Ai-assisted label-free optical platform to characterize NANO and micro-vesicles and biological tissues

Abstract

A platform for characterizing biomolecules, biological tissue sections, viruses, or other particles of interest in a non-invasive, high-throughput, cost-effective, and label-free manner. The platform includes a merged set of imaging modalities including and capable of interferometric microscopy and wide-field or confocal microscopy (such as, for example, a wide-field super-resolution surface enhanced Raman spectroscopy, SERS), although some aspects may be employed with data sets only received from wide-field or confocal microscopy. Related devices, systems, methods, techniques, and variations to the same are also provided.

Description

AI-ASSISTED LABEL-FREE OPTICAL PLATFORMTO CHARACTERIZE NANO AND MICRO-VESICLES AND BIOLOGICAL TISSUESCROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 643,119 entitled "Al-Assisted Label-Free Optical Platform to Characterize Nano and Micro-Vesicles and Biological Tissues" 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] This disclosure relates to compositions, devices, processes, methods, and systems designed for the rapid and accurate optical fingerprinting, identification, and sequencing of nano and micro-vesicles and their contents. Additionally, this disclosure relates to an interferometric imaging approach to localize extracellular vesicles (EVs) and wide-field superresolution Raman microscopy to determine the 3D structure of the genetic biomarkers, thereby enhancing the capabilities of the optical device in molecular diagnostics and genetic research.BACKGROUND

[0004] Recently, extracellular vesicles have been the focus of many diagnostics and theragnostic studies since their biogenesis was discovered as they are cel l-to-cell communication agents. They carry cellular materials like small extracellular RNAs, namely exRNAs, cytokines, heat shock, and transmembrane proteins. The composition of the cellular material is diverse depending on the size, origin, and target of the secreted EVs. Their size distribution is 30-150 nm, below the optical diffraction limit. Thus, the optical localization of these biological particles is challenging or usually needs sophisticated setups like superresolution microscopes. A simple and accurate EV localization method is lacking.SUMMARY

[0005] Challenges remain in biomedical research and pharmaceutical development in characterizing biomolecules like proteins and nanovesicles that are crucial for disease diagnosis, treatment monitoring, and drug delivery. In particular, methods are needed to efficiently quality control and characterize nanovesicle cargo in a non-invasive, cost-effective, high-throughput, and label-free manner. As determining the EV cargo is a crucial task, several optical methods have been proposed to differentiate EVs from different origins. However, the exRNA profiling of the EVs is essential to find novel markers of the diseases, especially by miRNA mutations. There are several efforts to classify miRNAs from each other optically. However, no optical method simultaneously has reached the single nucleotide sensitivity and molecular specificity.

[0006] As such, for nano- and micro-vesicle content, there is an increasing need and demand to rapidly interrogate their nucleic acid and protein structure without lysing them, label-free, and with a fully optical approach.

[0007] Additionally, ensuring the quality and purity of protein samples is important in successful biochemical and structural studies, particularly in academic research and pharmaceutical production. With the increasing prevalence of biologic drugs, maintaining protein purity is paramount for the efficacy and safety of therapeutic interventions.

[0008] Still further, there is still an unmet need for a faster, more accurate and easier analysis methods to predict the clinical status of the tissue samples. Histopathological imaging methods, particularly Hematoxylin and Eosin (H&E) staining, have been pivotal in the diagnosis and characterization of various diseases, providing invaluable insights into tissue morphology and pathology. However, despite their widespread use and undeniable utility, these methods suffer from subjectivity, reproducibility and image quality. To overcome this, an additional analysis like sequencing is performed to prove or elaborate the H&E outcomes. These techniques usually take weeks to give outputs, and the kits used in these analyses are expensive.

[0009] This disclosure provides optical systems, methods and algorithms for accurately, rapidly and straightforwardly analyzing vesicles and biological tissue sections with high sensitivity and molecular specificity. The novel method uses interferometric microscopy to localize single EVs and wide-field super-resolution surface enhanced Raman spectroscopy to determine the types and contents of different vesicles without any lysing operation. It is noted that the platform can operate without an interferometric imaging module; however, incorporating interferometric imaging provides unique advantages, which are elaborated further below. The present disclosure introduces novel methods and algorithms designed toanalyze tissue sections for metal content, employing standardized H&E color mapping for image enhancement, and leveraging interferometric imaging to expedite the scanning process. Additionally, this disclosure integrates wide field imaging techniques and super-resolution methods synergistically to enhance scanning efficiency without compromising spatial resolution. Moreover, this disclosure introduces a comprehensive analysis for quantifying protein and nucleic acid content within vesicles, showcasing its applicability in distinguishing nanovesicles associated with pancreatic cancer, drug toxicity, environmental hazards, diabetes, and Alzheimer's disease.

[0010] Imaging nanosized biological particles, such as EVs, without labels poses a significant challenge due to their sizes falling below the diffraction limit of light. The disclosed localization method is based on interferometry, a precise technique that can measure minute phase differences between light waves. In interferometric imaging, the interference between the scattering signal scattered by the nanosized particle and a known reference light can be leveraged. The total intensity ( / tot) is calculated as \ E5ca+ Eref\2= | Esca|2+ | Eref\2+ 2-ErefESCa-cos(0), where Escaand Eref represent the fields generated by the scattering from the particle and the reference field, respectively, and 0 is the phase difference between these two fields.

[0011] An on-axis interferometric microscopy configuration is employed, where the reference light is collected directly from the sample substrate. Here, a glass slide can be used simply as a sample substrate, or alternatively a specially designed, cost-effective thin film can be utilized that is structurally tuned to acquire a high interferometric signal. Given that the scattering field from EVs is small, and the background light is a static term, the contrast resulting from a single particle is fundamentally governed by the ratio ESca / Eref cos(p). To enhance sensitivity and enable the mapping of single EVs, defocused images are employed with tuned phase differences.

[0012] The optical system can be combined with a wide-field Raman spectroscopy module. This module can be able to illuminate a large area of the sample (wide-field) and create fingerprint images of the EV cargo materials within 100 milliseconds. The spectral resolution can be increased computationally by evaluating the surface-enhanced Raman spectroscopy (SERS) images using an open-source STORM (Stochastic Optical Reconstruction Microscopy)fitting algorithm. Thus, the intensities of the close peaks coming from different nucleotides in the miRNA content can be analyzed.

[0013] The spectral detection module of this novel method utilizes a silver-coated substrate that provides SERS hotspots, where the spectral enhancement factor dramatically increases and allows ultrasensitive detection up to a single molecule level. To predict the genomic structure of the localized intact EVs, it is proposed to create a library of individual nucleotide bases and sequences of nucleotides in different lengths. To obtain a more realistic Raman library of nucleotides, the library with nucleotide sequences can be created in two forms: i) free space, ii) encapsulated in the liposomes with 100 nm diameter.

[0014] The technology described above can be equipped with a comprehensive computational package that includes image acquisition, image analysis and neural network algorithms. Using the above-mentioned library that combines images and spectra from nucleotide bases inside liposomes, a training neural network model may be created to accurately identify contents of intact EVs. This novel approach uses the power of data augmentation by variational auto encoder (VAE) to extend the number of spectra in the dataset while still keeping the form of the spectra. Further, another neural network model can learn the interferometric images of the intact liposomes with known cargo. Prediction of the content structure of the individual intact EVs can be performed by weighting the prediction scores from the image-based and spectral models.

[0015] Thus, in some aspects, the present disclosure provides a comprehensive platform for the analysis of extracellular vesicles (EVs) and similar nanosized biological particles by integrating interferometric imaging and Raman spectroscopy within a single workflow. The system first employs wide-field interferometric imaging to localize individual vesicles across the sample substrate. Through this imaging step, the spatial position of each vesicle is mapped, allowing targeted subsequent spectroscopic interrogation.

[0016] In addition to simple localization, the interferometric imaging can provide quantitative physical characterization of each detected vesicle. Specifically, the contrast observed in the interferometric signal, which arises from the phase and amplitude differences between the scattered and reference light, correlates with the size and refractive index of the vesicle. By analyzing the interferometric contrast and phase shift, the system is capable ofestimating the size and volume of each individual vesicle without the need for labeling or destructive processing. This information can be used for scaling the acquired Raman spectrum.

[0017] Note that it is contemplated interferometric imaging may be omitted from the platform in some versions and the platform may still remain operable. So while interferometric imaging provides some unique advantages, it is also contemplated that the wide field or Raman spectroscopy aspects described below and herein could be employed as separate stand-alone improvements over the state of the art.

[0018] Following physical localization and sizing, Raman spectroscopy is selectively performed at the mapped vesicle positions. When mapping has occurred first, this targeted spectroscopic acquisition ensures that the collected Raman signals correspond specifically to intact structures, thereby significantly improving molecular specificity and eliminating contamination from free-floating cargo, background material, or lysed debris.

[0019] The integration of vesicle localization, size and volume estimation, particle counting, and targeted molecular fingerprinting results in substantially improved classification accuracy compared to conventional random-sampling approaches. The additional physical parameters— size, particle number, and spatial distribution— serve as auxiliary features for machine learning classification models, thereby enhancing the robustness, sensitivity, and specificity of disease diagnostics based on EV signatures.

[0020] This platform also offers detection and characterization of sample heterogeneity. Biological samples, particularly clinical specimens such as plasma, serum, or cyst fluid, inherently contain a heterogeneous mixture of extracellular vesicles (EVs) varying in size, molecular content, and cellular origin. Cancer-derived EVs coexist with vesicles secreted from normal, healthy cells, and their biochemical signatures often overlap with background molecules released during tissue turnover or inflammation. Furthermore, the physical dimensions of vesicles in these samples can range from approximately 30 nm to over 1000 nm, encompassing diverse subpopulations such as exosomes, microvesicles, and apoptotic bodies.

[0021] The present disclosure can address this heterogeneity through the combined use of interferometric imaging and targeted Raman spectroscopy. By imaging the vesicles prior to spectral interrogation, the platform captures physical heterogeneity, enabling the measurement of size distributions, vesicle concentration, and spatial clustering patterns. Theability to selectively acquire Raman spectra from intact, physically characterized vesicles allows molecular heterogeneity to be resolved at the single-particle level. This combination of physical and molecular profiling permits the differentiation of vesicles originating from healthy versus diseased (for example, cancerous) cells based on both their morphological attributes and biochemical fingerprints.

[0022] As a result, this disclosure enables more accurate and robust classification of disease states in complex clinical samples, reducing misclassification arising from overlapping or nonspecific background signals and significantly enhancing the biological interpretability of extracellular vesicle-based diagnostics.

[0023] This disclosure also enables the integration of size-resolved Raman spectroscopy with omics data for subpopulation analysis. In the literature, the size of extracellular vesicles (EVs) is closely linked to their cargo composition and biogenetic origin. Different EV subpopulations, even within a nominally homogeneous sample, exhibit distinct molecular profiles associated with their size, such as variations in nucleic acid, protein, lipid, and metabolite content. The platform of the present disclosure, by simultaneously acquiring precise size measurements via interferometric imaging and molecular fingerprints via Raman spectroscopy, enables the direct correlation of physical properties with molecular content at the single-vesicle level. When integrated with external omics datasets— including genomics, proteomics, metabolomics, lipidomics, and transcriptomics— the platform allows researchers and clinicians to link the Raman signatures and size distributions of individual EVs to broader molecular profiles. This correlation facilitates the functional annotation of EV subpopulations, providing insights into their biological roles, diagnostic potential, and therapeutic relevance. By combining size-resolved Raman data with omics analysis, the system creates a powerful multidimensional framework for the detailed characterization of EV heterogeneity.

[0024] Thus, according to one aspect, a platform is disclosed for characterizing particles including one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules and / or liposomes in a non-invasive, high-throughput, and label-free manner. The platform includes (a) an interferometric microscopy module configured to localize and size individual particles and (b) a wide-field or confocal microscopy module comprising a surface-enhanced Raman spectroscopy (SERS) system configured to acquire molecular signatures from localized particles.

[0025] In some forms of the platform, the surface-enhanced Raman spectroscopy (SERS) system may be a wide-field super-resolution surface enhanced Raman spectroscopy (SERS) device.

[0026] In some forms, the interferometric microscopy module may be used to localize single biomolecules (such as extracellular vesicles) and the wide-field or confocal microscopy module may be used to determine the type and / or contents of the biomolecules. The interferometric microscopy module may be used to estimate size and volume of one or more of the particles.

[0027] In some forms of the platform, the platform may further include a computational package that includes at least the following: image acquisition including interferometric microscopy data and wide-field or confocal microscopy data; image analysis including the interferometric microscopy data and the wide-field or confocal microscopy data; and neural network algorithms for characterization of the particles from integration and / or correlation of the interferometric microscopy data and wide-field or confocal microscopy data. Such characterization may include, but is not limited to identifying a size, shape, and location of one or more particles using interferometric microscopy data; identifying regions where one or more particles are clustered or isolated using interferometric microscopy data for further targeted wide-field or confocal microscopy analysis; processing wide field or confocal microscopy data to identify the molecular signatures of particles; and / or using machine learning algorithms to match spectra from the wide-field or confocal microscopy data to known structures from a preexisting database or data set.

[0028] According to another aspect, a device is disclosed including an integrated optical system capable of performing label-free, interferometric localization of nanovesicles and a wide-field super-resolution Raman microscopy to determine both physical and chemical structure of nanovesicles.

[0029] In yet another aspect of this disclosure, a diagnostic platform includes interferometric imaging and surface-enhanced Raman spectroscopy (SERS) to provide simultaneous localization, structural, and compositional analysis of extracelluar vesicles for usein molecular diagnostics and genetic research, and omics research including but not limited to lipidomic and proteomic metabolism studies.

[0030] According to another aspect, a method is disclosed for characterizing particles including one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules and / or liposomes in a non-invasive, high-throughput, and label-free manner. The method includes utilizing a merged set of imaging modalities including and capable of interferometric microscopy and wide-field or confocal microscopy and collecting images of the particles from the imaging modalities.

[0031] In some forms, a device for wide-field or confocal microscopy may be a wide-field super-resolution surface enhanced Raman spectroscopy (SERS). The wide-field superresolution surface enhanced Raman spectroscopy may provide rapid imaging of with high spectral resolution and single-molecule sensitivity.

[0032] In some forms, the particles including one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules and / or liposomes may be nanosized biological particles. The nanosized biological particles may be localized label-free using interferometric microscopy and, more particularly, the nanosized biological particles may be localized based on the measurement of phase differences between light waves scattered by the nanosized biological particles, molecules and a reference light. In some forms, the nanosized biological particles may be extracellular vesicles, viruses, or lipoproteins, although are not so limited.

[0033] In some forms, the method may further include detecting cellular response to environmentally toxic agents.

[0034] In some forms, the method may further include determining the liver hepatocyte toxicity in response to different drug concentrations.

[0035] In some forms, the method may further include detecting of pancreatic cancer risk via pancreatic cyst fluid analysis. This may involve using computer models to predict subtypes of pancreatic cyst fluids, providing insights into the potential malignancy of a cyst.

[0036] In some forms, the method may further involve utilizing the merged set of imaging modalities to forecast the correlation between Raman spectral responses of patient bloodsamples across diverse racial, weight, and age. This may result in the step of producing diagnostic suggestions based on that correlation. For example and below, this is described in the particular context in which the patients are type-two diabetics, but this certainly can be applied more widely.

[0037] According to another aspect, a process is disclosed for enhancing the resolution of interferometric imaging beyond the diffraction limit of light. The process includes employing techniques selected from a group consisting of confocal interferometric imaging, tomographic interferometric imaging, and structured illumination interferometric imaging.

[0038] According to another aspect, a system is disclosed for utilizing artificial intelligence to predict the biomarkers or nucleic acid composition of intact biological particles. The system may include a computational package for image acquisition, image analysis, and spectral data correlation via machine learning algorithms.

[0039] In another aspect, a method is disclosed for differential refractive index-based imaging and quantification of biological and synthetic particles. The method includes providing multiwavelength illumination to a substrate supporting the biological and synthetic particles and capturing a series of images that reflect differential refractive index effects.

[0040] According to yet another aspect, a method is disclosed for characterization of a molecular composition of extracelluar vesicle cargo by analyzing Raman spectra obtained from extracelluar vesicles or nanovesicles after localizing the extracellular vesicles or nanovesicles using interferometric microscopy. The method includes computational enhancing spectral resolution using a fitting algorithm.

[0041] According to still another aspect, a method is disclosed for enhancing the Raman signal of low-abundance particles on a substrate. The method includes obtaining an image using surface-enhanced Raman spectroscopy of the low-abundance particles on the substrate, the substrate being coated with a thin film to create enhanced Raman effects.

[0042] In some forms of this method, the substrate may be glass and the thin film may be SiC>2. In other forms, the thin film may be a metallic deposition, a dielectric metallic thin film, and / or a semiconductor thin film. In still another form, the substrate may be glass, the thin film may be SiOz, and the thin film may be further coated by silver or gold. In this form, the thin film of SiOz may be approximately 100 nm thick and the silver or gold on the thin film may beapproximately 5 nm thick. For those thickness and at that scale, "approximately" means those values can be taken to be within + / -20% of those exemplary thicknesses. In some forms, the thin film has a thickness below 1 pm, as more generally, the film thickness should be shorter than the light wavelength (that is, subwavelength)

[0043] In yet another aspect, a surface substrate is disclosed that is designed to enhance Raman spectroscopic analysis. The substrate includes various materials including an (ordinary) glass slide, a metal-coated surface with metals such as silver (Ag) or gold (Au) for surface- enhanced Raman spectroscopy (SERS), a thin film coating, or a structured surface featuring selfassembled nanoparticles.

[0044] In still another aspect, a method is disclosed to fast scan tissue samples for their biomolecular content. The method includes obtaining spatial and chemical composition of tissue slices.

[0045] In some forms of this method, fluorescent information from a spatial Raman tissue scan may be collected and correlated to gather additional information about the tissue sample.

[0046] In yet another aspect, a method is disclosed in which tissue and tissue slices are scanned to detect and quantify the presence of various metals.

[0047] In some forms of this method, the method may further include detecting local differences in tissue and tissue slices such as inflammation or cancer and a diagnosis may be established based on that detection. In some forms, the method may also provide histopathological staining-free imaging of the tissue and tissue slices. This may lead to spatial omic map of the tissues such as proteomic, genomic, metabolic or lipodomic maps. This may be likened to a label-free spatial transcriptomics machine. It may also be operated in labelled mode, but it would additionally give molecular data.

[0048] In yet another aspect, a method is disclosed including obtaining interferometric and Raman spectral signatures of a sample containing one or more types of vesicles; and, based on the interferometric and Raman spectral signatures obtained from the sample, distinguishing between different types of vesicles in the sample uniquely based on the interferometric and Raman spectral signatures.

[0049] In some forms of this method, a cargo content within biological and synthetic particles in the sample may be quantified.

[0050] According to still another aspect, a method is disclosed in which interferometric and Raman spectral signatures of a sample containing one or more types of viruses are obtained; and, based on the interferometric and Raman spectral signatures obtained from the sample, distinguishing between different types of viruses in the sample uniquely based on the Raman spectral fingerprints as well as potentially any morphological information made available via the interferometric microscope.

[0051] In some forms, distinguishing between different types of viruses in the sample uniquely based on the Raman spectral fingerprints may employ machine learning algorithms for accurate virus classification.

[0052] In yet another aspect, a method is disclosed of modeling drug-response using an algorithm using information collected from the general method and platform described above and herein. The method includes generating IC50 and EC50 curves algorithmically by leveraging regression model outputs in response to Raman fingerprints of extracellular vesicles after normalizing the intensity to the total number of particles counted by the interferometric microscopy module.

[0053] According to yet another aspect, a method is disclosed of establishing a protein Raman library using the platform described above and herein. This method includes producing the protein Raman library by collection of a set of images and quantifying correlations between sample content and elements within the library.

[0054] In still another aspect, a method is for providing computational scores to predict a nucleic acids composition from intact biological and synthetic particles using artificial intelligence. The method includes collecting a set of information from the merged set of imaging modalities from the general method; and applying artificial intelligence algorithms to the set of information to produce computational scores predictive of the nucleic acid composition from the intact biological and synthetic particles.

[0055] In some forms of this method, intact biological and synthetic particles may include one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules and liposomes.

[0056] According to yet another aspect, a method is disclosed for structural optical sequencing of nucleic acids, proteins, and their hybrids (aptamers). The method includes collecting a set of information from the merged set of imaging modalities from the general method; and optically sequencing the nucleic acids, proteins, and their hybrids (aptamers) based on the set of information.

[0057] According to another aspect, a method is disclosed for characterization of drug ligand interactions such as binding, affinity, avidity, association, dissociation kinetics using the general method.

[0058] According to still yet another aspect, a method is disclosed for studying extracellular vesicle dynamics including characterization of extracellular vesicle secretion kinetics and their cargo using the general method.

[0059] According to another aspect, a quantum-acoustic-enhanced interferometric Raman microscope is provided. The microscope includes quantum-enhanced Raman spectroscopy to reduce quantum noise and improve the signal-to-noise ratio (SNR); surface acoustic wave (SAW)-based nanoparticle trapping to confine and dynamically manipulate particles within enhanced optical fields, increasing Raman interaction times without requiring plasmonic substrates; and optionally, interferometric imaging to provide simultaneous real-time size, shape, and molecular characterization.

[0060] In some forms of this quantum-acoustic-enhanced interferometric Raman microscope, interferometric imaging may also be employed in addition to Raman imaging. A combination of the quantum-enhanced Raman spectroscopy and interferometric imaging may provide simultaneous label-free molecular fingerprinting and high-sensitivity detection at the single-particle level. However, it should be appreciated that while single-particle level resolution is obtainable, that the range of this technology is capable of imaging a single or few particle to many particles, even in the millions of particles.

[0061] In some forms, the quantum-enhanced Raman spectroscopy may reduce quantum noise and improve the signal-to-noise ratio (SNR) by employing squeezed light and quantum correlations to increase Raman signal-to-noise ratio while minimizing photodamage.

[0062] According to another aspect, a method for characterizing biomolecules, biological tissue sections, viruses, or other particles of interest in a non-invasive, high-throughput, andlabel-free manner is disclosed. The method includes confining and / or dynamically manipulating biomolecules, biological tissue sections, viruses, or other particles of interest using surface acoustic wave (SAW)-based nanoparticle trapping; and imaging the biomolecules, biological tissue sections, viruses, or other particles of interest using quantum-enhanced Raman spectroscopy.

[0063] In some forms, the method further includes imaging the biomolecules, biological tissue sections, viruses, or other particles of interest using interferometric imaging.

[0064] According to still another aspect, a method for particles including one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules and / or liposomes in a non-invasive, high-throughput, and label-free manner is disclosed. The method includes obtaining a set of wide-field or confocal microscopy images and applying artificial intelligence algorithms to the set of wide-field or confocal microscopy images to produce a predictive library.

[0065] In some forms, the method may further involve obtaining and analyzing a subsequent wide-field or confocal microscopy image and comparing the subsequent wide-field or confocal microscopy image to the predictive library to assess and classify the subsequent wide-field or confocal microscopy image.

[0066] In some forms, the method may further involve suspending particles in a fluid during at least part of the step of obtaining a set of wide-field or confocal microscopy images. The fluid may be a suspension medium and the suspension medium may be subjected to mechanical or acoustic actuation. The mechanical or acoustic actuation can facilitate averaging during dynamic Raman readout during obtaining a set of wide-field or confocal microscopy images. In some other forms, the particles suspended in a fluid may be flowed on a surface during at least part of the step of obtaining a set of wide-field or confocal microscopy images. In yet some other forms, the particles suspended in a fluid may be flowed through a channel (which in some more particular forms could be a nano or microfluidic channel) during at least part of the step of obtaining a set of wide-field or confocal microscopy images.

[0067] In some forms, the method may include two operational modes for imaging including (a) a dry surface mode in which isolated particles are immobilized onto a solidsubstrate and (b) a suspension dynamic mode in which particles remain within a fluid medium during measurement, allowing real-time detection of free-floating or loosely bound particles.

