Photoacoustic label-free detection and classification of targets with a whispering gallery mode resonator

Abstract

A system and corresponding method of an optofluidic, high throughput, optical microresonator sensor with an extended sensing range, enabling real-time, label-free detection and interrogation of analytes in their native medium environments. The system includes a sensor such as a high-quality, WGM microresonator and / or a high-quality optical resonator, such as Fabry-Perot resonator, connected to a microfluidic channel that allows sample media to flow inside while being optically stimulated by a modulated light source to generate acoustic waves through photoacoustic effects detected by such optical microresonator sensors.

Description

CTPHOTOACOUSTIC LABEL-FREE DETECTION AND CLASSIFICATION OF TARGETS WITH A WHISPERING GALLERY MODE RESONATORCROSS-REFERENCE TO RELATED APPLICATION

[0001] This application claims the benefit of priority to U. S. Provisional Application No. 63 / 722,351, filed November 19, 2024, the content and disclosure of which are incorporated herein by reference in their entirety.FIELD OF THE DISCLOSURE

[0002] The field of the invention relates generally to sensors, sensor designs (e.g., thickwall design, microfluidic design), optical materials, microresonators, whispering gallery modes, and sensing systems empowered by artificial intelligence, for use in a variety of applications, including classification of nanoparticles and other biological materials (e.g., blood cells) in connection with biological sensing applications, as well as other materials.BACKGROUND OF THE DISCLOSURE

[0003] Particles in micro- and nano-scale are abundant in nature and play critical roles in diverse biological and environmental processes. Beyond their natural presence, these tiny particles are also fundamental in a wide range of scientific research and industrial applications. Due to their small size, they possess unique physical, chemical, and biological characteristics markedly distinct from their bulk counterparts. Therefore, detecting, characterizing, and identifying these particles accurately and efficiently is crucial. In the field of medicine, for instance, nanoparticles are indispensable in targeted drug delivery systems, therapies, and diagnostics. They can interact with biological systems at the molecular level, enhancing treatment precision and efficacy. In photothermal therapy, nanoparticles are used as agents that absorb light and convert it into heat, specifically targeting and destroyingCT cancer cells. In material science, nanomaterials exhibiting properties such as increased strength, chemical reactivity, or electrical conductivity open up new avenues for innovation in electronics, optics, and energy storage. Given that the properties of particles depend on their composition, size, and shape, characterizing these particles becomes vital for understanding and leveraging the structure-property relationship.

[0004] Spectroscopic techniques like Raman, photoacoustic (PA), and UV-Vis spectroscopy are widely used for characterizing particles and molecules. Raman spectroscopy, for instance, relies on the Raman shift, which results from photons interacting with molecular vibrations and other low-frequency modes in the sample. The Raman spectrum can provide a direct molecular fingerprint that reflects the composition of materials. However, the detection efficiency in Raman spectroscopy can be low, posing a challenge, especially for nanoscale objects. Raman scattering from water can obscure signals from target analytes in aqueous solutions, posing a challenge for conventional spectroscopic measurements. To address such limitations, various Artificial Intelligence (Al) techniques have recently been explored to enhance the performance of conventional sensing and spectroscopy techniques. In parallel, PA spectroscopy offers a complementary approach, combining optical excitation with acoustic detection, where short-pulsed lasers probe targets through photon absorption and subsequent thermoelastic expansion. Analyzing the frequency domain features of PA signals provides insights into the shape, size, and orientation of microstructures, and acoustic scattering properties, as well as various biophysical properties of the samples, making it a powerful tool for detailed characterization. However, currently, this technique is mainly limited to nanoparticles with strong absorption properties, which are used as contrast agents for PA imaging, due to the lack of highly sensitive sensors capable of detecting very weak acoustic waves generated by nanoparticles on the nano- or micro-scales.CT

[0005] On the other hand, optical microsensors have emerged as a promising sensing technology in recent years, demonstrating great potential in small particle sensing due to their high sensitivity, rapid response, versatility, and affordability. The fundamental mechanism of detection involves analyzing alterations in optical signals as particles interact with light confined in the sensor. This provides information on various aspects of the target, such as its existence, concentration, and size. Enhancing the light-matter interaction enables higher sensitivity; strategies including resonator-based sensing, plasmonic enhancement, and microlasers have been widely exploited, demonstrating the most demanding biosensing tasks such as the detection of single virus particles and other nanoparticles. However, in some cases, targets need to be in close proximity to or in direct contact with the microsensor’s surface for detection. For example, optical resonator sensors use evanescent fields to probe the surrounding media, and plasmonic-enhanced sensors rely on “hot spots,” posing a challenge in sample collection. Away from the surface, free-flowing particles in the sample medium remain undetected (as shown in FIG. 13 at element 1302). This means that only a small fraction of the target particles in the sample can be captured and analyzed, thereby limiting their detection efficiency and capacity for high-throughput sensing. These limitations highlight that current microsensor approaches remain fundamentally constrained by their reliance on near-surface interrogation, leaving free-flowing particles in the bulk medium largely undetected. In addition, measurements based on light intensity (phase) and resonance shift provide little identifiable difference in the properties of different molecules or particles. While individual nanoparticles can be characterized by tracking their Brownian motion, the throughput is limited. To achieve selectivity (specificity) in label-free sensing, target molecules need to bind to receptors that are functionalized on the limited sensing surface. The need for surface binding and surface functionalization can introduce additional complexities and challenges, including surface fouling (non-specific binding), massCT transport limitations (slow random diffusion rate, limited the speed, sensitivity, and throughput), and surface regeneration (bound particles must be removed to reuse).

[0006] Moreover, to date, it is challenging for most optical microsensors, which are vulnerable to contamination to obtain accurate measurements or detect specific targets in complex samples such as whole blood. The presence of numerous interfering substances, such as metabolites, electrolytes, and other biomolecules, can interact with the optical mode via evanescent fields, leading to false-positive or false-negative results and compromising the specific and sensitivity of the detection system. To ensure accuracy, samples often necessitate the implementation of intricate, multi-step sample purification protocols, including incubation and washing. Such procedural complexities pose significant challenges to their practical deployment in clinical and industrial environments.

[0007] In addition, the complex nature of the sample matrix can cause signal attenuation, optical absorption and scattering, or background noise, further obscuring the target signal and making it difficult to obtain accurate quantitative measurements. Therefore, there is a strong need for label-free sensing technologies that can detect target particles directly in complex solutions, without the need for surface binding and purification. Such techniques would enable the detection and analysis of target particles in their native environment, with benefits including higher throughput, simpler and faster detection, improved sensitivity, and greater overall efficiency in sensing applications such as biosensing.

[0008] Additionally, conventional whispering gallery mode (WGM) sensing platforms rely heavily on the direct interaction between light and target, such as in the detection of biomolecules bound to the sensor area. The small sensing area restricts their practical applications in various sensing scenarios. The WGM field (which is the sensor area) is very small in these devices, ranging in area from 0.002 mm2to 0.004 mm2.

[0009] What is needed is improved: 1) detection and classification of (e.g., gold) nanoparticles or molecules (e.g., free-flowing in solution, anchored on surface, or free gasCT molecules), from their photoacoustic signals inside of a thick-walled microbubble resonator (MBR); 2) specific detection and classification of different types of materials, including biological materials such as red blood cells and / or chemical / mechanical materials such as gold nanoparticles and / or dye molecules (e.g., Rodomine 6G dye molecules) by its / their PA signals without the addition of PA contrast agents; 3) design of microbubble resonators with a thick wall to fully protect high-quality optical modes (e.g., within a thick wall), which is critical for the sustained effectiveness of the sensors in various sensing applications over an extended period; (4) detect acoustic waves generated by particles in solution flowing freely inside a capillary structure, which helps improve specificity and reliability by making it immune to dispersive (refractive index changes) and dissipative changes (loss) in the solution; (5) improved (e.g., increased) effective sensing area of a sensor device to increase the ability of (e.g., WGM MBR, WGM hollow core, WGM microsphere, microring, or other types of optical resonators that can enhance light-matter interactions through their high- quality optical modes) sensors to detect biological targets (e.g., improve their biological detection sensitivity by means of identifying specific frequency components of photoacoustic targets of interest (e.g., gold nanoparticles and animal red blood cells); and (6) specific detection of photoacoustic signals for purposes of spectroscopy in both liquid and gaseous phase (e.g., the specific detection and classification of gas molecules for applications in environmental, health and industrial monitoring).

[0010] This background section is intended to introduce the reader to various aspects of art that may be related to various aspects of the present disclosure, which are described and / or claimed below. This discussion is believed to be helpful in providing the reader with background information to facilitate a better understanding of the various aspects of the present disclosure. Accordingly, it should be understood that these statements are to be read in this light, and not as admissions of prior art.CTSUMMARY

[0011] In one aspect, a sample medium testing system is provided. The sample medium testing system includes a sensor. The sensor includes a body including a microfluidic channel therein, a microresonator, and a sample medium contained within the microfluidic channel, the sample medium including at least one target of interest provided therein. The sample medium testing system further includes a modulated light source configured to emit modulated light to provide optical stimulation to at least one target of interest present within the sample medium. The sample medium testing system further includes a detector configured to detect transmitted light signals from the microresonator that carry information associated with properties of acoustic waves associated with the at least one target of interest. The sample medium testing system further includes a control device including one or more processors programmed to control operation of the modulated light source and to acquire data and analyze results from testing of the sample medium. The control device is configured to control the modulated light source to optically stimulate the at least one target of interest via the modulated light. The sensor is configured to allow the sample medium to flow inside the microfluidic channel while being optically stimulated by the modulated light source from outside the microfluidic channel to generate acoustic waves through photoacoustic effects. The photoacoustic effects are based at least in part on material properties of the at least one target of interest. The control device classifies the at least one target of interest based on analysis of detections of the acoustic waves obtained from the detector.

[0012] In another aspect, a non-transitory computer-readable recording medium having computer executable instructions stored thereon is provided. The instructions, when executed by a processor of a sample solution testing system, cause the processor to control operation of a modulated light source associated with the sample solution testing system. The instructions, when executed by the processor of the sample solution testing system, further cause the processor to acquire data associated with results from a test of a sample solutionCT using the modulated light source, the modulated light source being used in association with a sensor of the sample solution testing system for the test of the sample solution. The sensor includes a body including a microfluidic channel, the sample solution being contained within the microfluidic channel during the test and including at least one target of interest and a whispering gallery mode (WGM) microresonator. The modulated light source is configured to emit modulated light to provide optical stimulation to the at least one target of interest. The instructions, when executed by the processor of the sample solution testing system, further cause the processor to control the modulated light source to optically stimulate the at least one target of interest via the modulated light from outside the microfluidic channel to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one target of interest. The instructions, when executed by the processor of the sample solution testing system, further cause the processor to classify the at least one target of interest based on analysis of the acoustic waves.

[0013] In yet another aspect, a computer-implemented method of testing a sample via a sensor, the sensor including a body including a microfluidic channel therein, a whispering gallery mode (WGM) microresonator, and the sample being contained within the microfluidic channel, the sample including at least one target of interest provided therein. The computer-implemented method includes providing the sensor. The computer- implemented method further includes providing a modulated light source, the modulated light source configured to emit modulated light toward the body of the sensor. The computer- implemented method further includes testing the sample, including optically stimulating, via the modulated light, the at least one target of interest to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one target of interest. The computer-implemented method further includes classifying the at least one target of interest based on analysis of at least one of the photoacoustic effects and the acoustic waves.CT

[0014] Additional aspects are listed below.

[0015] In one additional aspect, a sensor testing system including a sensor. The sensor includes a capillary including a microfluidic channel therein; a microresonator; and a sample medium contained within the microfluidic channel, the sample medium including at least one target of interest provided therein. The sensor testing system further includes a modulated light source, the modulated light source configured to emit modulated light to provide optical stimulation to at least one target of interest; and a control device including one or more processors programmed to control operation of the modulated light source and to acquire data and analyze results from testing of the sensor. Therein the control device is configured to control the modulated light source to optically stimulate the at least one target of interest via the modulated light, the sensor is configured to allow the sample medium to flow inside the microfluidic channel while being optically stimulated by the modulated light source from outside the microfluidic channel to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one target of interest, and the control device classifies the at least one target of interest based on analysis of at least the acoustic waves.

[0016] In another additional aspect, a non-transitory computer-readable recording medium having computer executable instructions stored thereon, which when executed by a processor of a sensor testing system, cause the processor to: control operation of a modulated light source of the sensor testing system and to acquire data and analyze results from testing of a sensor via the sensor testing system. The sensor includes (a) a capillary including a microfluidic channel therein; (b) a whispering gallery mode (WGM) microresonator; and (c) a sample solution contained within the microfluidic channel, the sample solution including at least one target of interest provided therein. The modulated light source is configured to emit modulated light to provide optical stimulation to the at least one target of interest. The processor is further caused to: (i) control the modulated light source to optically stimulateCT the at least one target of interest via the modulated light from outside the microfluidic channel to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one target of interest, and (ii) classify the at least one target of interest based on analysis of at least the acoustic waves.

[0017] In yet another additional aspect, a method of testing a sensor, the sensor including (i) a capillary including a microfluidic channel therein; (ii) a whispering gallery mode (WGM) microresonator; and (iii) a sample solution contained within the microfluidic channel, the sample solution including at least one target of interest provided therein. The method includes: (a) providing the sensor; (b) providing a modulated light source, the modulated light source configured to emit modulated light; (c) exciting, via a laser beam of a probe laser, whispering gallery modes in the WGM microresonator; (d) optically stimulating, via the modulated light laser beam, the at least one target of interest to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one target of interest; and (e) classifying the at least one target of interest based on analysis of at least the acoustic waves.

[0018] In yet another additional aspect, a sensor testing system including: (i) a first laser configured to emit a laser beam having a first wavelength to excite WGMs; (ii) a sensor, the sensor including a capillary including a microfluidic channel therein, a whispering gallery mode (WGM) microresonator including a core connected to the microfluidic channel, a sample medium contained within the microfluidic channel, the sample medium including at least one contrast agent suspended therein; (iii) a second laser different than the first laser, the second laser configured to emit a pulsed laser beam to provide optical stimulation to the targets of interests or structures attached to at least one contrast agent; and (iv) a computing system including one or more processors programmed to control operation of the first laser and the second laser and to analyze results from testing of the sensor. The computing system is configured to control the second laser to optically stimulate the at least one contrast agentCT via the pulsed laser beam, the sensor is configured to allow the sample medium to flow inside the microfluidic channel while being optically stimulated by the pulsed laser beam from outside the microfluidic channel to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the sensing targets or a contrast agent attached to the targets of interest, and the computing system classifies the sensing targets based on analysis of at least the acoustic waves.

[0019] In yet another additional aspect, a non-transitory computer-readable recording medium having computer executable instructions stored thereon, which when executed by a processor of a sensor testing system, cause the processor to: control operation of a first laser and a second laser of the sensor testing system and to analyze results from testing of a sensor via the sensor testing system, wherein: (i) the first laser is configured to emit a laser beam having a first wavelength; (ii) the sensor includes a capillary including a microfluidic channel therein, a whispering gallery mode (WGM) microresonator including a core connected to the microfluidic channel, and a sample medium contained within the microfluidic channel, the sample medium including at least one contrast agent suspended therein; (ii) the second laser is configured to emit a pulsed laser beam to provide optical stimulation to the at least one contrast agent; and (iii) the processor is further caused to control the second laser to optically stimulate the at least one contrast agent via the pulsed laser beam from outside the microfluidic channel to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one contrast agent, and classify the at least one contrast agent based on analysis of at least the acoustic waves.

[0020] In yet another additional aspect, a method of testing a sensor, the sensor including: (i) a capillary including a microfluidic channel therein; (ii) a whispering gallery mode (WGM) microresonator including a core connected to the microfluidic channel; and (iii) a sample medium contained within the microfluidic channel, the sample mediumCT including at least one contrast agent suspended therein. The method further includes: providing the sensor; providing a first laser configured to emit a laser beam having a first wavelength; providing a second laser different than the first laser, the second laser configured to emit a pulsed laser beam to provide optical stimulation to the at least one contrast agent; exciting, via the laser beam, whispering gallery modes in the WGM microresonator; optically stimulating, via the pulsed laser beam, the at least one contrast agent to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one contrast agent; and classifying the at least one contrast agent based on analysis of at least the acoustic waves.

[0021] Various refinements exist of the features noted in relation to the above-mentioned aspects. Further features may also be incorporated in the above-mentioned aspects as well. These refinements and additional features may exist individually or in any combination. For instance, various features discussed below in relation to any of the illustrated embodiments may be incorporated into any of the above-described aspects, alone or in any combination.DESCRIPTION OF DRAWINGS

[0022] The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.

[0023] The accompanying drawings, which are incorporated in and form a part of the specification, illustrate the embodiments of the present disclosure and together with the description, serve to explain the principles of the disclosure. Those of skill in the art will understand that the drawings, described below, are for illustrative purposes only. The drawings are not intended to limit the scope of the present disclosure in any way.CT

[0024] FIG. 1 A illustrates a schematic of an example sensing system using photoacoustic(PA) contrast agents in a whispering gallery mode (WGM) microbubble (MBR) resonator (e.g., WGM-MBR resonator) according to one embodiment of the disclosure.

[0025] FIG. IB illustrates an example photoacoustic process inside of gold nanoparticles used in conjunction with a resonator according to one embodiment of the disclosure.

[0026] FIG. 1C illustrates example output signals of a resonator vibrated by acoustic waves according to one embodiment of the disclosure.

[0027] FIG. ID illustrates a schematic of additional aspects of the sensing system shown in FIG. 1 A, according to one embodiment of the disclosure.

[0028] FIG. 2A illustrates example temporal and spectral measurements of nanospheres according to one embodiment of the disclosure.

[0029] FIG. 2B illustrates example temporal and spectral measurements of nanorods according to one embodiment of the disclosure.

[0030] FIG. 2C illustrates example temporal and spectral measurements of nanocubes according to one embodiment of the disclosure.

[0031] FIG. 2D illustrates example temporal and spectral measurements of nanoshells according to one embodiment of the disclosure.

[0032] FIG. 3A illustrates example time domain photoacoustic signals with calculated frequency domain of red blood cells of a first animal species (pig) according to one embodiment of the disclosure.

[0033] FIG. 3B illustrates example time domain photoacoustic signals with calculated frequency domain of red blood cells a second animal species (sheep) according to one embodiment of the disclosure.

[0034] FIG. 3C illustrates example time domain photoacoustic signals with calculated frequency domain of red blood cells of a third animal species (goat) according to one embodiment of the disclosure.CT

[0035] FIG. 3D illustrates example time domain photoacoustic signals with calculated frequency domain of red blood cells of a fourth animal species (turkey) according to one embodiment of the disclosure.

