Quantum dot fluorescent encoded multi-target detection chip system and method

Abstract

The application relates to the field of food safety detection, and discloses a quantum dot fluorescence coded multi-target detection chip system and method, the system comprising: a microfluidic detection chip, the chip being provided with an array of reaction detection zones, and quantum dot fluorescence probes being fixed in each micro-reaction unit; a portable signal acquisition module, which is used for coupling an intelligent terminal and an adapter of the microfluidic detection chip; an intelligent terminal, which realizes light path alignment with the microfluidic detection chip through the adapter; and application software, which analyzes the fluorescence image to obtain a detection result. Through cooperation between the microfluidic detection chip and the adapter of the portable signal acquisition module, detection can be carried out without relying on expensive equipment or professional detection instruments, the hardware cost of the system is reduced, the entire detection system is portable, and the system is independent of professional laboratories and can be applied in ordinary laboratories, field environments and even home scenarios.

Description

Technical Field

[0001] This invention relates to the field of food safety testing, specifically to a quantum dot fluorescently encoded multi-target detection chip system and method. Background Technology

[0002] In vitro diagnostic technologies, especially fluorescence-based molecular detection methods, play a crucial role in fields such as healthcare, food safety, and environmental monitoring. Currently, technologies capable of achieving high sensitivity and specificity typically rely on large, sophisticated benchtop instruments, such as ELISA readers, flow cytometers, fluorescence microscopes, or real-time quantitative polymerase chain reaction (qPCR) systems. While these instruments offer superior performance, they generally suffer from high cost, large size, complex operating procedures, and the need for specific laboratory environments.

[0003] This strong reliance on centralized laboratories and specialized operators constitutes a fundamental limitation of traditional testing technologies. It requires samples to undergo a series of processes, including collection, preservation, and transportation, before being sent to an institution with testing capabilities for analysis. This entire process is time-consuming and cannot meet the growing demand for point-of-care testing (POCT) that demands immediacy and portability. Therefore, simplifying and miniaturizing complex testing processes to free them from the constraints of large instruments and laboratory environments is a key direction for development in this field.

[0004] However, despite miniaturization of the reaction components, signal reading and analysis remain a bottleneck for achieving true portability and low cost. Most existing high-sensitivity microfluidic detection systems, after completing the biochemical reaction on the chip, still require imaging and analysis on an external desktop fluorescence microscope or a dedicated, non-portable signal reader. This means that the overall portability of the system is limited by the bulkiest and most expensive reading device, and it has not fundamentally eliminated its dependence on specialized detection instruments. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this invention provides a quantum dot fluorescence-encoded multi-target detection chip system and method, which solves the technical problem that existing fluorescence detection systems rely on large and expensive professional instruments for signal reading, resulting in high costs, poor portability, and inability to meet the needs of on-site real-time detection.

[0006] To achieve the above objectives, the present invention provides the following technical solution: a quantum dot fluorescence-encoded multi-target detection chip system, comprising: Microfluidic detection chips, portable signal acquisition modules, smart terminals, and application software installed in the smart terminals.

[0007] The microfluidic detection chip has at least one reaction detection area. In one embodiment, the substrate of the chip is made of a biocompatible polymer material, which includes a three-dimensional computer-aided design model of a pre-defined microstructure, and is integrally manufactured using an additive manufacturing process. The additive manufacturing process can be digital light processing 3D printing or stereolithography 3D printing. This process can directly convert the digital model into a physical object without relying on traditional photolithography processes, significantly shortening the research and development cycle and reducing manufacturing costs.

[0008] At least one quantum dot fluorescent probe is immobilized within the reaction detection region. To achieve multi-target detection, various different quantum dot fluorescent probes can be immobilized. To ensure effective differentiation of multiple signals, the central emission wavelengths of any two quantum dots used for encoding must meet the following condition: ; in: and The first species and first The central emission wavelength of a quantum dot; and These are the full width at half maximum (FWHM) of the emission spectrum of the corresponding quantum dot; This is a preset dimensionless separation coefficient used to ensure spectral resolution, and its value is usually greater than 0.5.