[0068] In some forms, the method may employ a self-assembled plasmonic substrate composed of gold or silver nanoparticles that have been deposited on the sensing surface via controlled drying, solvent evaporation, or chemical functionalization techniques, allowing them to spontaneously organize into densely packed nanoclusters.

[0069] In some forms, the method may involve characterizing and differentiating between HIV clades and to stratify patient-specific viral load levels using optical fingerprinting. In some forms, the method may involve characterizing extracellular vesicles and detecting changes overall in the extracellular vesicles across a patient which extracellular vesicles are changed by a presence of disease or virus in a patient. This is to say, the disease changes the overall EV characteristics across the patient that leads to the difference that can be detected; thus, it is possible not only to detect a particular virus (for example, HIV) or a disease, but also through EVs detection to see the impact overall on all body secreted EVs of the virus or disease. In other forms, more generally, the method may involve characterizing and differentiating between different types of viruses (or virus sub-types). In still other forms of the method, the method may involve characterizing and distinguishing between extracellular vesicles, lipoproteins, and other blood components.

[0070] In some forms, the method may involve characterizing and identifying patient- derived extracellular vesicles as part of a non-invasive biomarker platform for early lung nodule characterization. In some forms, for better decision-making, nodules may be detected by integrating, in a machine learning format, radiomics features of medical images (such as PET and MRI scans) as well as -omic data of the patient tissue or liquid biopsy samples. In some forms, the method may involve obtaining further clinical imaging data corresponding to the set of wide-field or confocal microscopy images; and, in this form, applying artificial intelligence algorithms to the set of wide-field or confocal microscopy images to produce a predictive library may involve further providing the clinical imaging data corresponding to the set of wide- field or confocal microscopy images to the artificial intelligence algorithms. This may improve or enhance the accuracy and predictive power of the algorithms and machine learning involved. It is contemplated that the further clinical data may be radiomics data. That radiomics datamay be - but is not limited to - one or more of PET / CT and / or PCCT lung nodule imaging (at least in the case lung cancer detection).

[0071] In some forms, the method may involve direct detection of extracted RNA in a dry assay format.

[0072] In some forms, the method may include detecting, not only extracellular vesicles, but all extracted components including all cargo components including but not limited to membranes, proteins, lipids, RNA, DNA that are isolated as a part of extracellular vesicle isolation.

[0073] In some forms, the method may further include subjecting the particles to a controlled actuation waveform; acquiring a z-stack of Raman spectra synchronously with the actuation waveform, capturing temporal variations in the Raman signal intensity at each wavenumber; and analyzing the acquired z-stack of Raman spectra by correlating the intensity fluctuations at each spectral position with the applied actuation waveform. This dynamic correlation strategy allows for decomposition of Raman data into distinct sample and background spectra, leading to improved detection sensitivity and molecular specificity.

[0074] In some forms, the method may further include determining, estimating, and / or calculating the volume, surface area, and / or mass of the particles being detected per area. That volume, surface area, and / or mass of particles being detected per area may be provided to an artificial intelligence algorithm. The artificial intelligence algorithm may be applied to the set of wide-field or confocal microscopy images along with the corresponding volume, surface area, and / or mass of particles to produce the predictive library.

[0075] In some forms, the method may further include determining, estimating, and / or calculating, the volume, surface area, and / or mass of particles being detected per area. Then a concentration of the particles in a subsequently prepared sample may be adjusted to optimize the Raman signal in the subsequently prepared sample.

[0076] In some forms, the method may further include determining, estimating, and / or calculating the volume, surface area, and / or mass of the particles being detected per area. Then, the Raman signal in the area being scanned can be scaled based on the determined, estimated, and / or calculated volume, surface area, and / or mass of the particles detected per area.

[0077] In some forms, the method may further include obtaining an interferometric image from a same sample and same location as that of a corresponding wide-field or confocal microscopy image. It is contemplated that obtaining the interferometric image from the same sample and same location as a corresponding wide-field or confocal microscopy image occurs may sequentially or simultaneously with the obtaining of the wide-field or confocal microscopy image.

[0078] In some forms, the interferometric image may also be supplied to an artificial intelligence algorithm producing the predictive library as a parameter to assess and classify the subsequent wide-field or confocal microscopy image.

[0079] In some forms of the method, the method may further include obtaining a plasmonic image from the same sample and same location as a corresponding wide-field or confocal microscopy image.

[0080] In some forms, the method may further involve plasmonically activating a surface on which the particles are located. The plasmonic activation of the surface may be an additional parameter provided to an artificial intelligence or machine learning algorithm to develop the predictive library and / or may be used to enhance a Raman signal during obtaining of the wide- field or confocal microscopy image and provide an improved signal to noise ratio.

[0081] In some forms the method may further involve obtaining an interferometric image from the same sample and same location as a corresponding wide-field or confocal microscopy image and plasmonically activating a surface on which the particles are located. Obtaining a set of wide-field or confocal microscopy images, obtaining an interferometric image from the same sample and same location as the corresponding wide-field or confocal microscopy image, and plasmonically activating a surface on which the particles may be at least three of the components that are provided to an artificial intelligence algorithm that are used to produce a predictive library.

[0082] According to yet another aspect, a method is disclosed in which particles are flowed or transported through a channel having an excitation region, the particles are excited in the excitation region, and Raman spectra are continuously or periodically obtained while the particles are in the excitation region.

[0083] In some forms of this method, the particles may be sorted based on their characterization from the Raman spectra after the particles have been further flowed or transported from the excitation region.

[0084] In some forms, exciting particles in the excitation region may involves one or more of acoustic, electromagnetic, mechanical, and antenna-based excitation that affects the sensing surface in any shape or form, causing a fluctuation in a Raman signal that allows better sensitivity, specificity and improved signal to noise ratio as well as different forms of averaging and resolution enhancement algorithms.

[0085] 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

[0086] FIG. 1A shows interferometric images of EVs and FIG. IB shows the size distribution of the detected particles from FIG. 1A.

[0087] FIG. 2A shows interferometric images depicting both empty lipid nanoparticles and lipid particles containing iron particles and FIG. 2B shows the particle contrast distribution of individual particles revealing distinct contrast value distributions, highlighting differences between empty and filled particles from FIG. 2A (with the empty lipids peak being on the left and the lipids and iron peak being on the right).

[0088] FIG. 3A shows interferometric images of Hepatitis B viruses (HBVs) isolated from clinical plasma samples and FIG. 3B shows the size distribution of the whole field of view from FIG. 3A.

[0089] FIG. 4 provides a schematic overview of a multimodal interferometric-fluorescence imaging system and corresponding images of captured polystyrene particles, with and without fluorescence on the left and right side panels, respectively, and as labelled.

[0090] FIG. 5 provides Raman spectra of polystyrene (PS) particles (100 nm diameter) immobilized on a glass slide, a Si / SiOz thin film, and a Si / SiOz / Au film. The right side panelil lustrates the Raman peaks corresponding to PS particles are enhanced in the case of the thin film and Au-coated film.

[0091] FIG. 6 provides a diagram of a multimodal Raman-interferometric microscope.

[0092] FIG. 7A provides a schematic of wide-field image from a field of view from a mixture of exosomes isolated from different sources and FIG. 7B illustrates an artificial intelligence algorithm or approach used in SERS analysis of pancreatic cyst fluid samples.

[0093] FIGS. 8A-8D depict SERS prediction of pancreatic cyst fluids subtypes. FIG. 8A shows the averaged normalized Raman spectra of the measured pancreatic cyst fluid subtypes.FIG. 8B shows the principal component analysis (PCA) score plot for the first two component to show the clustering of the dataset. FIG. 8C shows a comparison of different machine learning approaches on the dataset (in which CNN: Convolutional neural networks, NN: neural networks, RF: Random forests, QDA: Quadratic discriminant analysis, XGB: XG Boost). FIG. 8D shows the binary classification performance using random forest classification to predict cancerous types from the benign types.

[0094] FIGS. 9A-9F shows the classification of normal weight diabetes patients and the chemical fingerprints related to this classification. FIG. 9A shows a comparison of BMI-splitted normalized data. FIG.9B shows a comparison of BMI and race-splitted data. The first two PCA scores were plotted to visualize clustering of the data for BMI-splitted in FIG. 9C and BMI and race-splitted data in FIG. 9D. FIGS. 9E and 9F show the fatty acid and lipoprotein levels obtained by Raman spectroscopy and their relationship with the HblAc levels.

[0095] FIGS. 10A-10F shows the prediction of toxin levels of in vitro hepatocyte cell cultures.

[0096] FIGS. 11A-11E shows the prediction of environmental toxins exposure to ovarian cancer cell lines like and OVCAR-3.

[0097] FIG. 12 shows an acquired Raman spectra of two different viruses (HIV and KSHV) showing distinct features for identification.

[0098] FIG. 13 provides a characterization of different iron forms using Raman spectroscopy in which a Raman scan of clinical Alzheimer tissue section mapped for a selected peak location at 631 cm4. This map was collocated with the previously scanned X-Ray image, confirming the iron deposited regions on the tissue section.

[0099] FIG. 14 provides a comparison of the fingerprints of microvesicles isolated from quiescent and senescent cell lines.

[0100] FIG. 15 illustrates schematically the operation and outcomes of the AI-SER.

[0101] FIG. 16A shows averaged normalized Raman spectra of different types of cyst fluids. In FIG. 16B, principal component analysis visualization using the first most important components, PC 1 and PC 2, are plotted. In FIG. 16C, a comparison of model accuracies applied on SERS data from cyst fluid derived EVs is shown (in which CNN: Convolutional neural networks, NN: neural networks, RF: Random forests, QDA: Quadratic discriminant analysis, XGB: XG Boost). FIG. 16D shows the binary classification performance using random forest classification to predict cancerous types from the benign types.

[0102] FIG. 17A shows quantification of bio / nanovesicles of known concentrations (top, blue curve) vs. negative control (bottom, red curve) and FIG. 17B shows interferometric images of HBV (Scale: 1 pm).

[0103] FIG. 18A shows Raman spectra of vesicles without and with protein cargo; FIG. 18B shows principal component analysis (PCA) of Raman spectra shown in FIG. 18A; and FIG. 18C shows exosomes isolated from pancreatic cyst fluid to determine subclass of the cyst [Pseudocyst (PC), Serous Cystic Adenocarcinoma (SCA), Low-Grade MCN (MLG), High-Grade IPMN (IHG), Low-Grade IPMN (ILG)].

[0104] FIG. 19 illustrates the concept of the Quantum-Acoustic-Enhanced Interferometric Raman (Q-AIR) Microscope. Q-AIR integrates quantum-enhanced Raman spectroscopy, surface acoustic wave (SAW) trapping, and interferometric imaging for high-sensitivity, label-free single-particle characterization. The bottom center of FIG. 19 shows nanoparticles are confined using SAW trapping on a substrate, the upper left of FIG. 19 shows squeezed light suppresses quantum noise, reducing shot-noise limitations, the top center of FIG. 19 shows high-resolution imaging enables real-time nanoparticle detection, and the top right of FIG. 19 shows noise suppression enhances signal detection, improving the signal-to-noise ratio (SNR).

[0105] FIG. 20 illustrates quantum squeezing in phase space. The circle represents balanced noise in a classical coherent state, while the ellipse shows reduced noise in the amplitude quadrature and increased noise in the phase quadrature in a squeezed state. This will improve Raman signal-to-noise ratio without increasing laser power.

[0106] FIG. 21 provides an experimental setup for quantum-enhanced Raman spectroscopy. The setup includes an SPDC-based squeezed light source, a balanced homodyne detection system, and a Raman excitation and detection module). The squeezed light source is generated using a 405 nm CW pump laser and a nonlinear crystal (PPKTP). The Raman signal is collected using a high-NA objective and analyzed through a spectrometer. The homodyne detection system improves the SNR by suppressing quantum noise.

[0107] FIG. 22 provides a schematic of the experimental setup for Raman-Interferometric setup, featuring the NIR spectrometer, Raman module, and interferometric module involving the CMOS camera and laser. The setup includes a galvo scanner for Raman beam scanning, filters, and optical elements for signal collection and analysis. Squeezed light excitation is omitted for simplicity.

[0108] FIG. 23 illustrates SAW trapping frequency versuse particle size showing that the frequency increases as particle size decreases.

[0109] FIGS. 24A and 24B provide Raman spectral and PCA-based classification, respectively, of HIV types, and FIGS. 24C and 24D provide Raman spectral and PCA-based classification, respectively of patient samples. The average SERS spectra of HIV Clade A, Clade C, and SHIV reveal distinct molecular fingerprints and PCA shows clear separation of viral types, demonstrating the platform's ability to subtype HIV based on label-free spectral features.

[0110] FIGS. 25A shows averaged, normalized SERS spectra for healthy controls, benign- nodule patients and early-stage lung cancer patients in which dashed lines show the difference spectra for healthy versus all samples and benign versus cancer, with shaded bands indicating ±1 standard error of the mean. FIG. 25B shows principal component analysis of individual SERS spectra projected onto PCI (75.3 % variance) and PC2 (4.8 %), with linear discriminant analysis boundaries (dashed lines) and background shading delineating healthy (upper right side of chart), benign (bottom left side of chart) and cancer (upper left side of chart) clusters. FIG. 25C is a heatmap of per-subject SERS intensities across representative Raman shifts (rows) for each healthy (H1-H9) and patient (P1-P14) sample, demonstrating clear separation between nodule- free and nodule-bearing cohorts. FIG. 25D shows an additional principal component analysis of individual SERS spectra projected onto PCI (53.3 % variance), PC2 (11.7 %), and PC3 (6.9%), with a QDE boundary encapsulating benign data points and excluding cancerous data points.FIG. 25E is a heatmap of SERS intensities taken from EVs derived from patient diagnosed as benign (B1-B4) and cancer (C1-C10.

[0111] FIG. 26 is a schematic of dynamic correlation Raman spectroscopy for background rejection and sample signal extraction. A z-stack of Raman spectra is acquired during controlled sample actuation (e.g., defocus oscillation). Correlation analysis between the spectral intensity variations and the actuation waveform yields a correlation spectrum, enabling separation of sample-specific Raman peaks from stationary background signals. The resulting decomposition generates a high-purity sample spectrum and an isolated background spectrum, improving signal-to-noise ratio and enhancing detection sensitivity for weak biological targets.

[0112] FIGS. 27A through 27D show dynamic correlation Raman analysis for background suppression and signal recovery. FIG. 27A shows a Raman spectrum acquired at nominal focus, showing strong background contributions alongside sample peaks. FIG. 27B shows correlation spectrum obtained by dynamic actuation analysis, highlighting features that modulate coherently with sample movement, while background signals are suppressed. FIG. 27C shows a background-subtracted Raman spectrum derived from correlation analysis, showing improved peak clarity and signal-to-noise ratio. FIG. 27D shows a reconstructed focused Raman spectrum with background removed, demonstrating substantial enhancement of true sample features compared to the initial nominal measurement.

[0113] FIG. 28A and 28B show patterned surface design approaches including, in FIG. 28A, a concentric-ring design and, in FIG. 28B, a chessboard design.

[0114] FIG. 29 illustrates the surface concentration-dependent SERS response measured at 1576 cm-1across a dilution series (log scale). The signal intensity exhibits a nonlinear trend, increasing with particle concentration up to an optimal density, followed by a decline due to plasmonic saturation, inter-particle coupling, and vesicle stacking effects. This highlights the precise surface quantification enabled by interferometric imaging, which allows accurate particle enumeration and normalization of Raman signals on a per-particle basis, ensuring assay reliability and reproducibility.DETAILED DESCRIPTION

[0115] Determining types and cargo of vesicles has not possible in a label-free way. Herein ways of using optical means are disclosed to determine for example, and not limited to, what avesicle is, whether it is a loaded RNA vaccine or empty for quality control, whether it is a virus and what type of a virus, whether it is a cancer vesicle or not. The classification and identification of vesicles in a label free manner enables unprecedented tools and applications in diagnostics, infectious diseases, cardiovascular diseases, cancer, vaccine delivery, drug delivery and many areas broadly.

[0116] The disclosed technology described herein encompasses compositions, devices, processes, methods, and systems specifically designed for the rapid and accurate optical fingerprinting, identification, and sequencing of nano and micro-vesicles and their contents. Additionally, this disclosure introduces a novel interferometric imaging approach to localize EVs and wide-field super-resolution Raman microscopy to determine the 3D structure of the genetic biomarkers, thereby enhancing the capabilities of the optical device in molecular diagnostics and genetic research.Example I: Biological Particle Localization and Physical Characterization

[0117] Imaging nanosized biological particles, such as EVs, without labels poses a significant challenge due to their sizes falling below the diffraction limit of light. The disclosed localization method is based on interferometry, a precise technique that can measure minute phase differences between light waves. In interferometric imaging, the interference between the scattering signal scattered by the nanosized particle and a known reference light can be leveraged. The total intensity ( / tot) is calculated as \ Esca+ Eref\2- | Esco |2+ | £re / |2+ 2-ErefESca-cos(0), where ESCa and Eref represent the fields generated by the scattering from the particle and the reference field, respectively, and 0 is the phase difference between these two fields.

[0118] An on-axis interferometric microscopy configuration is employed, where the reference light is collected directly from the sample substrate. Here, simply a glass slide can be used as a sample substrate, or alternatively a specially designed, cost-effective thin film can be utilized that is structurally tuned to acquire a high interferometric signal. Given that the scattering field from EVs is small, and the background light is a static term, the contrast resulting from a single particle is fundamentally governed by the ratio EsCalEref cos(p). To enhance sensitivity and enable the mapping of single EVs as depicted in FIGS. 1A and IB, defocused images are employed (with tuned phase differences).Wide Field Super-Resolution Integration

[0119] Confocal interferometric imaging provides focused optical sectioning, enhancing depth localization and reducing background noise.

[0120] Tomographic interferometric imaging employs projection images at various angles to reconstruct three-dimensional representations of EVs, offering detailed structural insights.

[0121] Structured illumination interferometric imaging utilizes patterned light to double the spatial resolution limit, allowing for the detailed observation of sub-200 nm EVs.Differential Refractive Index-Based Imaging and Quantification Using Multiwavelength Illumination

[0122] The differential refractive index-based imaging component of the system leverages the principle that the refractive index of biological particles, such as EVs, varies with the wavelength of the incident light. By illuminating the sample with light at multiple wavelengths, the system can capture a series of images that reflect these differential refractive index effects. Analyzing these images allows for the quantification of specific properties of the EVs, including size, concentration, and compositional information, without the need for external labeling.

[0123] The multiwavelength illumination is achieved through a specialized light source capable of emitting at predefined wavelengths, optimized for maximal contrast between different biological materials. The interferometric imaging module is adapted to work seamlessly with this light source, capturing high-resolution images that are then processed by the computational unit to extract quantitative information based on the differential refractive index measurements.Nanosized biological particle refractive index analysis (i.e cargo load detection) using Interferometric Detection

[0124] The intensity of light scattered by a particle is directly proportional to both the particle's volume and refractive index. This correlation between the scattering signal and refractive index is consistent across various biomolecules such as glycoproteins, nucleic acids, and lipids, making interferometry a valuable tool for analyzing biomolecular cargo and detecting cargo loads. FIG. 2A displays interferometric images of lipid nanoparticles, both empty ones and those containing iron particles, along with the corresponding particle contrast distribution in FIG. 2B. Despite their similar sizes, the variation in interferometric contrast,attributed to differing refractive indices, allows the assessment of the cargo content of individual nanoparticles and to spatially locate them on the sample substrate for subsequent Raman analysis.Single Virus Imaging and Localization

[0125] Viruses are biological entities that exists on the nanoscale. Leveraging the capabilities of the disclosed imaging platform, the detection and analysis of viral particles have been demonstrated. In a demonstrative application, Hepatitis B viruses (HBVs) have been isolated from human plasma and imaged using this system as shown in FIGS. 3A and 3B. The platform captured the interferometric images of HBVs, achieving precise localization, important for subsequent spectral analysis as discussed below and with reference to FIG. 12.Multimodal Interferometric- Fluorescence Detection for Surface Protein Content Analysis

[0126] To provide a comprehensive biological nanoparticle analysis, a multimodal microscope capable to provide detailed nanoparticle characterization was developed. FIG. 4 illustrates a dual-modality setup that combines interferometric and fluorescence detection methods to analyze protein content on surfaces with high sensitivity and specificity. In FIG. 4, in the upper left panel, a schematic of an interferometric setup is shown. A coherent light source, such as an LED, illuminates the sample surface, with the light passing through a beamsplitter and an objective lens to focus on the surface of interest, typically a silicon wafer with a thin layer of silicon dioxide (SiOz). The reflected light from the sample and the reference beam from the beamsplitter create an interference pattern, which is captured by a CMOS camera. This interference pattern, such as the one depicted in the lower left panel of FIG. 4, can be analyzed to determine particle location as well as physical properties such as size and refractive index of the particle. In the second modality in the upper right side panel of FIG. 4, a fluorescence detection setup is presented. Here, the LED light is filtered through a high pass filter to ensure only the excitation wavelengths reach the sample. The proteins of interest are tagged with a fluorescent dye that emits light at a different wavelength when excited. The emitted fluorescence passes through the beamsplitter and is captured by the CMOS camera. The intensity and distribution of the fluorescent signal provide information about the quantity and spatial distribution of the proteins on the surface. Output images of polystyrene beads withfluorescence dye and without dye is given as example measurement, i.e. discrete bright spots on a dark background for interferometry and bright red emissions on a dark background for fluorescence in the lower right panel of FIG. 4.

[0127] When used in tandem, these techniques offer a powerful approach for analyzing surface-bound proteins. Interferometry provides nanoscale information on the protein layer's physical properties, while fluorescence gives insight into the molecular identity and abundance. The combination of these techniques into a single, integrated system allows for simultaneous acquisition of complementary data, enhancing the ability to characterize biological nanoparticles.Thin Film Enhancement of Raman Signature

[0128] Fluorescence detection, while widely used for analyzing surface-bound proteins, comes with inherent limitations. The requirement of labeling target molecules with fluorophores can introduce issues such as photobleaching, where the fluorescent signal fades with time under the light exposure used to excite the fluorophores. Additionally, this labeling can perturb the natural state of the proteins, potentially altering their function. Furthermore, the finite observation times imposed by photobleaching and the limited multiplexing capabilities due to spectral overlap of fluorophores can restrict the range of analysis.

[0129] In contrast, Raman spectroscopy is a label-free imaging technique that circumvents these limitations. It allows for the observation of molecules in their native state without the need for fluorescent labels. This technique's non-invasive nature ensures that there is no photobleaching effect, thus enabling prolonged observation times and repeated measurements, which are incredibly valuable for dynamic studies.

[0130] Interferometric microscopy involves a highly flat sample substrate in order to acquire highly sensitive images, which limits the utilization of nanopatterned SERS substrates. Herein, utilization of interferometric films in Raman microscopy is demonstrated to enhance the acquired signal. FIG. 5 illustrates the performance of Raman spectroscopy, further enhanced by a SiCh thin film substrate. The comparison of Raman spectra of polystyrene (PS) particles on standard glass versus those on a glass slide coated with an SiO2 thin film reveals a significant enhancement of the Raman signal in the latter in FIG. 5. This enhancement isespecia I ly noticeable in the characteristic peaks of PS, which are substantially more pronounced on the SiCh film.

[0131] The amplification of the Raman signal on the thin film is attributable to the surface- enhanced Raman scattering (SERS) effect. The SiC>2 thin film modifies the local electromagnetic field, leading to an increased signal intensity that is especially beneficial for detecting low- abundance molecules.

[0132] This thin film enhancement of Raman spectroscopy presents a compelling alternative to fluorescence detection for surface protein content analysis. By overcoming the limitations of fluorophore labeling and enabling a more robust, label-free detection method, Raman spectroscopy with thin film enhancement holds promise for advancing the accuracy and reliability of molecular detection and characterization.Simplified Flow Chart for Data Acquisition and Processing of the Integrated Platform

[0133] The integrated platform employs a data acquisition phase which is then followed by a data processing phase.