[0036] FIG. 3E illustrates example time domain photoacoustic signals with calculated frequency domain of red blood cells of a fifth animal species (llama) according to one embodiment of the disclosure.

[0037] FIG. 3F illustrates example Fast Fourier Transform (FFT) spectra of individual measurements of red blood cells from a same animal species (sheep) according to one embodiment of the disclosure.

[0038] FIG. 3G is a diagram illustrating a schematic overview of the classification process and convolutional neural network (CNN) with a learning structure according to one embodiment of the disclosure.

[0039] FIG. 3H is a diagram illustrating principal component analysis (PCA) visualization of samples without machine learning-based feature extraction according to one embodiment of the disclosure.

[0040] FIG. 31 is a diagram illustrating PCA visualization of samples using features extracted by a machine learning algorithm.

[0041] FIG. 4A illustrates example WGM spectra when a core of a sensor is filled with various solutions according to one embodiment of the disclosure.

[0042] FIG. 4B illustrates an example simulation of field distribution of a WGM according to one embodiment of the disclosure.

[0043] FIG. 5A is a schematic illustrating scanning of a pulsed laser along a microfluidic channel, with whispering gallery modes (WGMs) being excited near the MBR equator while a sample medium flows through its core, according to one embodiment of the disclosure.

[0044] FIG. 5B illustrates example time delays observed between a pulse trigger and a detected peak of PA signals as a laser beam spot traverses along the microfluidic channel inCT accordance with the schematic shown in FIG. 5A, according to one embodiment of the disclosure.

[0045] FIG. 5C illustrates example corresponding peak-to-peak amplitudes of the measured signals during the scanning in accordance with the schematic shown in FIG. 5A, according to one embodiment of the disclosure.

[0046] FIG. 6A illustrates an example PA signal response of gold nanospheres as the concentration is varied from a first concentration to a second concentration according to one embodiment of the disclosure.

[0047] FIG. 6B illustrates an example PA signal response of gold nanospheres at a certain concentration under increasing power delivered by a pulsed laser according to one embodiment of the disclosure.

[0048] FIG. 7 is a block diagram schematically illustrating an example computing system in accordance with one embodiment of the disclosure.

[0049] FIG. 8 illustrates an example component configuration of a computing device according to one embodiment of the disclosure.

[0050] FIG. 9 illustrates an example configuration of a remote or user computing device according to one embodiment of the disclosure.

[0051] FIG. 10 illustrates an example configuration of a server system according to one embodiment of the disclosure.

[0052] FIG. 11 illustrates one machine learning (ML) approach according to one embodiment of the disclosure.

[0053] FIG. 12 is a flow diagram of an example method according to one embodiment of the disclosure.

[0054] FIG. 13 is a diagram illustrating a principle of long-range acoustic-assisted sensing according to one embodiment of the disclosure.CT

[0055] FIG. 14 is a diagram illustrating a schematic of a sensing platform using PA in an optofluidic microresonator, including a photoacoustic process of analytes according to one embodiment of the disclosure.

[0056] FIG. 15 is a diagram illustrating selective detection of red blood cells in a complex blood matrix according to one embodiment of the disclosure.

[0057] FIG. 16A is a diagram illustrating time domain photoacoustic signals and corresponding frequency spectra of whole blood samples from pigs according to one embodiment of the disclosure.

[0058] FIG. 16B is a diagram illustrating time domain photoacoustic signals and corresponding frequency spectra of whole blood samples from sheep according to one embodiment of the disclosure.

[0059] FIG. 16C is a diagram illustrating time domain photoacoustic signals and corresponding frequency spectra of whole blood samples from turkeys according to one embodiment of the disclosure.

[0060] FIG. 16D is a diagram illustrating time domain photoacoustic signals and corresponding frequency spectra of whole blood samples from goats according to one embodiment of the disclosure.

[0061] FIG. 16E is a diagram illustrating time domain photoacoustic signals and corresponding frequency spectra of whole blood samples from horses according to one embodiment of the disclosure.

[0062] FIG. 16F is a diagram illustrating PC A visualization of samples using features extracted by the machine learning algorithm according to one embodiment of the disclosure.

[0063] FIG. 17A is a diagram illustrating conventional resonance shift sensing mechanism according to one embodiment of the disclosure.

[0064] FIG. 17B is a diagram illustrating acoustic-mediated sensing according to one embodiment of the disclosure.CT

[0065] FIG. 18 is a diagram illustrating a machine learning pipeline for classifying PA signals according to one embodiment of the disclosure.

[0066] FIG. 19A is a diagram illustrating a confusion matrix of a machine learning models for gold nanoparticles according to one embodiment of the disclosure.

[0067] FIG. 19A is a diagram illustrating a confusion matrix of a machine learning models for red blood cells according to one embodiment of the disclosure.

[0068] FIG. 20A is a diagram illustrating a confusion matrix on red blood cells according to one embodiment of the disclosure.

[0069] FIG. 20B is a diagram illustrating a result with prototype learning according to one embodiment of the disclosure.

[0070] FIG. 21A is a diagram illustrating PCA without prototype learning, including input features from red blood cell signals according to one embodiment of the disclosure.

[0071] FIG. 2 IB is a diagram illustrating PCA without prototype learning, including input features from whole blood signals according to one embodiment of the disclosure.

[0072] FIG. 22A is a diagram illustrating deviations in PA signals for whole blood according to one embodiment of the disclosure.

[0073] FIG. 22B is a diagram illustrating deviations in PA signals for a red blood cell sample of a goat according to one embodiment of the disclosure.

[0074] FIG. 23 is a diagram illustrating a microring resonator on a substrate according to one embodiment of the disclosure.

[0075] FIG. 24 is a diagram illustrating an on-chip sensor for droplet sample sensing according to one embodiment of the disclosure.

[0076] FIG. 25 is a diagram illustrating an on-chip sensor integrated with microfluidics according to one embodiment of the disclosure.

[0077] FIG. 26 is a diagram illustrating a wearable sensor using flexible photonic structures according to one embodiment of the disclosure.CT

[0078] FIG. 27 is a diagram illustrating a microsphere fiber probe according to one embodiment of the disclosure.

[0079] FIG. 28A is a diagram illustrating an endoscopic probe that may be implemented as a PA sensor, such as a microsphere fiber probe scanning through a sample medium, according to one embodiment of the disclosure.

[0080] FIG. 28B is a diagram illustrating an endoscopic probe that may be implemented as a PA sensor, such as a microsphere fiber probe scanning through a medium / vessel such as gas pipes or blood vessels to detect PA signals, according to one embodiment of the disclosure.

[0081] FIG. 29A is a diagram illustrating an endoscopic probe that may be implemented as a PA sensor, such as a microsphere fiber probe for environmental monitoring in water, according to one embodiment of the disclosure.

[0082] FIG. 29B is a diagram illustrating an endoscopic probe that may be implemented as a PA sensor, such as a microsphere fiber probe scanning through a water pool, according to one embodiment of the disclosure.

[0083] FIG. 30 is a diagram illustrating a WGM resonator sensor including a microbubble structure with WGM.

[0084] FIG. 31 is a diagram illustrating a WGM resonator sensor including a capillary structure with WGM.

[0085] FIG. 32 is a diagram illustrating a WGM resonator sensor including a microbubble array including microbubble structures with corresponding WGMs according to one embodiment of the disclosure.

[0086] FIG. 33 is a diagram illustrating a WGM resonator sensor including a capillary structure with multiple WGMs configured as a sensor array according to one embodiment of the disclosure.CT

[0087] FIG. 34 is a diagram illustrating an on-chip sensor array for multiplexing according to one embodiment of the disclosure.

[0088] FIG. 35 is a diagram illustrating a sensor (e.g., a Fabry-Perot resonator / interferemoter sensor) according to one embodiment of the present disclosure.

[0089] There are shown in the drawings arrangements that are presently discussed, it being understood, however, that the present embodiments are not limited to the precise arrangements and are instrumentalities shown. While multiple embodiments are disclosed, still other embodiments of the present disclosure will become apparent to those skilled in the art from the following detailed description, which shows and describes illustrative aspects of the disclosure. As will be realized, the invention is capable of modifications in various aspects, all without departing from the spirit and scope of the present disclosure. Accordingly, the drawings and detailed description are to be regarded as illustrative in nature and not restrictive.DETAILED DESCRIPTION

[0090] Described herein are systems and methods for detecting properties associated with tiny particles (e.g., micro- and / or nano-scale particles), such as particles that are present within solutions, media, and / or other matrices, including for purposes of spectroscopy in both liquid and gaseous phase. In one embodiment, the systems and methods described herein include designing and implementing sensors and / or sensor systems that can capture properties such as subtle acoustic signals generated by targets of interest within a solution from the absorption of pulsed light energy, providing photoacoustic spectroscopy information for real-time, label-free detection and interrogation of targets such as particles and / or other materials such as cells in their native solution environments across an extended sensing volume. Leveraging the unique optical absorption properties of the target particles / materials, the techniques described herein selectively detect and classifyCT particles / materials flowing through sensors / sensor systems. As an example, the systems and methods include designing optofluidic, high throughput, ultra-sensitive optical microresonator sensors that can be implemented to capture properties such as subtle acoustic signals generated by tiny particles from the absorption of pulsed light energy, providing photoacoustic spectroscopy information for real-time, label-free detection and interrogation of particles and materials such as cells and the like in their native solution environments across an extended sensing volume. This is described and showcased herein by certain experiments, such as the measurement of gold nanoparticles with diverse geometries and different species of red blood cells in the presence of other cellular elements and a wide variety of proteins. These particles / materials are effectively identified and classified based on their photoacoustic fingerprint that captures particle shape, composition, molecule properties, and / or morphology features. By way of the systems and methods described herein, new avenues to achieve rapid, reliable, and high-throughput particle and / or cell identification in clinical and industrial applications are opened up, offering a valuable tool for understanding complex biological and environmental systems.

[0091] More specifically, described herein are various types of optical resonators, including various geometries, including but not limited to a Fabry -Perot resonator, a racetrack resonator, a microsphere, a microring resonator, and / or a whispering-gallery-mode (WGM) microbubble resonator (MBR), that can be used as a photoacoustic (PA) sensor, with biological and / or particle sensing applications, different from conventional optical WGM sensors which detect refractive index changes near the sensing surface. The microbubble resonators disclosed herein may range in size from 30 pm to 1 cm in diameter (some embodiments may utilize a range of 350 pm to 800 pm) and may be made from fused silica capillaries. Additional details about (e.g., WGM) micro-resonators and their construction, etc., are found in: US Patent No. 8,704,155; US Patent No. 9,733,125; US Patent No.9,766,402; US Patent No. 10,782,289; and US Patent Publication No. US 2023 / 0010794,CT published on January 12, 2023, each of which are hereby incorporated by reference herein in their entireties.

[0092] Micro- and nano-scale particles have played crucial roles across diverse fields, from biomedical imaging and environmental processes to early disease diagnosis, influencing numerous scientific research and industrial applications. Their unique characteristics demand accurate detection, characterization, and identification. However, conventional spectroscopy and microscopy commonly used to characterize and identify tiny objects often involve bulky equipment and intricate, time-consuming sample preparation. Over the past two decades, optical microsensors have emerged as a promising sensor technology with their high sensitivity and compact configuration. However, their broad applicability is constrained by the requirement of surface binding for selective sensing and the difficulty in differentiating between various sensing targets, which limits their application to detecting targets in their native state in complex biological samples. Developing label-free and immobilization-free sensing techniques that can directly detect target particles in complex solutions is crucial for overcoming the inherent limitations of current microsensors.

[0093] To transition from typical refractive index biosensing to PA assisted biosensing, a photoacoustic contrast agent may be needed. A contrast agent is a substance that may be used to enhance the visibility of structures or functions within biological material such as biological tissues, and which plays a crucial role in PA imaging, as such agents may significantly enhance the ability to visualize and analyze physiological and molecular processes in real-time, aiding in various biomedical applications. Contrast agents work by absorbing light and converting it into ultrasonic waves, which can then be detected to create detailed images. Contrast agents are substances that are deliberately introduced into the body to improve the visibility of internal structures during imaging procedures. These agents enhance the contrast between different tissues or between tissues and fluids, allowing for better visualization and more accurate diagnosis. There are two main types of contrast agents,CT including: (i) natural (e.g., internal) contrast agents, which may include naturally occurring substances (e.g., hemoglobin, melanin, lipids, and water) within the body of certain species that absorb light at specific wavelengths, making them useful for imaging without the need for external agents; and (ii) external (e.g., introduced) contrast agents, which are externally introduced substances designed to improve imaging contrast, and may include organic dyes, semiconducting polymeric nanoparticles, gold nanoparticles, and other inorganic nanoparticles. Such external contrast agents are chosen for their strong light absorption properties, particularly in the near-infrared region, which allows for deeper tissue penetration and clearer images. For the systems and methods described herein, the photoacoustic (PA) signal is produced by the absorption of light by specific targets of interest. If, for example, the targets of interest have poor light absorption, a contrast agent can be attached to indicate the presence of the sensing targets through the PA effects. Modulated light from modulated light sources can be used with a wavelength that matches with the absorption peak of the target. Modulated light sources may include a laser, a light emitting diode (LED), a flash lamp (e.g., for gas sensing), and the like. A contrast agent is needed in situations where the target of interest is transparent, and / or there is no modulated light source available that matches with the absorption peak. Modulated light sources such as lasers may be used in connection with contrast agents. Various laser configurations may be implemented. For example, a second laser beam can come from a same source as a first laser beam, such as by way of the light from a same laser being split into two paths. One beam may be modulated to become a pulsed laser to provide optical stimulation to excite acoustic signals from analytes, and the other beam may be used as a probe to probe the changes in WGMs induced by acoustic waves. Splitting may be accomplished by a beam splitter (not shown) or the like.

[0094] One application of the systems and methods described herein is to tag biological targets with specific PA contrast agents for specific identification of the target. Another application is to utilize the systems and methods described herein to increase the effectiveCT sensing area of the device to increase the ability of WGM MBR sensors to detect biological targets. For starters, what is needed is to look at the identification and classification of the PA contrast agents, such as nanoparticles, and in particular gold nanoparticles of different geometries, including: nanospheres, nanorods, nanocubes and nanoshells. To insulate the WGM field from refractive index changes while still being sensitive to PA ultrasonic waves, a thick-walled WGM resonator is employed. Thick-walled resonators may include any hollow core WGM resonator where the wall thickness is much greater (e.g., >10x) than the wavelength of light used to excite the WGM field.

[0095] The systems and methods described herein provide for detection via microbubble resonators, more specifically detection of naturally occurring and external contrast agents via WGM microbubble resonators, some examples of which include: 1) detection and classification of gold nanoparticles, from their photoacoustic signals inside of a thick-walled microbubble resonator; 2) specific detection and classification of different types of red blood cells by their PA signals without the addition of PA contrast agents; and 3) design of microbubble resonators with a thick wall to fully protect high-quality optical modes, which is critical for the sustained effectiveness of the sensors in various sensing applications over an extended period.

[0096] As discussed above, micro- and nano-scale particles have played crucial roles across diverse fields, ranging from biomedical imaging and environmental processes to early cancer diagnosis, influencing numerous scientific research and industrial applications. Their unique characteristics demand accurate detection, characterization, and identification. However, conventional methods like spectroscopy and microscopy often involve bulky equipment and intricate, time-consuming sample preparation. Optical microsensors, with their rapid response and compact configuration, have shown promise in particle sensing. Yet, their broad applications are hindered by limitations such as surface-bound sensing requirements and evanescent field constraints.CT

[0097] Described and demonstrated herein is an optofluidic, high throughput, and label- free optical microresonator sensor with an extended detection volume, enabling real-time, label -free detection and interrogation of particles and cells in their native solution environments. The system is based on a high-quality (Q) bubble-shaped whispering gallery mode (WGM) microresonator, featuring a hollow core directly connected to a microfluidic channel that allows one or more sample media to flow inside while being optically stimulated by a pulsed laser from the outside to generate acoustic waves through PA effects. The acoustic waves carrying the PA spectroscopic information of the samples can propagate through the solution and reach the WGM sensor, which could be positioned close (or far) from the site where the acoustic waves are initially generated. In this way, the approach described herein enables fast and simple detection and measurement of micro- and nanoscale objects across the microfluidic channel (see FIG. 13 at element 1304). In addition to exploiting photoacoustic effects for detection, the system and methods described herein demonstrate, for the first time, the ability to detect and differentiate free-flowing particles that are beyond the reach of the evanescent field near the surface of an optical sensor, which is typically a limitation under conventional detection schemes. It is also worth noting that the resonator’s hollow core with a thick- wall design is ideally suited for integration with microfluidics. In this design, high-Q WGMs are fully protected within the thick wall, avoiding any direct overlap between the sample and the optical field of the resonant mode. This optical isolation from the sample ensures that both the quality factor and the SNR remain unaffected by the absorbing and scattering properties of the sample. Since the PA process produces ultrasound in the MHz frequency range, the detection is background-free and is not affected by the environment. The exceptional stability and background-free detection of the novel sensing platform described herein not only ensure reliable and consistent results across multiple measurements, but also enable direct target detection in complex sample matrices, such as whole blood, without the need for laborious sampleCT purification. This is leveraged into the novel sensing platform described herein to facilitate high-throughput photoacoustic identification of particles and cells in liquid media using advanced machine learning approaches. This approach enables sensitive and specific detection of particles and cells within a small sample volume (~pL), making for an ideal solution for applications requiring rapid, accurate, and cost-effective analysis in complex matrices such as complex biological matrices.

[0098] As described and demonstrated herein, different nanoparticles, including gold nanoparticles (AuNPs) with diverse geometries and different species of red blood cells, are tested. These particles are effectively identified and classified based on their photoacoustic fingerprint that captures particle shape and morphology features. The systems and methods disclosed herein open up new avenues to achieve rapid, reliable, and high-throughput particle and cell identification in clinical and industrial applications, offering a valuable tool for understanding complex biological and environmental systems.

[0099] In various aspects, the performance of WGM resonators may be assessed using any suitable existing analysis method, without limitation, for specific detection of photoacoustic signals for purposes of spectroscopy in both liquid and gaseous phase (for example, the specific detection and classification of gas molecules for applications in environmental, health, and industrial monitoring).Optofluidic PA WGM sensor

[0100] Details of certain embodiments and / or implementations of photoacoustic WGM sensors are described herein.

