithms may facilitate the efficient and effective analysis of the recorded SERS spectra, potentially enhancing the diagnostic accuracy of the platform. Furthermore, the development and validation of the platform with patient samples mayenhance the specificity and sensitivity of the platform in detecting and analyzing brain tumors, potentially improving patient prognosis.

[0073] In some embodiments, the diagnostic platform may be configured to update the spectral library with new spectral data collected from additional healthy and diseased samples. This feature may allow for the continuous enrichment of the spectral library with new spectral data, potentially improving the accuracy and reliability of the platform's diagnostic capabilities. The spectral library may be updated with spectral data collected from various types of samples, including but not limited to, tumor tissue, corresponding blood and / or blood derivatives, or biofluid samples obtained pre, intra, and / or post operatively, and non-tumorous tissue from tumor-free and healthy volunteers. This comprehensive collection of spectral data may provide a rich source of information for the classification of SERS spectra, potentially improving the specificity and sensitivity of the platform in detecting and analyzing brain tumors.

[0074] In other embodiments, the spectral library may be enriched with distinct types of tumors and specified for particular conditions or tumor subtypes for higher accuracy. This feature may allow for the adaptation of the platform to evolving diagnostic challenges, potentially enhancing its long-term usability and effectiveness in the field of brain tumor diagnostics. The ability to update and enrich the spectral library with new spectral data may provide a dynamic and adaptable diagnostic tool that can keep pace with advances in brain tumor research and clinical practice.

[0075] The use of an updatable spectral library in the diagnostic platform may provide several technical advantages. For instance, the continuous enrichment of the spectral library with new spectral data may enhance the accuracy and reliability of the platform's diagnostic capabilities. Furthermore, the comprehensive collection of spectral data from various types of samples may enhance the specificity and sensitivity of the platform in detecting and analyzing brain tumors. Additionally, the updatable nature of the spectral library may allow for the adaptation of the platform to evolving diagnostic challenges, potentially enhancing its longterm usability and effectiveness in the field of brain tumor diagnostics.

[0076] In some embodiments, the diagnostic platform may be developed for and validated with patient samples. These patient samples may include a wide variety of biofluids (i.e., blood samples, tissue samples, cerebrospinal fluid (CSF) samples, interstitial tissue fluid samples, orsolid organ tumor samples, among others). The platform may be designed to analyze these samples and categorize SERS spectra based on an updatable proteomics and SERS spectral library encompassing distinct brain tumor types. This feature may enhance the specificity and sensitivity of the platform in detecting and analyzing brain tumors, potentially improving patient prognosis.

[0077] The development and validation of the diagnostic platform with patient samples may provide several technical advantages. For instance, the use of actual patient samples may ensure that the platform is capable of accurately detecting and analyzing brain tumors in a real- world clinical setting. This may enhance the reliability and clinical relevance of the platform, potentially improving its adoption and use in the field of brain tumor diagnostics. Furthermore, the ability of the platform to analyze a wide range of sample types, including blood, tissue, CSF, interstitial tissue fluid, and solid organ tumor samples, may enhance its versatility and applicability in various clinical and research settings. This may allow for the comprehensive detection and analysis of brain tumors, potentially improving patient prognosis.

[0078] In some embodiments, the diagnostic platform may include an integrated compartment for easy in-cavity application to assist neurosurgeons during surgery. This integrated compartment may be designed to accommodate the removable fiber probe, facilitating its easy insertion and removal during surgical procedures. This feature may allow for the precise positioning of the fiber probe within the surgical cavity, potentially improving the accuracy of tumor boundary identification and enhancing surgical outcomes.

[0079] The integrated compartment may be designed to be easily accessible and user- friendly, allowing for quick and efficient in-cavity application of the fiber probe. This may reduce the time and effort involved in positioning the fiber probe within the surgical cavity, potentially improving the efficiency of surgical procedures. Furthermore, the integrated compartment may be designed to securely hold the fiber probe in place during surgery, potentially reducing the risk of probe displacement and ensuring consistent and reliable spectral data collection.

[0080] The use of an integrated compartment in the diagnostic platform may provide several technical advantages. For instance, the easy in-cavity application of the fiber probe may facilitate the precise positioning of the probe within the surgical cavity, potentially improvingthe accuracy of tumor boundary identification and enhancing surgical outcomes. Furthermore, the user-friendly design of the integrated compartment may improve the efficiency of surgical procedures, potentially reducing surgical time and improving patient outcomes. Additionally, the secure holding of the fiber probe in the integrated compartment may ensure consistent and reliable spectral data collection, potentially enhancing the diagnostic accuracy of the platform.

[0081] In some embodiments, the diagnostic platform may be applied for cerebrospinal fluid (CSF), interstitial tissue fluid, or solid organ tumor in addition to the blood and blood derivatives such as serum, plasma, platelets etc. This versatility in sample types may enhance the applicability of the platform in various clinical and research settings, potentially improving the detection and monitoring of brain tumors. The platform may be designed to analyze these samples and categorize SERS spectra based on an updatable proteomics and SERS spectral library encompassing distinct brain tumor types. This feature may enhance the specificity and sensitivity of the platform in detecting and analyzing brain tumors, potentially improving patient prognosis.

[0082] The use of the diagnostic platform for various types of samples, including CSF, interstitial tissue fluid, solid organ tumor, and blood derivatives, may provide several technical advantages. For instance, the ability to analyze a wide range of sample types may enhance the versatility and applicability of the platform in various clinical and research settings. This may allow for the comprehensive detection and analysis of brain tumors, potentially improving patient prognosis. Furthermore, the categorization of SERS spectra based on an updatable proteomics and SERS spectral library may enhance the specificity and sensitivity of the platform in detecting and analyzing brain tumors. This may lead to more accurate and reliable diagnostics, potentially improving patient outcomes.

[0083] In some embodiments, the diagnostic platform may include an updatable spectral library that can be enriched with distinct types of tumors and specified for particular conditions or tumor subtypes for higher accuracy. In some embodiments, the diagnostic platform may include an updatable spectral library composed of tumor tissue and / or corresponding blood and / or blood derivatives or biofluid samples obtained pre, intra, and / or post operatively, non- tumorous tissue, from tumor free and healthy volunteers. This spectral library may comprise Raman spectral signatures of diseased tissue, blood derivatives, or biofluids, and / or EVsisolated from patient biofluids and tissues. The spectral library may be designed to be updatable, allowing for the continuous enrichment of the library with new spectral data collected from additional healthy and diseased samples. This feature may facilitate the efficient and effective analysis of the recorded SERS spectra, potentially enhancing the diagnostic accuracy of the platform.