[0134] In the data acquisition phase, interferometric data and Raman spectral data are acquired. For the interferometric data acquisition, the optical platform is initialized to capture interferometric images and high-resolution images are collected to locate particles within the sample. For the Raman spectral data acquisition, the Raman spectroscopy module is utilized to illuminate located particles and the resulting Raman spectra for each particle or region of interest is recorded.

[0135] In the data processing phase, the interferometric data analysis and Raman spectral data analysis are performed. During the interferometric data analysis, the interferometric images are processed to quantify particle size, shape, and location and regions are identified where particles are clustered or isolated for targeted Raman analysis. During the Raman spectral data analysis, Raman spectra are analyzed to identify the molecular signatures of particles and machine learning algorithms are used to match spectra to known molecular structures from a database.

[0136] Following data acquisition and processing, the platform can integrate the data and provide reporting. The spatial data from interferometric analysis is integrated with themolecular data from Raman spectroscopy. Then, a report is generated detailing the characteristics and distribution of the particles, supported by both imaging and spectroscopic data.Example II: Extracellular vesicle content analysis

[0137] Raman spectroscopy represents a potent technology for ascertaining the chemical composition of materials. The intricacy arises with extracellular vesicles, particularly exosomes, due to their diminutive size and significant heterogeneity, complicating the identification of their origins. Recently, a novel approach has been pioneered to traditional surface-enhanced Raman spectroscopy by integrating a neural network methodology. In this innovative technique, a supervised algorithm was employed to train classes with known labels. Subsequently, measurements from mixtures of known classes with progressively increasing concentrations were forecasted utilizing the trained model.

[0138] Herein, it is further disclosed (i) to provide a widefield Raman imaging system capable of swiftly capturing extracellular vesicle images across a substantial field of view (200 pm X 200 pm). It is further disclosed (ii) to enhance the neural network methodology to identify exosomes within entire images and predict RNA content. This is achieved by leveraging a pretrained library of signals derived from nucleotides in powdered form, aqueous solution, and liposomes with a diameter of 100 nm. It is additionally disclosed (iii) to demonstrate the clinical applicability of the method and the algorithms on pancreatic cyst fluid analysis, liver hepatocyte toxicity, type two diabetes characterization, virus type determination, prediction of effects of environmentally toxic materials on ovarian cell cultures and iron content determination of tissue sections taken from patients with Alzheimer's disease. A scheme of widefield image from a field of view from a mixture of exosomes isolated from different sources and the use of an artificial intelligence algorithm used in SERS analysis of pancreatic cyst fluid samples are exemplary depicted in FIG. 7A and 7B.

[0139] The artificial intelligence (Al) pipeline used in the content analysis of the EVs is summarized as below. The data is pre-processed including cosmic ray removal, calibration, baseline correction, and normalization. Then, the sample number is explored, and the data is augmented using Generative Adversarial Networks (GAN). The number of groups (n) and data clustering are explored using principal component analysis (PCA) and the Al pipeline is decidedupon depending on how the samples overlap in the PC space. If n=2 and the overlap score=l, then the Support Vector Machines for binary classification is applied without augmentation. If n=2 and the overlap score=2, then the Convolutional Neural Networks (CNN) is applied without augmentation. If n > 2 and the overlap score = 1, the Random Forest (RF) is applied. If n > 2 and the overlap score = 2, the CNN is applied with augmentation. For CNN, convolutional layers are used to detect local patterns, pooling layers are used to reduce dimensions, and activation functions are used for complexity.

[0140] As shown in FIG. 7B, the first layer is a one-dimensional convolutional layer, with thirty two filters, kernel size twenty, and stride equal to one. Attached to it is a batch normalization and ReLU activation function. The second convolutional layer is made of the same elements, except with sixty four filters. Next, the model is made of two linear layers of size eight hundred and thirteen times sixty four nodes, and twenty nodes respectively. Their activation functions are ReLU for the first linear layer, and softmax for the last one. The output is thus seven nodes with values ranging between zero and one, with each node representing a cyst type.Prediction of Pancreatic Cyst Fluids

[0141] The classification of the different pancreatic cyst fluid (PCF) types has been demonstrated for the first time, as shown in FIGS. 8A through 8D. Fifty-one human cyst fluid samples were tested. EVs were isolated from these cyst fluids and these samples were accurately classified using the EV signatures as Low-Grade IPMN (ILG), Moderate-Grade IPMN (IMG), High-Grade IPMN (IHG), Serous Cystic Adenocarcinoma (SCA), Pseudocyst (PsC), Moderate-Grade MCN (MMG), and Low-Grade MCN (MLG).

[0142] After isolating the EVs from 100 pL of a clinical PCF sample, 961 SERS spectra were acquired from each type of cyst fluid, and their vector-normalized spectra were compared as shown in FIG. 8A. FIG. 8B shows their Principal Component Analysis (PCA) score graph for the first two components that provides a method to explore how the measurements are clustered along different groups. This plot shows that each PCF type clusters distinctly as a class that shows the potential to identify unknown test samples. The SCA cluster presents a sparse distribution, which is expected since this group has a large in-group variance naturally.Furthermore, the high grade IPMN (IHG) group contains subtypes. In FIG. 8C, the accuracies are illustrated of diverse machine learning models applied to cyst fluid-derived EV data. An aim is to utilize models for accuracy and minimum overfitting. Upon completion of testing, the chosen pretrained classification model can be integrated into the device. When this approach was applied to patient-to-patient variance, the concordance was very within each PCF type as shown in the Receiver Operator Characteristics (ROC) curve in FIG. 8D. The area under the curve for the binary classification is 0.98, which shows the potential of this method for differentiating pancreatic cancer and non-pancreatic cancer groups from PCF samples using EVs, hence, its clinical translation potential and utility in identifying clinically relevant and actionable information as a decision guide.Biophotonic Classification of Normal Weight Diabetes using SERS Analysis of Plasma EVs

[0143] A methodological framework was developed to discern cases of type 2 diabetes mellitus based on their associations with biochemical profiles. The approach involved utilizing surface-enhanced Raman spectroscopy (SERS) to analyze the molecular signatures of clinical plasma extracellular vesicles (EVs). Initially, samples were stratified into two groups based on body mass index (BMI), categorizing them as either above or below 25 kg / m2. The comparison of SERS signatures between these groups as shown in FIG. 9A revealed generally similar responses, albeit with slight variations at specific wavenumbers. However, when the averaged spectra of groups stratified by BMI and race were examined as depicted in FIG. 9B, more pronounced differences in signatures emerged.

[0144] Further analysis, as depicted in FIG.9C and FIG. 9D, demonstrated a significant reduction in overlap in principal component analysis (PCA) scores when considering different racial groups separately. This underscores the importance of accounting for racial diversity in such analyses.

[0145] Additionally, the method enabled the characterization of specific diabetes-related compounds and their relationships with key patient parameters, such as HblAc levels. As illustrated in FIGS. 9E and 9F, an inverse relationship was observed between fatty acid levels, as measured by SERS, and HblAc levels. Conversely, the lipoprotein curve exhibited a trend consistent with the HblAc curve. These findings shed light on the potential utility of SERS inel ucidati ng biochemical correlates of type 2 diabetes mellitus, emphasizing its capacity to discern subtle variations associated with patient demographics and clinical parameters.Determination of Drug Exposure to in vitro Liver Hepatocyte Cell Cultures

[0146] This disclosure demonstrates the efficacy of exosomal surface-enhanced Raman spectroscopy (SERS) coupled with multivariate statistical analysis for classifying hepatic drug dose responses. Surface-enhanced Raman spectroscopy (SERS) intensity differences was quantified in hepatic cultures exposed to four doses (0, 10, 20, and 40 mM) of Acetaminophen. Differences were observed on the day of drug administration and 24 hours later as shown in FIGS. 10A and 10B. Group differences were more apparent post-administration from FIG. 10B, while average spectra were similar across concentrations before administration in FIG. 10A. Principal component analysis (PCA) scores confirmed group distinctions in FIGS. 10C and 10D, indicating early differentiation of exosomal cargo. A regression model quantified dosedependent Raman response changes, with higher residuals observed on the dose-application day, suggesting no initial dose-based response. Conversely, measurements after 24 hours aligned better with the linear model, indicating SERS sensitivity to Acetaminophen dose administration. These findings suggest that this method, alongside proposed advancements, may offer rapid and comprehensive exosomal fingerprinting for sensitive cytotoxicity prediction, complementing, or substituting detailed -omics analyses.Classification and of Environmental Toxins to Ovarian Cancer Cell Lines

[0147] Exposure to per- and poly-fluoroalkyl substances (PFAS) is linked to significant impacts on female reproductive health, including menstrual cycle disruptions, altered timing of menarche and menopause, fertility issues, and heightened risk of ovarian cancer. Assessing these effects via liquid biopsy samples is crucial, yet unmet. This disclosure demonstrates the utility of SERS responses from EVs isolated from ovarian cancer cell cultures (CAOV3 and OVCAR3) in assessing PFAS exposure. FIGS. 11A-11E illustrates progressive alterations in band intensities across various concentrations of PfHpA administration, accompanied by corresponding shifts in band wavenumbers indicative of PFAS exposure. Notably, even though the application of methanol decreases the average Raman intensity dramatically, PFASexposure is further affecting the Raman responses of the cells, proving the vulnerability of the ovarian cancer cell lines to per- and poly-fluoroalkyl substances.Comparison of Different Virus Types

[0148] Preliminary experiments were performed on viruses using single particle imaging. The label-free and reagent-free detection and identification of viruses, i.e., HIV and KSHV viruses are shown in FIG. 12, highlighting the distinctions in spectral profiles between these two virus types.Alzheimer's Disease Prediction by Iron Content Raman Mapping using Histopathological Tissue Sections

[0149] As shown in FIG. 13, the tissue and exosomes are studied with comprehensive, quantitative characterization of iron using Raman spectroscopy and microscopy. With Raman spectroscopy, signal ratios were standardized based on known iron standards and inexpensive, high throughput exosomal oxidation state analysis were performed as depicted in the left panel A of FIG. 13. Then, these peaks were utilized to map the clinical tissue sections to profile the iron deposition inside the tissue. The scan results were collocated with the previously measured X-Ray images to prove the compatibility of the outcomes as depicted in the right panel B of FIG. 13.Comparison of Quiescent and Senescent Origin Vesicles

[0150] The comparison between quiescent and senescent origin oncosomes reveals distinct cellular behaviors and molecular compositions. In FIG. 14, a Raman spectroscopy analysis is presented comparing the biomolecular signatures of vesicles derived from quiescent and senescent cells. The spectral lines demonstrate distinct peaks and intensities, indicative of the molecular differences between the two vesicle types. This comparative analysis can shed light on the vesicles' roles in cellular communication and the aging process within the tumor microenvironment.Example III: Al -SER - Next Generation Pancreatic Cyst Fluid Analysis

[0151] With reference being had to U.S. Patent No. 11,073,511 which is incorporated herein by reference in its entirety, ExoTIC technology, backed by substantial evidence, optimizes EV isolation. However, to be clear, both in this example and at any point in thisapplication referencing ExoTIC isolation device (in this or other examples or description), any other form of EV or particle isolation may be used instead as well as other forms of obtaining EV or particles already in a form suitable for downstream processing. Thus, any reference to ExoTIC is meant to be illustrative only and not restricting or limiting in any way on the process.

[0152] Development of a label-free and rapid diagnostic tool to analyze pancreatic cyst fluids is a need in the clinical management of pancreatic disorders. Extracellular vesicles (EVs) have been investigated in recent years to assess their potential utility in disease management, such as in cancers. The uniqueness of the secreted exosomes and their cargo help towards more comprehensive profiling of the bodily fluids and may provide a new diagnostic parameter. Preliminary results with surface-enhanced Raman (SER) spectroscopy (SERS) show good capability to differentiate pancreatic cyst samples using small EVs. To further improve this study, an advanced approach called AI-SER can be used, which isolates EVs in a microfluidic chip rapidly. This will isolate exosomes using the high-yield ExoTIC tool, profile their chemical fingerprints using surface-enhanced Raman spectroscopy, and image the single exosomes using interferometric scattering imaging (iSCAT) to drastically increase the Raman scanning speed. This may be helpful to the classification of different human pancreatic cyst fluid types as a clinical decision aid.

[0153] There is an unmet need in classification of pancreatic cyst fluid types immediately after they are collected in the clinics. Having this capability may offer significant benefits for patients, such as reduction of unnecessary surgeries, misdiagnoses, and anxiety. Herein, an Al- assisted imaging-based solution is disclosed to detect the types of the pancreatic cyst fluids from clinical samples.

[0154] Pancreatic cancer (PCa) remains a significant health concern, impacting an estimated 64,050 individuals in the United States alone in 2024. With a five-year survival rate of only 10%, early detection is paramount, as most cases are diagnosed in advanced stages. Yet, a rapid, reliable and accurate diagnostic tool for the determination of pancreatic cyst fluid types is not available currently. Exosomes are classes of membrane-bound extracellular vesicles (EVs) that are involved in intercellular communications. Exosome research has gained significant importance, particularly in cancer due to their potential as a new type of bioanalyte. However,the isolation of EVs and their downstream analyses present significant challenges due to their small size and heterogeneity.

[0155] To address these needs, the instant disclosure offers an Artificial Intelligence- assisted platform technology capitalizing on the unique, high-yield exosome isolation tool, interferometric imaging and Surface-Enhanced Raman spectroscopy (AI-SER). The AI-SER can be built to acquire morphological information from single EVs and provide data on their size and content (proteins, lipids, and nucleic acids). This can utilize EVs to be rapidly profiled by interferometric imaging and SERS. By utilizing the AI-SER, chemical and morphological differences can be detected between EVs from different sources such as different cyst types from clinical human samples.

[0156] Even low-volume (100 pL) cyst fluids have been shown to be rich in EVs containing sufficient RNA as well as proteins suitable for downstream multi-omics analyses. Molecular fingerprints of the five pancreatic cyst fluid types tested have been shown to have distinct Raman signatures due to their differences in EV cargo. It has also been demonstrated that source of origin of EVs from different cultured in vitro cells can be determined by using a similar approach. Preliminary results establish a strong indication that the AI-SER approach could provide fundamental biochemical information about their biological properties that allows differentiation between various cyst fluid types.

[0157] Three innovative aspects can be leveraged to develop, test, and demonstrate AI-SER for the detection of pancreatic cyst fluid types. First, Raman scan speed can be increased up to 50-fold compared to the commercial Raman systems with the addition of single EV AI-SER maps using interferometric imaging capability. With this enhancement in scan speed, AI-SER can allow label-free rapid PCF analysis within 30 minutes from a small-volume (100 pL) of clinical sample. Second, AI-SER's spatial resolution is ~30 nm, which allows single-EV detection and SERS characterization. Third, when these two capabilities are combined to analyze different pancreatic cyst fluid types, different malignancy risk levels may be differentiated by using EVs.

[0158] One aim is the development of Single-EV Characterization Platform or "AI-SER" and this involves integration and optimization interferometric-SERS microscope and further involves validation of sorting and microscope performance using selected nanoparticles and EVs.Another aim is the validation of the AI-SER platform using different pancreatic cyst fluid types.This involves (i) isolating, sorting, and characterizing EVs from clinical human PCF samples, (ii) obtaining single exosome images and their coordinates on the surface, (iii) analyzing the SERS signals of single-EVs, and (iv) building and optimizing a predictive model of pancreatic cyst fluid types for clinical applications.

[0159] This can provide a wholly-integrated bioanalytical platform including an interferometric- SERS and an Al analysis pipeline that is trained with known cyst fluids and predicts the unknown PCF type with high accuracy. This can advance this innovative tool towards automation for use in research and diagnostic laboratories. This unique technology offers advancements in the rapidly expanding field of EV research and broad applications that spans over various biomedical fields from metabolic diseases to cancer.

[0160] There is great clinical importance of pancreatic cyst analysis. Pancreatic cancer (PCa) is one of the deadliest cancers. A recent report by the American Cancer Society estimated 64,050 new PCa cases in the US, with an expected mortality rate of 79% in 2023 within the civil and military population. Globally, PCa has a five-year survival rate of 10% because most cases are diagnosed late, making early detection essential. A unique opportunity for early detection of PCa exists in specific high-risk populations, such as patients with precancerous pancreatic cysts. Patients with cysts with a high risk for malignant transformation are optimally managed surgically. Those with lower-risk cysts may be followed within active surveillance programs depending on risk stratification. Pancreatic cysts are common, often identified incidentally during clinical care when abdominal imaging is obtained. There are various etiologies with correlating clinical significance. Some are congenital or acquired and benign in nature with little clinical significance. Some are inflammatory and require active investigation for causes of such inflammation with associated treatment. Some cysts warrant active management and treatment. Accordingly, diagnosing pancreatic cyst type is of paramount clinical significance.

[0161] In terms of tools for pancreatic cysts fluid classification, cyst fluid analysis is an essential tool in managing high-grade dysplasia and improving patient outcomes. The process entails collecting a fluid sample using a fine needle under imaging guidance and analyzing it in a laboratory for abnormal cells, enzymes, and proteins. Molecular analysis is increasingly important in assessing cysts and genetic mutations, as cytology has low sensitivity in detecting types of pancreatic cysts per Table 1 below:Table 1Unfortunately, no accurate tool is available to differentiate between cyst types, and the use of EVs and their cargo as a diagnostic target has yet to be thoroughly studied.

[0162] Also, more generally, there remains great importance in EV research. EVs are membrane-bound vesicles, and their cargo is involved in intercellular communications. EV research has gained significant importance, particularly in cancer, since it has the potential to enable a method that can bypass the lengthy histopathology processes of tissue biopsy and analysis and replace the stress-inducing tests with cheap, disposable, and self-administered tests that potentially give rapid results.

[0163] Various methods have been used for the characterization of EVs, including ultrastructural examination, flow cytometry, size-based techniques, and fluorescence-based methods. Although flow cytometry has proven itself in cell sorting, it may be insufficient in characterizing nanovesicles.

[0164] Among the optical methods, fluorescence labeling, which is commonly used, often requires complex protocols, and the antibodies used may not bind to desired regions in some cases, providing misleading information about surface proteins. On the other hand, Raman spectroscopy, being a label-free and non-destructive vibrational spectroscopic method without sample preparation protocol, rapidly provides information specific to the analyzed sample. It gives chemical fingerprint information about the sample being analyzed. Hence, in recent years, Raman spectroscopy has been increasingly used in exosome studies.

[0165] EVs are heterogeneous populations of nano to micro-sized membrane derived vesicles released by essentially all cells. Cells release large numbers of EVs into their environment, cells' modulating the recipient behavior differently. This diversity indicates that each EV may exert multiple functions. As shown in FIG. 15, AI-SER aims to utilize the predictionability of molecular spectroscopy to accurately classify pancreatic cyst fluid-derived EVs. Single- EV integration is tested to provide deeper insight from the measured EVs, which will also speed up the spectral measurements and prediction ability. This single-EV morphometric / chemometric analysis approach has not been demonstrated for pancreatic cyst fluid analysis before.

[0166] This example utilizes a unique single-EV-sensitive interferometric module that will allow the creation a spatial map of the EVs on the surface. This advance will speed up SERS scanning up to 50 times by avoiding unnecessary measurements from locations without EVs. With this enhancement in scan speed, AI-SER can allow label-free, rapid PCF analysis within 30 minutes from a small volume (100 pL) of clinical sample. AI-SER's spatial resolution is currently down to 30 nm, which allows single-EV detection and SERS characterization. This may be the first demonstration of pancreatic cyst fluid (PCF) analysis with SERS.

[0167] When the capabilities above are combined to analyze different pancreatic cyst fluid types, different malignancy risk levels can be differentiated by using EVs. This is the first time that cyst differentiation analysis is believed to have been performed using the EVs isolated from PCFs.

[0168] Exosome biogenesis begins with endocytosis of the cell membrane leading to multivesicular body formation by the Ras protein family, followed by exosome membrane decoration by the endosomal sorting complex required for transport protein family. Biogenesis of exosomes produces different sizes and cargo. Furthermore, similar findings have also been shown in past studies involving proteomic cargo of different EV sizes. EVs derived from cell cultures, plasma, urine, saliva, and sperm samples previously have been isolated and characterized.

[0169] A single-EV microscope has been built using an interferometric scattering approach that uses a depth scan correlation algorithm to boost the contrast of the images. Note the same pattern of the exosomes in both images, showing that the optical interferometric image was taken exactly from the same particles measured with SEM.

[0170] The classification of the different PCF types has been demonstrated for the first time, as shown in FIGS. 16A through 16D. Fifty-one human cyst fluid samples were obtained from Dr. Walter Park's (MPI) clinical practice. EVs were isolated from these cyst fluids and thesesamples were accurately classified using the EV signatures as Low-Grade IPMN (ILG), Moderate- Grade IPMN (IMG), High-Grade IPMN (IHG), Serous Cystic Adenocarcinoma (SCA), Pseudocyst (PsC), Moderate-Grade MCN (MMG), and Low-Grade MCN (MLG) (see Table 1 for PCF types). After isolating the EVs from 100 pL of a clinical PCF sample, 961 SERS spectra were acquired from each type of cyst fluid and compared their vector-normalized spectra in FIG. 16A. FIG. 16B shows their Principal Component Analysis (PCA) score graph for the first two components that provides a method to explore how the measurements are clustered along different groups. This plot shows that each PCF type clusters distinctly as a class that shows the potential to identify unknown test samples. The SCA cluster presents a sparse distribution, which is expected since this group has a large in-group variance naturally. Furthermore, the high grade IPMN (IHG) group contains subtypes. In FIG. 16C, the accuracies of diverse machine learning models applied to cyst fluid-derived EV data are illustrated. Upon completion of testing, the chosen pretrained classification model can be integrated into the device. When this approach was applied to patient-to-patient variance, the concordance was very with-in each PCF type as shown in the Receiver Operator Characteristics (ROC) curve in FIG. 16D. The area under the curve for the binary classification is 0.98, which shows the potential of this method for differentiating pancreatic cancer and non-pancreatic cancer groups from PCF samples using EVs, hence, its clinical translation potential and utility in identifying clinically relevant and actionable information as a decision guide.

[0171] In machine learning classification, as the training dataset accumulates more data, the model's understanding of the underlying patterns and relationships in the data improves. This can lead to more accurate predictions for new samples, regardless of whether the class distributions converge to Gaussian distributions. Consequently, the platform can predict samples not present in the training dataset without requiring prior sample-specific knowledge, as the model becomes more adept at generalizing from the training data.

[0172] The potential of EV signatures in discerning human pancreatic cyst types can be explored using the AI-SER platform. Interferometric and SERS capabilities can be integrated into a unified instrument and optimizing with nanoparticles / nanovesicles of varied sizes and concentrations. iSCAT can facilitate precise mapping of individual EVs, allowing for targeted interrogation by SERS to obtain their unique chemical fingerprints, thereby significantlyenhancing SERS scanning efficiency. ExoTIC can be employed to isolate EVs from 100 pL of cyst fluid and analyze them on a surface conducive to both interferometric imaging and SERS contrasts. A range of machine learning algorithms can be evaluated on a subset of the dataset to identify the most accurate one, subsequently training the selected algorithm on the entire PCF SERS dataset to enhance its classification accuracy in identifying malignant IPMN. This can result in the development of a robust clinical decision tool aimed at enhancing the management of pancreatic cancer in future studies.Development of Single-EV Characterization Platform: "AI-SER"

[0173] An aim is to develop and optimize an iSCAT-SERS microscope that utilizes EVs. Thus, the content of different distributions of EVs can be analyzed from the same sample and across different pancreatic cyst types. The platform can rapidly sort EVs inside the PCF types and give ultrasensitive chemical content information that can be utilized for accurate malignancy assessments in the clinic.Integration and optimization of iSCAT-SERS microscope

[0174] Initial studies have been conducted to optimize interferometric imaging and SERS to optimize both methods for EV analysis individually. The single-EV sensitivity of iSCAT can be integrated with the molecular specificity of SERS in a single benchtop microscope.