[0009] The quantum dot fluorescent probe can be immobilized on the surface of the reaction detection region via chemical covalent coupling. This method involves modifying the surface of the microfluidic detection chip with a first active group and introducing a second active group that can react with it onto the quantum dot fluorescent probe molecule, allowing the two to react and form a stable covalent bond, thereby achieving firm immobilization of the probe.

[0010] The portable signal acquisition module is an adapter for coupling a smart terminal and a microfluidic detection chip. This adapter is made of an opaque polymer material to create a dark chamber that isolates ambient light. The adapter includes a chip positioning mechanism, such as a groove or slot matching the chip's shape, to physically position the microfluidic detection chip and ensure precise alignment of its detection area. To optimize the optical path, optical filtering elements can also be integrated within the adapter. Specifically, an excitation filter is placed between the LED flash (which serves as the excitation source) and the microfluidic detection chip to purify the excitation light; and an emission filter is placed between the microfluidic detection chip and the rear camera lens of the smart terminal to filter out excitation light interference, allowing only emitted fluorescence to pass through.

[0011] The smart terminal is equipped with a rear camera and an LED flash, and achieves optical path alignment with the microfluidic detection chip through the adapter.

[0012] The application software, installed within the smart terminal, performs the following functions: controlling the LED flash to excite the reaction detection area; controlling the rear camera to acquire a fluorescence image containing the reaction detection area, and locking key parameters of the rear camera, including sensitivity, exposure time, focus mode, and white balance, before acquisition to ensure consistent acquisition conditions. The acquired fluorescence image is saved in the original sensor data format, thereby bypassing the uncontrollable algorithmic modifications of the image signal processor inside the smart terminal and obtaining linear light intensity information closest to the sensor output.

[0013] Therefore, the core of this invention lies in the fact that the application software incorporates an artificial intelligence decoding model for analyzing the acquired fluorescence images. The function of this decoding model is to convert the raw image data acquired by the rear camera, which contains signal distortion specific to the device, into a standardized, corrected signal strength that is device-independent.

[0014] Specifically, the data collection process of different smart terminals can be abstracted as follows: ; ; In the formula, and These represent the physical response functions of the CMOS sensors and their optical components in mobile phones A and B, respectively. and These represent a complex set of non-linear image processing functions executed by the image signal processors of mobile phone A and mobile phone B, respectively. and These represent the random noise introduced by each of the two systems; This is a real signal; The artificial intelligence decoding model is a convolutional neural network. The construction of this network relies on training on a specific training dataset that includes device noise. This training dataset is constructed as follows: First, a series of standard fluorescent chips of known concentrations are imaged using a calibrated research-grade fluorescence imaging device to obtain baseline true signal intensity. Then, using a large number of consumer-grade smart terminals covering different brands, models, and systems on the market, raw image data of the same batch of standard fluorescent chips were collected under the same conditions. Therefore, a large number of [structures] were constructed. The data pairs that make up the data.

[0015] The training process of this convolutional neural network is to solve the following optimization problem to minimize the mean square error between the network's predicted output and the baseline true value: ; In the formula, For parameters The convolutional neural network model for input images The given predicted signal strength; This is the device-independent reference true signal strength corresponding to the image; These are the optimized network parameters obtained after training. This optimization process typically employs gradient descent-based algorithms, such as the Adam optimizer, to progressively adjust the network parameters through multiple iterations on the dataset. This continues until the loss function converges to a minimum value; This represents the entire dataset used to train the model; This represents the square of the L2 norm of the difference.