[0101] FIGS. 1 A-1D illustrate aspects of an example label-free all-optical photoacoustic microresonator-based sensor and system according to one embodiment. FIG. 1 A illustrates a schematic of aspects of an example sensing system 100 including a sensor 102 such as a whispering gallery mode (WGM) microbubble resonator (MBR) sensor and use of photoacoustic (PA) contrast agents 104 in sensor 102. Contrast agents may include particlesCT and / or cells and may be referred to as particles 104 and / or cells 104. Sensor 102 includes a body, and in some embodiments, the body of sensor 102 may include or be fabricated in-line with a capillary 106, such as a silica capillary with a 250 pm inner diameter and 360 pm outer diameter. Diameters of 75 (inner) / 125 (outer) may also be utilized in some embodiments. A microbubble 108 of sensor 102 includes a wall 110 of a certain thickness and may be formed in the capillary 106. By controlling the heating power and the applied pressure during formation of microbubble 108, the diameter as well as the wall-thickness of wall 110 can be controlled. Wall 110 may be a thick-wall, which may be categorized as a wall thickness that is much greater (e.g., >10x) than the wavelength of light used to excite the WGM field. Elements 108, 110, 122, in combination, may be referred to herein as a microresonator (MBR).

[0102] One or more WGMs 112 are excited in the sensor 102 using, for example, a commercially available (e.g., from Thorlabs) tapered 780HP optical fiber 114. A laser 116 such as an external cavity diode laser (e.g., also referred to as probe laser and / or an excitation laser and / or an excitation / pump laser) centered around 780 nm connected to a function generator (shown in FIG. ID) is used to scan the wavelength of the laser around 40 pm at a rate of 60 Hz. The probe laser (e.g., 116) may be configured in a spectral window different from 780 nm. The criterion for choosing the wavelength of the probe laser may include minimizing the absorption loss in the resonator. An optical attenuator (shown in FIG. ID) and polarization controller (shown in FIG. ID) may be used to adjust the light intensity and polarization before coupling into the sensor 102 to optimize the WGM spectrum. The transmitted intensity (e.g., transmission spectrum) from the sensor 102 is detected by and / or otherwise sent to a high-speed photodetector (shown in FIG. ID) connected to an oscilloscope, where the wavelength may be fixed at an optical resonance by eliminating the scanning signal. Another laser 118 may be used to provide for achieving photoacoustic excitation 120 (e.g., excitation spot(s), or just spots) of photoacoustic (PA) contrast agentsCT104, where laser 118 may be a Q-switched 532 nm pulsed laser at a repetition rate of 60 Hz. For example, photoacoustic excitation may be achieved using a Q-switched 532 nm pulsed laser with a pulse duration of 20 ns at a repetition rate of 60 Hz. Sensor 102 may include a hollow core 122 connected (e.g., directly) to a microfluidic channel 124a that allows a sample medium 124b (e.g., sample solution) to flow inside while being optically stimulated by laser 118 from the outside to generate acoustic waves through PA effects. In some embodiments laser 116 and 118 may be the same laser as described herein.

[0103] Mode Field Radial Distribution of the WGMs 112 in sensor 102 is shown in plot 128 and graph 130, illustrating distributions including (i) an outside environment distribution 132 (e.g., an air environment), (ii) a wall distribution 134 (e.g., of a (silica) wall such as 110) of sensor 102, and (iii) an inner cavity distribution 134 of sensor 102 (e.g., in relation to core 122). FIG. 1A illustrates three different cases 138, 140, 142 for detecting PA waves 144, 146, 148 (also referred to as PA Signals 144, 146, 148) in microbubble 108 of sensor 102. Case 138 produced a result showing the strongest PA signal 144 (e.g., where the contrast agent 104 is close to the WGM 112 of sensor 102). Case 140 produced a result showing no PA signal 146 in the location where there is laser excitation but no contrast agent. Case 142 produced a result showing a weaker PA signal 148 detected from a contrast agent far from the WGM 112 field. FIG. 1 A illustrates the working principle of the sensor system described herein, including sensor 102. Further regarding probe laser 116, the present disclosure encompasses a pump and probe approach by pumping the target to generate a PA signal and use the probe light (e.g., 116 shown in FIG. 1 A) to detect the PA signal.

[0104] In one embodiment, the sensor 102 disclosed herein is associated with or built into capillary 106 to collect acoustic waves. The optical field of WGMs (e.g., 112) in the 780 nm band is confined within a silica wall (e.g., 110). Solution samples (e.g., 124b) containing particles (e.g., 104) flow through the fluidic channel connected to sensor 102 (e.g., a WGM MBR). In some embodiments, samples flow through a core of sensor 102 and are exposedCT to laser pulses (e.g., from laser 118) at 532 nm for PA excitation. This sensing process can be elucidated from two aspects: namely, acoustic wave generation by sensing targets, e.g., nanoparticles, via the PA process, and detection of acoustic waves by WGMs 112. Sensor 102 may include packaging material (not shown) that encapsulates all or part(s) of sensor 102. These nanometer wavelengths are for example only and other wavelengths may be utilized. For example, in some embodiments, a wavelength of greater than 500 nm may be used, such as the 780 nm described herein, or an 800 nm wavelength. The wavelength that is used may be determined based on whether a target of interest has a strong absorption of the light being used. In some embodiments, multi wavelength PA spectroscopy, where multiple wavelengths are applied, may be implemented. For example, if a target molecule “A” has absorption on red light, and a target molecule “B” has absorption on green light, using multiple wavelengths allows for the composition of molecule A and B in a mixture system to be determined.

[0105] FIG. IB illustrates the PA effect of a particle (e.g., 104). More specifically, FIG. IB illustrates the photoacoustic process inside of gold nanoparticles. When a particle 104 flowing freely in the fluid is exposed to a pulsed laser at 532 nm (e.g., via spots 120), the energy absorbed by the particle 104 causes a rapid thermoelastic expansion 150, resulting in the generation of an acoustic wave 152. These acoustic waves 152 then propagate through the solution (e.g., 124b) within the capillary (e.g., 106) to reach the sensor 102. When an acoustic (e.g., 152) wave interacts with a WGM 112 field, it alters the effective optical path length, and consequently modulates the optical resonance (as shown in FIG. 1C). In this way, PA signals carrying information about the particles can be read out from the continuous wave laser coupled to the sensor 102. The acoustic wave generation by sensing targets, e.g., nanoparticles, via the PA process, and the detection of these acoustic waves by WGMs are two notable features of the sensing process described herein. Additional advantages of these aspects of the techniques described herein are label-free and immobilization-free detection,CT significantly enhanced detection volume, and direct target detection in complex matrices such as complex biological matrices.

[0106] FIG. 1C is a graph 154 illustrating how acoustic waves (e.g., 152) vibrate the resonator (e.g., 102) structure and result in a frequency shift of the WGM mode, producing output signals (e.g., where the PA Signal shown in FIG. 1C is similar to signals 144, 148 shown in FIG. IB).

[0107] FIG. ID is a schematic illustrating additional aspects of sensing system 100. Sensing system 100 may include, in addition to (or alternative to) the components shown in FIG. 1A, a function generator 156, one or more lasers 158 (e.g., realized as lasers 116, 118 shown in FIG. 1A), a polarization controller 160, an attenuator 162, one or more photodetectors 164, an oscilloscope 168, a data acquisition unit (e.g., DAQ) 170 that is operatively connected to and in communication with a computer device 172 which includes one or more processors 174 in communication with one or more memory devices 178. A machine learning (ML) model 180 may be integrated within or otherwise in operative communication with computer device 172 so that results obtained by system 100 can be utilized in and / or executed on by model 180. Model 180 may be used in conjunction with classifying elements found within a sample that was tested, including but not limited to elements such as particles, cells, etc. Model 180 may be trained on data such as collected nanoparti cl e / cell images and / or other data obtained via system 100 and / or outside data. Each of components 156, 158, 160, 162, 164, 168, and 170 may be operatively coupled to computer device 172, where such coupling may include a wired (e.g., network cable, USB, etc.), wireless (e.g., wireless network), or other remote (e.g., cloud) connection between computer device 172 and components 156 to 170. Sensor 102 shown in FIG. ID may include related components such as a fiber (e.g., 114 shown in FIG. 1 A).

[0108] Computer device 172 may be configured to provide user control of system 100, store data associated with system 100 and provide other computing functions and / orCT computing resources in association with system 100. For example, computer device 172 may control power and / or pulses / duration of a beam emitted from laser(s) 158, a type of polarization applied by polarization controller 160, an amount of attenuation by attenuator 162, and other associated controls with other components (e.g., photodetector 164) of system 100. Laser(s) 158 may include one or more lasers and one or more components such as one or more beam splitters configured to manipulate and / or otherwise control laser beams emitted by the lasers. For example, laser(s) 158 may be one laser where the beam emitted therefrom is split into two, one beam used for excitation and the other for probing. More generally, the outputs from the various light sources described herein may be controlled / manipulated by various components to arrive at desired light for use in the testing described herein. Alternatively, computing device 172 may include a plurality of computing devices each with their own memory devices and processors, for example a dedicated computer for each of components 156 to 170 (as well as model 180) or a dedicated computer for various subsets of components 156 to 170 (as well as model 180).

[0109] Additional aspects of computer device 172 are described in more detail herein. DAQ 170 may include a combination of hardware and software used to collect, measure, and analyze data (e.g., from oscilloscope 168, for example and without limitation). Components such as polarization controller 160 and attenuator 162 may have their order swapped within system 100, as needed, or be optional. A syringe / pump combination 182 may be used to handle distribution of solution (e.g., 124b) in connection with a channel (e.g., 124a) of sensor 102. In some embodiments, a control device such as a computer, a function generator, an FGPA, and / or other equipment may be implemented to perform computing and / or other tasks, for example, to control flow using the syringe pump. For example, a control device may include all of part of system 100. In some embodiments, the control device may be computer device 172 shown in FIG. ID, and / or computer device 172 including one or more components shown in FIG. ID such as function generator 156, without limitation. ForCT detection, a lock-in amplifier may be applied to further amplify the signal. The control device may be configured to obtain data and / or analyze results using a machine learning model and / or signal processing algorithms as described herein.

[0110] Sensing system 100 disclosed herein is different from conventional optical WGM sensors, which detect refractive index changes near the sensing surface. Instead, sensing system 100 detects acoustic waves generated by particles in solution flowing freely inside a capillary structure, which significantly improves the sensing capability, throughput, and speed. Additional improvements may include improvements to specificity and reliability by immunity to dispersive (refractive index changes) and dissipative changes (loss) in the solution. As shown in FIG. 1 A at distributions 132, 134, 136, the radial mode field is strongly confined within the thick wall (e.g., 110) of sensor 102. With little penetration into the core with the solution (e.g., 124b) sample, the optical WGMs 112 are protected from either refractive index changes or potential absorption or scattering loss in the sample. Even if the core is filled with a solution such as a black dye solution which has strong absorption, no apparent resonance shift or Q-factor deterioration is observed (shown in FIGS. 4A and 4B, for example). In this way, the noise and false signal can be greatly mitigated and consequently improve reliability and robustness.

[0111] Furthermore, unlike conventional methods that rely on surface binding via random diffusion, sensing system 100 can actively scan the pulse laser across the microfluid channel and search for target particles. Sensing system 100 may also be configured to detect distant particles (e.g., 104) that are far away from the sensor 102 where the WGM 112 resides, where the pulse laser 118 can be scanned across the optofluidic channel (e.g., 124a) to search for (excite) potential PA signals in the fluid(s). As shown by signals 144, 146, 148 in FIG. 1A, by scanning along the capillary 106, particles such as particle 104 far away from the sensor 102 (e.g., from the MBR where the WGM resides) can also be detected. The resulting PA signal exhibits a delay between the pulse excitation and its optical readout, dueCT to the distance between the optical mode and the particles. The PA signals were measured from particles at various distances from the MBR, indicating extended detection volume being achieved. In addition, by selecting the wavelength of the pulse laser that overlaps with the optical absorption of the particle of interest, the target particles can be selectively and effectively detected even in a complex fluidic medium / solution. As shown in FIG. 15, for example, a whole blood sample is injected into the sensor. In such a complex medium as whole blood, red blood cells can be selectively detected in the presence of other cellular elements as well as proteins, due to the unique absorption wavelength of hemoglobin in red blood cells. In some embodiments, the PA signals from particles at various distances from the sensor 102 were experimentally measured and exhibited / achieved an extended sensing range (as shown in FIGS. 5A-5C, for example).Nanoparticle sensing

[0112] FIGS. 2A-2D illustrate example temporal (time domain) and spectral (frequency domain) measurements of nanoparticles, in particular gold nanoparticles (also referred to herein as AuNPs) that were tested in accordance with the systems and methods of the present disclosure. FIGS. 2A-2D show photoacoustic time domain signals collected by sensor 102 filled with each geometry and their corresponding photoacoustic spectra (e.g., frequency domain spectrum) for each of four AuNP geometries tested. Accompanying each time and frequency domain plot are scanning transmission electron microscope (STEM) images of each geometry. The STEM images include zoomed-in insets. To demonstrate the capability to detect and identify nanoparticles, four different geometries of gold nanoparticles (AuNPs) were tested: nanospheres (FIG. 2A), nanorods (FIG. 2B), nanocubes (FIG. 2C), and nanoshells (FIG. 2D). Each solution was injected into the microfluidic channels (e.g., 124a). Care was taken to excite each AuNPs at the same position in order to standardize the collection of each signal.CT

[0113] FIG. 2A shows nanosphere test data 200, including nanosphere STEM image 202, time domain graph 204 (x-axis: Time (ps), y-axis may be an arbitrary unit (e.g., a.u.) for intensity), and frequency domain graph 206 (x-axis: Frequency (MHz), y-axis may be an arbitrary unit (e.g., a.u.) for intensity).

[0114] FIG. 2B shows nanorod test data 208, including nanorod STEM image 210, time domain graph 212 (x-axis: Time (ps), y-axis may be an arbitrary unit (e.g., a.u.) for intensity), and frequency domain graph 214 (x-axis: Frequency (MHz), y-axis may be an arbitrary unit (e.g., a.u.) for intensity).

[0115] FIG. 2C shows nanocube test data 216, including nanocube STEM image 218, time domain graph 220 (x-axis: Time (ps), y-axis may be an arbitrary unit (e.g., a.u.) for intensity), and frequency domain graph 222 (x-axis: Frequency (MHz), y-axis may be an arbitrary unit (e.g., a.u.) for intensity).

[0116] FIG. 2D shows nanoshell test data 224, including nanoshell STEM image 226, time domain graph 228 (x-axis: Time (ps), y-axis may be an arbitrary unit (e.g., a.u.) for intensity), and frequency domain graph 230 (x-axis: Frequency (MHz), y-axis may be an arbitrary unit (e.g., a.u.) for intensity).

[0117] The PA signal variations among the four geometries of AuNPs are notably distinct. This indicates that different shapes of nanoparticles, even when composed of the same material, produce unique PA signals. This spectral information in the PA signal further provides a measure of the shape property of the particles.Identification of Cells from Different Species and Machine Learning for Cell Classification from PA Spectra

[0118] FIGS. 3A-3F illustrate example results testing the feasibility of using photoacoustic fingerprinting to identify red blood cells according to the systems and methods of the present disclosure. FIGS. 3A-3E show the time-domain and (e.g., calculated)CT frequency-domain PA signals of five different types of red blood cells that were tested: pig (FIG. 3A); sheep (FIG. 3B); goat (FIG. 3C); turkey (FIG. 3D); and llama (FIG. 3E). These cells were diluted to 1%, by volume in a phosphate-buffered saline (PBS) solution (e.g., solution 124b), injected into sensor system 100, and flown through the core (e.g., 122) of the capillary (e.g., 106) connected to sensor 102. A pulse energy of ~40 nJ was used in the experiments to avoid photodamage. The temporal and frequency domain characterizations were obtained simultaneously, showing a SNR exceeding 30dB. The measurements may be ultrafast and repeatable, and the Fast Fourier Transform (FFT) spectra show very little difference between the different kinds of cells. The specificity and reliability are improved by making the sensor described herein immune to dispersive (refractive index changes) and dissipative changes (loss) in the solution. The geometry of the microresonator was designed such that the radial mode is strongly confined within the thick wall.

[0119] FIG. 3 A shows pig red blood cell test data 300, including time-domain graph 302 (x-axis: Time (ps), y-axis: Intensity (a.u.)) and frequency domain graph 304 (x-axis: Frequency (MHz), y-axis: Frequency spectra (a.u.)).

[0120] FIG. 3B shows sheep red blood cell test data 306, including time-domain graph 308 (x-axis: Time (ps), y-axis: Intensity (a.u.)) and frequency domain graph 310 (x-axis: Frequency (MHz), y-axis: Frequency spectra (a.u.)).

[0121] FIG. 3C shows goat red blood cell test data 312, including time-domain graph 314 (x-axis: Time (ps), y-axis: Intensity (a.u.)) and frequency domain graph 316 (x-axis: Frequency (MHz), y-axis: Frequency spectra (a.u.)).

[0122] FIG. 3D shows turkey red blood cell test data 318, including time-domain graph 320 (x-axis: Time (ps), y-axis: Intensity (a.u.)) and frequency domain graph 322 (x-axis: Frequency (MHz), y-axis: Frequency spectra (a.u.)).CT

[0123] FIG. 3E shows llama red blood cell test data 324, including time-domain graph 326 (x-axis: Time (ps), y-axis: Intensity (a.u.)) and frequency domain graph 328 (x-axis: Frequency (MHz), y-axis: Frequency spectra (a.u.))

[0124] FIG. 3F is a graph 330 showing offset FFT spectra 332 of 10 individual measurements 332 of red blood cells from the same species, in this case sheep (the x-axis of graph 330 is Frequency (MHz), the y-axis may be an arbitrary intensity (e.g., Frequency spectra (a.u.))). The vibration frequencies in the PA signal remain consistent across all measurements. Graph 330 also includes inset 334 showing measurements 332 which shows a close-up of the peak frequency 336 occurring near 3.9 MHz across different measurements (without offset). As shown in FIG. 3F, repeatability can be found in 10 measurements of red blood cells from the same species. In contrast with the results of AuNPs, the photoacoustic spectra of the 5 distinct cells look similar, with a major spectral peak near 4 MHz preserved across the species.

[0125] Although there are some slight differences between different species, these differences may not be apparent to identify, which may pose challenges for distinguishing the different species of cells directly. To tackle these challenges, machine learning (as described herein) may be deployed to analyze the differences at every single frequency component, which may assist in identifying the differences with high accuracy.

[0126] FIG. 3G is a diagram 338 illustrating a schematic overview of the classification process and convolutional neural network (CNN) with prototype learning structure. FFT is performed on PA signals, followed by processing through the CNN. The CNN learns specific signal features during training, and then these features are used for classification by finding the nearest prototype embedding in the latent space. The model is eventually evaluated on the test set. Principal Component Analysis (PCA) is employed for feature visualization by simplifying the data into lower dimensions.CT

[0127] FIG. 3H is a diagram 340 illustrating PCA visualization of samples without machine learning-based feature extraction. The resulting data points are intermingled, making them difficult to distinguish.

[0128] FIG. 31 is a diagram 342 illustrating PCA visualization of samples using features extracted by the machine learning algorithm. The data points from each species form distinct, well-separated clusters. The results indicate that the machine learning-extracted features are effective in differentiating between the blood cells of various species compared to using raw signal features.