[0175] First, the ExoTIC device can be fabricated using laser etching and assembly methods in the literature and above referenced patent. This tool can then be used to isolate varying concentrations of nanoparticles and EVs from known-to-be benign cyst samples, as part of a series of spike-in experiments. The nanoparticles can be tested, in a field of view of 200 pm X 200 pm area and an image analysis tool can be created to count them and measure their sizes. The software can also create a matrix of the coordinates of the imaged nanoparticles. Next, the iSCAT maps can be utilized to maximize the scanning speed by excluding the points that do not include nanoparticles, which would result in an imaging scan speed up to 6400 times (assuming a field of view of 200 pm X 200 pm scanned with 80X80 measurement points). Then, using this map, the same surface can be scanned with SERS without removing the sample.Validation of sorting and microscope performance using selected nanoparticles and EVs

[0176] The single-EV capability of the iSCAT images can be validated by SEM scans of the same surface. The SEM and iSCAT images can also be used as references for SERS scans. The size distribution of the sorted particles can be validated by NTA.

[0177] A single EV map can be created using interferometric imaging that optimized the scanning routes of Raman scanning to increase the scanning speed up to 6400-fold. The whole operation of sample to decision can take less than 30 minutes including isolation, optical image collection, and spectroscopic characterization with Al learning models.

[0178] In the event the sample is not be sparse enough to allow iSCAT to create a map with a single-EV resolution, the sample can be diluted down and the same iSCAT mapping can be repeated as this process can be very fast. There may be an inconcordance between iSCAT and SERS images due to mechanical vibrations or XYZ positioners. In that case, a software module for auto-alignment might be used.Validation of theAI-SER Platform Using Different Pancreatic Cyst Fluid Types

[0179] Pancreatic cyst fluids are rare and precious samples and their study can have the potential to give early signals for patients showing signs of pancreatic disorders. Pancreatic cyst fluid types are different in content and are rich in EVs. The AI-SER platform can offer a potential solution to this unmet need. As shown in FIGS. 16A through 16D, seven different EV types were classified from n=51 clinical human PCFs. This can be expanded with a larger cohort of 100 clinical samples using the integrated platform for further validation and studies.Isolating, sorting and characterizing EVs from clinical human PCF samples

[0180] One hundred previously collected and deidentified PCF samples from Dr. Walter Park's group clinically are classified as IPMN (high and low grade), MCN (high and low grade), SCA, and PsC. EVs can be isolated from these PCF samples and can be characterized using NTA, flow cytometry (using CD9, CD 63, and CD 81 tetraspanin markers), and TEM.Obtain single exosome images and their coordinates on the surface

[0181] A large benchtop interferometric scattering microscope can predict particles with diameters as small as 30 nm. This technique can be utilized to create single-EV maps to boost SERS scanning speed. A cartesian map of the locations of the particles can be created to be fedinto the scanning algorithm to enhance the resolving power of the SERS. An image analysis software (MATLAB) can be applied to count the EVs using an edge detection algorithm and estimate the size distribution ranges.Analyze the SERS signals ofsingle-EVs

[0182] The current benchtop Raman spectroscopy setup can be integrated with the interferometric microscope to build a powerful single-EV analysis platform, AI-SER. A unified single substrate can be utilized to get single-EV images and SERS signals from the EVs. A Raman signal-to-noise ratio (SNR) larger than 20 dB is targeted, which has been shown to be possible with the SERS surfaces previously. Furthermore, the 200 X 200 pm field of view can be scanned in 5-10 minutes, depending on the sparsity of the EVs. To adjust for the surface-to-surface variance, the measurements can be repeated with three different surfaces for each patient's sample and take the average values.Build and optimize a predictive model of PCF types for clinical applications

[0183] The best-performing machine learning algorithm based on the Raman measurements collected from the EVs can be used to find a model to differentiate cyst types with high accuracy in an n=100 clinical sample cohort. 20% of the patient data can be separated and will not be inserted into the training set to make sure they will be tested without any bias. The predictive model can be expected to give real-time cyst type outcomes for the measured spectra with an accuracy greater than 90% for PCF classification. User-friendly software that provides guidance and reporting for the measurements and predictions can be provided via a simple interface.

[0184] AI-SER can be expected to result in the isolation of 108- 109particles / mL concentration of EVs and these EVs were characterized using NTA, TEM and flow cytometry. All types of pancreatic cyst fluids are expected to be classified accurately (>90%) using SERS data. The accuracy of the classification of the neural network model can be targeted to be higher than 90% for the assessment of the malignancy potential. A completion of prediction from sample to decision guidance can be achieved within 30 minutes and targeted to decrease this further down to 10 minutes.

[0185] The variability, and quality of samples can influence the response measured through Raman spectroscopy, thereby impacting the efficacy of the training set on new, untrained samples. To address this challenge, the model may be augmented with additional parameters that define EV quality, such as protein quantity measured via BCA assay. This rapid test, employing commercial kits and compatible with 96-well plates, can be completed in as little as 10 minutes. Consequently, performing these rapid BCA measurements on new samples not included in the training set becomes important for accurate model performance evaluation and prediction. The test time may not satisfy the clinical needs. In this case, a denoising algorithm can be used that will increase the scanning speed up to 10 times by decreasing the exposure time without compromising the SNR.

[0186] EVs derived from human clinical cyst fluids (n=100) that are known PCF cyst types can be isolated and characterized. A sample size of 100 will provide 90% power at one-sided 5% error to demonstrate a concordance coefficient of 0.9 given the null hypothesis of 0.8. The nonparametric Mann-Whitney-U test can be used with Bonferroni correction if normality may not be assumed for multiple comparisons. Variance inflation factor can assess whether multicollinearity (i.e., significant relationships between predictors) exists. Ridge regression can be performed if multicollinearity exists. This is a statistical technique that stabilizes the coefficients of regression equation by reducing the mean square error of regression coefficients.Example IV: Multimodal IPR Microscopy for Virus Detection

[0187] The rapid transmission of viral infections and the need for effective screening tools for viral diagnosis has become increasingly urgent in light of global pandemics such as COVID- 19. Current virus identification methods lack the required speed and widespread usability, making label-free and reagent-free optical bio-sensors an attractive solution. An interferometric - Plasmonic - Raman (iPR) microscopy platform is proposed that enables rapid, label-free and binding reagent-free detection and characterization of single viruses, revolutionizing virus analysis and diagnostics. By leveraging metasurface-based plasmonic enhancement, interferometric imaging, and Raman spectroscopy, single-particle level detection and characterization of viruses can be achieved. The innovative iPR approach offers the unique advantage of label-free detection and the ability to differentiate between different virus typeswithout the use of antibodies. The platform utilizes metasurfaces for surface patterns, enabling specific plasmonic modes activation, and incorporates interferometric imaging to visualize viruses at the single particle level. Additionally, Raman spectroscopy provides molecule-specific information, significantly enhancing sensitivity. Through the project, virus analysis and advance diagnostic applications can be revolutionized by providing both structural and chemical information simultaneously. In this project, two specific aims are addressed: (1) the development of a multimodal IPR microscopy platform using metasurfaces, and (2) the validation of the platform's performance with different types of viruses. Through comprehensive validation experiments, model library generation and neural network training the platform's sensitivity, specificity, and reliability can be assessed, and a library of Raman signatures built for different virus types as a first attempt to create such a library. This project holds the potential to transform virus detection and characterization, aiding in effective management and control of viral infections and it has broad applications in vesicle research and vesicle-based drug delivery science, and applications in disease early detection and monitoring.

[0188] There is an unmet need in public health for rapid and sensitive screening tools to diagnose viral infections. Current virus identification methods lack the required speed and suitability for widespread use. The disclosed Interferometric-Plasmonic-Raman (IPR) platform offers a portable, high-throughput, and label-free solution for detecting and differentiating various types of viruses without the use of antibodies, revolutionizing virus analysis and enabling advanced diagnostic applications and other applications in vesicle-based drug delivery science.Clinical Significance

[0189] The global COVID-19 pandemic caused by the SARS-CoV-2 has highlighted the need for rapid and efficient detection of viruses in clinical and point-of-care (POC) settings. Various laboratory-based diagnostic techniques, including paper-based lateral flow assays (LFAs), reverse-transcription polymerase chain reaction (RT-PCR), enzyme-linked immunosorbent assays (ELISA), and isothermal amplification have been used for virus detection. PCR is considered the gold standard technique for viral molecular diagnostics due to its high sensitivity in detecting viral genomes. Although there are portable PCR devices available commercially,these are still not affordable and readily available enough for wide use. Thus, the requirement for centralized PCR facilities limits this technique's use in remote or field locations. ELISA techniques, while useful, do not provide information on the level of viral genome replication or viral load. Furthermore, existing diagnostic approaches, despite their utility, require specific knowledge about the pathogen (its genome and proteins) and the availability of precise reagents (antibodies, primers, probes), which can delay critical response times in managing outbreaks and the rapid emergence of viral variants, as occurred in during the COVID-19 pandemic. Hence, there is a need for label-free, rapid, reusable, and sensitive detection technologies suitable for widespread use.Technical Significance

[0190] Metasurfaces are nanostructured interfaces consisting of thin layers of plasmonic or dielectric materials. These engineered structures manipulate light by modifying its phase, amplitude, and polarization, extending the photonic behavior of natural materials.Metasurfaces have unique optical phenomena such as optical magnetism, negative reflection, electromagnetic-induced transparency, bi-anisotropy, and Fano resonance. They enable precise optical control by inducing rapid changes in phase and light polarization, useful for applications in flat lenses, and photonics. They have shown promise in biomolecular sensing due to their ability to achieve high sensitivity through spectral shifts in response to molecular oscillations and perturbations, as prior work has shown. Recently, plasmonic metasurfaces with Fano resonances and asymmetric spectral lines, originating from the response of isolated metal atoms, have been described. However, current Fano-resonant metasurfaces require complex nanofabrication techniques and expensive equipment like Fourier transform infrared (FTIR) spectrometers, limiting their application at POC or primary care settings.

[0191] Interferometric imaging and Raman spectroscopy have been shown and developed earlier as well-established optical techniques. However, their integration via a unique resonance matching metasurface design as proposed here to specifically boost the interferometric signal simultaneously with the Raman excitation create a whole new avenue of nano / biosensing applications with high specificity. The disclosed surface design and integratedoptical tool can lead to innovative applications in virus imaging, classification and detection that were otherwise previously not achievable.

[0192] Herein, a microscopy platform (iPR) is proposed that combines interferometry, plasmonics and Raman spectroscopy for imaging and characterization of biological nanosized particles (bio / nanoparticles). As best known, such a platform does not currently exist. Further, metasurfaces are designed and fabricated, which metasurfaces induce unique optical Fano responses to enhance interferometric imaging and Raman spectroscopy, enabling the visualization, localization, and chemical composition analysis of viruses. This is the first platform that can provide both structural and chemical information of bio / nanoparticles without using any labels. Furthermore, by incorporating a machine learning (ML) approach into the Raman modality, the disclosed iPR microscope offers rapid and sensitive detection and identification capabilities at the nanoscale. Unlike high-end imaging platforms like electron microscopy, atomic force microscopy, confocal Raman or CARS microscopy which typically cost over $500K, the disclosed platform is designed to be more affordable (<$10K). The disclosed iPR platform provides game-changing capabilities, enabling rapid detection and identification of different viruses and nanovesicles for biomedical use.Clinical Utility and Use-Cases

[0193] IPR serves as a versatile platform technology with numerous applications that are uniquely enabling and not possible otherwise. One such use case involves the detection of both infectious and non-infectious types of Hepatitis B Virus (HBV). Mature or infectious HBV virions contain the complete viral genome and are fully assembled with viral envelope proteins. These particles can infect hepatocytes (liver cells) and replicate within the host, leading to the generation of new viral particles. Conversely, empty HBV particles, also known as defective or non-infectious virions, lack the viral genome but still retain viral envelope proteins. However, due to the absence of genetic material, they are incapable of infecting cells or propagating further.

[0194] This distinction is particularly significant as serological diagnostic tools struggle to differentiate between active and non-active chronic infections. The disclosed approach utilizes Raman signals, which are sensitive to the presence of nucleic acids in fully infectious HBVparticles, enabling effective differentiation between these two types of virus vesicles. In another use case of human coronaviruses, the correlation between amino acid composition is explored, particularly in the receptor binding domain, and the infectivity and transmissibility of various coronavirus strains is also explored. Raman chemical fingerprinting capability of the iPR platform presents notable advantages over extensive, costly, time-consuming, and intricate proteomic analysis, resulting in the rapid development of diagnostics agnostic to prior knowledge of the protein sequence.Single Particle High Resolution Imaging of Biological Nanoparticles, i.e., Viruses and Nanovesicles

[0195] Direct optical imaging of viruses without labels is challenging because of their subwavelength size. Above, an interferometric microscopy method was described for visualizing and detecting bio / nanoparticles such as viruses and vesicles at a single nanoparticle level. Unlike fluorescence-based methods that require labeling and may affect the particle's functionality, this label-free approach utilizes de-focused images, which is termed as Depth Scanning Correlation (DSC), to enhance nanoparticle visibility in interferometric microscopy, improving on the detectable size limit. The experiments show successful detection of single nanovesicles ~30 nm. This high-throughput technique provides a powerful way of sensing and characterization of biological particles within the range of a few tens to a few hundred nanometers, where conventional methods fall short. Utilizing widefield interferometric imaging over a 300 x 200 pm2field and employing raster scanning, an entire 2 x 2 cm2sensor area can be imaged. This technique can process a total sample volume of 200 pL in a single measurement. The method involves depositing a drop of sample onto the substrate and allowing it to dry for immobilization and subsequent detection. The preliminary results show that 103particles / mL can be detected processing a 200 pL sample volume with an SNR>10 as illustrated in FIG. 17A. Therefore, an aim is to further enhance our detection sensitivity and specificity by further introducing the Raman signal. Interferometric initial image of HBV is given in FIG. 17B.Label-free, Spectral Identification of Vesicles

[0196] Extracellular vesicles (EVs) are bio / nanoparticles that are similar in size, structure, and composition to viruses. A novel approach was employed to classify these bio / nanoparticlesusing surface-enhanced Raman spectroscopy (SERS) via machine learning (ML) algorithms. While label-free imaging interferometry and plasmonics provide structural information, spectral analysis facilitates identification. The ML-assisted SERS method successfully differentiated healthy versus cancer bio / nanoparticles.Label-free, Spectral Identification of Viruses

[0197] Building upon these principles tested on EVs, preliminary experiments were performed on viruses using the single particle imaging as discussed above. The label-free and reagent-free detection and identification of viruses was demonstrated, such as for example, HIV and KSHV viruses as in FIG. 12, highlighting the distinctions in spectral profiles between these two virus types.Development and Optimization of a Cost-Effective Plasmonic Metasurface Platform

[0198] An objective herein is to leverage the large-scale active area and uniform surface patterns present on metasurfaces to develop a cost-effective method for biosensing applications. The goal is to excite asymmetric plasmonic modes and enable tunable optical Fano resonance allowing for the detection of multiple targets using the visible wavelength range. Through efficient layer-by-layer metal deposition and surface engineering plasmonic metasurfaces can be fabricated. Multilayered and periodic structures are designed and simulated (COMSOL) and then are fabricated and tested as detailed below.Simulations

[0199] The simulation model is based on the three-dimensional geometry of a grating structure with specific dimensions (in the range of width = 300-500 nm, height = 5-40 nm, and period = 600-800 nm). By leveraging the planar symmetry in the x-y plane and the periodic nature of grating structure, a single unit cell is simulated, considering periodic boundary conditions, an input port for light excitation, and an output port for light measurement. Similar studies in the literature have previously reported the simulation of single unit cell structures in periodic metasurfaces. The grating structure can be fabricated using polycarbonate (with a refractive index of 1.58) coated with different types of deposition metals that are available such as titanium (Ti), silver (Ag), and gold (Au).Design and Development of Interferometric Imaging Module for Virus Detection

[0200] Plasmonic detection offers resonant detection with a high signal-to-noise ratio (SNR), providing valuable information about thickness of the different biolayers, yet single particle detection is not achievable. To address this need interferometry is employed, a technique that measures minute phase differences between light waves. Interferometric imaging utilizes the interference between the scattering signal from the nanosized particle and a known reference light. The total intensity ( / tof) is given by:where Esand Errepresent the fields due to scattering from the particle and reference field, respectively, and 0 is the phase difference between these two fields. An on axis interferometric microscopy configuration is to be utilized where reference light will be collected directly from the sample substrate. Considering that the scattering field is small compared to the wavelength of the light, and the background light is a static term, the contrast due to a single virus is essentially determined by the ratio Es / Ef cos(0). In order to improve the image contrast, first Esis increased, which depends on the interaction between the incoming light and the viral particle, by engineering the surface to enhance plasmonic effects. The scattered field, which directly correlates with the size of the smallest detectable particle in the system, scales with the volume of the particle. Based on the initial simulations, the goal is to achieve a surface field enhancement of at least 50, which would result in a four-fold improvement in the minimum detection limit. Consequently, an aim is to demonstrate the detection of single viruses and other biological particles with the smallest feature size of 20 nm. Second, the optimization of Erfacilitates achieving maximum contrast while maintaining the interferometric signal and this can be accomplished by engineering the Fano resonance curve. Since the resonance frequency shift is a function of the thickness of the metal layers, the phase difference can be fine-tuned by nanoengineering the layer thicknesses.Design and Development of Raman Module for Virus Identification and Subtyping

[0201] By employing interferometric detection, nanometric particles are accurately localized, which sets the stage for the Raman spectroscopy to subsequently identify the viralparticles. While the modules developed in the previous steps provide structural information about the viruses (for example, existence of surface projections / spikes of the virions, shape, size), they do not offer spectral chemical information fingerprints for identification. To address this, a Raman module is integrated into the system which comprises a laser source and an inexpensive handheld spectrometer. By modulating the gold layer thickness and pattern during sample substrate preparation, SERS is leveraged for enhanced signal detection. Raman spectra of each virus type should be unique due to the DNA, RNA composition, chemical signatures (for example, lipids, carbohydrates), viral proteins and other properties. To explore this, SERS datasets are collected from various virus samples. Depending on the size of the dataset, a ML algorithm is selected to classify the samples. The random forest classifier is effective for smaller datasets with known labels, while the neural network performed better for larger, unsupervised datasets. The details of the algorithms are given below. Finally, all the software modules can be combined into a user-friendly smart-device app to control the submodules of the iPR. The software can include instrument communication (motor, laser, and camera), spectra collection, smart spectrum preprocessing, and real-time ML prediction using datasets trained on known samples and can report the virus ID and counts.Expected Outcomes

[0202] An objective is to develop a compact fully integrated multimodal microscopy prototype which utilizes a metasurface sample substrate for plasmonic enhancement, interferometric label and biorecognition reagent free imaging and Raman spectroscopy. A label- free detection limit of 20 nm for single biological nanoparticles and a spectral resolution of approximately 4 cm1, accompanied by a SNR >20 dB is expected. An Al-based classification algorithm can be employed to identify individual viruses solely based on interferometric size, shape and morphology information coupled by the Raman fingerprint signal, eliminating the need for any labels or recognition reagents.

[0203] To ensure the single particle level detection of viruses, various optical techniques (tomographic illumination / collection, polarization enhancement etc.) can be adapted to the interferometric imaging module. It is contemplated that various periodicity and film thickness geometries may be explored to improve SERS signal uniformity and quality if poor output isobserved. The plasmonic substrates are tested as an initial development basis due to their unique nano-architectures, which are naturally instrumental in plasmonic resonance enhancement. This choice is driven by the substrates' ability to offer a scalable, cost-effective method for inducing strong plasmonic enhancements. Additionally, alternative nanofabrication techniques, such as nanoimprint lithography, can be adapted. These methods allow further customization of the nanostructures, increasing the platform's sensitivity and specificity for virus detection.Validation of the platform using distinct viruses

[0204] Rapid and reliable identification of nano-bioparticles, i.e., viruses, is valuable for multiple biomedical applications, including disease management and containment of infectious agents. The disclosed iPR platform offers unique capabilities by combining three imaging modalities: interferometry for morphology and structure, plasmonics for field enhancement, and Raman spectroscopy / imaging for chemical composition, nanoscale mapping and virus identification. This integrated approach, coupled with ML-based classification, modeling, and library building, enables rapid, label-free, and capture reagent-free detection of unknown entities, for example viruses, with high reliability. iPR imaging and spectroscopy of nanovesicles of varying size ranges and nucleic-acid content

[0205] To assess the performance of the system, nanovesicles are utilized since they are available readily in assorted sizes, in varying nucleic acid and protein content, in large quantities. This assessment begins by depositing 10 pL dispersions of nanovesicles, ranging in size from 30 to 200 nm, in serial dilutions. Subsequently, controlled mixtures of nanovesicles are created with varying sizes and they are classified and counted. The iPR then provides interferometric, plasmonic, and Raman information in a sequential manner and reports virus identification and concentration. In the preliminary studies, lipid-based bio / nanovesicles were engineered and fabricated with specific size range and protein cargo comparable to viruses. Here, protein content is analyzed using Raman as shown in FIG. 18A. As shown in FIG. 18B, principal component analysis (PCA) clearly exhibited the discrimination capability of our method across different nanoparticles. Here, the same approach can be leveraged for optimization studies of the biosensor with known vesicle types and cargo.iPR imaging and spectroscopy of different virus types of varying shapes (spherical, pleomorphic), sizes (40-160 nm) and viral genetic content of RNA and DNA

[0206] iPR imaging and spectroscopy are performed on serially diluted virus samples for imaging and chemical fingerprinting, as described above. The iPR is used for imaging and raster scanning to determine viral particle counts per surface area of the substrate. This measurement is converted to concentrations per ml. Subsequently, proportion-controlled mixtures of the virus types are created in simulated clinical samples and ML-based modeling and Al-assisted predictive analysis are employed for virus type classification. All original virus samples are prepared using sterile, prefiltered PBS, distilled water using heat- and gamma irradiation- inactivated viral stock solutions. The viruses used in this study are obtained from the BEI Resource Center (NIH-NIAID). To fully assess the utility of iPR platform, a variety of human viruses with distinct sizes and genomic content are examined as given in Table 2.An aim is to visualize the morphological features of these particles at the single-entity level and obtain their concentration. Additionally, the system can provide Raman fingerprints that enable subtyping of different strains within a specific virus family, for example coronaviruses. The identification and counting process can be enhanced by ML algorithms with a comprehensive spectral database, designed to improve the specificity and accuracy of our system even subtle variations in Raman spectra that allow precise subtyping of viruses.Validation of iPR in relevant clinical sample types

[0207] The sample types mentioned above are studied, including spiked human plasma, simulated saliva, and viral collection and transport media, to assess the practical applicability of the iPR in field-like conditions. To establish a comprehensive imaging and spectroscopic library, structural and chemical fingerprints of these samples are compiled within relevant field and medically oriented contexts. Standard ML techniques, such as PCA, random forests, linear / polynomial discriminant analysis, and support vector machines, can be employed forclassification purposes, with relevant features extracted to construct as in our previous published models.

[0208] The device performance measures include a compact size (50 x 30 x 30 cm3) and lightweight design (<2 lb). It will also be capable of differentiating between bio / nanovesicles from mixtures (30-200 nm size range nanovesicles and four virus types) and within 30 minute assay time. The iPR platform can also have the ability to classify virus subtypes. It is believed this can result in an affordable, disposable metasurface (<$5), accompanied by a portable iPR microscope (<$10K).

[0209] It is possible that this model, originally trained on one type of virus, may not accurately classify other viruses. In this case, the model may be retrained using viruses and tested on additional samples not included in the original model. If the model remains inaccurate, the variance within the virus samples can be re-evaluated, the virus samples can be classified into subgroups, and the model retrained accordingly.