[0016] To achieve final quantification, the application software also includes a built-in standard curve that describes the target concentration value. Corrected signal strength compared to the output of the decoding model The functional relationship between them. In one specific implementation, the standard curve is fitted using a four-parameter logistic regression model, the mathematical expression of which is: The application software stores and calls the inverse function of this function to convert the output of the AI ​​model. Substituting the values ​​into the calculation, the final concentration of the target sample is obtained: ;

[0017] The user interface of this application software is responsible for presenting the calculated concentration values ​​to the user in a clear and intuitive way, such as directly displaying the numerical value and its unit (e.g., ng / mL, g / kg, or ppm). In some application scenarios, the software can also perform qualitative judgments on the results based on built-in threshold standards, such as displaying information like qualified, exceeded, negative, or positive.

[0018] This invention also provides a quantum dot fluorescence-encoded multi-target detection method, comprising the following steps: Step a, System Assembly and Alignment: A microfluidic detection chip is placed in the positioning mechanism of a portable signal acquisition module. The chip has a reaction detection area array composed of multiple spatially isolated micro-reaction units, and each micro-reaction unit has a specific quantum dot fluorescent probe fixed inside. Then, the module is coupled to a smart terminal so that the LED flash of the smart terminal is optically aligned with the reaction detection area of ​​the chip through an excitation filter, and the rear camera is optically aligned with the reaction detection area of ​​the chip through an emission filter. Step b, Excitation and Image Acquisition: Using the application software on the smart terminal, control the LED flash to excite the reaction detection area; at the same time, after locking the sensitivity, exposure time, focus mode and white balance parameters of the rear camera, control it to acquire the raw (RAW) format fluorescence image of the reaction detection area. Step c, Signal Decoding and Standardization: The pre-built artificial intelligence decoding model in the application software is invoked to analyze the original format fluorescence image, thereby converting the image data containing inherent distortions of the device into a standardized fluorescence signal intensity that is independent of the device hardware and built-in image processing algorithms. Step d, Quantitative calculation and result output: Substitute the standardized fluorescence signal intensity into the inverse function of the standard curve pre-stored in the application software to calculate the concentration value of the target to be tested, and display the concentration value on the smart terminal interface.

[0019] This invention provides a quantum dot fluorescence-encoded multi-target detection chip system and method. It has the following beneficial effects: 1. This invention enables detection without relying on expensive equipment or professional testing instruments by using a microfluidic detection chip and an adapter for a portable signal acquisition module. This replaces the expensive dedicated fluorescence reading equipment required for traditional detection with a low-cost hardware combination, which not only reduces the hardware cost of the system, but also makes the entire detection system portable, free from dependence on professional laboratories, and can be applied in ordinary laboratories, field environments and even home scenarios.

[0020] 2. This invention can process raw data through an artificial intelligence decoding model, reversely correct signal distortion caused by the inherent differences in image sensors and signal processors of different models of smart terminals, and output standardized signal strength independent of the device, fundamentally solving the problem of incomparable detection results due to device differences in the existing technology.

[0021] 3. This invention integrates a functionalized microfluidic detection chip, a proprietary optical adapter, a consumer-grade smart terminal, and its built-in AI analysis software to construct a complete, self-contained detection system. This system achieves, for the first time, accurate and quantitative on-site detection of multiple targets based on quantum dot fluorescence encoding without requiring any specialized external equipment, fundamentally resolving the contradiction between equipment dependence and result accuracy in portable detection. Attached Figure Description

[0022] Figure 1 This is a schematic diagram of the system architecture of the present invention; Figure 2 This is a schematic diagram of the microfluidic detection chip structure of the present invention; Figure 3 This is a schematic diagram of the portable signal acquisition module of the present invention; Figure 4 This is a schematic diagram of the optical path of the present invention; Figure 5 This is a schematic diagram illustrating the spectral principle of quantum dot fluorescence encoding in this invention; Figure 6 This is a schematic diagram of the base structure of the present invention; Figure 7 This is a schematic diagram of the adapter portion of the present invention; Figure 8 This is a schematic diagram of the internal structure of the adapter of the present invention. Detailed Implementation

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] To better understand the present invention, the above content will be described in detail below with reference to specific embodiments.