[0129] To analyze the unique PA signals and extract the features associated with different species of red blood cells, comprehensive measurements were carried out, compiling a substantial dataset of PA signals for each species. Given the notable similarities in the PA signals and spectra across various species, it is crucial to identify the subtle characteristics of each type of blood cell. As a powerful tool to extract meaningful information from complex datasets, machine learning techniques were applied to learn from the PA signals by automatically identifying and extracting relevant features. The extracted features are more effective in distinguishing different types of red blood cells than the raw features.

[0130] Referring back to FIG. 3G, diagram 338 shows that PA signals 344 are input into model 346 (e.g., a backbone model) that includes a plurality of convolutional neural networks including one or more convolution layers 348 and one or more pooling layers 350. Model 346 may be configured as a core neural network architecture responsible for feature extraction from input data (e.g., including but not limited to PA signals 344), and to function as a foundational part of a larger Al system (e.g., see FIG. 11). For example, model 346 may have a main function of feature extraction, including transforming raw data into a set of features and / or feature representations. Backbone model 346 may be implemented to generate a fully connected layer 352, which is a layer where every neuron is connected toCT every neuron in the previous layer (e.g., every input neuron is connected to an output neuron) and may have corresponding weights associated with the connections. Fully connected layer 352 may be implemented as part of a classification operation 354 (e.g., classification process), described below. Backbone model 346 and / or fully connected layer 352 may be referred to herein as a feature extractor (e.g., a machine-leaming-based feature extractor).

[0131] To train the machine-leaming-based feature extractor, the data was split into a training set (80%) and a testing dataset (20%). FIG. 3G shows the input features that were first converted from the time domain to the frequency domain using the Fast Fourier Transform (FFT). A one-dimensional convolutional neural network (CNN) is used to extract the local dependencies among neighboring frequency components (e.g., spatial dependencies) in the frequency domain of PA signals. The classification process begins by feeding the PA signals into the CNN, where the convolutional layer (e.g., 348) uses filters to extract features such as signal patterns. A pooling layer (e.g., 350) then reduces the data dimensionality, maintaining essential information as labeled in dashed red lines. The subsequent fully connected layer (e.g., 352) integrates these features, forming a comprehensive understanding to make predictions. To ensure robust predictions and mitigate the noise in signals, prototype learning (e.g., via 354) was further introduced to enhance the similarity of the features extracted from the same cell species while reducing the chance of misclassification. The prediction depends on the sample’s distance to the prototype embeddings in the latent space of each cell category. The model calculates the shortest distance from the learned feature to the prototype embeddings and classifies the sample as the cell category of its closest prototype embedding. Once trained, the machine learning model can analyze new, unseen PA signals and accurately classify them into one of the five species categories based on the features it has learned.

[0132] To illustrate the effectiveness of the machine learning approach described herein in classifying PA signals, principal component analysis (PCA) was utilized to visualize theCT testing data with both raw features and the features extracted by the machine learning model. PCA is a well-established technique for visualizing datasets by reducing their dimensionality. However, with only the raw features, the data of different species overlap considerably, making it challenging to differentiate the different species of red blood cells (see FIG. 3H). In contrast, with the features extracted by the machine learning model of the present disclosure, PCA reveals distinct decision boundaries between different categories, clearly separating the different red blood cell species, as shown in FIG. 31. This demonstrates the significant improvement achieved by machine learning in classifying different types of blood cells, and the unique photoacoustic signatures can provide sufficient information to enable reliable classification.

[0133] These results demonstrate that the approach described herein can not only detect the presence of free-flowing micro / nano particles in their natural environment but also obtain their photoacoustic properties. The frequency domain features of photoacoustic signals extracted by machine learning can be used to detect and classify different species of cells without the need for incubation, culturing, labeling, and imaging. It is also possible to characterize the PA signals on nanoparticle samples at varying concentrations. The PA signal intensity was shown to have increased with higher nanoparticle concentrations, which can be used for quantitative detection. For example, in some embodiments, the PA signal intensity increased with higher nanoparticle concentrations, exhibiting a slope of 2.5 mV mL fM'1in the linear region. This makes the technique described herein highly suitable for applications requiring precise detection of concentration gradients. With these findings, a rapid, PA detection and characterization that allows for particle / cell identification without the need for full spectroscopic and / or microscopic analysis, which typically relies on bulky equipment and complicated sample preparation is able to be realized.Additional TestingCT1. Excellent stability for reliable measurement

[0134] FIGS. 4 A and 4B illustrate testing of an example thick wall for mode protection. FIG. 4A shows a WGM spectra graph 400 when the core (e.g., 122) of the sensor (e.g., 102) is filled with a solution (e.g., 124b) including deionized (DI) water 402 and black dye 404, respectively. FIG. 4B shows simulation 406 (e.g., COMSOL simulation) of field distribution of a WGM (e.g., 112), including for a liquid core 408 (e.g., a realization of core 122, see also distribution 136 in FIG. 1A), wall 410 (e.g., a realization of wall 110, see also distribution 134 in FIG. 1 A), and air 412 (e.g., a realization of outside environment, see also distribution 132 in FIG. 1 A). When the wall 410 is 5 pm thick, for example, there is little overlap between the WGM (e.g., 112) and the liquid core 408 in the sensor 102.

[0135] Since the light field is confined within the thick wall of the microbubble resonator (MBR), direct interaction between whispering gallery modes (WGMs) and the sample medium (e.g., 124b) is prevented. Consequently, when the core filling switches from DI water to black dye, there is negligible change observed in the WGM spectra. This means that the Q-factor, along with the SNR, and therefore the overall sensing performance, remain unaffected by the composition and refractive index of the sample media itself. This characteristic is critical for the sustained effectiveness of the sensors in various sensing applications over an extended period.2. Extended sensing range

[0136] FIGS. 5A, 5B, and 5C illustrate an example extended sensing range of sensor 102. FIG. 5A is a diagram 500 illustrating the scanning of the pulsed laser (e.g., 118) along the microfluidic channel (e.g., 124a), with WGMs (e.g., 112) being excited near the MBR center 502 (e.g., MBR equator) while the sample medium (e.g., 124b) flows through its core (e.g., 122). FIG. 5B illustrates a plot 504 showing time delays observed between the pulse trigger and the detected peak of the PA signals as the laser beam spot traverses along theCT channel (e.g., 124a). FIG. 5C illustrates a plot 506 corresponding peak-to-peak amplitudes of the measured signals during the scanning.

[0137] By adjusting the beam spot position of the pulsed laser (e.g., 118), an extended sensing range is achieved, enabling the detection of particles even at distances far from the microresonator. FIG. 5 A illustrates the scanning of the pulse laser (e.g., 118) along the transparent capillary (e.g., 106) that serves as the microfluidic channel (e.g., 124a). To experimentally validate this expanded sensing range, the pulsed laser's beam spot (e.g., 120) was moved along the microfluidic channel (e.g., 124a) while keeping the laser current constant. The distance between the WGM (e.g., 112, shown in FIG. 1A) and the particles (e.g., 104, shown in FIG. 1A) causes a delay between the pulse excitation and the optical readout. The measured delay time changes linearly with distance of the pulse laser excitation location from the sensor 102, as shown in FIG. 5B. This delay time can be used as an indicator providing us with the position information of the particles. The signal amplitude, characterized as the peak-to-peak amplitude, drops at a larger distance. Even with the presence of loss, measurable signals at a very large distance of 6 mm were still able to be obtained. This capability, despite the inherent loss typically associated with the mechanical property of sample media and the capillary, underscores the extended reach of the sensing approach described herein, making it particularly effective in scenarios where proximity to the target is a challenge.3. PA Signal amplitude as a function of nanoparticle concentration and input laser power

[0138] FIGS. 6 A and 6B illustrate the PA signal response of gold nanospheres as the concentration is varied from 1.7e10nps / mL to 3.28e10nps / mL. (FIG. 6A), and the PA signal response of gold nanospheres at a concentration of 3.28e10nps / mL under increasing power delivered by the pulsed laser (FIG. 6B).CT

[0139] The photoacoustic response observed from the gold nanoparticles varies with the concentration of the nanoparticles in the solution and the input power delivered by the pulsed laser. Plot 600 of FIG. 6A shows that the photoacoustic signal amplitude increases with increasing concentrations of gold nanospheres. The dilutions were made from the original concentration of 3.28e10nanoparticles / mL (nps / mL) and the concentration of the diluted samples was confirmed with UV / Vis spectroscopy. The power of the pulsed laser was held constant at 37 pW and a repetition rate of 60 Hz. There is a near linear relationship between the nanosphere concentration and the PA signal amplitude until about 3e10nps / mL where saturation of the detector response is observed. Plot 602 of FIG. 6B shows the dependence of the PA signal amplitude on a sample of gold nanospheres at a constant concentration of 3.28e10nps / mL under increasing power from the pulsed laser.Methods - Materials

[0140] WGMs 112 are excited in the sensor 102 using a commercially available (e.g., from Thorlabs) tapered 780HP optical fiber (e.g., 114). An external cavity diode laser (e.g., 116) centered around 780 nm connected to a function generator (e.g., 156) is used to scan the wavelength of the laser (e.g., 116) around 40 pm at a rate of 60 Hz. An optical attenuator (e.g., 162) and polarization controller (e.g., 160) are used to adjust the light intensity and polarization before coupling into the sensor 102 to optimize the WGM spectrum. The transmission or reflection spectrum from the sensor 102 is detected by or otherwise transmitted to a high-speed photodetector (e.g., 164). For example, measurements from reflected spectrum may be made by measuring reflected light from the sensor. Photoacoustic excitation was achieved using a Q-switched 532 nm pulsed laser (e.g., 118) at a repetition rate of 60 Hz.Methods - Sensor Fabrication

[0141] Microbubble resonators are fabricated in-line with a silica capillary (250 pm inner diameter and 360 pm outer diameter). Capillaries with different diameters may be used. First,the polymer coating on the silica capillary is stripped away using a butane torch and clean the silica window with isopropyl alcohol. Next, one end of the capillary is sealed with epoxy to allow the build-up of internal air pressure inside the capillary. The capillary is then placed onto a commercially available (e.g., Vytran) precision glass processing station (e.g., machine) and internally pressurized with air. The bare silica capillary can be locally heated and inflated into a spherical geometry and form a microbubble resonator. For example, the precision glass machine may be used to locally heat the bare silica, and with the combination of internal air pressure and local heating, the capillary inflates into a spherical geometry - resulting in a microbubble resonator. The built-in microscope on the precision glass machine is used to monitor the fabrication process for quality control. By controlling the heating power and the applied pressure, the diameter as well as the wall-thickness of MBRs can be controlled.Methods - Experimental Setup

[0142] FIG. 1 A shows the experimental schematic for the MBR photoacoustic detection system (e.g., sensing system 100) described herein. The built-in fluidic channels (e.g., 124a) of the MBR are used to deliver the sample medium to the WGM resonator by means of a syringe pump (e.g., 182). The pump is set to withdraw the solution through the MBR at a rate of 50 pL I min. This flow rate ensures stable operation and reliable signal acquisition. No significant change in the PA signals was observed when the flow rate was varied. Additionally, in some embodiments, in the other cross-axis of the MBR is the adiabatically tapered optical fiber (minimum waist diameter ~ 1 pm) used to couple light in and out of the resonator. The transmission spectrum of the MBR is detected via a photodetector (e.g., 164) which is read out with an oscilloscope (e.g., 168). A custom software (e.g., Lab VIEW) script (which may be integrated with DAQ 170) is used to collect the oscilloscope data.

[0143] More generally, photodetector 164 may be configured to detects results from testing of samples / sample solutions / sample mediums as described herein, includingCT detecting results from and / or associated with photoacoustic effects and / or acoustic waves associated with testing of the samples / sample solutions / sample mediums. This may include, for example, photodetector 164 being configured to detect transmitted light signals from a sensor including a microresonator, where the light signals carry information associated with properties of acoustic waves associated with a target of interest (e.g., targets including analytes, particles, and / or other materials such as biological materials including blood, cells, etc.). Photodetector 164 may be used in conjunction with the various probes (e.g., probe lights, probe lasers, etc.) described herein and / or outputs from and / or associated with these various probes. For example, photodetector 164 may be used in association with one or more probes to locate the target of interest at different positions within the sample medium within the sensor during testing of the sample medium. Outputs from both the probes and photodetector 164 may be implemented to make determinations as to the type of parti cle(s) present within the sample medium, using, for example, photoacoustic properties of the particles. Outputs from the probes and / or photodetector 164 may be used in association with one or more AI / ML models described herein. For example, the AI / ML model(s) may be trained on and / or executed on data associated with outputs from the probes and / or photodetector 164 for purposes of detecting targets such as particles. In some embodiments, photodetector 164 may include one or more detectors, and / or detectors of different types other than photodetectors, including but not limited to transducers and / or other acousticbased detectors. That is, various detectors that detect light, motion, etc. may be implemented to detect properties of targets such as particles that are tested by the systems and methods described herein.

[0144] The 532 nm pulsed laser 118 is used to excite the photoacoustic effect inside of the AuNP. This laser pathway is completely free space and focused on to the MBR outer surface with an objective lens. An objective lens coupled to a CCD camera may be used for optical alignment. This objective lens can also be scanned along the capillary axis, whichCT means target detection is not limited to just inside the MBR, but inside the capillary as well.A spot diameter (FWHM) may be set to 4 pm under the objective lens.Methods - Sample preparation

[0145] Four different geometries of gold nanoparticles (AuNPs) were used in this study, namely nanospheres, nanorods, nanocubes, and nanoshells, as shown in FIGS. 2A-2D. The AuNPs ranged in size from 10 nm to 50 nm and were suspended in a deionized water buffer with a citrate capping agent. The concentration of the colloidal AuNP ranged from 1010nanoparticles / mL (spheres, rods and cubes) to 1012nanoparticles / mL (shells).

[0146] The five different species of whole blood as well as washed red blood cells (as shown in FIGS. 3A-3E) were purchased from a commercial supplier (e.g., LAMPIRE Biological Laboratories). In some embodiments, the suspensions of these cells were diluted to 1%, such as in phosphate-buffered saline (PBS) solution (e.g., 124b) and then delivered into the core (e.g., 122) of the sensor (e.g., 102) using a small syringe with a syringe pump (e.g., 182). In some embodiments, the suspensions of washed red blood cells may also be diluted to 1% in PBS solution.Methods - Scanning transmission electron microscopy and imaging

[0147] For scanning transmission electron microscopy (STEM), 6 pL of different AuNPs solutions were deposited on a pure carbon film having a mesh size of 400 and were then dried for 60 minutes at ambient temperature (-21° C). The carbon surface adsorbed the AuNP particles and the water was evaporated. This carbon support film for STEM is very thin (15- 25 nm) and highly transparent to electrons. The dried film is then mounted on the STEM sample holder for imaging (e.g., mounted to the STEM (Field Emission Electron Microscope) and then imaged). This electron microscope has dual modality and it can take both transmission electron microscopy (TEM) and STEM images (e.g., 202, 210, 218, 226). STEM uses a focused beam that scans the sample line by line, it provides a much better resolution than TEM which uses a broad, parallel beam.CTMethods - Machine learning of the PA signals

[0148] Additional details and / or aspects of the particular datasets of the present disclosure and machine learning analysis thereof are described below. While these details / aspects are described with respect to the particular AI / ML models of the present disclosure, they are applicable to any suitable AI / ML model.

[0149] For Al analysis for feature learning and particle classification, in some embodiments, a 1 -dimensional Convolutional Neural Network as a feature extractor to learn useful information about the photoacoustic signal in the frequency domain was deployed. In the convolutional layer, multiple convolutional kernels stride along the vectors of the input, where each kernel can capture a unique local pattern in the spectrum, and sub-sequent pooling layers distill essential features. Then the feature maps, after the pooling process, are flattened into a vector and propagated into a fully connected layer with the activation function of Rectified Linear Unit (ReLU). This learned feature is usable for making predictions based on prototype learning.Methods - Prototype learning for robust prediction

[0150] A potential limitation of CNNs is that they tend to learn surface statistical regularities in the dataset rather than higher-level abstract concepts. For highly sensitive optical sensors like WGM, trivial environment changes can be detected, and a well-trained CNN may misclassify with slight perturbations of the spectrum. Therefore, a robust machine learning model may facilitate and / or improve assisting in the detection and / or sensing applications described herein.

[0151] In prototype learning, prototype embeddings represent different classes in the latent space. The classification of a sample is simply implemented by finding the nearest prototype embedding using Euclidean distance in the latent space. The prototype embeddings may be denoted as myj where y e { 1, 2, ... C] represents the index of the classes and j eCT{ 1, 2, ... K} represents the index of the prototype embeddings in each class. In the present techniques described herein, the number of prototype embeddings of each category is set to be K = 1, assuming that for each class there is only one representative embedding myin the latent space. These prototype embeddings, with a dimensionality equivalent to the feature space, can be initialized with trainable parameters so that they can be simultaneously updated with the model parameters in the training process.

[0152] For feedforward propagation, the prediction is determined by the distance between the sample x and the prototype embedding myin the latent space instead of calculating the Softmax function:

[0153] where x, q) denotes the output of the machine learning model (learned features) with trainable parameter q. Then, for the loss function, the probability of the prediction result is proportional to the negative distance— | | / (x, q) — my| | . Considering the non-negative of the probability and sum-to-one properties, this prediction can be written as:

[0154] Therefore, a cross-entropy (CE) loss based on prototype learning can be described as: CE = — logp(y|x)

[0155] Moreover, the robustness can be interpreted as the learned feature close to the prototype embedding in the latent space, which indicates that the model can neglect the noise and only preserve the key features of the class. Therefore, the loss function of prototype loss (PL) can be defined as minimizing the distance between the learned feature and the prototype embedding with the correct category: PL = | \f(x, q — my112.CTEventually, the total loss can be defined as the sum of cross entropy loss (CE) and the prototype loss (PL): loss = CE + x PL, where is the hyperparameter that governs the influence of Prototype loss on feature extraction and decision boundary formation. As A gets larger, the embedding of extracted features will get closer to the prototype. In the experiment, the hyperparameter is set to be 0.1.

[0156] Additional aspects of computing components / devices / systems implemented as part of the systems and methods described herein are provided below.