[0210] Thus, an iPR platform is contemplated that can enable label-free imaging and spectral characterization of single viruses without recognition reagents. The automation of this platform for virus detection may be initiated in clinical work-flow for future primary care or POC use through the following steps: (i) development of sample isolation cartridge, (ii) integration of the isolation system (iii) development of a fully automated fluidic-pump system to control. Broad potential applications of this tool for other bio / nanoparticles in liquid biopsy, make the iPR a promising technology in broad fields of research including disease diagnosis and management.Example V: Specialized Consumables

[0211] This technology has the potential to be integrated to advanced technologies spanning particle science, lipid nanoparticles, and sophisticated surface methodologies that presents a promising frontier. The fusion of interferometric imaging and Raman spectroscopy across diverse substrates holds immense potential for multifaceted applications. This convergence paves the way for the development of groundbreaking consumables tailored to various scientific and other application domains.

[0212] One avenue lies in the creation of specialized consumables designed for bioanalytical purposes. Leveraging the combined power of interferometric and Ramantechniques, consumables could be engineered to facilitate precise molecular analysis at the interface of biological systems and synthetic materials. These consumables can manifest as innovative platforms capable of interrogating intricate biomolecular structures with unprecedented sensitivity and specificity. The incorporation of lectin and aptamer arrays into these consumables represents a significant advancement in biomolecular recognition and analysis. By harnessing the affinity of lectins for carbohydrates and the specificity of aptamers for diverse molecular targets, these consumables may enable precise profiling of complex biological samples with unparalleled resolution and efficiency. A notable innovation lies in the augmentation of Raman collection through target-binder-coated microwells equipped with microlens arrays. Herein, the target-binder abbreviation refers to any target recognition element or elements that can be immobilized, printed, conjugated or functionalized on the prepared substrate surface. These recognition agents could be aptamers, single or double stranded nucleic acid molecules, locked or modified nucleic acids, peptide nucleic acids, lectins, proteins, peptides, antibodies and their derivatives such as affibodies, nanobodies and others. This integration not only enhances the sensitivity of Raman spectroscopy but also facilitates high-throughput screening and analysis, revolutionizing the landscape of biomolecular assays and diagnostics.

[0213] Moreover, the incorporation of additional modalities, including infrared spectroscopy, can complement the label-free molecular information obtained from the microscope. Other than this, addition of ultrashort pulses to interrogate the higher-harmonic responses from the biological samples with a spatial resolution less than diffraction limit will elucidate the inner and secondary structure of proteins.

[0214] With respect to the sample substrate, the sample substrate is a part of a disposable that can integrate multiple modalities of imaging described in the application. The substrate can be an open substrate or in well format or covered with fluidic channels or reservoirs on top to allow the flow and transfer and recovery of the samples and introduction of different types of reagents or chemicals to the sensor to allow multiple types of multimodal measurements in parallel or in series.

[0215] One of the aspects of the disclosure is to create a multimodal imaging system [interferometric - Raman and fluorescence (optional)]. Such a system would employ a uniquesample substrate for high performance. Interferometric microscopy demands a highly flat substrate to acquire sensitive images, which typically limits the use of nanopatterned substrates optimized for Surface Enhanced Raman Scattering (SERS). However, the utilization of interferometric thin films in Raman microscopy is shown to significantly enhance the acquired signal.

[0216] The surfaces can also involve metal, organic or non-organic nanoparticles or film coatings to allow for various modalities for sensing and imaging. The nanoparticles may potentially be on the top or at the bottom or supported in intermediate layers. The structure can be specialized and / or adapted so that both Raman and interferometric to work together.

[0217] Referring back to FIG. 5, FIG. 5 illustrates the enhanced performance of Raman spectroscopy when using a SiO2thin film substrate. Comparing the Raman spectra of polystyrene (PS) nano particles on standard glass to those on a glass slide coated with an SiO2thin film, a substantial enhancement of the Raman signal on the SiO2-coated substrate is observed. This enhancement is particularly notable in the characteristic peaks of PS, which are markedly more pronounced on the SiO2film. Further enhancements are achieved by coating a very thin layer of gold film on top of the SiO2thin film (depicted in the bottom left layer structure). This configuration enhances the Raman signal, facilitating the detection of low- abundance molecules. This design strategy can be used in developing the substrate for this multimodal imaging system, leveraging the unique properties of each layer to enhance overall performance across different imaging modalities.Example VI: Q-AIR

[0218] The ability to detect and characterize individual biomolecules, biological nanoparticles, and synthetic nanocarriers with high sensitivity and specificity is essential for biomedical research, diagnostics, and nanomedicine. However, existing analytical techniques remain fragmented, with each method offering only partial insights while suffering from inherent trade-offs that limit their applicability to real-time, single-particle characterization. Current approaches such as biochemical and mass-based methods such as mass spectrometry (MS) and enzyme-linked immunosorbent assays (ELISA) offer molecular identification but require extensive sample preparation, labeling, or destructive processing, making them unsuitable for high-throughput, label-free applications. Optical and microscopic techniquessuch as fluorescence microscopy, cryo-electron microscopy (cryo-EM), and Raman spectroscopy provide high sensitivity or chemical specificity but often demand labeling, intensive sample processing, or struggle with inherently weak signals, limiting their use for real-time, live imaging. For example, Raman spectroscopy offers molecular fingerprinting but suffers from weak intrinsic signals, requiring high-intensity laser illumination or enhancement techniques such as surface-enhanced Raman spectroscopy (SERS), which depends on complex nanofabrication and lacks reproducibility. Interferometric and scattering-based approaches such as interferometric scattering microscopy (iSCAT) or mass photometry enable highly sensitive, label-free detection but lack chemical specificity, preventing detailed molecular analysis.

[0219] These fundamental limitations create a critical gap in biomedical research, where no single platform can simultaneously provide real-time, high-sensitivity, label-free detection with molecular specificity at the single-particle level. Addressing this unmet need requires a multimodal approach that integrates the strengths of existing techniques while overcoming their limitations. A Quantum-Acoustic-Enhanced Interferometric Raman ("Q-AIR") Microscope is provided herein, a fundamentally new imaging platform that unifies (i) quantum-enhanced SERS to reduce quantum noise and improve the signal-to-noise ratio (SNR), (ii) surface acoustic wave (SAW)-based nanoparticle trapping to confine and dynamically manipulate particles within enhanced optical fields, increasing Raman interaction times without requiring plasmonic substrates, and (iii) interferometric imaging to provide simultaneous real-time size, shape, and molecular characterization. By leveraging these advances, Q-AIR aims to establish an innovative, fully optical, label-free single-particle characterization technique that will significantly impact biomedical research, nanomedicine, and molecular diagnostics.

[0220] A first aim is to establish the feasibility of quantum-enhanced interferometric Raman spectroscopy for single-particle detection. This involves developing a quantum-enhanced Raman excitation system including implementing a quantum-enhanced Raman excitation scheme using squeezed light and quantum correlations to increase Raman signal-to-noise ratio while minimizing photodamage. This further involves integrating interferometric microscopy with Raman spectroscopy including designing and constructing a multimodal imaging platform that combines interferometric microscopy with Raman spectroscopy, enabling simultaneouslabel-free molecular fingerprinting and high-sensitivity detection at the single-particle level. This aim finally involves validating feasibility using well-characterized nanoparticles and biomolecules including testing the integrated system using dielectric nanoparticles (polystyrene, silica, and so forth), as well as biologically relevant vesicles (extracellular vesicles and virus, and proteins), to benchmark performance against conventional Raman and interferometric techniques.

[0221] A second aim is to implement SAW trapping to enhance Raman and Interferometric sensitivity. This involves designing and fabricating a SAW-based trapping system for Q-AIR integration including engineering and optimizing a surface acoustic wave (SAW) trapping system to enable label-free, dynamic particle confinement for coupling with quantum- enhanced interferometric Raman spectroscopy. This further involves quantifying SAW effects on interferometric Raman enhancement including measuring the impact of SAW-induced confinement on Raman and interferometric signals, optimizing acoustic parameters to maximize sensitivity. Finally, this second aim involves integrating SAW trapping with the quantum-enhanced system and evaluating real-time performance including integrating SAW trapping into the Q-AIR platform and assessing its ability to provide real-time, high-throughput, label-free analysis of single nanoparticles and biomolecules.

[0222] The ability to detect and characterize individual biomolecules, biological nanoparticles, and synthetic nanocarriers with high sensitivity and specificity is crucial for advancing biomedical research, molecular diagnostics, and nanomedicine. Current analytical techniques remain fragmented, with each approach offering only partial insights while suffering from inherent trade-offs that limit their applicability for real-time, single-particle characterization. The disclosed Quantum-Acoustic-Enhanced Interferometric Raman (Q-AIR) Microscope overcomes these limitations by integrating quantum-enhanced Raman excitation, surface acoustic wave (SAW)-based nanoparticle trapping, and interferometric imaging into a single, multimodal platform. The successful development of Q-AIR significantly advances the state-of-the-art in label-free single-particle detection and characterization, enabling new capabilities in biomedical research and clinical applications.

[0223] There are many current limitations in the field. For example, despite advancements in molecular and biological nanoparticle characterization (e.g., vesicles, viruses, lipoproteins).existing analytical techniques remain fragmented, each offering only partial insights while suffering from trade-offs that limit their real-time, single-particle applicability. These limitations hinder key applications such as rapid pathogen detection, extracellular vesicle (EV) characterization, and nanoparticle-based drug delivery assessment. Biochemical and massbased methods like mass spectrometry (MS) and enzyme-linked immunosorbent assays (ELISA) provide molecular identification but require extensive sample preparation, labeling, and destructive processing, making them unsuitable for high-throughput, label-free applications. Their complex workflows further limit scalability for live-cell or dynamic systems. Optical and microscopic techniques, including fluorescence microscopy, cryo-electron microscopy (cryo- EM), and Raman spectroscopy, offer high sensitivity or chemical specificity but often require labeling, intensive sample processing, or struggle with weak signals, restricting their use in realtime live imaging. Fluorescence-based methods rely on tags that can alter biological processes, while cryo-EM requires freezing, preventing dynamic analysis. Raman spectroscopy, despite its molecular fingerprinting capability, suffers from weak signals requiring high-intensity laser illumination or enhancement techniques such as surface-enhanced Raman spectroscopy (SERS). However, SERS depends on complex, inconsistent nanofabrication, making reproducibility a challenge for standardized biomedical applications. Interferometric and scattering-based approaches like interferometric scattering microscopy (iSCAT) and mass photometry enable highly sensitive, label-free detection by exploiting light scattering. However, they lack chemical specificity, making it difficult to differentiate biomolecules with similar optical properties, limiting their utility in molecular identification.

[0224] Q-AIR introduces a novel approach to single-particle characterization, overcoming key methodological barriers. It establishes the feasibility of integrating quantum-enhanced Raman excitation, acoustic trapping, and interferometric microscopy for the first time.

[0225] The Quantum-Acoustic-Enhanced Interferometric Raman (Q-AIR) Microscope introduces a paradigm-shifting, high-risk technology for single-particle characterization that integrates three cutting edge approaches never before combined: quantum-enhanced Raman spectroscopy, acoustic trapping, and interferometric microscopy. This label-free multimodal system directly addresses bottlenecks in molecular analysis that have hindered progress across multiple biomedical domains. As to quantum-enhanced Raman excitation, this is believed torepresent the first-ever application of squeezed light to enhance Raman signal-to-noise ratio in a SERS framework, achieving superior sensitivity while eliminating the need for destructive laser intensities or variable nanostructures that plague current approaches. As to surface acoustic wave (SAW)-based trapping, this implements a gentle, non-invasive, label-free single-particle isolation technique that dramatically extends viable integration time while preserving sample integrity, overcoming fundamental limitations of both optical and mechanical trapping methods. With respect to interferometric imaging, this delivers unprecedented sub-nanometer sensitivity for simultaneous, real-time measurement of particle morphology, refractive properties, and molecular fingerprinting, capabilities currently requiring multiple separate instruments. This also provides transformative multimodal integration, creating the first platform to synergistically combine optical, mechanical, and quantum techniques in a unified system, enabling a step-change in both throughput and resolution for single-particle characterization.

[0226] To achieve single-molecule sensitivity in label-free molecular characterization, two complementary techniques are developed and integrated including (i) quantum-enhanced Raman excitation, which leverages squeezed light to suppress quantum noise and enhance Raman signals without requiring high-intensity lasers or traditional SERS substrates, and (ii) surface acoustic wave (SAW) trapping, which provides label-free, non-invasive single-particle confinement, extending interaction times and improving signal acquisition. While these techniques are designed to work synergistically in the Q-AIR Microscope, each can function independently, offering new pathways for high-sensitivity molecular analysis. The integration of these technologies into a multimodal platform represents a fundamentally new approach to real-time, high-throughput, label-free single-particle detection, addressing longstanding challenges in biomedical research and molecular diagnostics.

[0227] FIG. 19 illustrates the concept of the Quantum-Acoustic-Enhanced Interferometric Raman (Q-AIR) Microscope. Q-AIR integrates quantum-enhanced Raman spectroscopy, surface acoustic wave (SAW) trapping, and interferometric imaging for high-sensitivity, label-free single-particle characterization. The bottom center of FIG. 19 shows nanoparticles are confined using SAW trapping on a substrate, the upper left of FIG. 19 shows squeezed light suppresses quantum noise, reducing shot-noise limitations, the top center of FIG. 19 shows high-resolutionimaging enables real-time nanoparticle detection, and the top right of FIG. 19 shows noise suppression enhances signal detection, improving the signal-to-noise ratio (SNR).

[0228] This approach verifies the feasibility of quantum-enhanced interferometric Raman Spectroscopy for single-particle detection. While squeezed light has been applied in quantum optics and metrology, it has not been used to enhance SERS sensitivity. Introducing quantum squeezing to SERS addresses key challenges such as signal instability, reliance on complex nanofabrication, and single-particle detection difficulties. Integrating quantum-enhanced SERS with interferometry adds further innovation by combining ultra-sensitive particle size and shape measurements with molecular specificity. This synergy between quantum photonics and advanced microscopy enables high-precision, label-free molecular characterization at the single-particle level.

[0229] This approach involved the development of a quantum-enhanced Raman excitation system. Raman spectroscopy is a powerful tool for molecular fingerprinting but suffers from weak signal strength and high background noise, limiting single-particle detection. Traditional approaches rely on high-intensity lasers, increasing photodamage risks, especially in biological samples. To overcome these limitations, a quantum-enhanced Raman excitation system leverages squeezed light and quantum correlations to improve SNR while reducing laser power. By using non-classical states of light, quantum fluctuations can be suppressed, shot noise can be lowered, and Raman signal acquisition can be enhanced.

[0230] The theoretical framework underlying this idea is adapted from quantum metrology, based on squeezed light states, where noise in one quadrature is reduced below the shot-noise limit. FIG. 20 illustrates the concept of quantum squeezing in phase space, which plays a crucial role in the approach to quantum-enhanced Raman excitation. In classical optics, a laser beam in a coherent state exhibits fundamental quantum noise, known as shot noise, that is evenly distributed between two quadratures as shown in FIG. 19: amplitude quadrature, which governs intensity fluctuations, and phase quadrature, which governs phase fluctuations. This balanced noise distribution is represented by the circular region in FIG. 19 and sets the standard quantum limit (SQL) for detection sensitivity in any optical detection technique. Quantum squeezing modifies this noise distribution by reducing uncertainty in one quadrature at the expense of increased uncertainty in the other. The elliptical region in FIG. 19 represents asqueezed vacuum or squeezed coherent state, where noise is compressed in the amplitude quadrature and expanded in the phase quadrature. By squeezing the amplitude quadrature, intensity fluctuations are suppressed, leading to a higher SNR in Raman detection without increasing the excitation laser power. However, since the total quantum uncertainty must remain constant, squeezing in results in increased noise in the phase quadrature. Importantly, this additional phase noise will not affect Raman signal detection, making this trade-off beneficial for Raman spectroscopy. The impact of this noise suppression is quantified by the squeezing parameter r, where an increase in r leads to exponential improvement in Raman signal strength. For instance, with a conservative r = 1.5, the expected Raman signal enhancement follows IR(C,)= IR- e2r, corresponding to approximately a 4.5-fold increase in Raman signal intensity without increasing laser power. This quantum enhancement not only improves signal detection but also minimizes photodamage, making it particularly suitable for studying fragile biological samples such as extracellular vesicles (EVs) and protein aggregates.

[0231] The experimental setup includes of three main components as shown in FIG. 21: a spontaneous parametric down-conversion (SPDC)-based squeezed light source, a balanced homodyne detection system, and a Raman excitation and detection module. Instead of using a commercial optical parametric oscillator (OPO), an SPDC-based approach was employed to generate squeezed light in a more cost-effective manner. A 405 nm continuous-wave (CW) pump laser is used to drive spontaneous parametric down-conversion in a periodically poled potassium titanyl phosphate (PPKTP) or lithium niobate (PPLN) crystal. The nonlinear crystal is temperature-stabilized to ensure phase-matching for efficient down-conversion, generating squeezed vacuum states. The phase of the squeezed light is actively stabilized using a local oscillator (LO) beam derived from the same pump laser. To enhance Raman signal detection, a balanced homodyne detection (BHD) system is employed to suppress shot noise and improve the SNR. The squeezed vacuum field is interfered with the Raman-scattered light at a 50:50 beam splitter, allowing quantum noise suppression in the amplitude quadrature. A balanced homodyne detector (BHD) using high-efficiency avalanche photodiodes (for example, Thorlabs APD430A) records the squeezed-state interference, while a phase-locking feedback loop stabilizes the LO phase to maintain optimal squeezing conditions. This noise suppression mechanism enables high-fidelity Raman acquisition without increasing the laser power. ForRaman excitation and signal collection, a 785 nm laser is focused onto the sample via a high numerical aperture (NA = 1.4) objective, optimized for SERS. The Raman-scattered light is collected through the same high-NA objective and passed through a notch filter to remove Rayleigh scattering. A high-throughput spectrometer coupled with an electron-multiplying charge-coupled device (EMCCD) detector records the Raman spectra, allowing for high- sensitivity detection. The collected signal is analyzed both with and without quantum enhancement to quantify improvements in SNR, photodamage threshold, and Raman intensity. This quantum-enhanced SERS using SPDC-based squeezed light provides a cost-effective alternative to commercial OPO-based systems while maintaining significant improvements in Raman signal acquisition. To validate this approach, benchmarking experiments can compare quantum-enhanced Raman excitation to conventional coherent laser excitation.

[0232] To achieve high-sensitivity, real-time, label-free molecular characterization at the single-particle level, interferometric microscopy is integrated with quantum-enhanced SERS. This multimodal approach combines interferometric ultra-sensitive nanoparticle detection with SERS's molecular fingerprinting for enhanced SNR. Key challenges with this approach include optical alignment and substrate compatibility to ensure both techniques operate effectively on a shared platform.

[0233] Interferometric microscopy, particularly interferometric scattering (iSCAT) microscopy, has emerged as a powerful tool for detecting nanoparticles, proteins, and single biomolecules with high spatial and temporal resolution. iSCAT exploits the interference between scattered light from a nanoparticle and a reference wave, enabling highly sensitive, label-free detection of nanoscale objects. Despite its advantages in detecting size and refractive index variations, iSCAT lacks chemical specificity, which limits its application in molecular characterization. By integrating it with quantum-enhanced SERS, both spatial precision and molecular information are gained. The system can use dual-wavelength illumination (532 nm for iSCAT, 785 nm for SERS) with a shared objective and a dichroic beam splitter for efficient signal separation. Shot-noise suppression and Rayleigh scattering mitigation enhance Raman signals.

[0234] The integration of these modalities requires a carefully designed optical system that allows simultaneous interferometric imaging and Raman signal acquisition as shown in FIG. 22(for the sake of clarity, the quantum enhancement module is not shown). The same objective lens is used to achieve diffraction-limited resolution and efficient signal collection. The system employs dual-wavelength illumination, where a shorter wavelength (for example, 532 nm) will be used for interferometric contrast imaging, and the pre-optimized 785 nm laser will excite SERS. A dichroic beam-splitting system will be implemented to efficiently combine and separate these illumination sources while minimizing background interference. Rayleigh scattering suppression and homodyne detection techniques enhance the Raman signal while maintaining shot-noise suppression.

[0235] One aspect of this integration is substrate optimization, ensuring compatibility between iSCAT contrast mechanisms and SERS enhancement. Traditional plasmonic SERS substrates, such as gold or silver nanostructures, provide strong Raman enhancement but often introduce optical distortions that degrade interferometric contrast. Alternatively, dielectric nanostructured substrates offer improved reproducibility and compatibility with interferometric detection while still supporting Raman enhancement. To determine the optimal substrate, finite-difference time-domain simulations and experimental validation are performed, assessing both interferometric contrast and SERS signal strength.

[0236] To validate the feasibility of our quantum-enhanced SERS and interferometric microscopy platform, its sensitivity, specificity, and reproducibility can be systematically assessed using nanoparticles, extracellular vesicles, and single proteins. The system is first benchmarked using polystyrene and silica nanoparticles (5-500 nm), ensuring optimal conditions for quantum-enhanced SERS and interferometric contrast. Key metrics include %95 interferometric sizing accuracy and <1 nm SERS spectral resolution, targeting a 3-5x increase in Raman SNR with an enhancement factor of 106-108. Next, biologically relevant nanoscale particles are evaluated, focusing on EVs (50-200 nm), which pose challenges due to weak Raman signals and molecular heterogeneity. The system is tested at >103particles / mL, aiming for higher sensitivity than flow cytometry. Finally, single-protein detection is assessed, a key challenge due to low intrinsic Raman signals. BSA (~66 kDa), amyloid-beta aggregates (4-500 kDa), and enzymatic proteins are analyzed, with a <10 pM detection limit and the ability to resolve oligomerization states via <5 cm-1spectral shifts. Spectral variations are maintained at <10% to ensure reproducibility. This validation phase establishes Q-AIR's multimodaladvantages in enhancing Raman sensitivity, molecular specificity, and reducing photodamage, demonstrating its potential as a high-sensitivity, label-free tool for biomedical research, diagnostics, and nanomedicine.

[0237] SAW trapping is implemented to enhance Raman and interferometric sensitivity. SAW trapping is implemented to enhance molecular interrogation in the Q-AIR system. While quantum-enhanced SERS improves Raman SNR by reducing shot noise, SAW trapping extends interaction times, boosting sensitivity and reproducibility. By minimizing diffusion and increasing throughput, SAW trapping enables precise nanoparticle manipulation using tunable acoustic potentials. Integrating this approach with Q-AIR improves single-particle characterization, spatial resolution, and molecular specificity. A SAW-based system is designed, fabricated, and evaluated to assess its impact on Raman enhancement and interferometric sensitivity.

[0238] A SAW-based trapping system is designed and fabricated for Q-AIR integration. To enable dynamic, label-free confinement of nanoparticles and biomolecules, a SAW trapping system is designed and fabricated compatible with the hybrid substrates. The trapping system may utilize a piezoelectric lithium niobate (Li N bO3) substrate with interdigital transducers (IDTs) that generate standing or traveling SAWs. By adjusting the SAW frequency (ranging from 8 MHz to 800 MHz) and power, the acoustic trapping forces can be optimized to effectively confine dielectric nanoparticles, extracellular vesicles, and proteins within high-intensity optical regions.

[0239] For theoretical background, the acoustic radiation force, F, acting on a nanoparticle in a standing SAW field is given by:F = (4n / 3) R3O(p,P)VE where R is the particle radius, (p,P) is the acoustic contrast factor, dependent on the density p and compressibility 0 of the particle and surrounding medium, and VE is the gradient of the acoustic energy density. For dielectric nanoparticles such as polystyrene and silica, the contrast factor is positive, meaning that they experience a force directing them toward the pressure nodes of the standing wave field, leading to stable trapping. For softer biological vesicles, such as extracellular vesicles (EVs) or lipid nanoparticles, tuning the SAW frequency and pressure amplitude ensures sufficient acoustic contrast to enable controlled manipulation. In travelingSAW fields, an additional drag force from acoustic streaming contributes to particle transport. This effect can be leveraged for dynamic positioning, sorting, or directed transport of biomolecules toward the SERS hotspots, further enhancing molecular interrogation times. The SAW frequency plays a meaningful role in determining trapping efficiency and particle response. Effective SAW trapping occurs when the acoustic wavelength Xsaw matches or exceeds the particle diameter, creating stable potential wells that confine the particles. The SAW wavelength is given by, A.sAw=vsAw / fsAw, where VSAW is the SAW propagation velocity, typically around 4000 m / s for lithium niobate (LiNbO3), and fSAw, is the applied SAW frequency. For nanoparticles ranging in size from 5 nm to 500 nm, the required SAW frequencies typically fall within the range of 8 MHz to 800 MHz as shown in FIG. 23. By carefully tuning the SAW frequency, the system can achieve optimal trapping conditions, enabling precise manipulation of nanoparticles and biomolecules for enhanced Raman and interferometric measurements. SAW trapping enhances Raman signals by confining nanoparticles within SERS-active regions, extending their interaction time from milliseconds to seconds. This prolonged residence boosts excitation efficiency, improving signal intensity and signal-to-noise ratio. Additionally, SAW trapping influences molecular alignment, optimizing interactions with the electromagnetic field for more reproducible spectra. These effects enhance the sensitivity, stability, and efficiency of label-free molecular detection in biomedical and nanotechnology applications.