[0025] Please see the appendix Figure 1 -Appendix Figure 8 This invention provides a quantum dot fluorescence-encoded multi-target detection chip system and method. The system includes a microfluidic detection chip, a portable signal acquisition module, and a smart terminal. The smart terminal includes, but is not limited to, a smartphone. The smart terminal also has application software installed to analyze fluorescence images to obtain detection results.

[0026] In one embodiment: The core biochemical reaction and signal generation processes of the detection system provided by this invention are completed on a disposable microfluidic detection chip. Specifically, its main body is composed of biocompatible polymer materials, such as polymethyl methacrylate (PMMA), polycarbonate (PC), or specific resin materials formed by photopolymerization 3D printing process. Furthermore, the overall structure of the chip can be a multi-layer composite structure, with its core functional layer being a substrate containing a preset microstructure, which is sealed by a transparent cover plate, thereby forming a closed fluid pathway system.

[0027] Specifically, the chip has at least one sample inlet. This sample inlet is a microstructure for receiving the liquid sample to be tested, and its specific implementation can be a micropore or a micro reservoir located on the chip surface. In a preferred embodiment, the sample inlet can also be designed in a funnel shape or with a certain taper, which facilitates sample addition using a pipetting device and effectively reduces the possibility of air bubbles being introduced during sample addition due to liquid surface tension, ensuring that the liquid can smoothly enter the subsequent flow channels.

[0028] Connected to the sample inlet is a microfluidic network inside the chip. This network consists of micrometer-scale channels used to guide and transport liquid samples along predetermined paths within the chip. The cross-sectional shape of the microchannels can be rectangular, trapezoidal, or semi-circular, with width and depth dimensions ranging from 10 micrometers to 500 micrometers. The layout of the microfluidic network must ensure that the liquid is delivered from the sample inlet to the downstream functional areas of the chip at a stable and controllable flow rate.

[0029] In some embodiments requiring sample mixing or dilution, the microfluidic network also integrates mixing chambers, microstructures designed to facilitate rapid and uniform mixing of two or more liquids. Specifically, this can be achieved through an elongated, serpentine channel—a serpentine mixer—that increases fluid contact area and time to promote diffusion mixing; or through an array of microstructures with specific geometries, such as herringbone grooves, to induce chaotic convection in laminar flow, thereby achieving efficient passive mixing.

[0030] Simultaneously, the terminals of the microfluidic network are connected to one or more reaction detection regions, which are the core functional areas of the chip, where specific biomolecule recognition reactions occur and subsequent fluorescence signals are generated. In a preferred embodiment, the reaction detection region can be an array, which can consist of multiple spatially isolated, independent microreaction units. These microreaction units can take the form of micropores, micropillar array surfaces, or specific regions defined on a flat substrate, and each unit is used to immobilize a specific quantum dot fluorescently encoded probe.

[0031] Specifically, the design of the reaction detection array needs to consider avoiding optical signal crosstalk. The center-to-center distance between two adjacent microreaction units must meet specific conditions to ensure that the fluorescence signal generated in one unit does not interfere with the accurate interpretation of the signal from the adjacent unit. This condition can be expressed as: ; in: The center-to-center distance between two adjacent microreaction units; The characteristic dimensions of a single microreaction unit, such as diameter or side length; The purpose of this safety isolation distance, set according to the optical system, is to prevent signal overlap due to light scattering or diffraction effects. In one specific implementation, The value of is related to the magnification, pixel size, and optical crosstalk tolerance of the imaging system. Therefore, this formula can ensure that when using imaging devices such as smartphone cameras to acquire images, the signal of each micro-reaction unit can be clearly and independently distinguished, which is the structural basis for achieving simultaneous and accurate quantification of multiple targets.