[0157] FIG. 7 is a block diagram schematically illustrating a system in accordance with one aspect of the disclosure. FIG. 7 illustrates a simplified block diagram of a computing system 700 for implementing the methods described herein. As illustrated in FIG. 7, the computing system 700 may be configured to implement at least a portion of the tasks associated with disclosed method using the disclosed resonator-based sensors (e.g., for displacement sensing). Computer system 700 may include a computing device 702 (where computer device 172 shown in FIG. ID may be a realization of computing device 702). In one aspect, the computing device 702 is part of a server system 704, which also includes a database server 706. Computing device 702 is in communication with a database 708 through database server 706. Computing device 702 is communicably coupled to a system 710 such as sensing system 100 and / or components thereof (e.g., as shown in and described in connection with FIG. ID) and a user computing device 712 of a user 714 through a network 716. Network 716 may be any network that allows local area or wide area communication between the devices. For example, network 716 may allow communicative coupling to the Internet through at least one of many interfaces including, but not limited to, at least one of a network, such as the Internet, a local area network (LAN), a wide area network (WAN), an integrated services digital network (ISDN), a dial-up-connection, a digital subscriber line (DSL), a cellular phone connection, and a cable modem. User computing device 712 may beCT any device capable of accessing the Internet including, but not limited to, a desktop computer, a laptop computer, a personal digital assistant (PDA), a cellular phone, a smartphone, a tablet, a phablet, wearable electronics, smart watch, or other web-based connectable equipment or mobile devices. In other aspects, computing device 702 is configured to perform a plurality of tasks associated with the operation of a resonator-based sensor and / or a system incorporating the resonator-based sensors including, but not limited to the displacement systems described herein.

[0158] FIG. 8 depicts a component configuration 800 of a computing device 802 associated with a user 804. Device 802 includes database 806 along with other related computing components. In some aspects, computing device 802 is similar to computing device 702 (shown in FIG. 7). User 804 may access components of computing device 802. In some aspects, database 806 is similar to database 708 (shown in FIG. 7).

[0159] In one aspect, database 806 includes obtained photoacoustic (PA) data 808 and algorithm data 810. Non-limiting examples of suitable PA data 808 may include data such as shown in and described in connection with FIGS. 1C, 2A-2D, 3A-3F, 4A, 5B, 5C, 6A, and 6B, without limitation. Non-limiting examples of suitable algorithm data 810 include any values of parameters defining the operation of the WGM resonator-based sensors, and sensing system 100. Additional non-limiting examples of suitable algorithm data 810 includes any algorithms and any values of parameters defining the algorithms associated with the disclosed method as described herein and / or any PA algorithms used to reconstruct or predict PA effects as described herein, and / or in connection with biological material and / or particle sensing as described herein. PA data 808 and / or algorithm data 810 may be used in conjunction with a machine learning model (e.g., 180, etc.) as described herein.

[0160] Computing device 802 also includes a number of components that perform specific tasks. In the example aspect, computing device 802 includes data storage device 812, PA component 814, sensor component 816, and communication component 818. DataCT storage device 812 is configured to store data received or generated by computing device 802, such as any of the data stored in database 806 or any outputs of processes implemented by any component of computing device 802.

[0161] Communication component 818 is configured to enable communications between computing device 802 and other devices (e.g., user computing device 712 and system 710, shown in FIG. 7) over a network, such as network 716 (shown in FIG. 7), or a plurality of network connections using predefined network protocols such as TCP / IP (Transmission Control Protocol / Internet Protocol).

[0162] FIG. 9 depicts a configuration of a remote or user computing device 900, such as user computing device 712 (shown in FIG. 7). Computing device 900 may include one or more processors 902 for executing computer-readable / -executable instructions. In some aspects, executable instructions may be stored in a memory area(s) of one or more memory devices 904. Processor 902 may include one or more processing units (e.g., in a multi-core configuration). Memory 904 may be any device allowing information such as executable instructions and / or other data to be stored and retrieved. Memory 904 may include one or more computer-readable media (e.g., hard drive, RAM, ROM, and the like).

[0163] Computing device 900 may also include at least one media output component 906 for presenting information to a user 908. Media output component 906 may be any component capable of conveying information to a user 908. In some aspects, media output component 906 may include an output adapter, such as a video adapter and / or an audio adapter. An output adapter may be operatively coupled to processor 902 and operatively coupled to an output device such as a display device (e.g., a liquid crystal display (LCD), organic light emitting diode (OLED) display, cathode ray tube (CRT), or “electronic ink” display) or an audio output device (e.g., a speaker or headphones). In some aspects, media output component 906 may be configured to present an interactive user interface (e.g., a web browser or client application) to user 908.CT

[0164] In some aspects, computing device 900 may include an input device 910 for receiving input from user 908. Input device 910 may include, for example, a keyboard, a pointing device, a mouse, a stylus, a touch sensitive panel (e.g., a touch pad or a touch screen), a camera, a gyroscope, an accelerometer, a position detector, and / or an audio input device. A single component such as a touch screen may function as both an output device of media output component 906 and input device 910.

[0165] Computing device 900 may also include a communication interface 912, which may be communicatively coupled to a remote device. Communication interface 912 may include, for example, a wired or wireless network adapter or a wireless data transceiver for use with a mobile phone network (e.g., Global System for Mobile communications (GSM), 3G, 4G or Bluetooth) or other mobile data network (e.g., Worldwide Interoperability for Microwave Access (WIMAX)).

[0166] Stored in memory 904 are, for example, computer-readable / -executable instructions for providing a user interface to user 908 via media output component 906 and, optionally, receiving and processing input from input device 910. A user interface may include, among other possibilities, a web browser and client application. Web browsers enable users 908 to display and interact with media and other information typically embedded on a web page or a website from a web server. A client application allows users 908 to interact with a server application associated with, for example, a vendor or business. Users 714, 804, 908 may be the same.

[0167] FIG. 10 illustrates an example configuration of a server system 1000. Server system 1000 may include, but is not limited to, database server 706 and computing device 702 (both shown in FIG. 7). In some aspects, server system 1000 is similar to server system 704 (shown in FIG. 7). Server system 1000 may include one or more processors 1002 for executing instructions. Instructions may be stored in a memory area(s) of one or moreCT memory devices 1004, for example. Processor 1002 may include one or more processing units (e.g., in a multi-core configuration).

[0168] Processor 1002 may be operatively coupled to a communication interface 1006 such that server system 1000 may be capable of communicating with a remote device such as user computing device 712 (shown in FIG. 7) or another server system 1000. For example, communication interface 1006 may receive requests from user computing device 712 via a network 716 (shown in FIG. 7).

[0169] Processor 1002 may also be operatively coupled to a storage device 1008. Storage device 1008 may be any computer-operated hardware suitable for storing and / or retrieving data. In some aspects, storage device 1008 may be integrated in server system 1000. For example, server system 1000 may include one or more hard disk drives as storage device 1008. In other aspects, storage device 1008 may be external to server system 1000 and may be accessed by a plurality of server systems 1000. For example, storage device 1008 may include multiple storage units such as hard disks or solid-state disks in a redundant array of inexpensive disks (RAID) configuration. Storage device 1008 may include a storage area network (SAN) and / or a network attached storage (NAS) system.

[0170] In some aspects, processor 1002 may be operatively coupled to storage device 1008 via a storage interface 1010. Storage interface 1010 may be any component capable of providing processor 1002 with access to storage device 1008. Storage interface 1010 may include, for example, an Advanced Technology Attachment (ATA) adapter, a Serial ATA (SATA) adapter, a Small Computer System Interface (SCSI) adapter, a RAID controller, a SAN adapter, a network adapter, and / or any component providing processor 1002 with access to storage device 1008.

[0171] Memory 904 (shown in FIG. 9) and 1004 may include, but are not limited to, random access memory (RAM) such as dynamic RAM (DRAM) or static RAM (SRAM), read-only memory (ROM), erasable programmable read-only memory (EPROM),CT electrically erasable programmable read-only memory (EEPROM), and non-volatile RAM (NVRAM). The above memory types are example only and are thus not limiting as to the types of memory usable for storage of a computer program.

[0172] FIG. 11 is a schematic diagram illustrating details of a machine learning (ML) model module 1100 which takes the input of frequency components of the photoacoustic signals and predicts the classification of the sample (where an output thereof may be realized as model 180 as shown in FIG. ID). Model module 1100 may be implemented on computer device 1102 (which may be realized as computer device 172 as shown in FIG. ID, or otherwise in operative communication with computer device 172), where computer device 1102 may communicate with other components of system 100, such as database server 706 and database 708 (each shown in FIG. 7), database 806 (shown in FIG. 8), and / or another (e.g., third-party) server via a network 1104 (which may be realized as network 716 shown in FIG. 7). Computer device 1102 may include and / or be in communication with a database 1106 that stores data 1108 including data such as PA data 808 and algorithm data 810 (each shown in FIG. 8). Data 1108 received from network 1104 may be stored in database 1106. Computer device 1102 may be configured to use data 1108 to generate model module 1100 for generating and providing a model (e.g., 180) for analyzing and making determinations about PA data (e.g., 808) and / or other data obtained by system 100 and / or used in conjunction with particle and biological sensing, as described herein.

[0173] In exemplary embodiments, computer device 1102 includes a training set builder module 1110 configured to submit one or more queries 1112 to database 1106 to retrieve subsets 1116 of data 1108, and to use those subsets 1116 to build training data sets 1114 for generating the model (e.g., 180) via the model module 1100. For example, query 1112 may be configured to retrieve certain fields from data 1108 in connection with obtained results such as shown in FIGS. 2A-2D and 3A-3F, including but not limited to time / frequency spectra data, delay data such as shown in FIGS. 5B and 5C, intensity (and / or amplitude) dataCT such as shown in FIGS. 6A and 6B, and image data such as STEM images 202, 210, 218, 226 shown in FIGS. 2A-2D, and the like.

[0174] In exemplary embodiments, training set builder module 1110 may be configured to derive training data sets 1114 from retrieved subsets 1116. Each training data set 1114 corresponds to historical data 1108 (“historical” in this context means completed in the past, as opposed to completed in real-time with respect to the time of retrieval by training set builder module 1110). Each training data set 1114 may include “model input” data fields along with at least one “result” data field representing a historical outcome associated with the model input. The model input data fields represent factors that may be expected to, or unexpectedly be found during model training to, have some correlation.

[0175] In exemplary embodiments, the model input data fields in training data sets 1114 may be generated from data fields in subsets. In other words, a machine learning model is trained based on prototype loss. With the set of model parameters trained by the model optimizer, the model is capable of making robust predictions based on input values that can be generalized to wider variations.

[0176] After training set builder module 1110 generates training data sets 1114, training set builder module 1110 passes the training data sets 1114 to model trainer module 1120. Within the trainer module, an optimizer based on gradient descent algorithm will update the model parameters in an iterative way until the error between the output and real label falls below a suitable threshold. “Machine learning” refers broadly to various algorithms that may be used to train and recognize patterns in existing data in order to facilitate making predictions for subsequent new input data. Here, the machine learning model of the present disclosure is based on prototype learning to make robust classifications.

[0177] Model trainer module 1120 trains the machine learning model to accurately predict the class of the given samples. To select the best model for tasks as described herein, the models were evaluated with different hyper-parameters on a separate validation set andCT the model with the least error is considered to the be best model to be implemented in real application. In exemplary embodiments, model trainer module 1120 may be configured to simultaneously train multiple candidate machine learning models and to select the best performing candidate for each result data field, as measured by the “error” between the at least one output and the corresponding result data field, to upload to model module 1100. Modules 1110, 1120, and 1126 may be implemented as part of model 1100 or as separate modules.

[0178] In certain embodiments, the one or more machine learning models may include one or more neural network layers, such as a convolutional layer, a multi-layer perceptron, or the like, and / or as described herein. The model parameters adjusted during training may be respective weight values applied to one or more inputs to each node to produce a node output. In other words, the nodes in each layer may receive one or more inputs and apply a weight to each input to generate a node output. The node inputs to the first layer may correspond to the model input data fields, and the node outputs of the final layer may correspond to the at least one output of the model, intended to predict the at least one result data field. One or more intermediate layers of nodes may be connected between the nodes of the first layer and the nodes of the final layer. In this fashion, the machine learning model is trained to produce output that reliably predicts the corresponding result data field. Alternatively, the machine learning model may have any suitable structure.

[0179] The computer device 1102 of the present disclosure is configured to operate on input data related to PA data 808 and other data obtained from testing of a sensor 102 within system 100, an / or other (e.g., sensing) data as described herein. In one exemplary embodiment, the computer device 1102 executes the model module 1100 programmed to learn, without limitation, outcomes of particle and / or biological sensing, images such as STEM images, and the like.CT

[0180] To facilitate this learning, the computer device 1102 includes one or more databases 1106 at which the data is stored. This data becomes one or more input training sets used by the training set builder 1110. Model outputs can be formatted for presentation or review as visual representations of recommendations, as text-based or natural language recommendations, and the like (which may be presented to a user such as user 714, 804, 908 via computing device 702, 802, 900, respectively). In exemplary embodiments, model module 1100 may compare feedback, and may route a comparison result 1122 generated by comparing recommendation 1124 to the feedback to a model updater module 1126 of the computer device 1102. Model updater module 1126 is configured to derive a correction signal 1128 from comparison results 1122 received for one or more recommendations 1124 and to provide correction signal 1128 to model trainer module 1120 to enable updating or “re-training” of the at least one machine learning model to improve performance. The retrained at least one machine learning model 1118 may be periodically re-uploaded to model module 1100. An output 1130 from model module 1100 may be used as model 180 shown in FIG. ID. For example, output 1130 may be a trained model that is ready for deployment. When combined with large datasets, the assistance provided to sensor system 100 by model 180, enables rapid scanning and identifying of aspects of a sample, such as a sample including cells in a patient sample and recommendations for an initial diagnosis, in one step. It can also provide valuable particle information for nanomaterial characterization or environmental assessment without needing to wait for a culture or purification and random diffusion step.

[0181] FIG. 12 is a flow diagram of an example method 1200 according to one embodiment of the disclosure. Method 1200 includes constructing 1202 the sensor (e.g., 102), for example in the manner described herein. Method 1200 includes testing the sensor (e.g., 102) for expected performance, including determining 1206 an accuracy of the sensorCT(e.g., 102). This may include comparing properties of the sensor to known quantities. Method 1200 further includes deploying 1208 the sensor (e.g., 102) for sensing, such as sensing AuNP and / or (red blood) cell properties using the PA properties described herein. Method 1200 also includes analyzing 1210 the results obtained and generated by the sensor (e.g., 102). Method 1200 yet further includes making 1212 determinations and / or classifications based on the sensor results. For example, this may include determining characteristics of targets of interest (e.g., target particles and / or molecules) and / or classifying targets of interest as described herein.Additional Aspects and Embodiments

[0182] Additional aspects and embodiments of the present disclosure are provided below, including but not limited to using protected mode (e.g., no direct interaction with sample) to detect PA signals. These additional aspects and / or embodiments designs may use different geometries of resonators, different materials, different layouts, and / or different configurations to interact with samples. For example, single sensor and / or sensor arrays may be implemented.

[0183] FIGS. 13, 14, and 15 are diagrams illustrating label-free all-optical photoacoustic (PA) microresonator-based sensors according to embodiments of the present disclosure, and as an alternative to the embodiment shown in and described in connection with FIG. 1.

[0184] FIG. 13 is a diagram 1300 illustrating a principle of long-range acoustic-assisted sensing. Section 1302 (e.g., Section (i)) of diagram 1300 of FIG. 13 illustrates that conventionally, particle detection through optical sensors is achieved by directly perturbing the photon mode (or hybrid mode by coupling with a plasmonic mode). These evanescent- field-based methods require the binding of particles on the sensor. Analytes flowing in the fluid cannot be detected effectively. Section 1304 (e.g., Section (ii)) of diagram 1300 of FIG. 13 illustrates that in the approach described herein, upon exposure to a pulse laser, analytesCT with matching absorption generate photoacoustic waves propagating in the fluid medium (see also FIGs. 5A and 14 regarding pulse lasers and analytes). These PA waves can be efficiently captured by the high-Q optical mode confined within the dielectric structure. The acoustic phonons mediate a long-range interaction between light and the analyte particles flowing in the solution (e.g., in the microfluidic channel).

[0185] FIG. 14 is a diagram 1400 illustrating a schematic of a sensing platform using PA in an optofluidic microresonator (e.g., a WGM microbubble resonator)), including a photoacoustic process of analytes. Section 1402 (e.g., Section (i)) of diagram 1400 illustrates a PA signal from an analyte far away from WGM(s) (the graph in Section (i) plots transmitted intensity (y-axis) over time (x-axis)). Section 1404 (e.g., Section (ii)) of diagram 1400 illustrates no PA signal is detected in the absence of the analyte (the graph in Section (ii) plots intensity (y-axis) over time (x-axis)). Section 1406 (e.g., Section (iii)) of diagram 1400 illustrates a stronger PA signal from an analyte close / nearby to WGMs such as the WGM shown in FIG. 14 (the graph in Section (iii) plots transmitted intensity (y-axis) over time (x- axis)). FIG. 14 shows a partial resonator 1408, including a body thereof. Red particles 1410 indicate analytes flowing inside a microfluidic channel 1412 of resonator 1408 (e.g., in the form of an optofluidic sensor). A laser 1414 such as a green pulse laser excites analytes 1410 at different locations, which can be captured by a whispering gallery mode (WGM) 1416 that may be confined in the equator 1418 of the body of resonator 1408. A simulation result 1420 (e.g., a COMSOL simulation) of field intensity (e.g., field distribution) of WGM 1416 (more generally referred to as an optical mode), including for air 1422 (e.g., a realization of outside environment as described herein), wall 1424 (e.g., a realization of a wall as described herein), and a liquid core 1426 (e.g., a realization of a core as described herein) is also shown in FIG. 14. Data for one or more target particles 1428 may therefore be obtained using the resonator 1408, where a target particle 1428 may be target analytes (e.g., a target red bloodCT cell 1410). Legend 1430 illustrates pictorial representations of various states of a particle, including excitation, expansion, and ultrasound generation.

[0186] FIG. 15 is a diagram 1500 illustrating selective detection of red blood cells in a complex blood matrix. Left portion 1502 of diagram 1500 is a microscopic image when the microsensor is filled with whole blood. Right portion 1504 of diagram 1500 illustrates the specific excitation of PA signals in red blood cells 1506 in the presence of other cellular elements such as platelets 1508, white blood cells 1510, and proteins 1512 in sample whole blood via sample injection 1514 (e.g., via a syringe) due to the unique absorption of hemoglobin in red blood cells at 532 nm. For example, the syringe associated with sample injection 1514 may be a syringe such as syringe 182 shown and described in connection with FIG. ID. A laser 1516 as described herein (e.g., green laser) performs scanning for targets such as red blood cells 1506, as described herein.