[0240] The fabrication process begins with the deposition of gold IDTs on a 128° Y-cut LiNbO3wafer using standard photolithography and electron-beam evaporation. The IDTs are designed to produce tunable SAW fields, allowing control over particle positioning and aggregation. To ensure compatibility with interferometric imaging and SERS enhancement, the LiNbO3substrate is coated with a thin SiO2or Au layer as needed, balancing acoustic performance with optical transparency.

[0241] SAW effects on interferometric Raman enhancement are then quantified. SAW trapping's impact on Raman signal intensity and interferometric sensitivity is then systematically evaluated using nanoparticles and biological samples. SAW confinement extends interaction time in high-intensity optical regions, enhancing Raman signals and stabilizing interferometric contrast. Using polystyrene and silica nanoparticles (5-500 nm), a 200x Ramansignal enhancement and a 20% improvement in contrast imaging is aimed for by reducing diffusion-induced fluctuations. To assess molecular specificity, extracellular vesicles (EVs, 50- 200 nm) and protein aggregates are analyzed, targeting <10% variation in Raman spectral features to improve reproducibility. Control experiments optimize SAW parameters, such as frequency and power, ensuring stable trapping. This establishes SAW trapping as an enhancement strategy, complementing quantum-enhanced SERS for improved sensitivity and consistency in single-particle characterization.

[0242] SAW trapping is integrated with quantum-enhanced Raman excitation to enable real-time, high-sensitivity, label-free molecular characterization. This combined approach enhances SNR, extends interaction times, and improves spectral reproducibility. The SAW trapping system is synchronized with the quantum-enhanced Raman excitation. Meaningful benchmarks include a 200x Raman signal enhancement from SAW trapping, a 5x SNR improvement from quantum enhancement, and stable trapping exceeding 30 seconds per particle for extended molecular interrogation. To validate real-time performance, conventional Raman spectroscopy, quantum-enhanced Raman without SAW trapping, and the fully integrated Q-AIR system can be compared. Spectral resolution, molecular alignment, and photodamage thresholds, can be assessed demonstrating Q-AIR as a scalable, high-throughput tool for biomedical research, nanomedicine, and molecular diagnostics.

[0243] It is contemplated that, if interferometric contrast degrades on substrates, dielectric-based nanoantennae can be explored for improved compatibility. It is further contemplated that SAW trapping efficiency may vary with particle composition and, if necessary, alternative acoustic wave parameters or hybrid optical-acoustic confinement strategies can be employed and tested. If quantum-enhanced Raman excitation introduces excess phase noise, adaptive phase-locking techniques might be implemented to stabilize the squeezed light state.

[0244] In summation, this example provides exciting prospects for integrating quantum- enhanced SERS and SAW trapping to improve Raman and interferometric sensitivity. Q-AIR has the potential to evolve into a unified platform for both chemical and physical analysis at the single biological particle level, bridging molecular fingerprinting with high-precision structural characterization. This capability opens new doors in biomedical research, nanomedicine, andmolecular diagnostics, enabling high-throughput, label-free detection with unprecedented sensitivity. In the future, this approach could be further extended to real-time live-cell imaging, ultra-sensitive biomarker detection, and nanoscale drug delivery monitoring.Example VII: Alternative Detection Modes - Dynamic Surface and Suspension Sensing Strategies

[0245] Building upon the interferometric localization and wide-field Raman capabilities outlined above, innovative dynamic sensing strategies are introduced to enhance sensitivity, flexibility, and signal robustness.

[0246] One feature of the platform is the precise control over the surface density of metallic nanoparticles (for example, gold or silver) used to enhance Raman signals. The ratio between the nanoparticle density and the target vesicle concentration is a meaningful parameter that can be used to tune sensitivity, specificity, and dynamic range of detection. By adjusting surface concentrations, the system can be adapted for samples with varying vesicle concentrations, from low-abundance clinical fluids to highly concentrated biological samples.

[0247] Moreover, beyond static surface-based sensing, the present disclosure encompasses dynamic suspension-based sensing modes. In this modality, target nano or micro-sized biological particles (for example, vesicles, virus, cells, biological molecules) and metallic nanoparticles interact while remaining in fluid suspension. The detection can occur during active movement, swimming or after settled on the substrate. Notably, sensing is feasible regardless of whether the vesicles are transiently bound to particles or freely interacting in solution, greatly expanding operational flexibility.

[0248] To further enhance the dynamic sensing mode, the system can employ mechanical or acoustic actuation of the suspension medium. For instance, mild acoustic vibrations can be applied either to the substrate or the fluid reservoir, preventing nanoparticle aggregation and facilitating uniform interaction between vesicles and enhancing agents. This vibratory actuation enables dynamic Raman readout, a novel operational mode wherein averaged signals are collected over time from a moving sample, in contrast to traditional static measurements. Such dynamic averaging can improve signal robustness, reduce artifacts, and expand the detectable target populations.Example VIII: Alternative Detection Modes - Dual-Mode Sensing: Dry Surface Mode and Suspension Dynamic Mode

[0249] The platform can be engineered to operate flexibly in two distinct modes: dry surface mode and suspension dynamic mode. In dry surface mode, isolated vesicles are immobilized onto a solid substrate, and samples are dried under controlled conditions prior to interferometric imaging and Raman spectroscopy. This mode is configured for high spatial resolution, minimal fluid motion artifacts, and stable measurements, making it ideal for static analyses such as single-vesicle localization and detailed cargo mapping. In suspension dynamic mode, vesicles remain within a fluid medium during measurement, allowing real-time detection of free-floating or loosely bound vesicles. This dynamic approach enables the system to monitor transient interactions and potentially capture broader vesicle populations that might not otherwise immobilize efficiently. When combined with acoustic or mechanical actuation, the suspension mode provides a dynamic averaging readout that can improve signal robustness across heterogeneous samples. The dual-mode capability of the platform enables users to tailor the sensing configuration to the biological question at hand, offering maximum versatility for both research and clinical applications.Example IX: Alternative Detection Modes - Self-Assembled Plasmonic Substrates for Label- Free Detection

[0250] To enhance Raman sensitivity while maintaining scalability and manufacturing simplicity, this disclosure contemplates the use of self-assembled plasmonic substrates composed of gold or silver nanoparticles. These nanoparticles can be deposited on the sensing surface via controlled drying, solvent evaporation, or chemical functionalization techniques, allowing them to spontaneously organize into densely packed nanoclusters. The resulting nanogap-rich structures form SERS "hotspots" without the need for lithographic patterning, enabling uniform enhancement of Raman signals across large areas. These hotspots can be on top of a substrate, embedded locally or floating freely between pillars of both hotspot and nonhotspot characteristic, or can be in suspension.

[0251] This self-assembly strategy significantly reduces production cost, improves batch-to- batch reproducibility, and facilitates integration into microfluidic or portable diagnosticformats. Furthermore, this architecture is compatible with the interferometric imaging modality, as it preserves the planar nature of the substrate for phase-sensitive detection.Example X: Alternative Detection Modes - Active Surface Modulation via Acoustic, Magnetic and Electromagnetic Actuations

[0252] It is contemplated that, in some forms, the surfaces can be actively modulated via acoustic, magnetic and electromagnetic actuations. This may help localize the particles to locations before imaging begins.Example XI: Integration Alternatives - Integration of Isolation, Coating, and Detection within a Unified Platform

[0253] Another major innovation described herein is the integration of vesicle isolation, nanoparticle coating, and label-free detection within a single device platform. Currently, EV or virus research workflows typically require multiple disconnected instruments: one for particle isolation (for example, ultracentrifugation, microfluidic isolation), one for particle coating or functionalization, and another for optical detection. This segmented workflow leads to sample loss, increased contamination risk, and extended processing times.

[0254] The present disclosure offers a fully integrated system capable of automating vesicle isolation from complex biological samples (for example, plasma, cyst fluid, urine, whole blood), immediate nanoparticle decoration or functionalization if needed, and subsequent label-free interferometric and Raman detection, all within a unified platform. Integration across these traditionally disparate steps enhances throughput, minimizes user intervention, preserves sample integrity, and improves overall sensitivity and reproducibility. This streamlined "sample- to-answer" design would dramatically simplify vesicle analysis workflows, making high-precision molecular diagnostics accessible even to non-specialist users.Example XII: Applications - HIV Detection, Subtyping and Quantification

[0255] With reference being had to FIGS. 24A through 24D, the capability of the label-free SERS-based platform to differentiate between HIV clades and to stratify patient-specific viral load levels using optical fingerprinting was evaluated. As shown in FIG. 24A, Raman spectra were acquired from Clade A, Clade C, and SHIV viral particles, as well as, in FIG. 24B, from two patient-derived HIV isolates: one with a low viral load and one with a high viral load. The resulting spectra revealed distinct molecular fingerprints, with notable differences in peakpositions and intensities that reflect underlying biochemical variability among both viral subtypes and clinical isolates.

[0256] To further analyze these spectral distinctions, principal component analysis (PCA) was performed, which revealed clear clustering by group as depicted in FIGS. 24B (Clade A, Clade C, and SHIV ) and 24D (patient). For example, in FIG. 24B, clade-level separation was observed between HIV Clades A, C, and SHIV, highlighting the system's ability to resolve fine molecular differences across closely related viral lineages.

[0257] Similarly, nanosized particles were isolated from human plasma using a filtrationbased method and quantified HIV viral load using Raman spectroscopy. Patient samples with high and low viral loads, along with healthy controls, formed distinct, non-overlapping clusters in PCA space, indicating that viral load states can be reliably distinguished based on their spectral signatures as shown in FIG. 24D.

[0258] These results demonstrate the potential of the platform for non-invasive HIV subtyping and viral load monitoring. This approach holds significant promise for applications in clinical diagnostics, epidemiological surveillance, and real-time assessment of therapeutic response, all without the need for molecular labeling or amplification.Example XIII: Applications - Lung Cancer

[0259] The approach leverages label-free surface-enhanced Raman spectroscopy (SERS) of patient-derived extracellular vesicles (EVs) as a non-invasive biomarker platform for early lung nodule characterization. By isolating EVs from peripheral blood and recording their molecular "fingerprints" via a portable Raman-interferometric microscope hybrid instrument, both chemical and size-distribution signatures are captured without subjecting patients to invasive biopsies or repeated imaging procedures. In contrast to traditional tissue biopsies— which carry risks of hemorrhage, pneumothorax, and potential tumor seeding that can compromise patient outcomes— the SERS-EV assay requires only a simple blood draw and delivers real-time molecular insights.

[0260] In this example / study, the averaged, normalized SERS spectra of healthy controls, benign nodule patients, and early-stage lung cancer patients are compared as shown in FIGS. 25A through 25E. From FIG. 25A, characteristic peaks at ~850, ~1000, ~1330, and ~1450 cm-1exhibited distinct intensity patterns across the three groups, and difference spectra (healthy vs.all; benign vs. cancer) further highlighted subtle molecular shifts. With reference to FIG. 25B, principal component analysis (PCA) on the full spectral data revealed clear clustering: PCI (75.3% variance) effectively separated cancerous from benign and healthy samples, while PC2 (4.8% variance) distinguished benign from healthy. Linear discriminant analysis (LDA) boundaries overlaid on the PCA scorespace yielded robust class segregation, and a subject-level heatmap underscored the assay's potential to classify individual patients with high fidelity.

[0261] In summary, the non-invasive SERS-EV platform addresses a gap in the early detection of lung cancer, where current imaging modalities (CT, PET-CT) and invasive biopsies remain the standard despite their risks and costs. Few studies to date have demonstrated the feasibility of label-free, blood-based Raman signatures for lung nodule screening, and none have integrated multi-modal spectral and size profiling in a single point-of-care instrument. By offering rapid, risk-free molecular diagnostics, this technology has the potential to improve patient stratification, reduce unnecessary invasive procedures, and ultimately enhance outcomes through earlier therapeutic intervention.Example XIV: Extracellular Vesicle Markers of Lung Cancer that Correlate with Radiomics to Predict Malignancy

[0262] The collected spectral data for lung cancer described in the above example - as well as more generally for other cancers - can be combined with advanced imaging analytics such as, for example, PET / CT and Photon Counting CT (PCCT) radiomics, and comprehensive molecular profiling (genomics and transcriptomics) in a single workflow via the disclosed platform. This unique combination, which does not exist in any current diagnostic platform, transforms the traditional laboratory-bound analysis into a portable, bedside diagnostic tool that could also fit in a primary clinic. Unlike existing methods that require multiple separate instruments, complex sample preparation, and specialized laboratory settings, this technology uniquely enables simultaneous physical and molecular EV characterization while correlating these signatures with imaging features in real-time. By bringing together these typically separate analytical modalities into a portable platform, unprecedented diagnostic accessibility and accuracy are provided for lung nodule classification and early cancer detection at the point of care, eliminating the need for sample transport and reducing time-to-result from days to minutes. The final tool leverages molecular-level data through Raman spectroscopy, a cornerstoneanalytical technique recently enhanced by AI-ML advances, to classify complex clinical samples by integrating iSCAT SERS and imaging.

[0263] This accordingly introduces a new capability for interpreting indeterminate disease patterns more generally by combining Raman spectroscopy, traditional imaging analytics, and machine learning . With respect to lung cancer, in particular, it is the demonstrated above that healthy (nodule-free), benign and malignant groups can be effectively separated. So the immediate import of this example is to better discrimination between benign and malignant nodules assessing the malignancy status of indeterminate small pulmonary nodules. Early lung cancer detection biomarkers are developed and investigated that would directly impact the growing need to integrate imaging and non-invasive EV molecular diagnostics for indeterminate pulmonary nodules and allow physicians to avoid unnecessary invasive procedures with benign lung disease.

[0264] However, this integration of Raman data with -omics data (and radiomic data, specifically) with Al and machine learning is also more widely applicable. So the utility of Raman spectral to provide a data set is well established across a wide range of applications with artificial intelligence and machine learning models for predictive purposes, and the more general supplementation of that predictive model with other advanced imaging techniques or further clinical data is also contemplated herein.

[0265] In the case of lung cancer detection, lung cancer diagnostics currently rely on CT imaging, tissue biopsy, and liquid biopsy— each with major drawbacks. CT scans lack specificity, with a 96% false-positive rate in the NLST trial, leading to unnecessary invasive procedures; biopsies are risky (22% complication rate), costly, and not always feasible; and liquid biopsy methods offer promise but alone have remain insufficiently sensitive for early-stage detection. This platform can overcome these limitations through a portable, integrated platform combining extracellular vesicle (EV) analysis with advanced imaging correlation. This enables accurate, non-invasive lung nodule characterization at the point of care. It is contemplated that the SERS- and Al-enhanced EV platform disclosed herein can delivers results within 2-3 hours from a simple blood draw, at an assay cost below $100— dramatically faster and more affordable than standard approaches.

[0266] Among liquid biopsy components— EVs, circulating tumor DNA (ctDNA), and circulating tumor cells (CTCs)— EVs offer unique advantages. Actively secreted by viable tumor cells, EVs reflect real-time cellular activity, unlike ctDNA (from cell death) or CTCs (from advanced disease). EVs require only 0.3 mL of plasma, compared to 5-10 mL for ctDNA and 7- 10 mL for CTCs, and are stable at -80°C, unlike ctDNA (96-hour window) or CTCs (immediate processing). EVs are far more abundant (1O8-1O11vesicles / mL vs. 1-10 CTCs / mL), and yield higher DNA quantities (5-200 ng / mL). Unlike ctDNA and CTCs, EVs enable multi-omic profiling— including RNA, proteins, and lipids— from a single sample. They also cross the bloodbrain barrier and appear in early-stage disease (l-ll), while ctDNA and CTCs typically emerge later (I II— IV).

[0267] Existing EV characterization methods— DLS, NTA, electron / atomic force microscopy, and MS— have major limitations: low throughput, limited compositional data, and complex sample prep. These lab-bound techniques cannot resolve EV heterogeneity or support realtime, point-of-care use. This platform directly addresses these issues, offering a portable platform for high-resolution, comprehensive EV analysis. It unlocks new possibilities in early cancer detection, disease monitoring, and integration with clinical imaging, advancing the future of personalized, non-invasive diagnostics.

[0268] As described in this example, PET / CT and PCCT lung nodule imaging can integrated be with the EV transcriptomic, genomic, Raman spectroscopic and radiomic and deep-learning data to non-invasively identify biomarkers for high-risk and low-risk lung nodules at the disease onset and progression. This has the potential to transform the field of clinical biomarker discovery and improve the diagnosis, prognosis, and therapy response monitoring strategies for lung cancer and, importantly, to identify non-cancerous nodules to reduce the risk of unnecessary procedures. Again, this impact goes beyond lung cancer, as the platform can potentially be broadly applied to many other benign masses and lesions throughout the body that often require invasive methods for diagnosis including pancreatic cystic lesions, liver lesions and other nodules in the adrenal gland, kidneys, and breast often require invasive diagnostics or surgery. The EV blood biomarker assay can significantly reduce the population of patients undergoing invasive testing for many different types of tumors.

[0269] To fully characterize EVs and their molecular cargo, this example offers a threepronged approach: (i) a multi-modal imaging microscope that combines SERS (produces chemical finger print profiles) and iSCAT microscopy (produces morphometric, concentration, and size distribution related features), (ii) machine learning based algorithms to classify EVs isolated on a disposable microfluidic cartridge in a rapid and a cost effective manner, (iii) machine learning based approaches to connect the label-free EV data from the platform, DNA and RNA sequencing data from EVs, and PET / CT and Photon Counting CT images to improve diagnostic decision making at the time of imaging. SERS provides molecular-level characterization complementary to traditional genomic / proteomic analysis but in real-time without destructive sample processing. The validated IPRM approach simultaneously detects cancer-associated proteins, nucleic acids, and lipid signatures, with Al algorithms correlating spectral features to established -omic profiles— effectively complementing multiple laboratory workflows with a single measurement.

[0270] To demonstrate and validate this, EV profile analysis and biomarker discovery is first demonstrated. Extracellular vesicles (EVs) are isolated and characterized from plasma samples of 30 healthy controls and 40 individuals with lung nodules using the ExoTIC platform referenced elsewhere herein. EVs will undergo multi-modal characterization, including nanoparticle tracking analysis (NTA), transmission electron microscopy (TEM), ELISA, Western blotting, and label-free molecular and morphological profiling via SERS and iSCAT microscopy. To gain deeper insight into the molecular cargo, RNA -omics and DNA seq on isolated EVs are performed. Finally, EV-derived molecular and imaging signatures are integrated with radiomics features extracted from corresponding PET / CT scans to identify composite biomarkers, if present, that could be predictive of malignancy.

[0271] A multi-modal data integration and model is developed. A robust, Al-driven framework is developed to integrate diverse data modalities for accurate classification of lung nodule malignancy. First, radiomics and deep learning models are developed using PET / CT and PCCT imaging data. A hierarchical learning architecture is implemented that processes each data modality through specialized components: (i) convolutional neural networks (CNNs) for EV image and spectral data, (ii) attention-based networks for radiomics features, and (iii) graph neural networks for analyzing high-dimensional molecular data. The resulting models aretrained and validated using a cohort of up to 100 samples, 42 of which have already been collected with paired EV and imaging data, to ensure clinical relevance and generalizability.

[0272] In regards to the radiomic workflow, at present, a pilot has been run with 12 patients with demonstrated lung nodules as well as Positron Emission Tomography (PET) images from 11 of them were obtained and de-identified. For each patient, the CT and the PET studies that were closest to the diagnosis date were selected for analysis. CT and PET images from patients with lung nodules were analyzed. This analysis resulted in the extraction of 1657 radiomics features, with the top 50 selected based on variance, demonstrating an improved ability to distinguish malignant from benign lung nodules when combining multi-omic liquid biopsy models with PET / CT.

[0273] Multi-institutional clinical validation is performed to validate the diagnostic performance and generalizability of the platform in a real-world, multi-institutional setting. A total of 500 participants are recruited — 400 patients with nodules and 100 healthy controls — across three collaborating clinical sites. For each participant, paired blood and tissue biopsy samples are collected to enable side-by-side molecular and histopathological analysis. Comprehensive correlation studies integrating EV-based molecular profiles with imaging and biopsy results are conducted to assess diagnostic concordance and refine the predictive models.

[0274] This integrated diagnostic platform is developed and further optimized for wide- scale use. Initially, a compact, folded optical design is to be implemented that integrates the SERS and iSCAT modules for high-resolution, label-free EV analysis. The system can use a 50x objective (NA 0.8) to achieve ~375 nm resolution with iSCAT (488 nm illumination). Illumination and collection paths can be separated, with compact dichroic optics incorporated into a shared optical path to ensure stable alignment. The optical system might be housed in a portable enclosure smaller than 20x20 cm, incorporating a miniaturized translation stage for high- throughput scanning (<5-minute total measurement time). 4 cm1spectral resolution and 20 dB signal to noise ratio (SNR) are targeted. Following optical integration, a microfluidic cartridge is designed for efficient EV capture and compatibility with the optical system. The cartridge can utilize a 10 x 10 mm disposable substrate to support scalable, single-use diagnostics and enable rapid sample processing. The complete system is further optimized.

[0275] The overarching objective in this example is the profiling of RNA, DNA, SERS / interferometric imaging EV-signatures of individuals with and without cancer from their plasma and lung biopsies to establish "health and cancer" discrimination solely from plasma- derived EVs for the development of a fully integrated portable technology platform for whole blood EV analysis that inform and improve medical decision making. Again, the end-product and principles used here are expected to find numerous other uses in the clinical management of wide-ranging cancer types. This approach has the potential to revolutionize cancer care, leading to improved patient outcomes. The initial work on lung cancer demonstrates an improved ability to distinguish malignant from benign lung nodules when combining multi-omic liquid biopsy models with PET / CT.Example XV: Applications - Extracted EV RNA Measurement on Fabricated Microfluidic Cartridges

[0276] In addition to label-free interrogation of intact vesicles and tissues, the platform supports direct detection of extracted RNA in a dry assay format. Following on-chip lysis and nucleic acid extraction, the purified RNA is deposited onto a pre-patterned SERS-active region within a disposable microfluidic cartridge and dried in situ. Once dry, interferometric microscopy localizes the RNA film and surface-enhanced Raman spectroscopy (SERS) acquires nucleotide-specific vibrational fingerprints without interference from bulk solvent.

[0277] Dry-mode RNA measurement delivers several benefits: (1) elimination of water Raman-background improves signal-to-noise by >3x; (2) fixed RNA morphology enhances spatial reproducibility of hotspot sampling; (3) integration on the same cartridge as upstream vesicle isolation enables seamless "sample-to-answer" workflows; and (4) single-nucleotide sensitivity is achieved at sub-pg input levels, supporting rapid miRNA profiling. This capability extends the platform's molecular sequencing reach beyond intact vesicle analysis into direct genomic and transcriptomic assays— all within a portable, label-free optical instrument.Example XVI: Signal Enhancement Algorithm Using Depth Scanning - Dynamic Correlation Raman Spectroscopy for Background Rejection and Signal Enhancement

[0278] The present disclosure can also incorporate dynamic Raman acquisition methodologies to enhance signal discrimination and background suppression. During spectral acquisition, the sample is subjected to a controlled actuation signal, such as mechanicaloscil lation, acoustic defocusing, or optical modulation. A z-stack of Raman spectra is acquired synchronously with the actuation waveform, capturing temporal variations in the Raman signal intensity at each wavenumber.