[0032] In one embodiment, the aforementioned microstructure design can be implemented using additive manufacturing processes, specifically, for example, 3D printing technology. Specifically, the manufacturing process is as follows: First, the completed 3D computer-aided design (CAD) model of the chip, containing all microstructure details, is converted into a standard data format recognizable by additive manufacturing equipment, such as an STL file or a 3MF file.

[0033] Next, a high-precision additive manufacturing device based on the photopolymerization principle is used to perform the printing task. In a preferred embodiment, the device is a digital light processing (DLP) 3D printer or a stereolithography (SLA) 3D printer. Taking DLP technology as an example, its working principle is to use a high-resolution digital projector to precisely project a two-dimensional cross-sectional image of each layer of the chip as a mask onto the surface of a liquid photosensitive resin material. Then, the area irradiated by light of a specific wavelength (usually ultraviolet light) undergoes a photopolymerization reaction and solidifies into shape. Then, the printing platform is vertically displaced according to a preset layer thickness (e.g., 10 micrometers to 50 micrometers). Through layer-by-layer curing and stacking, a complete chip substrate with a complex three-dimensional structure including a sample inlet, microchannels, mixing chambers, and reaction detection areas is finally manufactured in an integrated manner.

[0034] To ensure the effectiveness and reliability of the chip in biochemical detection applications, the selected photosensitive resin material must meet stringent requirements. Specifically, these may include: First, the material must have good biocompatibility, such as conforming to the ISO-10993 standard, to ensure that it will not undergo non-specific adsorption or reaction with biomolecules (such as proteins, nucleic acids, etc.) in the sample, thereby interfering with the detection results; Secondly, the material needs to exhibit excellent optical properties after curing, especially high transmittance in the visible spectrum of quantum dot fluorescence emission (e.g., 400nm to 700nm), and its own spontaneous fluorescence intensity under excitation light irradiation is extremely low. This is a prerequisite for ensuring high signal-to-noise ratio fluorescence signal acquisition.

[0035] Third, the material must be chemically inert and able to withstand the buffer solutions and other reagents involved in the detection process.

[0036] After printing, the chip requires post-processing, which may include rinsing and ultrasonic cleaning with a specific solvent, such as isopropanol (IPA), to thoroughly remove any uncured resin remaining inside and on the surface of the microchannels. After cleaning, the chip is placed in a UV curing chamber for secondary deep curing to ensure complete polymerization of the material, resulting in stable and uniform chemical properties and mechanical strength, guaranteeing consistent chip performance during subsequent use and storage.

[0037] Finally, by employing the aforementioned 3D printing manufacturing process, digital models can be directly converted into physical entities without relying on expensive lithography machines, costly custom photomasks, or stringent environments such as cleanrooms. This shortens the development cycle from structural design to prototype verification, thereby reducing the manufacturing cost per chip.

[0038] After the physical structure of the chip is manufactured, in order to enable it to recognize and capture specific target molecules, the reaction detection region of the chip must be functionalized. This means fixing a variety of pre-prepared quantum dot fluorescent probes at specific positions in the region in a spatially resolved manner.

[0039] Specifically, by independently placing each probe solution containing a different fluorescent code into a corresponding microreaction unit within a reaction detection region array, the implementation may include: First, microspotting technology uses a computer-controlled automated device equipped with an array of micron-tipped spotting needles. During operation, the array of spotting needles is immersed in a microplate containing different quantum dot fluorescent probes to pick up the liquid. Then, it moves above the chip and spots nanoliter or picoliter-sized probe droplets onto preset cell positions in the reaction detection area using contact or non-contact methods. By changing the probe solution and repeating this process, a two-dimensional array containing multiple probes can be constructed.