[0084] In other embodiments, the spectral library may be updated with spectral data collected from various types of samples, including but not limited to, tumor tissue, corresponding blood and / or blood derivatives, or biofluid samples obtained pre, intra, and / or post operatively, and non-tumorous tissue from tumor-free and healthy volunteers. This comprehensive collection of spectral data may provide a rich source of information for the classification of SERS spectra, potentially improving the specificity and sensitivity of the platform in detecting and analyzing brain tumors.

[0085] The use of an updatable spectral library in the diagnostic platform may provide several technical advantages. For instance, the continuous enrichment of the spectral library with new spectral data may enhance the accuracy and reliability of the platform's diagnostic capabilities. Furthermore, the comprehensive collection of spectral data from various types of samples may enhance the specificity and sensitivity of the platform in detecting and analyzing brain tumors. Additionally, the updatable nature of the spectral library may allow for the adaptation of the platform to evolving diagnostic challenges, potentially enhancing its longterm usability and effectiveness in the field of brain tumor diagnostics.

[0086] In some embodiments, the diagnostic platform may be designed to distinguish tumor boundaries from healthy tissue regions based on SERS signatures of tumor biomarkers, allowing surgeons to intraoperatively see the tumor margins to make a maximum safe resection during operations. This feature may be facilitated by the removable fiber probe that is integrated into a Cavitron Ultrasonic Surgical Aspirator (CUSA) instrument with an embedded camera for intraoperative utilization. The fiber probe may be designed to collect SERS spectra from the surgical site, providing real-time, molecule-specific insights into the tissue being operated on. This may allow surgeons to accurately identify tumor boundaries and differentiate them from healthy tissue regions, potentially improving the precision of tumor resection and enhancing surgical outcomes.

[0087] In other embodiments, the diagnostic platform may include an integrated compartment for easy in-cavity application to assist neurosurgeons during surgery. This integrated compartment may be designed to accommodate the removable fiber probe, facilitating its easy insertion and removal during surgical procedures. This feature may allow for the precise positioning of the fiber probe within the surgical cavity, potentially improving the accuracy of tumor boundary identification and enhancing surgical outcomes.

[0088] The use of the diagnostic platform for intraoperative tumor boundary assessment may provide several technical advantages. For instance, the real-time, molecule-specific insights provided by the SERS spectra collected by the fiber probe may enhance the precision of tumor resection, potentially improving surgical outcomes and patient prognosis. Furthermore, the easy in-cavity application of the fiber probe facilitated by the integrated compartment may improve the efficiency of surgical procedures, potentially reducing surgical time and improving patient outcomes.

[0089] In conclusion, the diagnostic platform described herein may provide a rapid, non- invasive, and accurate method for detecting and analyzing brain tumors. The platform's ability to analyze a wide range of sample types, including CSF, interstitial tissue fluid, solid organ tumor, and blood derivatives and to analyze blood and tissue signatures and classify SERS spectra based on an updatable proteomics and SERS spectral library may enhance the accuracy of brain tumor detection and analysis. Furthermore, the platform's intraoperative tool, potentially including a removable fiber probe and an integrated compartment for easy in-cavity application, may assist in accurately identifying tumor boundaries during surgical procedures, potentially improving surgical outcomes and patient prognosis. The integration of a portable fiber SERS module, a removable fiber probe, disposable nanoplasmonic cartridges, and an updatable spectral library may enable the platform to provide comprehensive and reliable diagnostics for brain tumors. The development and validation of the platform with patient samples may enhance its reliability and clinical relevance, potentially improving its adoption and use in the field of brain tumor diagnostics.

[0090] Thus, the diagnostic platform described herein may offer several benefits and technical advantages. For instance, the platform's ability to provide a rapid, non-invasive, and accurate method for detecting and analyzing brain tumors may potentially improve patientprognosis and outcomes. The platform's ability to analyze blood and tissue signatures and classify SERS spectra based on an updatable proteomics and SERS spectral library may enhance the accuracy of brain tumor detection and analysis. This could potentially lead to earlier detection and treatment of brain tumors, which may improve patient survival rates.

[0091] Furthermore, the platform's intraoperative tool, including a removable fiber probe and an integrated compartment for easy in-cavity application, may assist in accurately identifying tumor boundaries during surgical procedures. This could potentially improve surgical outcomes by enabling maximum safe resection of tumors and reducing the risk of leaving behind residual tumor tissue.

[0092] The integration of a portable fiber SERS module, a removable fiber probe, disposable nanoplasmonic cartridges, and an updatable spectral library may enable the platform to provide comprehensive and reliable diagnostics for brain tumors. This could potentially streamline the diagnostic process, reducing the time and cost associated with traditional diagnostic methods.

[0093] The development and validation of the platform with patient samples may enhance its reliability and clinical relevance, potentially improving its adoption and use in the field of brain tumor diagnostics. This could potentially lead to more widespread use of the platform in hospitals, clinics, and research laboratories, improving the overall standard of care for patients with brain tumors.

[0094] According to some aspects, a liquid biopsy technique is provided to assess whether there is a need for an expensive MRI or biopsy. Biomarkers may be identified that can cross blood brain barrier and these can be detected using extracellular vesicles from biofluids.

[0095] Using an analytical technique such as Raman spectroscopy that can detect biomarkers and give molecular fingerprint information about the given sample.

[0096] According to some aspects, in order to assess the exact tumor boundaries, and interoperative solid biopsy may be performed. This can involve obtaining molecule level information for accurate assessment of tumor boundaries, utilize Photonics Crystal Fiber Raman probe integrated CUSA, and involve deep tissue signal collection with wavefront shaping and spatially offset Raman spectroscopy.EXAMPLE I- A liquid biopsy approach: Al-based identification of brain tumor extracellular vesicles via their SERS signatures

[0097] Liquid biopsy has emerged as a promising non-invasive approach for tumor detection, offering significant potential to transform current diagnostic and monitoring practices for various cancers, including brain tumors. Traditional diagnostic methods, such as tissue biopsies, histopathology, and advanced imaging techniques, are often invasive, costly, and complex, necessitating specialized laboratory infrastructure and expertise. Although early cancer biomarkers have been explored extensively, clinically validated and readily available liquid biopsy techniques remain limited for most cancers, particularly GB.

[0098] There is an unmet clinical need to detect GB rapidly, via easily accessible and an accurate diagnostic tool. The current diagnosis of GB relies on the histopathological analysis of the biopsy / su rgica I ly removed samples in addition to the costly and time-consuming molecular genetic tests, which are not available in every hospital in the developing regions. This method requires highly invasive surgical procedures with significant morbidity and mortality risk to the patient. Having a minimal invasive GB diagnostic tool that can play a role as a frontline precision such as the one that is proposed in this study would be a game changer in terms of eliminating unnecessary invasive procedures including biopsies and surgeries if not necessary.