[0279] Subsequently, the acquired spectral stack is analyzed by correlating the intensity fluctuations at each spectral position with the applied actuation waveform. Peaks in the correlation spectrum correspond to sample features that respond coherently with the modulation, whereas incoherent background signals are suppressed. This approach enables the separation of sample-specific Raman features from stationary background signals (for example, substrate scattering, autofluorescence) and improves signal-to-noise ratios for weak biological samples such as extracellular vesicles. This dynamic correlation strategy further enhances the robustness and sensitivity of the platform, particularly in suspension-based detection modes where background contributions are otherwise substantial.

[0280] As illustrated in FIG. 26, this dynamic correlation strategy allows for the decomposition of Raman data into distinct sample and background spectra, leading to improved detection sensitivity and molecular specificity. Further, FIGS. 27A through 27D demonstrates the experimental result of this approach: following dynamic correlation and background subtraction, the reconstructed Raman spectrum exhibits significantly enhanced clarity of true sample peaks compared to the nominally focused raw measurement.

[0281] In addition to static surface-based detection, the present disclosure contemplates Raman interrogation of nanosized biological particles, such as extracellular vesicles, during flow through a microfluidic channel or suspension system. Vesicles are transported through the excitation region under controlled flow conditions while Raman spectra and / or interferometric signals are continuously or periodically acquired. The system optionally combines interferometric localization with Raman readout in real-time or near real-time, allowing dynamic spectral acquisition of vesicles without requiring immobilization. Flow-based operation enables high-throughput analysis of vesicle populations, facilitates rapid screening of clinical samples, and supports the dynamic study of vesicle behavior under varying environmental or biological conditions. Integration of size information (obtained via interferometric or optical scattering signals) with Raman spectral data during flow allows simultaneous classification of vesicle subpopulations based on both physical and molecular properties. This dynamic sensingapproach expands the platform's applicability to applications requiring rapid, continuous, and label-free molecular analysis.Example XVII: Signal Enhancement Algorithm Using Depth Scanning - Dynamic Raman Detection and Sorting of Flowing Vesicles in Microfluidic Channels

[0282] In addition to static surface-based detection, the present disclosure enables Raman interrogation of nanosized biological particles, such as extracellular vesicles, lipoproteins or viruses, during flow through a microfluidic channel or suspension system. Vesicles may be transported through the excitation region under controlled flow conditions while Raman spectra are continuously or periodically acquired. Importantly, the system accommodates vesicles that are rolling along, flowing across, or moving in the vicinity of the surface, rather than requiring full immobilization or free suspension. The system optionally combines interferometric localization with Raman readout in real-time or near real-time, allowing dynamic spectral acquisition of vesicles during motion near the surface. In certain embodiments, this can further incorporate active sorting mechanisms triggered by physical or molecular information obtained during flow. Vesicles may be sorted based on their Raman spectral features, size, refractive index, or a combination of these properties. Sorting modalities may include magnetic deflection (where vesicles are labeled with magnetic nanoparticles), acoustic manipulation (via surface acoustic waves or bulk acoustic waves), dielectrophoretic separation (using applied electric fields), or hydrodynamic flow focusing. Raman-based molecular signatures can be used to selectively enrich vesicle subpopulations of clinical or biological relevance, such as tumor-derived vesicles, pathogen-associated particles, or EVs enriched for specific nucleic acid cargo.

[0283] Flow-based operation enables high-throughput analysis and sorting of vesicle populations, facilitates rapid screening of clinical samples, and supports the dynamic study of vesicle behavior under varying environmental or biological conditions. Integration of physical properties and molecular signatures during flow allows simultaneous characterization and realtime sorting of EV subpopulations, expanding the platform's applicability for diagnostics, biomarker discovery, and therapeutic monitoring.Example XVIII: Signal Enhancement Algorithm Using Depth Scanning - Dynamic Multi-Point Raman Excitation and / or Collection Using Programmable Optics

[0284] The present disclosure further contemplates incorporating programmable optical control elements to dynamically configure the Raman excitation and / or collection geometry across the sample field. In certain embodiments, a digital micromirror device (DMD), spatial light modulator (SLM), or adaptive optics system is employed to selectively and simultaneously illuminate multiple regions corresponding to localized nanosized biological particles, such as extracellular vesicles. Based on prior interferometric imaging, the programmable optical elements may dynamically target excitation specifically to verified particle positions and / or selectively enhance Raman signal collection from these regions, enabling parallel spectral acquisition from multiple vesicles in a single measurement cycle.

[0285] This approach significantly increases throughput by avoiding traditional sequential point-by-point scanning, while also reducing photodamage and improving the overall signal-to- noise ratio. Adaptive optics may further be utilized to correct for optical aberrations introduced by the sample substrate, fluidic channel, or biological medium, thereby optimizing both excitation focus and collection efficiency across a wide field of view. The system can dynamically adjust the excitation and / or collection patterns in real-time based on vesicle size, distribution, or molecular properties, allowing selective interrogation, enrichment, or exclusion of particular subpopulations.

[0286] By integrating programmable optical excitation and / or collection with simultaneous Raman spectral acquisition, this disclosure provides an adaptable, high-throughput platform for detailed molecular characterization of heterogeneous vesicle populations in both static and flow-based operational modes.Example XIX: Signal Enhancement Algorithm Using Depth Scanning - Differential Surface Imaging for Detection of Particle Binding Events on Patterned Substrates

[0287] In certain embodiments, the present disclosure contemplates employing differential imaging strategies to detect the binding of nanosized biological particles, such as extracellular vesicles, onto rough or patterned substrates used for surface-enhanced Raman spectroscopy (SERS). Due to the inherent optical roughness and scattering background from structured surfaces (for example, nanopillars, nanogaps, metallic aggregates), direct detection of newlyarriving particles can be challenging using static imaging. To overcome this limitation, the system continuously or periodically images the surface over time, capturing sequential frames of the sample region.

[0288] By analyzing the temporal differences between consecutive frames or averaged frame sets, subtle localized changes indicative of new particle binding events can be detected with high sensitivity, even against noisy or irregular backgrounds. Techniques such as differential contrast enhancement, phase change detection, or intensity subtraction may be utilized to identify binding sites. Once particle landing events are confirmed, Raman spectral acquisition can be selectively targeted to these dynamically identified locations, thereby enhancing molecular detection specificity and improving SERS signal averaging. This approach extends the capability of the platform to monitor dynamic vesicle-surface interactions and enables high-sensitivity label-free molecular analysis on complex, high-surface-area substrates.Example XX: Signal Enhancement Algorithm Using Depth Scanning - Patterned surface for bimodal interferometric microscope and SERS usage

[0289] In one embodiment, the sample support comprises a lithographically defined multilayer pattern of alternating gold (Au) and silicon dioxide (SiO2) regions arranged in a concentric-ring geometry, as shown in FIG. 28A. A planar SiO2substrate is first deposited or thermally grown to a thickness of 100-300 nm, providing an optically transparent background with low intrinsic scattering. Successive concentric square rings of Au (20-50 nm thickness, 1-5 pm ring width, 1-10 pm spacing) are then patterned on top of the SiO2layer. The Au rings serve as localized plasmonic "hot spots" for surface-enhanced Raman scattering (SERS), while the surrounding SiO2regions produce a stable reference signal for interferometric microscopy. Incident laser light (X ~ 785 nm) excites both the plasmonic resonance in the Au rings for SERS enhancement and the refractive-index contrast at the Au / SiO2interfaces for interferometric detection, thereby enabling correlated chemical and mass-indexed measurements on the same field of view.

[0290] In another embodiment as illustrated in FIG. 28B, discrete Au islands are arranged in a periodic square array (chessboard design) embedded within a continuous SiO2matrix. Here, Au squares (5-10 pm side length, 30-50 nm thickness) are separated by SiO2gaps of 2-10 pm, fabricated via lift-off or subtractive etching processes. The array pitch can be tuned to optimizenea r-field coupling for SERS and far-field scattering for interferometric microscopy. The SiO2background minimizes non-specific adsorption and provides uniform optical properties, while the Au islands concentrate the electromagnetic field to generate strong Raman enhancement. By scanning the sample stage or by patterned selection of individual islands / rings, one can perform high-throughput, multiplexed assays in which molecular fingerprints (via Raman) are co-registered with label-free interferometric measurements of nanoparticle binding or mass accumulation.

[0291] While FIGS. 28A and 28B show square and concentric ring geometries, respectively, other geometries are certainly contemplated. For example, circles, triangles, hexagonal and other higher packing density geometries and various shapes such as nanopillar surfaces could be implemented. Thus, while illustrative, the geometries are certainly not limited to those depicted in the figures.

[0292] Fabrication of both embodiments may employ standard photolithography or electron-beam lithography, followed by physical vapor deposition of Au and chemical vapor deposition of SiO2. The thicknesses and lateral dimensions of the Au and SiO2regions can be varied to match the excitation wavelength and target analyte, and to balance SERS enhancement against interferometric imaging contrast. This dual-functional patterned surface thus provides a robust platform for simultaneous chemical and biophysical characterization of, for example, extracellular vesicles, protein aggregates, or nanoparticle-biomolecule complexes in biomedical diagnostics and materials science applications.Example XXI: Utilization of Artificial Intelligence to Correct Known Experiment Issues of SERS - Batch to Batch Non-uniformity Correction

[0293] In one embodiment, a machine-learning-based calibration module is employed to correct for fabrication-induced non-uniformities across successive batches of patterned Au / SiO2surfaces— whether in concentric-ring or chessboard geometries— to ensure reproducible SERS signal intensities from biological samples. A reference analyte (for example, a self-assembled monolayer of benzenethiol or a standard Raman dye) is deposited uniformly across each freshly fabricated substrate. High-resolution SERS maps are acquired over the full extent of the pattern, generating a spatial dataset of raw intensity values, lraw(x,y), at one or more diagnostic Raman peaks. These maps serve as calibration targets for training a regression model (forexample, a convolutional neural network, gradient boosting regressor, or Gaussian process) that predicts a pixel-wise correction factor, C(x,y), such that lcalibrated(x,y) — I raw (x,y) x C(x,y) where lCaiibrated(x,y) approximates a batch-invariant reference intensity profile.

[0294] In another embodiment, the Al calibration model takes as input both image-based features of the patterned substrate (for example, local ring width, island spacing, edge roughness) and process metadata (for example, deposition rate, lithography dose, etch time). A supervised learning workflow is implemented: known reference intensities from multiple fabrication runs are paired with their corresponding raw SERS maps and substrate features to train the model. Once trained, the model predicts C(x,y) for new substrates without requiring a full reference scan— only a rapid low-resolution "pre-scan" is needed to extract features. The correction factors are applied in real time to experimental SERS measurements of extracellular vesicles, harmonizing batch variations and reducing the coefficient of variation of peak intensities to below 10%.

[0295] In a further embodiment, the system continuously refines its correction model via online learning: periodically, fresh reference scans are incorporated into the training set, and model weights are updated to capture subtle drifts in fabrication and environmental conditions. Corrected SERS data from multiple patterned regions (for example, all rings in the concentric design or all squares in the chessboard array) are aggregated to generate a composite diagnostic spectrum for each sample. This Al-driven normalization ensures that biological assay results— such as relative concentrations of disease biomarkers— are comparable across substrates, days, and fabrication lots, thereby enhancing the robustness of downstream diagnostic decision-making.Example XXII: Utilization of Artificial Intelligence to Correct Known Experiment Issues of SERS - Al-Based Classification of Cancer-Related Vesicles Using Fuzzy Logic

[0296] In one embodiment, the system integrates interferometric microscopy and surface- enhanced Raman scattering (SERS) data streams to perform automated classification of extracellular vesicles (EVs) into cancerous, pre-cancerous, benign, and healthy categories. Raw image frames and Raman spectra acquired over each field of view are preprocessed to extractkey features— namely vesicle count, size distribution, volumetric estimates (from interferometric microscope), and spectral intensity vectors across multiple Raman bands (from SERS). These features are then fed into a fuzzy-logic inference engine which assigns, for each vesicle or pixel region, a continuous membership value p_cancer G [0,1] reflecting the degree of "cancer-likeness."

[0297] The fuzzy logic engine comprises a set of membership functions defined over the extracted feature dimensions (for example, vesicle diameter, peak Raman intensity at biomarker wavenumbers, spectral shape coefficients). The resulting fuzzy membership values are aggregated— via fuzzy union or weighted average operators— to yield, per field of view, a diagnostic score S = f(p_cancer, p_pre-cancer, p_benign). Scores exceeding a predetermined threshold T_cancer trigger a "high probability of cancer" indication, while intermediate scores fall into an uncertainty band recommending additional scanning or sample preparation.

[0298] In a further embodiment, the inference engine is implemented within an Al framework (for example, a neural network or support vector machine) that is initially trained using labeled EV datasets derived from clinical plasma or saliva samples. Fuzzy membership outputs serve both as target labels during supervised learning and as continuous feedback variables for adaptive rule refinement. As new sample data are acquired— either from additional fields of view, sequential scans of the same sensor area, or multiple sensor chips processed at different time points— the model continuously updates its membership functions and decision thresholds, thereby improving diagnostic accuracy and reducing false positives over time.

[0299] The system supports multiplexed analysis across an array of patterned Au / SiO2regions: by scanning individual islands or concentric rings, the platform can perform high- throughput, parallel classification of mixed populations of cancerous and healthy vesicles. Aggregated diagnostic results from multiple fields of view may be displayed as a heatmap of membership scores, enabling clinicians or end-users to visualize spatial heterogeneity in vesicle populations.

[0300] This Al-driven, fuzzy-logic-enhanced approach provides a robust mechanism for label-free, quantitative discrimination of cancer-related EVs in complex biological samples,facilitating real-time decision support in liquid-biopsy applications and early-stage disease detection.Example XXIII: Surface Charge Density Adjustment to Repel Lipoproteins

[0301] In one embodiment, the patterned Au / SiO2substrates (whether concentric-ring or chessboard geometries) are further modified to present a high surface charge density that selectively repels plasma lipoproteins while retaining extracellular vesicles (EVs) for analysis. Following the lithographic fabrication of Au and SiO2regions, each surface is coated with a selfassembled monolayer (SAM) of charged thiol or silane molecules [for example, 11- mercaptoundecanoic acid on Au; (3-aminopropyl)triethoxysilane on SiO2] to achieve a controlled negative zeta potential (-30 to -50 mV) at physiological pH. The high surface charge density generates an electrostatic barrier that inhibits adhesion of low-density and high-density lipoprotein particles (~10-100 nm diameter), which typically carry a net negative surface charge but have reduced Debye lengths in buffer conditions.

[0302] In another embodiment, surface charge tuning is performed via layer-by-layer deposition of polyelectrolytes [for example, poly(acrylic acid) and poly(allylamine hydrochloride)] onto the patterned regions. By varying the number of bilayers and the ionic strength during assembly, the net charge density on Au and SiO2features can be precisely adjusted. EVs— enveloped by a lipid bilayer and decorated with membrane proteins— exhibit distinct dielectric and steric interactions that overcome the electrostatic repulsion, enabling their selective capture at the patterned "hot spots" for interferometric imaging and SERS interrogation. Lipoproteins, by contrast, are sterically excluded or electrostatically repelled, reducing background signal and improving assay specificity.

[0303] In a further embodiment, the surface charge can be dynamically modulated by applying a small DC bias (±0.1-0.5 V) between the Au regions and a reference electrode in the sample buffer. Real-time control of the electrode potential shifts the double-layer capacitance and alters the local zeta potential, allowing on-demand tuning of lipoprotein repulsion during sample incubation. This dynamic adjustment ensures that only EV populations remain in close proximity to the plasmonic Au sites for SERS enhancement and to the refractive-index contrast boundaries for interferometric detection. As a result, the platform delivers high-purity EVmeasurements from complex biological fluids— such as plasma, serum, or saliva— without the need for extensive pre-purification steps.Example XXIV: Adaptive Spectral Sampling Enabled by Interferometric Priors

[0304] In the disclosed multimodal platform, interferometric imaging is applied either sequentially or in parallel with wide-field Raman spectroscopy on the same spatial region of the sample, ensuring precise spatial and molecular co-registration. The interferometric module provides quantitative localization, sizing, and refractive index contrast mapping of individual extracellular vesicles (EVs) with sub-diffraction precision, which is then used to define a region- of-interest (ROI) mask for subsequent Raman acquisition. To optimize data acquisition efficiency, the system integrates a compressive sensing framework, wherein the interferometric priors guide adaptive sampling of the Raman domain— acquiring high-dimensional spectral data only from ROIs exhibiting nontrivial phase contrast or structural complexity. This reduces the total spectral acquisition burden without compromising information integrity. The result is a high-throughput, label-free analysis pipeline capable of correlating structural and biochemical signatures at the single-particle level with enhanced spatiotemporal resolution.Example XXV: Surface Concentration-Dependent Raman Response and the Role of Interferometric Imaging

[0305] To assess the relationship between analyte surface concentration and Raman signal output, we performed a dilution series and measured the SERS intensity at 1576 cm-1across varying concentrations of vesicle suspensions immobilized on the substrate as depicted in FIG. 29. The resulting curve demonstrates a nonlinear response, characterized by an initial increase in signal intensity with concentration, followed by a decline beyond a critical threshold. This behavior is consistent with known SERS phenomena, where plasmonic hotspot saturation, inter-vesicle electromagnetic coupling, and multilayer stacking lead to destructive interference and signal quenching at high surface densities.

[0306] This nonlinear regime highlights the importance of precise particle enumeration and spatial resolution, which is enabled by the interferometric imaging subsystem. By providing quantitative maps of vesicle locations and surface coverage, interferometric imaging allows the platform to determine the actual number of surface-bound particles contributing to the Raman signal. This capability supports per-particle normalization and enables the exclusion ofoverloaded or undersaturated regions, thereby enhancing the reproducibility and interpretability of spectroscopic measurements.Example XXVII: Interferometric-Guided Control of Nanoparticle Surface Coverage for Linear SERS Quantification

[0307] In one exemplary embodiment, a method of characterizing nanoparticle coverage on a surface-enhanced Raman scattering (SERS) substrate comprises (1) preparing a series of suspensions of label-free nanoparticles at increasing dilution factors (l:x); (2) depositing each suspension onto a plasmonic substrate; and (3) measuring the Raman intensity at a characteristic band [as shown, for example, in the FIG. 29],

[0308] As the dilution factor increases, the density of particles on the substrate decreases, yielding three distinct regimes. A first regime is an overcrowded regime (low dilution; 1:1-1:4) in which particles stack and electromagnetically screen one another, abolishing the SERS enhancement and producing negligible Raman signal. A second regime is linear regime (moderate dilution; ~1:4-1:12) in which particles are sparsely distributed, and the measured Raman intensity falls linearly with surface particle density. A third regime is an underdense regime (high dilution; >1:12) in which insufficient particle coverage leads to loss of measurable SERS signal.

[0309] In one embodiment, interferometric imaging is performed simultaneously with SERS acquisition to monitor real-time particle coverage and hotspot distribution. By correlating the interferometric contrast with the Raman intensity curve, the operator can identify and maintain deposition within the linear regime— where signal scales predictably with particle number— for reliable quantitative analysis.Example XXVII: Integrated Platform for Multimodal Characterization Using Plasmonic Substrates, Surface-Enhanced Raman Spectroscopy, and Interferometric Imaging

[0310] In certain embodiments, the disclosed platform integrates a plasmonically active substrate, a Raman spectroscopic module, and an interferometric imaging system into a single, co-registered optical architecture for the comprehensive characterization of nanoscale biological and synthetic particles. The substrate comprises a planar or nanostructured surface- such as a SiO2thin film coated with a nanometrically controlled layer of silver or gold— engineered to support localized surface plasmon resonances (LSPR). These plasmonic hotspotsresult in electromagnetic field amplification, facilitating surface-enhanced Raman scattering (SERS) with sensitivity sufficient for single-molecule detection.

[0311] The Raman module— configured for either wide-field or targeted point-scan acquisition— leverages the SERS-active surface to extract molecular vibrational spectra with high signal-to-noise ratio (SNR) and minimal integration time. In parallel or sequence, an on-axis interferometric imaging subsystem is employed to generate quantitative phase-contrast maps of the same field of view, enabling high-resolution determination of particle localization, size, morphology, and refractive index contrast. This interferometric modality is particularly critical for distinguishing intact, surface-bound vesicles from free-floating cargo or debris and provides spatial priors to guide and optimize spectroscopic interrogation.

[0312] Together, this tri-modal configuration enables a synergistic workflow wherein interferometric imaging supplies quantitative biophysical metrics, the plasmonic substrate enhances molecular signal acquisition, and Raman spectroscopy provides label-free biochemical fingerprinting. The modalities operate on a common optical axis or spatial coordinate system, ensuring precise spatial co-registration and minimizing sample perturbation or misalignment. This integrated approach facilitates high-throughput, label-free, and multi-parametric analysis of extracellular vesicles, viruses, liposomes, and other nanoscale particles, thereby enhancing diagnostic accuracy, assay robustness, and translational applicability in clinical and research settings.Example XXVIII: Integration of Isolation and Detection Modules

[0313] In some exemplary forms, the platform can also include an integrated sample preparation in addition to sensing system designed to process complex biological fluids such as whole blood or plasma. This system enables the isolation of target particles including extracellular vesicles (EVs), nanosized vesicles, or viruses directly from the biological sample. The isolation process may involve one or more techniques such as filtration, centrifugation, immunoaffinity capture, or size exclusion chromatography to enrich the sample for particles of interest.

[0314] Following isolation, the enriched particles are introduced to the sensing substrate through a microfluidic delivery module. This microfluidic integration allows controlled, continuous, or batchwise introduction of the sample onto the sensing region. The sensingregion may include optically engineered surfaces optimized for interferometric imaging and / or surface-enhanced Raman spectroscopy (SERS), enabling simultaneous physical and chemical characterization.

[0315] The combination of microfluidics and integrated optical sensing supports high- throughput, label-free analysis of particles with minimal sample handling, making the platform suitable for diagnostic applications, liquid biopsy workflows, and point-of-care molecular profiling.Example XXIX: Fano Resonance Enhancement of SERS and Interferometric Signal

[0316] In this example, the platform incorporates Fano resonance-based plasmonic enhancement mechanisms to improve the sensitivity of both surface-enhanced Raman spectroscopy (SERS) and interferometric imaging. Fano resonance arises from the interference between a narrow discrete resonance, often supported by a localized plasmonic mode, and a broad spectral continuum, typically associated with a delocalized or propagating plasmonic mode or dielectric background. This interaction generates an asymmetric spectral line shape, which can be precisely tuned to amplify local electromagnetic fields.

[0317] For SERS, plasmonic nanostructures such as coupled nanoparticle dimers, metallic nanoclusters, or periodic metasurfaces are designed to exhibit strong Fano resonances. These configurations generate intense localized surface plasmon resonances (LSPRs) that create nanoscale electromagnetic "hot spots" at specific junctions or gaps. The resulting field enhancement significantly increases the Raman scattering cross-section of nearby particles, allowing for single-particle or single-molecule detection with high signal-to-noise ratio, even in complex biological environments.

[0318] In interferometric imaging, Fano-resonant plasmonic-dielectric hybrid structures or multilayer thin films can be used to modulate the amplitude and phase of scattered light from nanoscale objects. By tuning parameters such as incident wavelength, angle of illumination, or nanostructure geometry, the interference between resonant and non-resonant scattering pathways can be controlled, resulting in enhanced optical contrast and phase sensitivity. This allows for improved localization and sizing of weakly scattering particles, such as extracellular vesicles, viruses, or lipid nanoparticles.

[0319] The integration of Fano-resonant plasmonic structures into the sensing substrate provides dual-modal enhancement: increased molecular detection sensitivity through Raman enhancement and improved physical resolution through interferometric contrast optimization. This synergistic approach enables label-free, non-invasive, and high-throughput physical and chemical analysis of nanoscale biological particles with unprecedented precision.