[0187] Some key advantages of the embodiment shown in and described in connection with FIGS. 13-15 include label-free and immobilization-free detection, significantly enhanced detection volume, and direct target detection in complex biological matrices. Different from conventional optical WGM sensors, which detect refractive index changes near the sensing surface (see a comparison in FIGS. 17A, 17B), the present sensing system described herein detects acoustic waves generated by particles in solution flowing freely inside a capillary structure, which significantly improves the sensing capability, throughput, and speed. Furthermore, unlike conventional methods that rely on surface binding via random diffusion, the techniques described herein can actively scan the pulse laser across the microfluid channel and search for target particles. As shown in FIG. 14 Sections (i)-(iii), by scanning along the capillary, the particles far away from the MBR where the WGM resides can also be detected. The resulting PA signal exhibits a delay between the pulse excitation and its optical readout, due to the distance between the optical mode and the particles. Measurements of the PA signals from particles at various distances from the MBRCT were obtained and achieved an extended detection volume. In addition, by selecting the wavelength of the pulse laser that overlaps with the optical absorption of the particle of interest, the target particles can be selectively and effectively detected even in a complex fluidic medium. As shown in FIG. 15, a whole blood sample is injected into the sensor. In such a complex medium, red blood cells can be selectively detected in the presence of other cellular elements as well as proteins, due to the unique absorption wavelength of hemoglobin in red blood cells.

[0188] Additional aspects of detection of cells in whole blood are described below in connection with FIGS. 16A, 16B, 16C, 16D, and 16E. FIGS. 16A, 16B, 16C, 16D, and 16E illustrate photoacoustic fingerprinting of whole blood samples, and more specifically time domain photoacoustic signals and corresponding frequency spectra of whole blood samples from five different species of animals: pig, sheep turkey, goat, and horse. FIG. 16A illustrates time domain photoacoustic signals 1600 and corresponding frequency spectra 1602 of whole blood samples from pigs. FIG. 16B illustrates time domain photoacoustic signals 1604 and corresponding frequency spectra 1606 of whole blood samples from sheep. FIG. 16C illustrates time domain photoacoustic signals 1608 and corresponding frequency spectra 1610 of whole blood samples from turkeys. FIG. 16D illustrates time domain photoacoustic signals 1612 and corresponding frequency spectra 1614 of whole blood samples from goats. FIG. 16E illustrates time domain photoacoustic signals 1616 and corresponding frequency spectra 1618 of whole blood samples from horses. Diagram 1620 of FIG. 16F illustrates PC A visualization of samples using features extracted by the machine learning algorithm, indicating clear decision boundaries and clustering of each species in whole blood samples.

[0189] The design of mode protection within solid walls of the MBR ensures that the optical resonances for photoacoustic detection operate effectively without being affected by the solution's complexity. To explore the feasibility of sensing in complex matrix solutions, tests were performed on whole blood samples of five different species - pig, sheep, turkey,CT goat, and horse, by injecting the sample into the sensor and recording the corresponding PA signals. Whole blood is a complex biological fluid with colloidal properties, comprised of multiple components including water, cellular elements (including red blood cells, white blood cells, and platelets), various metabolites, dissolved electrolytes, a wide variety of proteins, and circulating hormones. The direct measurement of specific analytes in whole blood poses significant challenges for conventional sensor technologies. The presence of numerous biomolecules and cellular components can hinder the selective detection of target substances, leading to compromised accuracy and reliability in sensor performance. The blood samples tested in experiments of the present disclosure were diluted to 1% by volume, without any purification, labeling, or incubation process. FIGS. 16A-16E show the timedomain and frequency-domain PA signals of whole blood from five different species. The PA signal variations among them exhibit characteristic differences that are usable as fingerprints for identification. The spectral features in the PA signal are related to the optical absorption, density, thermal, and mechanical properties in the whole blood samples. Due to the complex composition of the blood matrix, relatively large deviations appear in the PA signal across measurements (see FIGS. 21B and 22A, 22B). To classify the different whole blood samples, robust algorithms are implemented to identify common features within the same kind of whole blood, while ensuring that these features are distinct enough to differentiate between different types of whole blood.

[0190] In whole blood, various factors, such as additional absorption, scattering, and the thermal properties of the fluid, may result in variations in the PA signal profile. To address this, separate machine learning models for whole blood and red blood cell samples were trained, respectively. The model trained on whole blood samples was specifically used for classification tasks involving whole blood. A prototype learning model was employed for the training process, where the model learns a uniform prototype representation for each species, ensuring that individual sample representations are aligned with the prototype. FIG.CT16F demonstrates the effectiveness of using PA signals obtained from whole blood samples for classification purposes. The clear and distinct decision boundaries between different categories highlight the ability to accurately differentiate between various sample types. Note that blood is a complex biological fluid containing various components that can interfere with conventional sensing techniques. However, the clear decision boundaries obtained from the PA signals suggest that this method is resilient to the inherent complexity of whole blood samples and allows for accurate classification of whole blood samples despite the inherent variability in the signal, illustrating the benefits of the systems and methods described herein for blood-based analysis.

[0191] Here, prototype learning was implemented, which maximizes the similarity of representations among samples from the same animal to extract decisive features for classification, but this is not limiting and other types of learning may be implemented.

[0192] FIGS. 17A and 17B are diagrams illustrating a comparison of sensing mechanisms in optical microresonators. Diagram 1700 of FIG. 17A illustrates conventional resonance shift sensing mechanism. The effective refractive index of the optical mode is modified by the binding of the analyte, resulting in the change of resonance wavelength. Diagram 1702 of FIG. 17B illustrates acoustic-mediated sensing. The acoustic wave modulates resonance and induces optical transmission modulation AT at a fixed wavelength.

[0193] Conventional resonance shift sensing is a straightforward method to measure the target of interest by tracking induced changes in the resonance wavelength (or frequency) of an optical resonator. FIG. 17A illustrates the resonance shift induced by the capture of target molecules on the resonator surface in many biosensing applications. This shift measurement allows for the extraction of quantitative and kinetic information about the binding of molecules. However, this method relies on random diffusion processes to bring particles to the sensing surface within the evanescent field, resulting in limited detection efficiency and randomized particle arrival locations. Consequently, only statistical or binary measurementsCT can be obtained, and computation-intensive techniques must be employed for more detailed information. Furthermore, the optical properties of the modes are directly modified by the analyte, making this method unsuitable for detecting large amounts of particles that exhibit scattering or absorption. Additionally, environmental factors can also induce resonance shifts, potentially affecting the sensing accuracy.

[0194] In contrast, acoustic-mediated sensing operates at a fixed wavelength and measures the transmission intensity changes induced by acoustic waves generated by the analyte, as shown in FIG. 17B. The acoustic waves propagate through the sample fluid and are detected by the optical mode, eliminating the reliance on random diffusion processes. The intrinsic photoacoustic and mechanical properties of the analyte are collected in the acoustic waves. In this method, the detection occurs through photon-phonon interaction, and the optical mode remains confined within the resonator without directly interacting with the analyte. As a result, acoustic-mediated sensing is not limited by the optical scattering and absorption of analytes or the fluidic medium. Moreover, since the photoacoustic process produces ultrasound in the MHz frequency range, the detection is background-free, providing a high signal-to-noise ratio and improved accuracy.

[0195] FIG. 18 is a diagram illustrating a machine learning pipeline 1800 for classifying PA signals. FIG. 18 includes aspects similar to FIG. 3B. At section 1802, the PA signals are preprocessed with Fourier Transform as input data. The learned prototype embeddings of the different classes are used for inference. At section 1804, the backbone model (e.g., as described and shown in connection with FIG. 3G) includes a multi-layer network such that includes a 2-layer CNN and a fully connected layer. The distance between the learned features of the sample and the prototype embeddings of each class is calculated in the latent space. Finally, the sample is classified as the class of its closest prototype embeddings.

[0196] More specifically, FIG. 18 shows trained prototype embeddings 1806 used as input 1808 (e.g., in connection with latent space 1814) and PA signals 1812 used as inputCT1812 (e.g., in connection with model 1816 such as a backbone model described in connection with FIG. 3G). With respect to latent space 1814, stars represent prototype embedding(s) and circles represent learned features of samples). An output 1818 of the overall machine learning model inference based on prototype learning aspects shown in sections 1802 and 1804 may include predictions, such as for distances, classifications and / or other related aspects.

[0197] Based on the prototype learning, the model shown in FIG. 18 is capable of capturing the similarities within the same species and focus on the dissimilarities between different species such that the model could generate robust classifications when the signal difference of different species is nuanced.

[0198] FIGS. 19A and 19B are diagrams illustrating confusion matrices of the machine learning models. Diagram 1900 of FIG. 19A illustrates a result on Au nanoparticles including for cubes, rods, shells, and spheres as described herein. Diagram 1902 of FIG. 19B illustrates a result on red blood cells including for goats, llamas, pigs, sheep, and turkeys as described herein.

[0199] The rows of the confusion matrix indicate the true label of the species, and the columns indicate the predictions. With large number on the diagonal elements, the machine learning models achieve excellent predictive performance in classifying the PA signals in both datasets. The details of the result for each category are shown in the confusion matrices of FIGS. 19A and 19B.

[0200] FIGS. 20 A and 20B are diagrams illustrating confusion matrices on red blood cells for goats, llamas, pigs, sheep, and turkeys as described herein. Diagram 2000 of FIG. 20A illustrates a result without prototype learning. Diagram 2002 of FIG. 20B illustrates a result with prototype learning.

[0201] Prototype learning was introduced to make robust classifications. To assess the effectiveness of prototype learning, the performance with and without prototype learningCT were compared. The confusion matrices of the two approaches are displayed in FIGS. 20A and 20B.

[0202] FIGS. 21A and 21B are diagrams illustrating PCA without prototype learning. Diagram 2100 of FIG. 21A illustrates PCA on input features from red blood cell signals. Diagram 2102 of FIG. 21B illustrates PCA on input features from whole blood signals. Without prototype learning, the resulting data points are intermingled, making them difficult to distinguish.

[0203] Principal Component Analysis (PCA) is a visualization method for highdimensional data. The high-dimensional data can be reduced to low dimensional-data via matrix decomposition techniques such that the most important information can be preserved and the irrelevant information is removed. PCA was performed directly on the PA signals of red blood cells and whole blood samples, without the aid of artificial intelligence. With only the raw features, the data of different species overlap considerably, making it challenging to differentiate the different species.

[0204] FIGS. 22A and 22B are diagrams illustrating deviations in the PA signals. PA signals from (a) whole blood (FIG. 22A, diagram 2200) and (b) red blood cell sample of goat (FIG. 22B, diagram 2202). The red curve represents the average of 2000 frame signals. The pink shaded area surrounding the red curve indicates the fluctuations in the PA signals, spanning from the minimum to the maximum values observed across the 2000 frames.

[0205] The PA signals obtained from whole blood samples exhibit larger deviations compared to those from red blood cell samples. These increased deviations can be attributed to the complex composition of whole blood. In addition to red blood cells, whole blood contains various other components such as white blood cells, platelets, and plasma, each of which may contribute to the PA signal in different ways. The presence of these additional components introduces more variability in the PA signals, leading to larger fluctuations and a wider range of signal intensities.CT

[0206] FIG. 23 is a diagram illustrating a microring resonator 2300 on a substrate 2302 such as a silicon substrate according to the techniques, systems, and / or methods described herein. Soft materials such as polymers (PDMS, PMMA, Polycarbonate, Polyurethane, Hydrogels, PTFE), Silicone Elastomers, Liquid Crystals may be used, for example, for a protection layer (e.g., encapsulation). Hard materials such as Silicon, Silicon Nitride, Silica, Gallium Arsenide, Indium Phosphide, Lithium Niobate, Glass, Aluminum Oxide may be used, for example, for substrate materials. A waveguide 2304 may be adjacent a portion of a microring 2306 of the microring resonator 2300. Additionally, or alternatively, other shapes / configurations may be implemented for the microresonator structure, including a microbottle shape / structure, a microsphere shape / structure, a microtoroid shape / structure, and the like.

[0207] FIG. 24 is a diagram illustrating an on-chip sensor 2400 for droplet (e.g., blood droplet) sample sensing according to the techniques, systems, and / or methods described herein. The on-chip sensor 2400 includes a sensor layer 2402 and a protection layer 2404 on which a droplet sample 2406 may be placed. On-chip sensor 2400 may include a waveguide 2408 and a microring 2410 and may be implemented as a PA sensor.

[0208] FIG. 25 is a diagram illustrating an on-chip sensor 2500 integrated with microfluidics according to the techniques, systems, and / or methods described herein. The on-chip sensor 2500 includes a sensor layer 2502, a protection layer 2504, and a microfluidic layer 2506. The microfluidic layer 2506 may include a microfluidic channel 2508 providing for sampling (e.g., sample in at one portion of the microfluidic channel 2508, and sample out at another portion of the microfluidic channel 2508). On-chip sensor 2500 may include a waveguide 2510 and a microring 2512 and may be implemented as a PA sensor.

[0209] FIG. 26 is a diagram illustrating wearable sensor 2600 using flexible photonic structures according to the techniques, systems, and / or methods described herein. The wearable sensor 2600 may include a wearable pad 2602 including a waveguide 2604 and aCT microring 2606 and may be implemented as a PA sensor. In this configuration, the wearable sensor 2600 can detect PA signals generated under the skin (e.g., skin 2608).

[0210] FIG. 27 is a diagram illustrating a microsphere fiber probe 2700, including an optical fiber 2702 and a microsphere resonator 2704, according to the techniques, systems, and / or methods described herein. In one embodiment, a U-shaped waveguide may be wrapped around a WGM resonator sensor to form a sensor probe suitable for applications with limited space for the sensor, such as endoscopic probes.

[0211] FIGS. 28 A and 28B are diagrams illustrating implementations of endoscopic probes that may be implemented as PA sensors, according to the techniques, systems, and / or methods described herein. FIG. 28A is a diagram 2800 illustrating a microsphere fiber probe 2802 scanning through a sample medium 2804. FIG. 28B is a diagram 2806 illustrating a microsphere fiber probe 2808 scanning through a medium / vessel 2810 such as gas pipes or blood vessels to detect PA signals. Probes 2802 and 2808 may be configured the same as or similar to that shown in FIG. 27.

[0212] FIGS. 29 A and 29B are diagrams illustrating implementations of endoscopic probes that may be implemented as PA sensors, according to the techniques, systems, and / or methods described herein. FIG. 29A is a diagram 2900 illustrating a microsphere fiber probe 2902 for environmental monitoring in water 2904 where the probe 2902 may be implemented as part of a water drone 2906. FIG. 29B is a diagram 2908 illustrating a microsphere fiber probe 2910 scanning through a water pool 2912. Probes 2902 and 2910 may be configured the same as or similar to that shown in FIG. 27.

[0213] FIGS. 30, 31, 32, and 33 illustrate diagrams of additional types of WGM resonator sensors according to the techniques, systems, and / or methods described herein, such as (e.g., capillary-based) WGM resonator sensors. FIG. 30 is a diagram illustrating a WGM resonator sensor 3000 including a microbubble structure 3002 with WGM 3004 (more specifically, a microbubble resonator with WGM). FIG. 31 is a diagram illustrating a WGMCT resonator sensor 3100 including a capillary structure 3102 with WGM 3104 (more specifically, a capillary resonator with WGM). FIG. 32 is a diagram illustrating a WGM resonator sensor 3200 including a microbubble array 3202 including microbubble structures 3204 with corresponding WGMs 3206 (more specifically, a microbubble resonator array with WGMs for multiplexing). FIG. 33 is a diagram illustrating a WGM resonator sensor 3300 including a capillary structure 3302 with multiple WGMs 3304 configured as a sensor array (more specifically, a capillary structure with multiple WGMs for multiplexing).

[0214] FIG. 34 is a diagram illustrating an on-chip sensor array 3400 for multiplexing, according to the techniques, systems, and / or methods described herein. The on-chip sensor array 3400 may include a sensor array layer 3402, a protection layer 3404, and a sample layer 3406. The sample layer 3406 may be liquid, solid, gel, etc. Target molecules / particles 3408 of a sample may be present / provided in the sample layer 3406. On-chip sensor array 3400 may include a plurality of waveguides 3410 and microrings 3412.1. Chip scale devices

[0215] Chip-scale optical devices, such as microring resonators, offer compact, integrated platforms capable of PA sensing. Their compact size allows for integration into portable and miniaturized devices, enabling real-time, on-site monitoring without the need for bulky equipment. This miniaturization also leads to lower material costs and reduced power consumption, making chip-scale sensors not only efficient but also cost-effective. They can be fabricated using materials compatible with standard semiconductor processes, allowing for mass production and easy integration with existing electronic and photonic systems. This scalability is crucial for developing high-volume, low-cost sensing solutions. Additionally, the ability to use a variety of materials - both hard (such as silicon, silicon nitride) and soft (such as PDMS or hydrogels) - enables customization of the sensor properties to suit specific applications, whether they require high rigidity and performance or flexibility and biocompatibility.CT

[0216] These devices can be fabricated from a variety of materials, both soft and hard, each tailored to specific application needs. Hard materials like silicon, silicon nitride, and silica are commonly used for microring resonators due to their excellent optical properties, high refractive indices, and compatibility with established semiconductor manufacturing processes. These materials enable robust, high-performance devices ideal for telecommunications, biosensing, and environmental monitoring. On the other hand, soft materials such as PDMS, polycarbonate, and hydrogels introduce flexibility and biocompatibility, making them suitable for wearable sensors or bio-integrated devices.

[0217] FIG. 23 shows a typical setup for a chip-scale optical sensor featuring a ring resonator coupled with a straight waveguide. The waveguide guides light from an input source through total internal reflection. As light propagates along the waveguide, a portion of it couples into the microring resonator positioned close to the waveguide. WGMs excited in the microring resonator can detect PA signals in the similar way in the microbubble resonator.2. Droplet sensing

[0218] FIG. 24 shows an on-chip sensor for droplet sample sensing operates by placing a small droplet of the sample directly onto the surface of the sensor chip, allowing real-time analysis with high sensitivity and precision. Typically, the sensor chip incorporates an optical waveguide or resonator, such as a microring resonator, which captures the acoustic signal generated by the PA process in the droplet. The substrate with the photonic structure is the sensor layer. To protect the optical mode from direct interaction with the sample / environment, a thin protection layer is applied on top of the sensor layer so the droplet can be placed on top of it. This layer should also exhibit low acoustic attenuation such that the PA signal generated from the droplet sample can travel through it and captured by the sensor.CT

[0219] This method is particularly useful for applications like biosensing, chemical detection, and environmental monitoring, where small sample volumes are required. The sensor's ability to function in a label-free and real-time manner provides efficient and rapid analysis with minimal sample preparation, making it ideal for point-of-care diagnostics or portable sensing devices.3. Integration with microfluidics

[0220] An on-chip sensor can be integrated with microfluidics in many ways, but not limited to, by adding a layer of microfluidics on top. It offers a powerful platform for precise, efficient, and automated sample analysis in a miniaturized environment. This system combines the sensing capabilities of chip-scale optical devices, such as waveguides or microring resonators, with the fluid handling precision of microfluidic channels, allowing for the controlled manipulation of tiny volumes of liquid or gas samples.