[0099] Glioblastoma (GB), the most aggressive and lethal type of brain tumors, remains a significant challenge in the field of neuro-oncology due to its rapid progression, infiltrative nature, and poor prognosis. The current diagnostic techniques often fall short in providing early and accurate identification, leading to delayed interventions and compromised patient outcomes. Traditional methods such as biopsy and surgical resection are time-consuming, costly, and invasive, so innovative approaches to revolutionize GB diagnosis are required. This study aims to explore the potential of a liquid biopsy technique in revolutionizing GB diagnosis and therapy management, addressing the urgent need for more effective and less invasive diagnostic tools in the field of neuro-oncology. Herein, we utilized the power of extracellular vesicles and artificial intelligence, to enable the rapid, minimal invasive, and high accuracy detection of GB through surface-enhanced Raman spectra analysis of blood extracellular vesicles.

[0100] GB accounts for the majority of primary brain tumors, presenting a dire healthcare challenge with a median survival of just over a year. The aggressive nature of GB stems from its ability to infiltrate adjacent brain tissues, making complete surgical resection near impossible. Despite advancements in treatment options, including surgery, chemotherapy, and radiation therapy, the prognosis remains grim due to the tumor's intrinsic heterogeneity and resistance to treatment. This underscores the urgency to develop novel diagnostic strategies that allow for early cancer detection, therefore enhancing survival rates and improving patients' quality of life.

[0101] Extracellular vesicles, or small extracellular vesicles secreted by cells, have emerged as promising candidates for non-invasive disease detection owing to their ability to carry a cargo of bioactive molecules reflective of their parent cells' physiological and pathological states. Extracellular vesicles are present in various bodily fluids, including blood, making them accessible for biomarker discovery. Recent research has shown their potential as carriers of information about the tumor microenvironment and molecular changes associated with diseases, including cancer.

[0102] Raman spectroscopy exploits the scattering of light to provide information about molecular vibrations and chemical compositions of materials. Surface-enhanced Raman spectroscopy (SERS) amplifies the Raman signal by orders of magnitude through the use of gold nanoparticles. This technique has shown promise in the analysis of biological samples, including extracellular vesicles, due to its ability to provide specific molecular-level signatures.

[0103] In this study, a liquid biopsy platform integrating SERS and Al is presented to accurately detect and classify EVs isolated from patient plasma. This approach is minimally invasive, rapid, and label-free, addressing critical unmet clinical needs by enabling early, cost- effective tumor detection and real-time monitoring of therapeutic responses. Importantly, this method circumvents the limitations associated with traditional diagnostics, including invasiveness and delayed result turnaround, thereby significantly reducing patient burden and healthcare-associated costs.

[0104] This example also seeks to investigate the potential of extracellular vesicles as disease indicators, focusing on GB. By analyzing surface-enhanced Raman spectra of isolated extracellular vesicles, the hope is to uncover disease-specific molecular signatures thatdistinguish GB-derived extracellular vesicles from healthy and non-glial tumor-derived extracellular vesicles like those from meningioma (MNG). Leveraging the power of neural networks and machine learning, subsets of artificial intelligence algorithms inspired by the human brain's functioning, we intend to develop a predictive model capable of accurately identifying GB based on Raman spectra patterns.

[0105] The integration of extracellular vesicles, surface-enhanced Raman spectroscopy (SERS), and artificial intelligence has the potential to improve GB diagnosis. By harnessing the unique characteristics of extracellular vesicles and the analytical power of SERS, this example aims to contribute to the development of a non-invasive, reliable, and early diagnostic tool for GB. Such an innovation has the potential to significantly impact patient care for the better, enabling timely interventions and improved overall outcomes for individuals afflicted by this disease.

[0106] A clinically applicable liquid biopsy technique that can distinguish different brain tumors and their inflammatory effects would reduce (i) the unnecessary imaging, (ii) invasive and risky surgical procedures including biopsy and / or craniotomy (surgical removal of the tumor), and (iii) imaging and operation costs, (iv) physiological and functional burdens that these operations may cause, (v) psychological effects of hospitalization on patients, (vi) number of malpractices and their sanctions to the hospital and the physicians, as well as damages to the patients. Besides this technique may benefit to (i) identify pseudo-progression, radiation necrosis, and tumor recurrence, (ii) monitor the inflammatory effects of benign or malign brain tumors in the patient metabolism, (iii) support neuropathologists and pathologists at the centers that are lack of expert neuropathologists to make these distinctions, (iv) support the diagnostic predictions during frozen sessions in the surgery to more accurately direct the operation. Therefore, the most frequent two type of benign and malign tumors are used to provide proof of concept to distinguish different type of brain tumors via their exosomal biomarkers.Methods - Sample Collection and Sample Inclusion Criteria

[0107] Blood samples were collected under the approval of Kog University InstitutionalReview Board (#2019.328. IRB2.106) with oral and written consent of the patient and / or legalguardians. Blood samples from glioblastoma (GB) and meningioma (MNG) patients were collected to EDTA tubes prior to tumor resection. Recurrent cases and / or cases with radiotherapy were not included to this study. Diagnoses were performed according to WHO 2021 CNS Tumor classification guidelines and standard operating procedures by the Kog University Department of Pathology through the evaluation of tumor samples obtained from the patients. Blood samples from healthy individuals were collected from volunteers without any known tumor or other malignancies.Methods - Blood Processing and Extracellular Vesicle Isolation

[0108] Blood samples were centrifuged at 2000 rpm and 4°C for 20 minutes, after which the upper part of supernatant (plasma) was collected without contacting the buffy coat. Plasma is aliquoted and kept at -80°C for further use. Before EV isolation, plasma samples were thawed extracellular vesicle. A volume of 500 pl of plasma was used for EV isolation using the Exosome Total Isolation Chip (ExoTIC) as a well-established tool for various applications, sample types and disease applications. The ExoTIC membranes were incubated in PBS, collected from the ExoTIC collection chamber, at 4°C overnight to facilitate EV release. ExoTIC system isolated EV and particles that range from 30-220nm from step b of FIG. 1. EVs suspended in phosphate- buffered saline (PBS, Cytiva) were collected the following day from the ExoTIC membranes and stored at -80°C until further use.Methods - Extracellular Vesicle Characterization

[0109] Extracellular vesicles were characterized with Transmission Electron Microscopy (TEM), Cryo- Electron Microscopy (Cryo-EM), Interferometric Scattering Microscopy (iSCAT), Nanoparticle Tracking Analysis (NTA), Bead-Captured Flow Cytometry (BC-FC), and Western Blotting (WB).TEM Analysis of Extracellular Vesicles

[0110] Extracellular vesicles were diluted to the optimal concentration (lOx-lOOx) and subjected to negative staining. A 10 pl aliquot of the EV suspension was pipetted onto a glow- discharged grid (Electron Microscopy Sciences - EM Sciences FCF-300-CU) and allowed to settle for 3 minutes. Subsequently, three drops of 1% uranyl acetate were applied sequentially, with each drop left on the grid for 1 minute. The excess amount of uranyl acetate was removed fromthe grid with filter paper. To remove residual uranyl acetate, three drops of water were applied to the grid, and the excess water was similarly blotted with filter paper. The grid is left to dry for approximately 10 minutes before imaging.