[0040] Second, piezoelectric inkjet printing technology uses different quantum dot fluorescent probe solutions as bio-inks, which are loaded into multiple independent ink cartridges in the printing device. Under program control, the print head deforms the piezoelectric element by applying voltage, thereby ejecting probe droplets from the nozzle and depositing them at designated locations on the chip substrate. This method is a non-contact operation with high throughput and high precision, and is especially suitable for large-scale, automated chip mass production.

[0041] Regardless of the allocation method used, the stable immobilization of probe molecules on the chip surface can be achieved through one or a combination of the following two approaches.

[0042] The first approach is physical adsorption, which involves hydrophobic interactions, van der Waals forces, or electrostatic attraction between probe molecules and the surface of the chip substrate material. In some embodiments, the chip surface can be modified by means of plasma treatment to adjust its hydrophobicity or surface charge, thereby enhancing the strength and stability of physical adsorption.

[0043] The second approach is chemical covalent coupling. This requires introducing reactive chemical functional groups onto both the chip substrate surface and the quantum dot fluorescent probe molecules. The specific steps are as follows: First, by chemically treating the chip or pre-doping specific monomers into the 3D printing resin material, a first active group, such as an amino group (-NH2) or a carboxyl group (-COOH), is modified onto the surface of the chip's reaction detection area. Second, it is ensured that the quantum dot fluorescent probe (whose main body is usually an antibody or other protein) carries a second active group that can react with it. For example, using the classic carbodiimide (EDC) and N-hydroxysuccinimide (NHS) chemistry, the carboxyl group can be activated, allowing it to react efficiently with the amino group on the protein to form a stable amide bond, thereby firmly covalently linking the probe molecule to the chip surface.

[0044] After the probe immobilization step is completed, sealing is required. Inert proteins, such as bovine serum albumin (BSA), can be used to cover and seal all exposed surfaces in the chip's reaction detection area that have not yet been occupied by probe molecules. This eliminates the adsorption of non-specific components in the sample to be tested onto the chip surface, thereby significantly reducing the background signal and improving the specificity and sensitivity of the detection. After sealing, cleaning, and drying, the chip has completed all functionalization and is finally a finished product that can be used for detection.

[0045] When a chip with its functional settings is used for actual testing, a device that can stably acquire its fluorescence signal is first needed. Therefore, this invention provides a portable signal acquisition module. Specifically, its core component can be an adapter for coupling a smartphone and the chip, and this adapter can also be manufactured quickly and at low cost using 3D printing technology.

[0046] Specifically, this adapter can be a shell structure, similar to a mobile phone case, used to snap onto the mobile phone and keep it coaxial with the camera. The adapter needs to be made of opaque polymer material, such as black acrylonitrile-butadiene-styrene copolymer (ABS) or black photosensitive resin dyed with dye, so as to ensure that the adapter can form a dark room environment inside to isolate ambient light interference when in use.

[0047] To achieve precise optical path alignment, a chip positioning mechanism is required at the bottom of the adapter. In one specific embodiment, this mechanism can be a groove or limiting slot that matches the chip's dimensions. When the microfluidic detection chip is inserted, the sidewall of the groove or slot will physically limit the chip's edge, ensuring that the chip's detection area is always positioned on a fixed, known two-dimensional coordinate within the adapter, thereby achieving positional correction between the phone, adapter, and chip.

[0048] Meanwhile, the main body of the adapter can also be provided with a mobile phone mounting area for accommodating and fixing the smartphone. At a certain position in this area, that is, directly above the corresponding chip reaction detection area, an imaging through hole is required to ensure that the rear camera lens of the mobile phone can be aligned with the detection area of ​​the chip through the through hole. Finally, in this way, the optical axes of the chip, the adapter and the smartphone camera can be calibrated and aligned.

[0049] After the smartphone completes its positioning with the chip via the adapter, it can then excite a light source, usually an LED flash used for supplemental lighting in photography. The on / off state and duration of the LED flash are controlled by the application software running on the smartphone through the corresponding interface provided by the operating system, thereby providing excitation energy to the quantum dots in the chip's detection area when needed, i.e., providing illumination.