[0320] Various additional aspects are now described which are contemplated as being workable variations on the disclosed concepts. As with the other aspects described above, they may be used alone or in combination with one or more of the various other aspects disclosed in various combinations and permutations to the extent they are workable.

[0321] According to one aspect, a device may comprise an integrated optical system capable of performing label-free, interferometric localization of nanovesicles [extracellular vesicles (EVs), virus etc.] and wide-field super-resolution Raman microscopy to determine both physical and chemical structure of nanovesicles.

[0322] According to another aspect, a diagnostic platform (device) is provided that combines interferometric imaging and surface-enhanced Raman spectroscopy (SERS) to provide simultaneous localization, structural, and compositional analysis of EVs for use in molecular diagnostics and genetic research.

[0323] According to another aspect, a method is provided for label-free localization of nanosized biological particles, such as EVs or viruses, using interferometric microscopy, wherein the particles are localized based on the measurement of phase differences between light waves scattered by the particles and a reference light.

[0324] According to another aspect, a process is provided for enhancing the resolution of interferometric imaging beyond the diffraction limit of light employing techniques selected from a group consisting of confocal interferometric imaging, tomographic interferometric imaging, and structured illumination interferometric imaging.

[0325] According to yet another aspect, a system is provided for utilizing artificial intelligence to predict the biomarkers or nucleic acid composition of intact biological particles, wherein the system comprises a computational package for image acquisition, image analysis, and spectral data correlation via neural network algorithms.

[0326] According to still another aspect, a method is provided for differential refractive index-based imaging and quantification of biological and synthetic particles using multiwavelength illumination to capture a series of images that reflect differential refractive index effects.

[0327] According to another aspect, a technique is provided for the characterization of the molecular composition of EV cargo by analyzing Raman spectra obtained from EVs or nanovesicles, wherein the technique includes the computational enhancement of spectral resolution using a fitting algorithm.

[0328] According to one aspect, a method is provided for enhancing the Raman signal of particles using a substrate coated with a thin film, such as SiC>2, to create enhanced Raman effects for the detection of low-abundance molecules.

[0329] Alternatively, in some forms, a surface substrate can be designed to enhance Raman spectroscopic analysis, wherein the substrate can be composed of various materials including an ordinary glass slide, a metal-coated surface with metals such as silver (Ag) or gold (Au) for surface-enhanced Raman spectroscopy (SERS), a thin film coating, or a structured surface featuring self-assembled nanoparticles.

[0330] According to another aspect, a method is provided to fast scan tissue samples for their cargo. Spatial and chemical composition of tissue slices can be obtained similar to a codex but giving chemical composition info additional to codex.

[0331] According to some aspects, fluorescent information from spatial Raman tissue scan can be correlated to specific fluorescent signals to gather more specific and extensive information about tissues and tissue slices.

[0332] According to some aspects, tissues and tissue slices can be scanned for various metals. As such a method for scanning tissue and tissue slices to detect and quantify the presence of various metals is contemplated.

[0333] In some forms, locally differences in tissues such as inflammation or cancer can be diagnosed.

[0334] According to some aspects, a method is disclosed to provide histopathological staining-free imaging of the sections.

[0335] According to some aspects, a method is provided for distinguishing between different types of vesicles, based on their unique interferometric and Raman spectral signatures.

[0336] According to another aspect, a method is provided for quantifying the cargo content within biological and synthetic particles, including the characterization of loading efficiency.

[0337] According to still another aspect, a method is provided for distinguishing between virus types from vesicles by analyzing the unique Raman spectral fingerprints and employing machine learning algorithms for accurate virus classification.

[0338] In some forms, this method can be operated in conjunction with various trapping modalities such as optical, acoustic, electrical and magnetic trapping. Utilizing optical trapping / tweezers, this method can enable the prediction of both the stiffness and size of trapped particles, while concurrently providing chemical fingerprint information.

[0339] In some forms, it can give information about single, a few or many nanoparticles.

[0340] According to some aspects, a method is provided for detecting cellular response to environmentally toxic agents, such as per- and poly-fluoroalkyl substances, utilizing extracellular vesicle. According to some other aspects, a method is provided for determining the liver hepatocyte toxicity in response to different drug concentrations.

[0341] According to another aspect, the prediction of pancreatic cancer risk via pancreatic cyst fluid analysis is contemplated. A method and artificial intelligence algorithm designed to predict subtypes of pancreatic cyst fluids is provided, providing operators with insights into the potential malignancy of the cyst.

[0342] An Al-based optimal prescription suggestion for diabetes via blood analysis is contemplated. A method and artificial intelligence algorithm devised to forecast the correlation between Raman spectral responses of patients across diverse racial and weight scales is provided. The classification algorithm can offer guidance for optimal prescriptions tailored to type two diabetes patients, recognizing that singular reliance on medication or exercise may not universally suit all populations.

[0343] Modelling drug-response using a novel algorithm is contemplated. An algorithm is contemplated for generating IC50 and EC50 curves by leveraging regression model outputs in response to Raman fingerprints of extracellular vesicles.

[0344] According to some aspects, a method is provided for establishing a protein Raman library and an algorithm for quantifying correlations between sample content and elements within said library.

[0345] According to yet another aspect, a method and algorithm provides computational scores to predict the nucleic acids (e.g. DNA, RNA) composition from intact biological and synthetic particles (such as vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, liposomes etc.) by artificial intelligence methods such as machine learning.

[0346] According to some aspects, a method is provided for structural optical sequencing of nucleic acids, proteins, and their hybrids (aptamers).

[0347] According to yet other aspects, wide-field rapid imaging of EVs with high spectral resolution and single-molecule sensitivity using super-resolution Raman.

[0348] For the surface, in various forms, it can be ordinary glass slide, it can be metal coated (Ag, Au) surface for SERS enhancement, it can be thin film, and / or it can be selfassembly nanoparticles.

[0349] According to some aspects, a method Is provided for characterization of drug ligand interactions such as binding, affinity, avidity, association, dissociation kinetics.

[0350] It is contemplated this can be used to study extracellular vesicle dynamics including characterization of EV secretion kinetics and their cargo.

[0351] According to some aspects, an optical or acoustic method is provided for characterization of electromechanics properties of biological and synthetic particles. In some aspects, a method and artificial intelligence algorithm is offered to provide tissue stiffness to predict tumor presence.

[0352] According to some aspects, a method and algorithm is provided to the use of the above properties for classification, sorting, enrichment, depletion of biological and synthetic particles of spherical and non-spherical morphologies in a microfluidic as well as meso-scale devices in combination with other methods or a single process.

[0353] All fiber portable Raman amplifiers are contemplated. According to some aspects, a method is provided to generate Raman spectra without any free-space element utilizing the feedback and cavity resonance mechanism of fiberoptics.

[0354] 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

1. CLAIMSWhat is claimed is:

1. A platform for characterizing particles including one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules, and / or liposomes in a non-invasive, high-throughput, and label-free manner, the platform comprising:(a) an interferometric microscopy module configured to localize and size individual particles; and(b) a wide-field or confocal microscopy module comprising a surface-enhanced Raman spectroscopy (SERS) system configured to acquire molecular signatures from localized particles.

2. The platform of claim 1, in which the surface-enhanced Raman spectroscopy (SERS) system is a wide-field super-resolution surface enhanced Raman spectroscopy (SERS) device.

3. The platform of claim 1, wherein the interferometric microscopy module is used to localize single biomolecules such as extracellular vesicles and the wide-field or confocal microscopy module is used to determine the type and / or contents of the biomolecules.

4. The platform of claim 3, wherein the interferometric microscopy module is used to estimate size and volume of one or more of the particles.

5. The platform of claim 1, further comprising a computational package that includes: image acquisition including interferometric microscopy data and wide-field or confocal microscopy data; image analysis including the interferometric microscopy data and the wide-field or confocal microscopy data; and neural network algorithms for characterization of the particles from integration and / or correlation of the interferometric microscopy data and wide-field or confocal microscopy data, which characterization may include, but is not limited to: identifying a size, shape, and location of one or more particles using interferometric microscopy data; identifying regions where one or more particles are clustered or isolated using interferometric microscopy data for further targeted wide-field or confocal microscopy analysis; processing wide field or confocal microscopy data to identify the molecular signatures of particles; and / or using machine learning algorithms to match spectra from the wide-field or confocal microscopy data to known structures from a preexisting database or data set.

6. A device comprising an integrated optical system capable of performing label- free, interferometric localization of nanovesicles and a wide-field super-resolution Raman microscopy to determine both physical and chemical structure of nanovesicles.

7. A diagnostic platform comprising interferometric imaging and surface-enhanced Raman spectroscopy (SERS) to provide simultaneous localization, structural, and compositional analysis of extracelluar vesicles for use in molecular diagnostics and genetic research, and omics research including but not limited to lipidomic and proteomic metabolism studies.

8. A method for characterizing particles including one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules, and / or liposomes in a non-invasive, high-throughput, and label-free manner, the method comprising: utilizing a merged set of imaging modalities including and capable of: interferometric microscopy; and wide-field or confocal microscopy, collecting images of the particles from the imaging modalities.

9. The method of claim 8, wherein a device for wide-field or confocal microscopy is wide-field super-resolution surface enhanced Raman spectroscopy (SERS).

10. The method of claim 9, wherein the wide-field super-resolution surface enhanced Raman spectroscopy provides rapid imaging of with high spectral resolution and single-molecule sensitivity.

11. The method of claim 9, wherein the particles including one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules and / or liposomes are nanosized biological particles.

12. The method of claim 11, wherein nanosized biological particles are localized label-free using interferometric microscopy and wherein the nanosized biological particles are localized based on the measurement of phase differences between light waves scattered by the nanosized biological particles, molecules and a reference light.

13. The method of claim 11, wherein the nanosized biological particles are extracellular vesicles.

14. The method of claim 11, wherein the nanosized biological particles are viruses.

15. The method of claim 11, wherein the nanosized biological particles are lipoproteins.

16. The method of claim 8, wherein the method further comprises detecting cellular response to environmentally toxic agents.

17. The method of claim 8, wherein the method further comprises determining the liver hepatocyte toxicity in response to different drug concentrations.

18. The method of claim 8, further comprising detecting of pancreatic cancer risk via pancreatic cyst fluid analysis.

19. The method of claim 18, further comprising using computer models to predict subtypes of pancreatic cyst fluids, providing insights into the potential malignancy of a cyst.

20. The method of claim 8, further comprising utilizing the merged set of imaging modalities to forecast the correlation between Raman spectral responses of patient blood samples across diverse racial, weight and age.

21. The method of claim 20, further comprising producing diagnostic suggestions for patients in which the patients are type-two diabetics.

22. A process for enhancing the resolution of interferometric imaging beyond the diffraction limit of light, the process comprising: employing techniques selected from a group consisting of confocal interferometric imaging, tomographic interferometric imaging, and structured illumination interferometric imaging.

23. A system for utilizing artificial intelligence to predict the biomarkers or nucleic acid composition of intact biological particles, the system comprising a computational package for image acquisition, image analysis, and spectral data correlation via statistical methods or machine learning algorithms.

24. A method for differential refractive index-based imaging and quantification of biological and synthetic particles, the method comprising: providing multiwavelength illumination to a substrate supporting the biological and synthetic particles; and capturing a series of images that reflect differential refractive index effects.

25. A method for characterization of a molecular composition of extracellular vesicle cargo by analyzing Raman spectra obtained from extracellular vesicles or nanovesicles after localizing the extracellular vesicles or nanovesicles using interferometric microscopy, the method comprising computational enhancing spectral resolution using a fitting algorithm.

26. A method for enhancing the Raman signal of low-abundance particles on a substrate, the method comprising: obtaining an image using surface-enhanced Raman spectroscopy of the low-abundance particles on the substrate, the substrate being coated with a thin film to create enhanced Raman effects.

27. The method of claim 26, wherein the substrate is glass and the thin film is SiO?.

28. The method of claim 26, wherein the thin film is a metallic deposition, a dielectric metallic thin film, and / or a semiconductor thin film.

29. The method of claim 26, wherein the substrate is glass, the thin film is SiCh, and the thin film is further coated by silver or gold.

30. The method of claim 29, wherein the thin film of SiCh is approximately 100 nm thick and the silver or gold on the thin film is approximately 5 nm thick.

31. The method of claim 26, where the thin film has a thickness below 1 pm.

32. A surface substrate designed to enhance Raman spectroscopic analysis, the substrate comprising various materials including an ordinary glass slide, a metal-coated surface with metals such as silver (Ag) or gold (Au) for surface-enhanced Raman spectroscopy (SERS), a thin film coating, or a structured surface featuring self-assembled nanoparticles.

33. A method to fast scan tissue samples for their biomolecular content, the method comprising: obtaining spatial and chemical composition of tissue slices.

34. The method of claim 33, further comprising collecting and correlating fluorescent information from a spatial Raman tissue scan to gather additional information about the tissue sample.

35. A method, the method comprising scanning tissue and tissue slices to detect and quantify the presence of various metals.

36. The method of claim 35, the method further comprising detecting local differences in tissue and tissue slices such as inflammation or cancer and establishing a diagnosis based on that detection.

37. The method of claim 35, wherein the method provides histopathological staining-free imaging of the tissue and tissue slices.

38. A method, the method comprising: obtaining interferometric and Raman spectral signatures of a sample containing one or more types of vesicles; and based on the interferometric and Raman spectral signatures obtained from the sample, distinguishing between different types of vesicles in the sample uniquely based on the interferometric and Raman spectral signatures.

39. The method of claim 38, further wherein a cargo content within biological and synthetic particles in the sample is quantified.

40. A method, the method comprising: obtaining interferometric and Raman spectral signatures of a sample containing one or more types of viruses; and based on the interferometric and Raman spectral signatures obtained from the sample, distinguishing between different types of viruses in the sample uniquely based on the Raman spectral fingerprints.

41. The method of claim 40, wherein distinguishing between different types of viruses in the sample uniquely based on the Raman spectral fingerprints employs machine learning algorithms for accurate virus classification.

42. A method of modeling drug-response using an algorithm using information collected from the method of claim 8, the method comprising generating IC50 and EC50 curves algorithmically by leveraging regression model outputs in response to Raman fingerprints of extracellular vesicles after normalizing the intensity to the total number of particles counted by the interferometric microscopy module.

43. A method of establishing a protein Raman library using the platform of claim 1, the method comprising producing the protein Raman library by collection of a set of images and quantifying correlations between sample content and elements within the library.

44. A method for providing computational scores to predict a nucleic acids composition from intact biological and synthetic particles using artificial intelligence, the method comprising: collecting a set of information from the merged set of imaging modalities from the method of claim 8; and applying artificial intelligence algorithms to the set of information to produce computational scores predictive of the nucleic acid composition from the intact biological and synthetic particles.

45. The method of claim 44, wherein intact biological and synthetic particles include one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules and liposomes.

46. A method for structural optical sequencing of nucleic acids, proteins, and their hybrids (aptamers), the method comprising: collecting a set of information from the merged set of imaging modalities from the method of claim 8; and optically sequencing the nucleic acids, proteins, and their hybrids (aptamers) based on the set of information.

47. A method for characterization of drug ligand interactions such as binding, affinity, avidity, association, dissociation kinetics using the method of claim 8.

48. A method for studying extracellular vesicle dynamics including characterization of extracellular vesicle secretion kinetics and their cargo using the method of claim 8.

49. A quantum-acoustic-enhanced interferometric Raman microscope, the microscope comprising: quantum-enhanced Raman spectroscopy to reduce quantum noise and improve the signal-to-noise ratio (SNR); surface acoustic wave (SAW)-based nanoparticle trapping to confine and dynamically manipulate particles within enhanced optical fields, increasing Raman interaction times without requiring plasmonic substrates; and optionally, interferometric imaging to provide simultaneous real-time size, shape, and molecular characterization.

50. The microscope of claim 49, wherein the interferometric imaging is present.

51. The microscope of claim 50, wherein a combination of the quantum-enhanced Raman spectroscopy and interferometric imaging provides simultaneous label-free molecular fingerprinting and high-sensitivity detection at the single-particle level.

52. The microscope of claim 49, wherein the quantum-enhanced Raman spectroscopy being used to reduce quantum noise and improve the signal-to-noise ratio (SNR) employs squeezed light and quantum correlations to increase Raman signal-to-noise ratio while minimizing photodamage.

53. A method for characterizing biomolecules, biological tissue sections, viruses, or other particles of interest in a non-invasive, high-throughput, and label-free manner, the method comprising: confining and / or dynamically manipulating biomolecules, biological tissue sections, viruses, or other particles of interest using surface acoustic wave (SAW)-based nanoparticle trapping; and imaging the biomolecules, biological tissue sections, viruses, or other particles of interest using quantum-enhanced Raman spectroscopy.

54. The method of claim 53, further comprising imaging the biomolecules, biological tissue sections, viruses, or other particles of interest using interferometric imaging.

55. A method for characterizing particles including one or more of vesicles, viruses, exosomes, extracellular vesicles, lipid nanoparticles, capsids, vaccine carriers, bacteria, protein aggregates, lipoproteins, nano-microplastics, biomolecules and / or liposomes in a non-invasive, high-throughput, and label-free manner, the method comprising: obtaining a set of wide-field or confocal microscopy images; and applying artificial intelligence algorithms to the set of wide-field or confocal microscopy images to produce a predictive library.

56. The method of claim 55, further comprising: obtaining and analyzing a subsequent wide-field or confocal microscopy image; and comparing the subsequent wide-field microscopy image to the predictive library to assess and classify the subsequent wide-field microscopy image.

57. The method of claim 55, further comprising suspending particles in a fluid during at least part of the step of obtaining a set of wide-field or confocal microscopy images.

58. The method of claim 57, wherein the fluid is a suspension medium and the suspension medium is subjected to mechanical or acoustic actuation.

59. The method of claim 58, wherein the mechanical or acoustic actuation facilitates averaging during dynamic Raman readout during obtaining a set of wide-field or confocal microscopy images.

60. The method of claim 57, wherein the method includes two operational modes for imaging including (a) a dry surface mode in which isolated particles are immobilized onto a solid substrate and (b) a suspension dynamic mode in which particles remain within a fluid medium during measurement, allowing real-time detection of free-floating or loosely bound particles.

61. The method of claim 57, further comprising flowing the particles suspended in a fluid on a surface during at least part of the step of obtaining a set of wide-field or confocal microscopy images.

62. The method of claim 57, further comprising flowing the particles suspended in a fluid through a channel during at least part of the step of obtaining a set of wide-field or confocal microscopy images.

63. The method of claim 56, wherein the method employs self-assembled plasmonic substrates composed of gold or silver nanoparticles that have been deposited on the sensing surface via controlled drying, solvent evaporation, or chemical functionalization techniques, allowing them to spontaneously organize into densely packed nanoclusters.

64. The method of claim 56, wherein the method involves characterizing and differentiating between HIV clades and to stratify patient-specific viral load levels using optical fingerprinting.

65. The method of claim 56, further comprising obtaining further clinical imaging data corresponding to the set of wide-field or confocal microscopy images; wherein applying artificial intelligence algorithms to the set of wide-field or confocal microscopy images to produce a predictive library involves further providing the clinical imaging data corresponding to the set of wide-field or confocal microscopy images to the artificial intelligence algorithms.

66. The method of claim 65, wherein the further clinical imaging data is radiomics data.

67. The method of claim 66, wherein the radiomics data is one or more of PET / CT and / or PCCT lung nodule imaging.

68. The method of claim 56, wherein the method involves characterizing extracellular vesicles and detecting changes overall in the extracellular vesicles across a patient which extracellular vesicles are changed by a presence of disease or virus in a patient.

69. The method of claim 56, wherein the method involves characterizing and differentiating between different types of viruses.

70. The method of claim 56, wherein the method involves characterizing and distinguishing between extracellular vesicles, lipoproteins, and other blood components.

71. The method of claim 56, wherein the method involves characterizing and identifying patient-derived extracellular vesicles as part of a non-invasive biomarker platform for early lung nodule characterization.

72. The method of claim 56, further comprising detecting nodules by integrating in a machine learning format radiomics features of medical images as well as -omic data of the patient tissue or liquid biopsy samples.

73. The method of claim 56, wherein the method involves direct detection of extracted RNA in a dry assay format.

74. The method of claim 56, comprising detecting not only extracellular vesicles, but all extracted components including all cargo components including but not limited to membranes, proteins, lipids, RNA, DNA that are isolated as a part of extracellular vesicle isolation.

75. The method of claim 56, further comprising: subjecting the particles to a controlled actuation waveform; acquiring a z-stack of Raman spectra synchronously with the actuation waveform, capturing temporal variations in the Raman signal intensity at each wavenumber; analyzing the acquired z-stack of Raman spectra by correlating the intensity fluctuations at each spectral position with the applied actuation waveform; wherein this dynamic correlation strategy allows for decomposition of Raman data into distinct sample and background spectra, leading to improved detection sensitivity and molecular specificity.

76. The method of claim 56, further comprising the step of determining, estimating, and / or calculating the volume, surface area, and / or mass of the particles being detected per area; providing the volume, surface area, and / or mass of particles being detected per area to an artificial intelligence algorithm; and applying the artificial intelligence algorithm to the set of wide-field or confocal microscopy images along with the corresponding volume, surface area, and / or mass of particles to produce the predictive library.

77. The method of claim 56, further comprising the step of determining, estimating, and / or calculating, the volume, surface area, and / or mass of particles being detected per area and then adjusting a concentration of the particles in a subsequently prepared sample to optimize the Raman signal in the subsequently prepared sample.

78. The method of claim 56, further comprising the step of determining, estimating, and / or calculating the volume, surface area, and / or mass of the particles being detected per area and scaling the Raman signal in the area being scanned.

79. The method of claim 56, further comprising obtaining an interferometric image from the same sample and same location as a corresponding wide-field or confocal microscopy image.

80. The method of claim 79, wherein obtaining the interferometric image from the same sample and same location as a corresponding wide-field or confocal microscopy image occurs sequentially with the obtaining of the wide-field or confocal microscopy image.

81. The method of claim 79, wherein obtaining the interferometric image from the same sample and same location as a corresponding wide-field or confocal microscopy image occurs simultaneously with the obtaining of the wide-field or confocal microscopy image.

82. The method of claim 79, wherein the interferometric image is also supplied to the predictive library as a parameter to assess and classify the subsequent wide-field or confocal microscopy image.

83. The method of claim 56, further comprising obtaining a plasmonic image from the same sample and same location as a corresponding wide-field or confocal microscopy image.

84. The method of claim 56, further comprising plasmonically activating a surface on which the particles are located.

85. The method of claim 84, wherein the plasmonic activation of the surface is an additional parameter provided to an artificial intelligence or machine learning algorithm to develop the predictive library and / or is used to enhance a Raman signal during obtaining of the wide-field or confocal microscopy image and provide an improved signal to noise ratio.

86. The method of claim 56, further comprising obtaining an interferometric image from the same sample and same location as a corresponding wide-field or confocal microscopy image and plasmonically activating a surface on which the particles are located.

87. The method of claim 86, wherein obtaining a set of wide-field or confocal microscopy images, obtaining an interferometric image from the same sample and same location as the corresponding wide-field or confocal microscopy image, and plasmonically activating a surface on which the particles are at least three of the components that are provided to an artificial intelligence algorithm to produce a predictive library.

88. A method comprising: transporting or flowing particles through a channel having an excitation region, exciting particles in the excitation region, and obtaining Raman spectra are continuously or periodically while the particles are in the excitation region.

89. The method of claim 88, further comprising: sorting the particles based on their characterization from the Raman spectra after the particles have been further flowed or transported from the excitation region.

90. The method of claim 88, wherein exciting particles in the excitation region involves one or more of acoustic, electromagnetic, mechanical, and antenna-based excitation that affects the sensing surface in any shape or form, causing a fluctuation in a Raman signal that allows better sensitivity, specificity and improved signal to noise ratio as well as different forms of averaging and resolution enhancement algorithms.

Images