[0111] Extracellular vesicle imaging was performed at lOOkV using a JEOL JEM 1400 120 kV TEM microscope (JEOL USA, Inc.), equipped with a Gatan UltraScan digital high-resolution camera (Pleasanton, CA) providing the TEM images in the left column of FIG. 2A.Cryo-TEM Analysis of Extracellular Vesicles

[0112] High-resolution imaging of isolated extracellular vesicle was done using cyro- electron microscopy (Cryo-EM). Quantifoil Holey Carbon 1.2 / 1.3 Cu 200 mesh grids were glow- discharged using PELCO easiGlow™ (TED PELLA Inc.) for 45 seconds at a current of 15 mA current. A 3 pL aliquot of the extracellular vesicle sample was applied to the glow-discharged grids and blotted using a Vitrobot Mark IV (FEI) for 3 seconds at a blot force of 3, with a 10- second wait time, at a temperature of 22°C and 100% humidity. The grids were then plunge- frozen in liquid ethane without drain time, clipped, and loaded into a ThermoFisher Scientific Glacios™ Transmission Electron Microscope (Cryo-TEM) operating at 200 kV.

[0113] The specimens were imaged at 36,000x magnification with a pixel size of 1.15 A, using a K3 direct electron camera (Gatan) and SerialEM software (University of Colorado, Boulder). The total accumulated electron dose did not exceed 60 e“ / A2, and the defocus range was set between -1.5 and -3 microns, with an exposure time of 3.995 seconds producing the Cyro-EM images of the right column of FIG. 2A.NTA Analysis of Extracellular Vesicles

[0114] Extracellular vesicle size and concentration distribution were evaluated using nanoparticle tracking analysis (NTA) on a Nanosight NS300 (Malvern Pa na lytica I, UK) equipped with NTA software version 4.2. A 532 nm green laser (50 mW) was used, with the camera level set between 8 and 10 and a detection threshold of 5. Extracellular vesicle samples were diluted to an optimal concentration (lOOx-lOOOx) to achieve a particle count of 20-100 particles per frame. NTA analyses were then performed. Those NTA results are provided in FIG. 2B in the left side column.Interferometric Scattering Microscope Analysis of Extracellular Vesicles

[0115] A custom interferometric scattering microscope specifically tailored for the quality control of isolated extracellular vesicles was developed. The microscope operates on a thin film substrate composed of SiC>2 / Si to capture high-resolution interferometric images. Illumination is provided by an LED (Thorlabs M53OL4, 530 nm) focused on the back focal plane of the objective lens (Nikon xyz) to ensure homogenous illumination. Scattered light from the extracellular vesicles, combined with the reflected light from the substrate, is captured by the objective lens and transmitted to a camera to generate interferometric images. A motorized stage (Standa) is employed to facilitate fine focus adjustments.

[0116] For imaging, extracellular vesicle samples were deposited onto a substrate consisting of a 100 nm SiO? thin film layer atop a Si base and allowed to dry for 15 minutes. iSCAT images of the extracellular vesicles were acquired over a large field of view (200 pm x 200 pm), enabling sensitivity to individual particles. Each particle within the field of view was detected using a custom MATLAB script, and the size distribution of the particles was calculated, as shown in the center column of FIG. 2B.Western Blot Analysis of EVs

[0117] Western Blot analysis of EVs was conducted following a previously described protocol. The analysis used Anti-Calnexin (#PAl-30197, Invitrogen, Carlsbad, CA, United States), Anti-TSGlOl (NB200-112, Novus Bio, Centennial, CO, USA), Flotillinl (#A6220, Abclonal, Woburn, MA, USA), and Anti-CD63 ((#556019, BD Pharmingen, San Jose, CA, USA) antibodies in accordance with the MISEV2023 guidelines. The Western Blot results are provided in FIG. 2C.Flow Cytometry Analysis of Extracellular Vesicles

[0118] EVs were incubated with aldehyde sulfate latex beads (ThermoFisher) for 15 minutes at room temperature (RT) on a roller. The extracellular vesicle-bead complexes were then blocked using 2% BSA for 2 hours, followed by a 30-minute incubation with 100 mM glycine to reduce nonspecific binding. The complexes were washed with cold PBS to remove unbound EVs and beads from the solution.

[0119] The EV-bead complexes were incubated overnight with primary antibodies, including anti-Calnexin (#PA1-3O197, Invitrogen), anti-CD81 (#SAB4700232, Sigma-Aldrich), and anti-HSPA8 (#NB100-41377, Novus Biologicals). The following day, the samples were centrifuged and washed with cold PBS to remove unbound antibodies. Secondary antibodies (#A28175, #A- 10931, and #31860, Invitrogen) were then added, and the samples were incubated for 2 hours. After centrifugation and washing with cold PBS, the samples were analyzed by flow cytometry using a Guava easyCyte flow cytometer, and data were processed with FlowJo vlO software.

[0120] Control samples, including only EVs, only beads, and beads with EVs, were used to identify the purest extracellular vesicle population, as recommended by MISEV2023 guidelines. To isolate the extracellular vesicle population, Calnexin-negative (Calnexin-) EVs were gated first. A multi-gating strategy was then applied to identify CD81-positive (CD81+) and HSPA8- positive (HSPA8+) populations, as well as double-positive and double-negative populations for these markers. These flow cytometry results are illustrated in FIG. 2D.Methods - Surface-Enhanced Raman Spectroscopy Analysis of Extracellular Vesicles Experimental Setup

[0121] A customized Raman spectroscopy system was developed to provide enhanced flexibility for multi-modal and multi-functional analysis while maintaining sensitivity comparable to state-of-the-art instruments. A diode laser (CrystaLaser, 785 nm, maximum output power: 130 mW) was used, with the output power adjusted via a half-wave plate and a linear polarizer to ensure optimal performance without damaging the sensitive gold (Au) substrate. See for example, FIG. 6.

[0122] The adjusted laser beam was directed to the sample plane of the microscope through a long-pass dichroic mirror (Thorlabs, DMLP 805) and a microscope objective (lOx, NA 0.22). A motorized XYZ stage (Standa) was used to facilitate lateral scanning of the sample surface and precise focal plane adjustments.