[0221] In the embodiment shown in FIG. 25, the microfluidic channels are etched or printed directly onto the chip, guiding the flow of the sample over or through the sensing region. The substrate with the photonic structure is the sensor layer. To protect the optical mode from direct interaction with the sample / environment, a thin protection layer is applied on top of the sensor layer. This layer should also exhibit low acoustic attenuation such that the PA signal generated from the droplet sample can travel through it and captured by the sensor. On top of the protection layer is the microfluidic layer. When a fluid sample, such as a biological specimen or chemical solution, is introduced into the microfluidic system, it does not directly interact with the integrated sensor like many other conventional designs. Instead, the sensor may only detect the PA signals.

[0222] This integration offers several advantages: (1) it allows precise control of the sample volume, enabling experiments with picoliter or nanoliter-sized droplets, which is crucial for applications where sample availability is limited; (2) the microfluidic channels can be designed to automate processes such as mixing, dilution, or delivery, providing a highCT degree of repeatability and efficiency in complex assays; (3) the combination of on-chip sensors and microfluidics allows for multiplexed sensing, where multiple sensors are integrated into the same chip to simultaneously detect different targets or perform various analyses in parallel, significantly increasing throughput. As shown in FIG. 34, the sensor layer may include multiple microring resonators as a sensor array. Each resonator on the array may have similar or entirely different geometry design and / or materials. They can provide similar response (e.g., provide multiple sampling) or different response (e.g., provide differential measurement or other purposes) to the same acoustic signal. Multiple microrings on the same chip may be implemented as a PA sensor array for more sophisticated measurements and / or PA imaging. To protect the optical mode from direct interaction with the sample / environment, a thin protection layer is applied on top of the sensor layer. The sample layer is on top of the protection layer, which can be fluidic channel, solution, solid, or gel matrix such as hydrogel. Target molecules in the sample, whatever the form of the sample is, can be detected from the PA signals. The probe laser can scan across the sample and locate the target molecule at different positions. This highly integrated system reduces the need for large equipment, simplifies sample handling, and provides real-time, label-free detection, paving the way for lab-on-a-chip devices suitable for point-of-care diagnostics, environmental monitoring, and biochemical analysis.4. Wearable Sensing Pad Based on Flexible Photonics

[0223] Using soft materials, photonic devices, such as microring resonators, can be built into a wearable format to enable highly sensitive and non-invasive detection of various sensing applications. The wearable pad can be placed in contact with the skin (e.g., FIG. 26), such as hand, forehead, and back. The flexibility of the pad allows it to conform seamlessly to the body, ensuring comfort and close contact with the skin.

[0224] A major application of this design lies in detecting disease-related biomarkers in body fluids such as sweat. Sweat contains a range of biochemical markers - such as glucose,CT lactate, electrolytes, and proteins - that can provide early warning signs of health conditions. For instance, by monitoring glucose levels in sweat, this sensor offers a non-invasive alternative for managing diabetes, reducing the need for painful finger-pricking tests. Similarly, it can detect inflammatory proteins or infection-related biomarkers, enabling early diagnosis of infections or monitoring immune responses. This makes it a powerful tool for tracking disease progression or response to treatment.

[0225] The sensing pad is also capable of non-invasive monitoring vital signs, including blood oxygenation (SpCh) and blood pressure, using PA signals. Modulated light source such as laser pulses are directed into the body, where they are absorbed by biological tissues and converted into ultrasound waves. The resulting signals provide rich information about blood oxygen levels and vascular health. By measuring oxygen saturation, the device can offer insights into respiratory and cardiovascular conditions, such as chronic obstructive pulmonary disease (COPD) or heart disease. This method can also provide blood pressure information.5. Optical fiber probes

[0226] FIG. 27 shows an optical fiber probe designed with a U-shaped bend and coupled to a microsphere resonator. The design takes advantage of the unique properties of both the optical fiber as the probe that can be inserted into certain target of interest such as pipes, reaction chambers, sample containers, rivers, and human bodies, and the microsphere resonator as a sensing element.

[0227] In this configuration, the optical fiber is bent into a U-shape, allowing the evanescent field — the portion of the light that extends outside the fiber — to couple into the microsphere resonator. The coupling between the fiber and the microsphere occurs at the U- bend, but it does not limit to a U-bend, can also be achieved by integrating grating structures such as Bragg gratings on the optical fiber, or other kinds of coupling configuration. WGMsCT in the microsphere resonator enables the detection of PA signals, and the signals can be delivered by the optical fiber over long distance.

[0228] As shown in FIGS. 28A and 28B, when the probe is immersed in a liquid sample medium, it can scan through the solution for rare molecule detection. As the microsphere passes through the solution, any PA signals generated by target molecules in the solution causes modulation in WGMs. The high sensitivity and label -free detection of the WGM allows for the detection of extremely low concentrations of molecules, together with the high specificity brought by optical absorption and PA fingerprinting, the probe is ideal for applications such as detecting biomarkers in biological fluids.

[0229] By placing the probe directly into river water, it can detect and analyze a wide range of pollutants, contaminants, and other chemical or biological substances with high sensitivity. The probe can be deployed directly into the river, either tethered to a stationary platform, buoy, or mounted on a mobile device like an autonomous water drone. The U- shaped fiber configuration ensures enhanced interaction between the evanescent field of the light in the fiber and the surrounding water, increasing the sensitivity of the probe to trace substances in, for example, a river.

[0230] As shown in FIGS. 29A and 29B, as the probe is mounted on a water drone in the river, it continuously scans the water for various contaminants. The probe can be used for detecting a wide range of substances, including heavy metals, pesticides, nitrates, phosphates, and even microbial life, such as bacteria or algae. The high Q-factor of the microsphere resonator allows it to detect even minute changes in the water composition, such as the presence of a few molecules of a pollutant.

[0231] FIG. 30 shows the single microbubble resonator used in experiments of the present disclosure, which a WGM excited near the equator of the bubble. FIG. 31 shows that capillary resonator can also support WGM for PA sensing. The single sensing element can further expand into sensor array, in many ways and not limited to the examples shown here.CTFor instance, FIG. 32 shows a microbubble resonator array where multiple microbubble resonators are fabricated in the same capillary. Each microbubble resonator can support its own WGMs and form a sensor array. In this case, it is possible to use each microbubble resonator to specifically detect a particular target of interest in the sample. FIG. 33 shows a similar configuration on the capillary, where multiple WGMs can be excited at different positions along the capillary. Each WGM can be used to specifically detect a particular target of interest in the sample.Sensors

[0232] FIG. 35 illustrates a configuration of a sensor 3500 such as a Fabry -Perot resonator / interferometer sensor, demonstrating aspects described herein and according to one embodiment of the present disclosure. Sensor 3500 may include a sample medium 3502 sandwiched between reflecting elements 3504 and 3506 (also referred to as reflecting surfaces), where element 3504 may be a first reflection element and element 3506 may be a second reflection element. The sample medium can also be outside the reflecting elements. It can be on one side of the reflecting element, or both sides. It can also be away from the two reflecting elements to minimize optical interference from the sample solution. For example, as long as the acoustic wave can induce detectable changes in the resonator / interferometer, it can be implemented as a PA sensor. Reflecting elements such as elements 3504 and 3506 may include two (or more) parallel (or semi-parallel) reflecting elements. FIG. 35 further illustrates target molecules 3508 within sample medium 3502, input light 3510 input into sensor 3500, and reflection light 3512 exiting from sensor 3500. Reflection light 3512 may result from light reflection 3514 within sensor 3500. Light such as that associated with input light 3510 is able to resonate in the Fabry-Perot cavity by reflection between two or more reflecting elements such as elements 3504 and 3506. The reflecting surfaces or elements (e.g., 3504, 3506) may be configured as thin mirrors, Brag gratings, photonic crystals, and / or fiber end surfaces, and have the sample medium (e.g.,CT3502, also referred to as a sample layer) between them. The sample layer (e.g., 3502) may include a microfluidic channel providing sample delivery in solution phase. The sample layer (e.g., 3502) may also be in a gas-phase, solid phase, or gel phase (such as hydrogel) configuration. The Fabry -Perot resonator / interferometer shown in FIG. 35 may be configured in the form of on-chip waveguides with Brag gratings, photonic crystals, and / or other forms of reflectors (e.g., scattering), and / or fiber structures with reflectors, and / or a microfluidic channel sandwiched by the micromirrors, and may be implemented as a photoacoustic (PA) sensor. In some embodiments, including but not limited to those shown in and described in connection with FIGS. 24, 25, and / or 35, the sample medium may be configured as a solid or a gel (e.g., Lipid bilayer) phase sample medium, capable of being placed on top of a packaged sensor using configurations such as those shown in FIGS. 24 and 25.Conclusions

[0233] Described herein are sensors and / or sensor systems / platforms for detecting properties of particles and / or other materials, learning properties of particles and / or other materials, and / or making predictions about particles and / or other materials based on learned properties. These properties may include PA properties as described herein. For example, some embodiments include an optofluidic sensing platform that integrates highly sensitive optical modes (e.g., for optical resonance) and the photoacoustic effect to achieve rapid, label-free, and high throughput measurements of particles. Unlike evanescent-field-based and SPR-based sensing platforms, the immobilization-free approach described herein can detect free-flowing particles in fluid (away from the sensing surface), with excellent specificity achieved using PA spectroscopic signatures captured by a high-Q MBR. By confining optical modes within the wall of the MBR sensor, this technique enables long- range acoustic-mediated measurements (e.g., resonant measurements) without reliance on random diffusion and spatially decouples the optical mode from the sensing volume,CT providing great resistance to potential contaminations in the solution. The measurement does not require external fluidic channels and chambers, lengthy culturing, expensive reagents, or thermal cycling equipment, and is robust to refractive index changes in the solution. The label-free and immobilization-free approach / approaches described herein help mitigate surface fouling, mass transport limitations, and surface regeneration challenges in conventional photoacoustic properties of particles can be obtained and implemented to classify the different morphologies of nanoparticles or different species of cells. Furthermore, other abundant information about the photoacoustic properties of particles can be obtained and implemented to classify the different morphologies of nanoparticles or different species of cells. The machine learning models described herein achieve accurate classification, effectively distinguishing different types of particles and cells.

[0234] The PA process described herein exhibits intrinsic sensitivity to both the functional and molecular composition of the sample, leveraging the extensive optical absorption contrast pre-sent in biological systems. By choosing appropriate wavelengths to excite PA signals, selective collection of signals from specific objects can be realized. This unique capability allows for the non-invasive monitoring of a substantial number of red blood cells in their native physiological state when a laser such as a 532 nm laser is used. The approach described herein circumvents the inherent complexities of whole blood, which often hinder conventional analytical techniques due to the potential for interference from the myriad of cellular and molecular components present in the blood matrix. In addition, the techniques described herein minimize the influence of external perturbations, such as temperature fluctuation and refractive index variations, to ensure an accurate assessment of the sensing targets through their PA signature. It is worth noting that the characteristics of the photoacoustic signals could be affected by various factors that influence the generation of the acoustic waves through transient thermoelastic effects induced by a pulsed laser. Photoacoustic generation in gold nano-particle colloidal suspensions differs fundamentallyCT from that in bulk solids or liquids. During laser exposure, the high thermal conductivity and nanoscale dimensions of AuNPs facilitate rapid heat transfer to the surrounding medium, which could affect the PA signals. The photoacoustic generation is also potentially affected by the laser pulse width, and the PA frequency of nanoparticles can be tuned from several MHz to hundreds of GHz by changing the pulse width. However, high-frequency components may attenuate significantly in the far-field, and current microresonator bandwidth limitations may restrict detection within this range. The present disclosure envisions sensor design that enables detection at even higher frequencies than those listed herein as well. While PA signals or opto-mechanical oscillations when the sensor was filled with water were not observed, mechanical resonances of the microresonator may help enhance the detection of the PA waves. When the photoacoustic frequency approaches by these resonances, the PA signals could be resonantly enhanced, potentially improving the detection sensitivity for weakly absorbing samples.

[0235] In principle, the techniques described herein can be applied to the detection and characterization of a wide range of particles and cells in their natural states and / or environment(s), (including particles and molecules in the gas phase, for example). Both artificial particles in nanomaterials or cells in biology can be measured with high throughput by choosing the proper wavelength of laser pulses or using a frequency comb in the PA process. This technique is not limited to liquid phase sensing and offers versatility beyond liquid phase sensing and is envisioned for use / application in detecting particles and / or molecules in the gas phase, providing a powerful sensing platform for a variety of applications. For instance, features in the PA spectra can be further used to study the status of red blood cells and differentiate diseased cells from healthy ones. This can be implemented to enable rapid disease diagnosis, hemoglobin C disease, for example, or hemoglobin S-C disease, sickle cell anemia, and various types of thalassemia. Highly automated measurement and data acquisition processes can be implemented for clinical and industrial applications.CTWhen combined with large datasets, the Al-assisted (e.g., ML-assisted) sensing platform presented herein is able to rapidly scan and identify cells in a patient sample and recommend an initial diagnosis in one step. The disclosed sensing platform can also provide valuable particle information for nanomaterial characterization or environmental assessment without needing to wait for a culture or purification and random diffusion step. Such a smart sensing platform holds significant potential for advancing diagnostics, nanomaterial industries, and environmental monitoring. The various AI / ML models described herein may be implemented as shown and described in connection with FIGs. ID and / or 11. For example, the various AI / ML models described herein may be implemented as one or more models 180 shown in FIG. ID and / or as one or more outputs 1130 shown in FIG. 11.

[0236] For all of the above-described embodiments and usages, any code and / or data or other information may be stored in a memory of the above-described system, and / or in a remote (e.g., cloud) storage system (e.g., in a dedicated database or other centralized storage mechanism). Embodiments of the invention may be described in the general context of computer-executable instructions, such as program modules, executed by one or more computers or other devices. The computer-executable instructions may be organized into one or more computer-executable components or modules. Generally, program modules include, but are not limited to, routines, programs, objects, components, and data structures that perform particular tasks or implement particular abstract data types. Aspects of the invention may be implemented with any number and organization of such components or modules. For example, aspects of the invention are not limited to the specific computer-executable instructions or the specific components or modules illustrated in the figures and described herein. Other embodiments of the invention may include different computer-executable instructions or components having more or less functionality than illustrated and described herein. Aspects of the invention may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linkedCT through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

[0237] In operation, a computer executes computer-executable code / instructions embodied in one or more computer-executable components stored on one or more computer- readable media to implement aspects of the invention described and / or illustrated herein. Code can include application source code, application object code, configuration data, additional input events, application states, assertion statements, validation results, and / or any other type of data. The order of execution or performance of the operations in embodiments of the invention illustrated and described herein is not essential, unless otherwise specified. That is, the operations may be performed in any order, unless otherwise specified, and embodiments of the invention may include additional or fewer operations than those disclosed herein. For example, it is contemplated that executing or performing a particular operation before, contemporaneously with, or after another operation is within the scope of aspects of the invention.

[0238] The raw and / or processed data and / or any related graphical or other representations of the data may be processed by the above-described computer system or the like and output for display on a display device such as a TV, monitor, mobile device (e.g., mobile phone or tablet) and the like such that a technician / practitioner / evaluator / therapist / user can view and / or manipulate the data (e.g., the data may be presented in a visual format for presenting certain aspects of the test results, for example as shown in the applicable above-noted figures). For example, a display monitor may be connected (e.g., wired or wirelessly) to the above- described computer system to provide a visual output on the computer system. The computer system may have an operating system with a graphical user interface capable of being used by a user to (i) input, view, execute and / or manipulate the above- described computer code and / or (ii) process theCT obtained sensor data and any related graphical representations of such data in the manners described above. The operating system may be capable of running commercially available software applications such as those described above for carrying out the above-described techniques and also any necessary post-processing and / or outputting of the obtained sensor data for viewing, such as for viewing by a therapist that is treating / diagnosing a patient / test subject. Additional software for other code / data manipulations and / or for generating other visuals relating to the data may also be present on the computer system.

[0239] In the present disclosure, all or part of the units or devices of any system and / or apparatus, and / or all or part of functional blocks in any block diagrams and flow charts may be executed by one or more electronic circuitries including a semiconductor device, a semiconductor integrated circuit (IC) (e.g., such as a processor), or a large-scale integration (LSI). The LSI or IC may be integrated into one chip and may be constituted through combination of two or more chips. For example, "processor" as used herein refers generally to any programmable system including systems and microcontrollers, reduced instruction set circuits (RISC), application specific integrated circuits (ASIC), programmable logic circuits (PLC), and any other circuit or processor capable of executing the functions described herein. The functional blocks other than a storage element may be integrated into one chip. The integrated circuitry that is called LSI or IC in the present disclosure is also called differently depending on the degree of integrations, and may be called a system LSI, VLSI (very large- scale integration), or ULSI (ultra large-scale integration). For an identical purpose, it is possible to use an FPGA (field programmable gate array) that is programmed after manufacture of the LSI, or a reconfigurable logic device that allows for reconfiguration of connections inside the LSI or setup of circuitry blocks inside the LSI. Furthermore, part or all of the functions or operations of units, devices or parts or all of devices can be executed by software processing (e.g., coding, algorithms, etc.). In this case, the software is recorded one or more non-transitory computer-readable recording media, such as one or more ROMs,CTRAMs (e.g., DRAM, SRAM), optical disks, hard disk drives, solid-state memory, servers, cloud storage, and so on and so forth, having stored thereon executable instructions which can be executed to carry out the desired processing functions and / or circuit operations. For example, when the software is executed by a processor, the software causes the processor and / or a peripheral device to execute a specific function within the software. The system / method / device of the present disclosure may include (i) one or more non-transitory computer-readable recording mediums that store the software, (ii) one or more processors (e.g., for executing the software or for providing other functionality), and (iii) a necessary hardware device (e.g., a hardware interface). Artificial intelligence in any and all types and formats may be utilized in any of the steps, techniques, protocols, analyses, and / or any other manipulation, generation, or other creation of data, results and / or any information described herein. This includes but is not limited to computer visions, machine learning, deep learning, neural networks, algorithms, and any data, models, and training needed for such. The above examples are example only, and thus are not intended to limit in any way the definitions and / or meanings of the terms.

[0240] Data conduits and any other communication or data transfer as described herein may include wired or wireless connections. For example, a wired network connection (e.g., Ethernet or an optical fiber), a wireless communication means, such as radio frequency (RF), e.g., FM radio and / or digital audio broadcasting, WiFi (e.g., IEEE 802.11 standards), WIMAX, a short-range wireless communication channel such as BLUETOOTH, a cellular phone technology (e.g., GSM), a satellite communication link, and / or any other suitable communication means. Such data conduits, in particular wired versions, can also be referred to as a system bus.

[0241] As will be appreciated based upon the foregoing specification, the abovedescribed embodiments of the disclosure may be implemented using computer programming or engineering techniques including computer software, firmware, hardware or anyCT combination or subset thereof. Any such resulting program, having computer-readable code means, may be embodied or provided within one or more computer-readable media, thereby making a computer program product, i.e., an article of manufacture, according to the discussed embodiments of the disclosure. The computer-readable media may be, for example, but is not limited to, a fixed (hard) drive, diskette, optical disk, magnetic tape, semiconductor memory such as read-only memory (ROM), and / or any transmitting / receiving medium such as the Internet or other communication network or link. The article of manufacture containing the computer code may be made and / or used by executing the code directly from one medium, by copying the code from one medium to another medium, or by transmitting the code over a network.

[0242] These computer programs (also known as programs, software, software applications, “apps,” or code) include machine instructions for a programmable processor and can be implemented in a high-level procedural and / or object-oriented programming language, and / or in assembly / machine language. As used herein, the terms “machine- readable medium” “computer-readable medium” refers to any computer program product, apparatus and / or device (e.g., magnetic discs, optical disks, memory, Programmable Logic Devices (PLDs)) used to provide machine instructions and / or data to a programmable processor, including a machine-readable medium that receives machine instructions as a machine-readable signal. The “machine-readable medium” and “computer-readable medium,” however, do not include transitory signals. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0243] As used herein, the terms “processor” and “computer” and related terms, e.g., “processing device”, “computing device”, and “controller” are not limited to just those integrated circuits referred to in the art as a computer, but broadly refers to a microcontroller, a microcomputer, a programmable logic controller (PLC), a reduced instruction set circuitCT(RISC), an application specific integrated circuit (ASIC), logic circuits, and any other circuit or processor capable of executing the functions described herein. The above examples are example only and are thus not intended to limit in any way the definition and / or meaning of the term “processor.”

[0244] As used herein, the terms “software” and “firmware” are interchangeable, and include any computer program stored in memory for execution by a processor, including RAM memory, ROM memory, EPROM memory, EEPROM memory, and non-volatile RAM (NVRAM) memory. The above memory types are example only and are thus not limiting as to the types of memory usable for storage of a computer program.

[0245] As used herein, the term “database” can refer to either a body of data, a relational database management system (RDBMS), or to both. As used herein, a database can include any collection of data including hierarchical databases, relational databases, flat file databases, object-relational databases, object-oriented databases, and any other structured collection of records or data that is stored in a computer system. The above examples are example only, and thus are not intended to limit in any way the definition and / or meaning of the term database. Examples of RDBMS’ include, but are not limited to including, Oracle® Database, MySQL, IBM® DB2, Microsoft® SQL Server, Sybase®, and PostgreSQL. However, any database can be used that enables the systems and methods described herein. (Oracle is a registered trademark of Oracle Corporation, Redwood Shores, California; IBM is a registered trademark of International Business Machines Corporation, Armonk, New York; Microsoft is a registered trademark of Microsoft Corporation, Redmond, Washington; and Sybase is a registered trademark of Sybase, Dublin, California.)

[0246] In another example, a computer program is provided, and the program is embodied on a computer-readable medium. In an example, the system is executed on a single computer system, without requiring a connection to a server computer. In a further example, the system is being run in a Windows® environment (Windows is a registered trademark ofCTMicrosoft Corporation, Redmond, Washington). In yet another example, the system is run on a mainframe environment and a UNIX® server environment (UNIX is a registered trademark of X / Open Company Limited located in Reading, Berkshire, United Kingdom). In a further example, the system is run on an iOS® environment (iOS is a registered trademark of Cisco Systems, Inc. located in San Jose, CA). In yet a further example, the system is run on a Mac OS® environment (Mac OS is a registered trademark of Apple Inc. located in Cupertino, CA). In still yet a further example, the system is run on Android® OS (Android is a registered trademark of Google, Inc. of Mountain View, CA). In another example, the system is run on Linux® OS (Linux is a registered trademark of Linus Torvalds of Boston, MA). The application is flexible and designed to run in various different environments without compromising any major functionality.

[0247] As used herein, an element or step recited in the singular and proceeded with the word “a” or “an” should be understood as not excluding plural elements or steps, unless such exclusion is explicitly recited. Furthermore, references to “example” or “one example” of the present disclosure are not intended to be interpreted as excluding the existence of additional examples that also incorporate the recited features. Further, to the extent that terms “includes,” “including,” “has,” “contains,” and variants thereof are used herein, such terms are intended to be inclusive in a manner similar to the term “comprises” as an open transition word without precluding any additional or other elements.

[0248] In some embodiments, numbers expressing sizes, quantities, and so forth, used to describe and claim certain embodiments of the present disclosure are to be understood as being modified in some instances by the term “about.” In some embodiments, the term “about” is used to indicate that a value includes the standard deviation of the mean for the device or method being employed to determine the value. In some embodiments, the numerical parameters set forth in the written description and attached claims are approximations that can vary depending upon the desired properties sought to be obtainedCT by a particular embodiment. In some embodiments, the numerical parameters should be construed in light of the number of reported significant digits and by applying ordinary rounding techniques. Notwithstanding that the numerical ranges and parameters setting forth the broad scope of some embodiments of the present disclosure are approximations, the numerical values set forth in the specific examples are reported as precisely as practicable. The numerical values presented in some embodiments of the present disclosure may contain certain errors necessarily resulting from the standard deviation found in their respective testing measurements. The recitation of ranges of values herein is merely intended to serve as a shorthand method of referring individually to each separate value falling within the range. Unless otherwise indicated herein, each individual value is incorporated into the specification as if it were individually recited herein. The recitation of discrete values is understood to include ranges between each value.

[0249] The terms “optional” or “optionally” means that the subsequently described event or circumstance may or may not occur, and that the description includes instances where the event occurs and instances where it does not.

[0250] The terms “comprise,” “have” and “include” are open-ended linking verbs. Any forms or tenses of one or more of these verbs, such as “comprises,” “comprising,” “has,” “having,” “includes” and “including,” are also open-ended. For example, any method that “comprises,” “has” or “includes” one or more steps is not limited to possessing only those one or more steps and can also cover other unlisted steps. Similarly, any composition or device that “comprises,” “has” or “includes” one or more features is not limited to possessing only those one or more features and can cover other unlisted features.

[0251] Use of ordinal terms such as “first,” “second,” “third,” etc., in the claims to modify a claim element does not by itself connote any priority, precedence, or order of one claim element over another or the temporal order in which acts of a method are performed, but are used merely as labels to distinguish one claim element having a certain name fromCT another element having a same name (but for use of the ordinal term) to distinguish the claim elements.

[0252] It should be noted that, as used herein, the term “couple” is not limited to a direct mechanical, electrical, and / or communication connection between components, but may also include an indirect mechanical, electrical, and / or communication connection between multiple components.

[0253] Furthermore, as used herein, the term “real-time” refers to at least one of the time of occurrence of the associated events, the time of measurement and collection of predetermined data, the time to process the data, and the time of a system response to the events and the environment. In the examples described herein, these activities and events occur substantially instantaneously.

[0254] All methods described herein can be performed in any suitable order unless otherwise indicated herein or otherwise clearly contradicted by context. The use of any and all examples, or exemplary language (e.g., “such as”) provided with respect to certain embodiments herein is intended merely to better illuminate the present disclosure and does not pose a limitation on the scope of the present disclosure otherwise claimed. No language in the specification should be construed as indicating any non-claimed element essential to the practice of the present disclosure.

[0255] Groupings of alternative elements or embodiments of the present disclosure disclosed herein are not to be construed as limitations. Each group member can be referred to and claimed individually or in any combination with other members of the group or other elements found herein. One or more members of a group can be included in, or deleted from, a group for reasons of convenience or patentability. When any such inclusion or deletion occurs, the specification is herein deemed to contain the group as modified thus fulfilling the written description of all Markush groups used in the appended claims.CT

[0256] In some embodiments, the system includes multiple components distributed among a plurality of computer devices. One or more components may be in the form of computer-executable instructions embodied in a computer-readable medium. The systems and processes are not limited to the specific embodiments described herein. In addition, components of each system and each process can be practiced independent and separate from other components and processes described herein. Each component and process can also be used in combination with other assembly packages and processes. The present embodiments may enhance the functionality and functioning of computers and / or computer systems.

[0257] The computer-implemented methods discussed herein can include additional, less, or alternate actions, including those discussed elsewhere herein. The methods can be implemented via one or more local or remote processors, transceivers, servers, and / or sensors (such as processors, transceivers, servers, and / or sensors mounted on vehicles or mobile devices, or associated with smart infrastructure or remote servers), and / or via computerexecutable instructions stored on non-transitory computer-readable media or medium. Additionally, the computer systems discussed herein can include additional, less, or alternate functionality, including that discussed elsewhere herein. The computer systems discussed herein can include or be implemented via computer-executable instructions stored on non- transitory computer-readable media or medium.

[0258] As used herein, the term “non-transitory” computer-readable media / medium is intended to be representative of any tangible computer-based device implemented in any method or technology for short-term and long-term storage of information, such as, computer-readable instructions, data structures, program modules and sub-modules, or other data in any device. Therefore, the methods described herein can be encoded as executable instructions embodied in a tangible, non-transitory, computer readable medium, including, without limitation, a storage device and / or a memory device. Such instructions, when executed by a processor, cause the processor to perform at least a portion of the methodsCT described herein. Moreover, as used herein, the term “non-transitory” computer-readable media / medium includes all tangible, computer-readable media, including, without limitation, non-transitory computer storage devices, including, without limitation, volatile and nonvolatile media, and removable and non-removable media such as a firmware, physical and virtual storage, CD-ROMs, DVDs, and any other digital source such as a network or the Internet, as well as yet to be developed digital means, with the sole exception being a transitory, propagating signal.

[0259] Additionally, or alternatively, the machine learning programs described herein may be trained by inputting sample data sets or certain data into the programs, such as images, statistics and information, and / or specific parameters, characteristics, properties, and the like. The machine learning programs may utilize deep learning algorithms that may be primarily focused on pattern recognition and may be trained after processing multiple examples. The machine learning programs may include Bayesian Program Learning (BPL), voice recognition and synthesis, image or object recognition, signal processing, optical character recognition, and / or natural language processing - either individually or in combination. The machine learning programs may also include natural language processing, semantic analysis, automatic reasoning, and / or machine learning.

[0260] Supervised and unsupervised machine learning techniques may be used. In supervised machine learning, a processing element may be provided with example inputs and their associated outputs and may seek to discover a general rule that maps inputs to outputs, so that when subsequent novel inputs are provided the processing element may, based upon the discovered rule, accurately predict the correct output. In unsupervised machine learning, the processing element may be required to find its own structure in unlabeled example inputs. In one embodiment, machine learning techniques may be used to determine types of particles and / or materials, based, for example, on PA properties determined by the AI / ML models described herein.CT

[0261] Based upon training / analyses, the processing elements of the model(s) may learn how to identify characteristics and / or patterns that may then be applied to analyzing data such as image data, model data, and / or other data such as PA data of unidentified particles, for identifying the unidentified particles. For example, the processing elements of the models may learn to identify particles and / or materials based on PA properties and / or other related properties such as those determined by the sensors described herein and use the learned aspects to execute on new datasets. The processing elements may also learn how to identify trends / patterns that may not be readily apparent based upon collected traffic data.

[0262] The embodiments were chosen and described in order to best explain the principles of the disclosure and its practical application to thereby enable others skilled in the art to best utilize the disclosure in various embodiments and with various modifications as are suited to the particular use contemplated. Aspects of the disclosed embodiments may be mixed to arrive at further embodiments within the scope of the invention.

[0263] As various modifications could be made in the constructions and methods herein described and illustrated without departing from the scope of the disclosure, it is intended that all matter contained in the foregoing description or shown in the accompanying drawings shall be interpreted as illustrative rather than limiting. Thus, the breadth and scope of the present disclosure should not be limited by any of the above-described example embodiments but should be defined only in accordance with the following claims appended hereto and their equivalents.

[0264] Various aspects of the embodiments described above may be used alone, in combination, or in a variety of arrangements not specifically discussed in the embodiments described in the foregoing and is therefore not limited in its application to the details and arrangement of components set forth in the foregoing description or illustrated in the drawings. For example, aspects described in one embodiment may be combined in any manner with aspects describe in other embodiments.CT

[0265] Having thus described several aspects of at least one embodiment, it is to be appreciated that various alternations, modifications, and improvements will readily occur to those skilled in the art. Such alterations, modifications, and improvements are intended to be part of this disclosure and are intended to be within the spirit and scope of the principles described herein. Accordingly, the foregoing description and drawings are by way of example only.

Claims

CTWHAT IS CLAIMED IS:

1. A sample medium testing system, comprising: a sensor, the sensor comprising: a body including a microfluidic channel therein; a microresonator; and a sample medium contained within the microfluidic channel, the sample medium including at least one target of interest provided therein; a modulated light source, the modulated light source configured to emit modulated light to provide optical stimulation to at least one target of interest present within the sample medium; a detector configured to detect transmitted light signals from the microresonator that carry information associated with properties of acoustic waves associated with the at least one target of interest; and a control device including one or more processors programmed to control operation of the modulated light source and to acquire data and analyze results from testing of the sample medium, wherein the control device is configured to control the modulated light source to optically stimulate the at least one target of interest via the modulated light, the sensor is configured to allow the sample medium to flow inside the microfluidic channel while being optically stimulated by the modulated light source from outside the microfluidic channel to generate acoustic waves through photoacoustic effects, the photoacoustic effects are based at least in part on material properties of the at least one target of interest, and the control device classifies the at least one target of interest based on analysis of detections of the acoustic waves obtained from the detector.

2. The sample medium testing system of claim 1, wherein the microresonator is a whispering gallery mode (WGM) microresonator.

3. The sample medium testing system of claim 2, wherein the WGM microresonator is at least one of: (i) a bubble-shaped WGM microresonator, and a core of the bubble-shapedCTWGM microresonator is a thick-wall hollow core directly connected to the microfluidic channel; (ii) a microbottle WGM microresonator; (iii) a microsphere WGM microresonator; (iv) a microtoroid WGM microresonator; (v) a microring WGM microresonator; and (vi) a Fabry -Perot resonator.

4. The sample medium testing system of claim 3, further comprising a probe laser configured to emit a laser beam having a wavelength of at least 780 nm and excite whispering gallery modes in the bubble-shaped WGM microresonator.

5. The sample medium testing system of claim 1, wherein the at least one target of interest includes one or more particles, and testing of the sample medium enables real-time, label-free detection and interrogation of the particles in a native solution environment of the sample medium.

6. The sample medium testing system of claim 5, wherein the particles are nanoparticles.

7. The sample medium testing system of claim 6, wherein the nanoparticles are gold nanoparticles.

8. The sample medium testing system of claim 5, wherein the sample medium is blood.

9. The sample medium testing system of claim 8, wherein the blood is whole blood and the particles are red blood cells within the whole blood.

10. The sample medium testing system of claim 1, further comprising an excitation / pump laser, the excitation / pump laser configured to emit a laser beam having a wavelength based on an absorption characteristic of the at least one target of interest to generate acoustic signals.CT11. The sample medium testing system of claim 1, wherein the modulated light has a wavelength of 532 nm.

12. The sample medium testing system of claim 1, wherein the control device is configured to analyze results from testing of the sample medium via at least one of: (i) a machine learning model associated with the control device; and (ii) signal processing algorithms.

13. The sample medium testing system of claim 12, wherein the machine learning model is trained at least on prior-obtained acoustic waves data.

14. The sample medium testing system of claim 1, wherein the control device is configured to analyze results from testing of the sample medium via a machine learning model associated with the control device.

15. The sample medium testing system of claim 14, wherein the machine learning model is configured to be executed on photoacoustic signal data associated with the photoacoustic effects and output a classification of the at least one target of interest based on the photoacoustic signal data.

16. The sample medium testing system of claim 1, wherein the sample medium includes liquid.

17. The sample medium testing system of claim 1, wherein the sample medium includes gas.CT18. The sample medium testing system of claim 1, wherein the modulated light source is configured to emit light that matches an absorption property of the at least one target of interest.

19. The sample medium testing system of claim 18, wherein the modulated light source is a light emitting diode.

20. The sample medium testing system of claim 18, wherein the modulated light source is a flash lamp.

21. The sample medium testing system of claim 20, wherein the flash lamp is configured for use in gas sensing via the sensor.

22. A non-transitory computer-readable recording medium having computer executable instructions stored thereon, which when executed by a processor of a sample solution testing system, cause the processor to: control operation of a modulated light source associated with the sample solution testing system; acquire data associated with results from a test of a sample solution using the modulated light source, the modulated light source being used in association with a sensor of the sample solution testing system for the test of the sample solution, wherein: the sensor comprises: a body including a microfluidic channel, the sample solution being contained within the microfluidic channel during the test and including at least one target of interest; and a whispering gallery mode (WGM) microresonator; and the modulated light source is configured to emit modulated light to provide optical stimulation to the at least one target of interest;CT control the modulated light source to optically stimulate the at least one target of interest via the modulated light from outside the microfluidic channel to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one target of interest; and classify the at least one target of interest based on analysis of the acoustic waves.

23. The non-transitory computer-readable recording medium of claim 22, wherein the processor is further caused to analyze results from the testing of the sample solution via a machine learning model.

24. The non-transitory computer-readable recording medium of claim 22, wherein the processor is further caused to train a machine learning model associated with the sample solution testing system using at least one of photoacoustic data, extracted features, and one or more prototype embeddings.

25. A computer-implemented method of testing a sample via a sensor, the sensor comprising (i) a body including a microfluidic channel therein; (ii) a whispering gallery mode (WGM) microresonator; and (iii) the sample contained within the microfluidic channel, the sample including at least one target of interest provided therein, the computer- implemented method comprising: providing the sensor; providing a modulated light source, the modulated light source configured to emit modulated light toward the body of the sensor; testing the sample, including optically stimulating, via the modulated light, the at least one target of interest to generate acoustic waves through photoacoustic effects, the photoacoustic effects being based at least in part on material properties of the at least one target of interest; and classifying the at least one target of interest based on analysis of at least one of the photoacoustic effects and the acoustic waves.CT26. The computer-implemented method of claim 25, wherein the providing of the sensor includes constructing the sensor.

27. The computer-implemented method of claim 25, wherein the WGM microresonator is at least one of: (i) a bubble-shaped WGM microresonator including a core, and the core of the bubble-shaped WGM microresonator is a thick-wall hollow core directly connected to the microfluidic channel; (ii) a microbottle WGM microresonator; (iii) a microsphere WGM microresonator; (iv) a microtoroid WGM microresonator; (v) a microring WGM microresonator; and (vi) a Fabry-Perot resonator.

28. The computer-implemented method of claim 25, further comprising classifying the at least one target of interest based on analysis of at least one of the photoacoustic effects and the acoustic waves based on an output from a machine learning model.

Images