[0123] The scattered photons were collected using the same objective lens in a 180-degree backscattering geometry. Elastic (Rayleigh) scattering was filtered out by the dichroic mirror, while inelastic (Stokes) scattered photons were directed to a multimode fiber via an ultra-steep long-pass filter (Semrock) and a fiber collimator (Thorlabs, F220SMA-780). The collected signal was then transmitted to a Raman spectrometer (StellarNet, HyperNova) equipped with a CCD camera (Andor iVAC 316), which was cooled to -60°C to minimize noise and enhance sensitivity.Sample Preparation

[0124] For the SERS measurements, hydrophobic Au-coated surfaces (Silmeco ApS, Copenhagen, Denmark) were utilized to optimize signal enhancement. A 2 pL aliquot of extracellular vesicle sample was deposited onto the Au surface and allowed to dry for 15 minutes. Each SERS surface was used only once to ensure consistent enhancement performance.Data Collection

[0125] A connection between the PC and the CCD spectrometer was established using the Andor MATLAB R2024A SDK. A custom MATLAB script was developed to automate surface scanning with predefined parameters. The laser output power was fine-tuned to 30 mW (approximately 20 mW at the sample plane). Real-time spectral feedback was employed to identify the optimal focal plane for data acquisition.

[0126] Once the focus was optimized, spectral acquisition was performed with the following parameters: an exposure time of 1 second, a scanning step size of 2.5 pm, and 21 steps along each axis (X and Y). This configuration yielded a total of 441 spectra per sample, ensuring robust and comprehensive spectral data for analysis.Employment of Artificial Intelligence Methods

[0127] A total of 32,193 spectra were collected from three classes: GB, MNG, and HC. The dataset was pre-processed using a series of steps implemented with the RamanSPy library in Python. First, the regions outside the fingerprint region were cropped from each raw spectrum. Cosmic rays were removed using the Whitaker-Hayes algorithm, and spectra were smoothed using the Savitzky-Golay algorithm. Fluorescence background was leveled and removed using the Improved Asymmetric Least Squares (IALS) baseline correction algorithm. Finally, each spectrum was normalized to its intensity, facilitating easier comparison of Raman profile shifts. The results are depicted in FIG. 3A.

[0128] After preprocessing, the data distribution was analyzed using t-Distributed Stochastic Neighbor Embedding (t-SNE), a non-linear dimensionality reduction technique. This method embeds high-dimensional data into two- or three-dimensional space (shown in FIG. 3C), allowing visualization of class distribution and potential similarities. Similar spectra formedclusters in the reduced-dimensional space, while dissimilar spectra were more dispersed, providing insights into class-specific patterns.

[0129] For patient state prediction, the performance of five pre-defined models were compared: Convolutional Neural Network (CNN) [Lecun, Y., Bottou, L., Bengio, Y. & Haffner, P. Gradient-based learning applied to document recognition. Proc. IEEE 86, 2278-2324 (1998)], Neural Network (NN) [Bishop, C. M. Neural Networks for Pattern Recognition. (Oxford University Press, Oxford, New York, 1996)], Random Forest Classifier [Breiman, L. Random Forests. Mach. Learn. 45, 5-32 (2001)], Support Vector Classifier (SVC) [Cortes, C. & Vapnik, V. Support-vector networks. Mach. Learn. 20, 273-297 (1995)], and Quadratic Discriminant Analysis (QDA) [Hastie, T., Tibshirani, R. & Friedman, J. The Elements of Statistical Learning. (Springer, New York, NY, 2009). doi:10.1007 / 978-0-387-84858-7], These models were evaluated to determine their accuracy and reliability in predicting patient states.Statistical Analysis

[0130] A Kolmogorov-Smirnov test was conducted on selected intense Raman peaks at 456, 587, 715, 761, 919, 1069, 1219, 1254, 1304, and 1422 cm ~1, revealing a non-normal distribution of the data. Consequently, non-parametric statistical methods were applied. To evaluate differences among the three groups in our study (GB, MNG, and HC), a Kruskal-Wallis test was performed on the selected peaks. The analysis showed significant differences in Raman intensity distributions among these groups (p < 0.001). The distributions of spectral data for the selected peaks are visualized using violin plots shown in FIGS. 3D, 3E, and 3F.Regression analysis

[0131] To assess the presence of amino acids in EGFR protein samples, the most and least frequent amino acids were first identified. A dataset was then constructed, incorporating Raman peak intensities from different concentrations of EGFR (#10001, Sino Biological Inc.) and EGFRvlll (#29662, Sino Biological Inc.) preparations and their constituent concentrations (v:v ratios). This dataset was used to develop a linear regression model. The data was split into training (80%) and testing (20%) sets, and k-fold cross-validation (k=5) was employed to validate the model.

[0132] Using this model, the presence of selected amino acids in the SERS spectra of extracellular vesicles (EVs) was validated by analyzing their most dominant Raman peaks as reported in the literature. The performance of the model was assessed by calculating the root mean square error (RMSE) values for the prediction.LC / MS analysis of plasma EVs

[0133] EVs were isolated using 50 mM ammonium bicarbonate for proteomics analysis. EV proteins were exposed by disrupting the lipid bilayer with 1.5% sodium dodecyl sulfate (SDS) before further processing. All analyses were performed with three technical replicates.

[0134] Four micrograms of tryptic peptides were loaded onto an Acclaim PepMap C18 trap column (Thermo Fisher Scientific) coupled to a Dionex Ultimate Rapid Separation Liquid Chromatography HPLC system (Thermo Fisher Scientific) at a flow rate of 5 pL / min for10 minutes. Separation of tryptic peptides was achieved using reversed-phase chromatography on a 25 cm-long C18 analytical column (New Objective) packed in-house with ReproSil-Pur 120 C18 AQ resin (Dr. Maisch GmbH). Peptides were eluted and ionized via a Nanospray Flex ion source (Thermo Fisher Scientific) with a 1.8 kV voltage and analyzed using an LTQ-Orbitrap Elite mass spectrometer (Thermo Fisher Scientific).

[0135] The chromatography gradient was programmed as follows: a flow rate of 0.4 pL / min was maintained throughout. Mobile phase A (0.1% formic acid in water) and mobile phase B (0.1% formic acid in acetonitrile) were used, starting at 98% A and 2% B for 10 minutes. This was followed by a gradual increase to 35% B over 100 minutes, then to 85% B over 2 minutes, with a 7-minute hold. Re-equilibration of the analytical column was performed before each subsequent injection. Each sample was analyzed in triplicate.

[0136] The top 10 most abundant ions from each MSI scan were selected for higher-energy collision-induced dissociation (HCD) at 35 eV in a data-dependent acquisition mode. MSI scans were performed with a resolution of 60,000, an FT AGC target of le6, and an m / z scan range of 400-1800. MS2 scans were acquired with an AGC target of 3e4, and dynamic exclusion was enabled for 30 seconds.Proteomics data analysis

[0137] Raw data files from each LC-MS run were analyzed using Byonic 4.0.12 (Protein Metrics, San Carlos, CA). Data were searched against the Swiss-Prot database, incorporating the 2024 reference human proteome with 20,692 protein entries. Search parameters specified trypsin as the digestion enzyme, allowing up to two missed cleavages. A precursor mass tolerance of 0.5 Da and a fragment mass tolerance of 10 ppm were applied. Fixed modifications included carbamidomethylation of cysteine, while variable modifications accounted for methionine oxidation and asparagine deamination.

[0138] To ensure high-confidence peptide identification, spectra with a false discovery rate (FDR) greater than 1% were excluded. All experimental conditions were performed in triplicate, with three technical replicates per condition. Protein quantification was based on the specific signal intensity of each protein, normalized to the average signal across all samples, enabling relative abundance determination within the dataset.

[0139] Normalization procedures were applied to adjust relative abundance values to a standard normal distribution with a mean of 0 and a standard deviation of 1, ensuring comparability across samples. Supervised clustering was then conducted using Pearson correlation across all biological conditions, including permutations. Only clusters showing significant correlations (p < 0.01) were retained for further analysis.

[0140] Significantly enriched clusters were analyzed using overrepresentation analysis via the WebGestalt tool [http: / / www.webgestalt.org / ]. This analysis identified enriched pathways and biological processes, providing insight into the functional implications of the proteomic data.Results - Characterization ofEVs via Transmission Electron Microscopy (TEM), Cryo-Electron Microscopy (Cryo-EM), Interferometric Scattering Microscopy (iSCAT) Analysis, Nanoparticle Tracking Analysis (NTA), Bead-Captured Flow Cytometry (BC-FC), and Western Blotting (WB)

[0141] Transmission Electron Microscopy (TEM), Cryo-Electron Microscopy (Cryo-EM), Interferometric Scattering Microscopy (iSCAT) Analysis, and Nanoparticle Tracking Analysis (NTA) were performed as physical characterization, and Bead-Captured Flow Cytometry (BCFC) and Western Blotting Analyses were performed as biological characterization of extracellular vesicles derived from blood samples from glioblastoma and meningioma patients, and healthyindivid ua Is. As shown in FIGS. 2A and 2B, the TEM, Cryo-EM and iSCAT images demonstrated that the majority of extracellular vesicles are within the diameter range of 120-160 nm, consistent with NTA results. The extracellular vesicles were further characterized with biological characterization methods Bead-Captured Flow Cytometry (BC-FC), and Western Blotting (WB) as described elsewhere herein. The BC-FC analysis resulted in 93% of particles were stained for Calnexin (-), 83.4% of particles were stained for Calnexin (-) and CD81 (+), 80.7% of particles were stained for Calnexin (-) and HSPA8 (+), and 68.9% were stained for Calnexin (-), CD81 (+), and HSPA8 (+), which aligns with the minimum requirements for EV characterization based on MISEV2023 guidelines. With reference being made to FIG. 2C, WB analysis was performed on extracellular vesicles using CD63, TSG101, Flotl, and obtained protein bands 50, 47, and 50 at kDa respectively. Calnexin was used as a cellular protein control and Calnexin bands were not obtained in EVs. These characterization results confirm that isolated EVs meet stringent quality criteria, ensuring that downstream SERS analyses accurately reflect pathological differences rather than methodological variability.Results - Surface-Enhanced Raman Spectroscopy (SERS) Data Analysis of Extracellular Vesicle

[0142] To analyze group-specific variations in Raman spectra of EV measurements, the average SERS spectra of the GB, MNG, and HC groups were first compared, as shown in FIG. 3A and 3B. Subsequently, Principal Component Analysis (PCA) was applied to the spectral dataset to reduce dimensionality and visualize group separation. With reference to FIG. 3C showing the 3D visualization, the 2D and 3D visualizations of PCA score plots demonstrate clear clustering of samples, with the first two principal components explaining 54.8%, 16.7%, and 11% of the variance, respectively.

[0143] Again, with reference being made to FIG. 3C, to further explore the spectral features contributing to group differentiation, the loading plot of the dataset was analyzed for the first three principal components. This analysis identified specific Raman shift regions that significantly contribute to separating the groups. Finally, the intensities at three key Raman peaks (1069 cm'1, 1304 cm'1, and 1422 cm'1) were analyzed using violin plots as illustrated in FIGS. 3D, 3E, and 3F, respectively. These plots provide a detailed comparison of the intensity distributions for each group, showing significant differences between GB, MNG, and HC.Distinct spectral shifts at these Raman peaks correspond to biomolecular differences such as nucleic acids, proteins, and lipid alterations, highlighting molecular heterogeneity between tumor types.Results - Surface-Enhanced Raman Spectroscopy (SERS) Measurement of EVs

[0144] Extracellular vesicles isolated from plasma samples collected from GB patients, MNG patients, and healthy individuals with no known disease were analyzed. The baseline corrected and L2-normalized average SERS spectra of extracellular vesicles from the abovementioned three groups in FIF. 3A. Notably, distinct spectral peaks were evident in extracellular vesicles from healthy individuals, suggesting unique molecular profiles. Conversely, extracellular vesicles from GB and MNG patients exhibited similar spectral trends, indicating potential molecular similarities.Results - SERS Data Analysis of Extracellular Vesicles

[0145] We performed t-SNE on the normalized data shown in FIG. 3C and explored the clustering behavior of the data using the low-dimensional components. We saw a clear distinction between HC group and the others, while GB and MNG have a level of overlap. One has to keep in mind that reducing high dimension vectors like those Raman profiles to such a low dimension space always leads to significant information loss, and thus this might not bother the algorithms in classifying the spectra.Results - SERS spectra analysis and Supervised Al Classification of EV SERS data from GB, MNG and HC

[0146] A total of collected and pre-processed 32,193 spectra from the GB, MNG, and HC classes were introduced to a comprehensive pipeline to build a classification model. Following pre-processing, PCA analysis from FIG. 3C revealed distinct clustering of the HC group, with certain overlapping spectra observed between the GB and MNG groups. This visualization suggested potential challenges in classification of GB and MNG.

[0147] With reference being had to the half of FIG. 4B, two approaches were employed to partition the spectral data into training and testing sets: splitting by individual spectra and splitting by patient. In the former, multiple spectra from the same patient may appear in both the training and testing sets, potentially inflating performance estimates by allowing the modelto learn patient-specific features rather than more generalizable patterns. Conversely, splitting by patient allocates all spectra from a given patient to either the training or the testing set, thereby eliminating any overlap at the patient level. This approach more closely mimics clinical conditions, in which the model must generalize to new, unseen patients, and thus provides a more accurate assessment of real-world performance.

[0148] To predict patient state, six classification models were built and tested: Convolutional Neural Network (CNN), Neural Network (NN), Random Forest (RF), Support Vector Machine (SVM), XGBoost, and Quadratic Discriminant Analysis (QDA). Models were evaluated using three train-test split ratios (80-20, 60-40, 40-60) to assess robustness.

[0149] When trained with an 80-20 split, the CNN model achieved the highest accuracy of 0.88, with specificity of 0.94 (BT vs HC), sensitivity of 0.85 (BT vs HC), and a weighted fl score of 0.87. The RF and SVM models followed with accuracy values of 0.79, while XGBoost, QDA, and NN showed lower accuracies of 0.76, 0.63, and 0.50, respectively. Table 1 below provides the full Al classification results for brain tumor (BT: GB and MNG) vs healthy control (HC) groups for different train-test split ratios. For each of the groups of 80 Train - 20 Test, 60 Train - 40 Test, and 40 Train - 60 Test and for each of the models at that split, the Area Under the Curve, Specificity, Sensitivity, and Weighted fl Score values are presented.TABLE 1

[0150] As the train-test split ratio shifted towards smaller training sets, all models showed a decline in performance. At a 60-40 split, CNN maintained robust performance with an accuracy of 0.73, outperforming other models. The RF and SVM models demonstrated competitiveaccuracy at 0.72 and 0.71, respectively. However, the NN model continued to underperform, with an accuracy of 0.26, highlighting its limited capacity to generalize with reduced training data. In the 40-60 split scenario, CNN remained the top performer with an accuracy of 0.67, while SVM and RF models exhibited comparable accuracies of 0.69 and 0.67, respectively. QDA showed the most significant drop in performance, with an accuracy of 0.38, indicating its sensitivity to smaller training datasets.

[0151] Overall, CNN consistently outperformed other models across all train-test split scenarios, demonstrating its robustness and reliability in predicting patient states using Raman spectral data. The observed model performance trends underscore the challenges of maintaining classification accuracy with reduced training data and highlight the strength of CNN in handling such complexities. The CNN model's high specificity (0.94) and sensitivity (0.85) suggest strong potential for accurate non-invasive brain tumor diagnosis, significantly reducing dependence on invasive biopsy methods.Results - EGFR Quantification in Extracellular Vesicle SERS Spectra Using Regression Analysis

[0152] To develop a quantitative prediction model, the average spectra corresponding to EGFR-wild type (wt), EGFRviii, GB and HC were compared to identify the Raman intensity differences. Then, a Mann-Whitney U test was applied to both datasets to select common differentiating regions, specifically at wavenumbers 1555 cm-1and 1440 cm .

[0153] Next, two linear calibration models were built using two selected wavenumbers (1440 and 1555 cm1), which effectively distinguish both EGFR-wt and EGFRvI 11 among GB, MNG, and HC. Finally, the model was built on the EGFR dataset was utilized to classify samples in terms of the predicted EGFR protein levels within EVs across three groups. Significant differences were found between GB-derived EVs and those from other groups. Quantitative predictions of EGFR status from EVs using SERS provide opportunities for non-invasive monitoring of tumor molecular subtypes, potentially informing personalized therapy selection.Results - Proteomics Analysis of Plasma EVs

[0154] LC-MS / MS analysis of plasma EVs derived from patients with GB (n=17) and MNG (n-20), and HC individuals (n-30) was performed. The same cohort with SERS study was used, except the extra three GB and three MNG samples that were used for later SERS validation.With reference being had to FIGS. 8A-8F, the proteins were investigated that may be related to the relevant pathology and their varying expression levels. Their distribution is illustrated through Venn diagram in FIG. 8A, volcano plot in FIGS. 8B and 8C, and a heatmap in FIG. 8D.

[0155] Proteomic analysis revealed significant differential expressions of proteins across MNG, GB, HC, and Brain Tumor groups. In the GB vs. MNG comparison, Clusterin (CLU) and Apolipoprotein C-lll (APOC3) demonstrated the highest significance (*****), indicating strong upregulation in GB samples. In the MNG group, proteins such as IGLV2-14, BTD, and LUM exhibited significant expressions (p < 0.001). Proteomic analyses revealing upregulated CLU and APOC3 suggest their potential utility as GB-specific prognostic markers, complementary to SERS-based EV profiling, thus enhancing diagnostic accuracy and providing deeper biological insights.

[0156] In this example, using a combination of SERS and neural networks, the potential of distinguishing between extracellular vesicles from GB patient plasma samples, MNG plasma samples, and healthy plasma samples was successfully demonstrated. The rationale behind choosing MNG as the control tumor group was due to the need to establish a proof of concept. By choosing a non-glial, benign tumor such as MNG, this aimed to validate the hypothesis that distinct tumors could be differentiated based on their exosomal molecular signatures using SERS. MNG is an ideal starting point due to its molecular dissimilarity to GB, which is a malignant glial tumor. This allows for a foundational understanding before progressing to differentiations between more closely related glial tumors.

[0157] The decision to focus on MNG as a control group was also influenced by its abundance. As the most common primary benign tumor, MNG is readily available for study, simplifying the process of obtaining samples. This successful discrimination between GB and MNG extracellular vesicles supports the validity of our methodology.

[0158] Moreover, the integration of SERS and Al demonstrated high diagnostic accuracy in distinguishing EV signatures derived from GB, MNG, and healthy controls. Notably, the CNN model achieved robust classification performance, with accuracy, specificity, and sensitivity values consistently superior to other tested models, highlighting the strength and generalizability of deep learning approaches for spectral data analysis. This underscores theimportance of patient-level data partitioning, which realistically simulates clinical conditions and avoids overly optimistic performance estimates that can arise from spectra-level splitting.

[0159] Proteomics analyses further validated the translation potential of our EV-based approach. Significantly expressed proteins in GB, such as CLU48 and APOC349, emerged as potential GB-specific prognostic biomarkers. Elevated CLU expression aligns with known roles in immune evasion, tumor survival, and resistance to therapy, reinforcing its potential diagnostic and prognostic relevance. APOC349, involved in lipid metabolism, was similarly upregulated in GB, suggesting metabolic reprogramming that supports rapid tumor cell proliferation. Conversely, proteins uniquely expressed in MNG samples, including IGLV2-1452, BTD, and Lumican (LUM), indicate differential tumor microenvironment interactions, particularly immune regulation and extracellular matrix remodeling.

[0160] Additionally, this example reveals systemic alterations in proteins such as F13A1 and ACTB, significantly differentiating healthy controls from brain tumor patients. Reduced expression of F13A1, a critical component of the coagulation cascade involved in extracellular matrix stabilization, has previously been linked with poorer survival in GB patients. ACTB is essential for maintaining cytoskeletal integrity and structural stability. Proteins like Retinol- Binding Protein 4 (RBP4) and Zinc-alpha-2-glycoprotein (AZGP1) demonstrated differential expression in brain tumor groups, highlighting their potential roles in metabolic regulation, retinoid signaling pathways, lipid metabolism, angiogenesis, and immune modulation. Elevated RBP4 expression is notably associated with GB proliferation, emphasizing its prognostic relevance. Similarly, AZGP1 has emerged as a relevant biomarker across multiple cancers, involved predominantly in lipid mobilization and immune response regulation. Although its role in GB is not yet extensively documented, the observation of differential AZGP1 expression underscores its potential broader significance within brain tumor pathology.

[0161] The analysis also identified Complement Component 9 (C9) and Leucine-Rich Alpha- 2-Glycoprotein 1 (LRG1), which may contribute to immune response regulation and aberrant vascular remodeling within brain tumors. C9, a crucial component of the complement cascade involved in innate immunity, aligns with emerging evidence linking complement activation to inflammatory modulation and tumor progression. LRG1, known for its role in promoting pathological angiogenesis, could further explain mechanisms underlying abnormalvascularization commonly observed in GB. Similarly, Serum Amyloid Al (SAA1), although demonstrating relatively lower statistical significance, has previously been proposed as a prognostic marker reflecting GB microenvironment dynamics. Collectively, these proteomic findings reinforce the complexity and heterogeneity of brain tumor biology, highlighting both well-established and new candidate prognostic protein signatures that warrant future validation studies.

[0162] These findings also more generally suggest significant translational potential for integrating EV-based SERS analysis with Al in clinical oncology. This method could particularly benefit clinical settings lacking specialized diagnostic infrastructure by offering a scalable, accessible tool for rapid tumor detection and monitoring. Thus, this EV-based liquid biopsy approach combining SERS and Al represents a significant step forward in the development of rapid, accurate, and minimally invasive diagnostic tools for brain tumors. This platform's scalability and adaptability highlight its broader potential for precision diagnostics across multiple diseases.

[0163] Moving forward, this disclosure can be strengthened by increasing the number of patients studied and expanding the spectral library. This will enhance the reliability of the predictive model, providing more accurate results for distinguishing between different tumor types. A subsequent focus may involve distinguishing between glial tumors with varying grades. Given the molecular similarities among these tumors, this represents a more intricate challenge that our methodology and neural network model can be developed to address. Furthermore, exploration into potential applications, such as predicting prognosis or monitoring treatment response, can broaden the utility of our tool clinically. Ultimately, the integration of extracellular vesicles, SERS, and neural networks holds promise in advancing the field of cancer early detection.

[0164] It should be appreciated that various other modifications and variations to the preferred embodiments can be made within the spirit and scope of the invention. Therefore, the invention should not be limited to the described embodiments. To ascertain the full scope of the invention, the following claims should be referenced.

Claims

CLAIMSWhat is claimed is:

1. A diagnostic platform for detecting and analyzing brain tumors, the platform comprising: a portable fiber surface enhanced Raman spectroscopy (SERS) module; a removable fiber probe for intraoperative use; disposable nanoplasmonic cartridges for individual or multiple sample loading; and an updatable spectral library collected from healthy and diseased samples, wherein the platform is configured to analyze blood and tissue signatures and employ artificial intelligence (Al) models to classify SERS spectra based on an updatable proteomics and SERS spectral library encompassing distinct brain tumor types.

2. The diagnostic platform of claim 1, wherein the portable fiber surface enhanced Raman spectroscopy (SERS) module is configured to amplify Raman scattering signals from molecules adsorbed on or in close proximity to engineered surfaces.

3. The diagnostic platform of claim 1, wherein the removable fiber probe is integrated into a Cavitron Ultrasonic Surgical Aspirator (CUSA) instrument with an embedded camera for intraoperative utilization.

4. The diagnostic platform of claim 1, wherein the disposable nanoplasmonic cartridges are designed for the isolation and deposition of extracellular vesicles (EVs) from patient biofluids and tissues.

5. The diagnostic platform of claim 1, wherein the updatable spectral library comprises Raman spectral signatures of diseased tissue, blood derivatives, or biofluids, and / or EVs isolated from patient biofluids and tissues.

6. The diagnostic platform of claim 1, wherein the platform is further configured to utilize artificial intelligence (Al) algorithms for the analysis and classification of the recorded SERS spectra.

7. The diagnostic platform of claim 1, wherein the platform is further configured to update the spectral library with new spectral data collected from additional healthy and diseased samples.

8. A method for early cancer detection, the method comprising: performing surface enhanced Raman spectroscopy on a sample containing extracellular vesicles to obtain a Raman spectra for the sample; analyzing the Raman spectra for the sample against a proteomics and SERS spectral library and providing a classification for the sample.

9. The method of claim 8, further comprising the step of obtaining the sample containing extracellular vesicles via liquid biopsy.

10. The method of claim 8, wherein analyzing the Raman spectra for the sample against the proteomics and SERS spectral library and providing the classification for the sample involves employing artificial intelligence models to perform the classification.

11. The method of claim 8, wherein providing a classification for the sample encompasses providing a distinct brain tumor type based on comparison of the Raman spectra for the sample against the proteomics and SERS spectral library.

12. The method of claim 8, wherein the proteomics and SERS spectral library is updatable.

13. The method of claim 12, further comprising adding the Raman spectra for the sample to the proteomics and SERS spectral library.

14. The method of claim 13, wherein adding the Raman spectra for the sample to the proteomics and SERS spectral library involves providing a classification for the sample which information is used to further develop an artificial intelligence model to perform classification.

15. The method of claim 8, wherein providing a classification for the sample encompasses providing a classification relating to a presence of cancer based on comparison of the Raman spectra for the sample against the proteomics and SERS spectral library.

16. The method of claim 8, wherein the classification relating to a presence of cancer includes information relating to a type of cancer.

17. The method of claim 8, further comprising the step of isolating extracelluar vesicles from a liquid biopsy sample before the step of performing surface enhanced Raman spectroscopy on a sample containing extracellular vesicles.

18. A nanoplasmonic cartridge for processing a liquid biopsy sample, the nanoplasmonic cartridge capable of collecting and isolating extracelluar vesicles from the liquid biopsy sample for surface enhanced Raman spectroscopy (SERS) spectral collection.

Images