[0050] In a preferred embodiment, to improve excitation efficiency and reduce background noise caused by broadband light sources, an excitation filter can be integrated inside the adapter. This excitation filter can be fixedly mounted between the LED flash and the microfluidic detection chip to selectively transmit light within a specific wavelength range that matches the excitation spectrum of the quantum dots used, while blocking stray light of other ineffective wavelengths, thereby obtaining purer excitation light.

[0051] The imaging unit used to collect the fluorescence signals generated on the chip utilizes the rear camera of a smartphone (see attached diagram for specific layout). Figure 6 -Appendix Figure 8 As shown in the figure, by converting the photon signals received by the CMOS image sensor into digital electrical signals, a two-dimensional digital image containing spatial and intensity information is formed.

[0052] To ensure the stability and repeatability of image acquisition and thus achieve accurate quantitative analysis, the application software must lock several key camera parameters before image acquisition to disable its automatic adjustment function. These parameters include, but are not limited to: ISO sensitivity (set to a fixed value); exposure time (shutter opening duration), also locked to a preset value; focus mode, set to manual focus and locked at the object distance corresponding to the chip's detection area surface; and white balance, set to a fixed color temperature value instead of automatic white balance. By locking these parameters, it is ensured that for different batches of tests, as long as the target concentration in the sample is the same, the acquired image signal intensity will be consistent, eliminating variables introduced by the phone's automatic algorithms.

[0053] In another preferred embodiment, to further improve the signal-to-noise ratio, an emission filter can be placed inside the adapter between the microfluidic detection chip and the mobile phone camera lens. The function of this emission filter is to filter out the excitation light reflected or scattered by the chip surface, and only allow the longer wavelength fluorescence signal emitted by the quantum dot to pass through and enter the camera sensor. By matching the spectral passband of the emission filter with the emission spectral range of the quantum dot fluorescent probe used, the interference of excitation light on imaging is reduced.

[0054] Furthermore, during the image data acquisition stage, the application software can also save images in the format of raw sensor data (RAW format, such as DNG files). This allows it to bypass the uncontrollable post-processing of images by the image signal processor (ISP) inside the smartphone, such as automatic noise reduction, sharpening, and color enhancement, thereby obtaining linear light intensity information that is closest to the direct output of the CMOS sensor and is uncompressed and algorithmically modified.

[0055] In one embodiment, the principle of quantum dot fluorescence encoding used in this invention lies in quantum dot semiconductor nanocrystals, which are typically composed of a core of group II-VI elements (such as CdSe, CdTe) or group III-V elements (such as InP), and are often coated with another wide-bandgap semiconductor material (such as ZnS) as a shell to improve their luminescence efficiency and chemical stability. Its core feature is that when the crystal size is reduced to the nanometer level (typically 2-10 nanometers), its scale is comparable to the exciton Bohr radius of the material itself, a significant quantum confinement effect will be generated.

[0056] Due to the quantum confinement effect, the electronic energy levels of quantum dots transform from a continuous energy band in bulk materials to a discrete, atom-like energy level structure. This physical effect leads to unique optical properties: the emission wavelength of quantum dots (… ) and its physical size ( There are dependencies between them, specifically: Smaller quantum dots have larger energy level gaps and emit shorter wavelength light (e.g., blue or green light) when excited; while larger quantum dots have smaller energy level gaps and emit longer wavelength light (e.g., orange or red light). Therefore, by controlling the size of the nanocrystals during the synthesis process, a series of quantum dot materials with continuously tunable emission wavelengths can be obtained.

[0057] The quantum dots used in this invention for encoding are mainly based on the following two key spectral characteristics.

[0058] First, its broad and continuous absorption spectrum means that quantum dots with different emission wavelengths (i.e. different colors) can be effectively excited by a single light source with a shorter wavelength. This makes it possible to use a single white or blue LED with a wide spectral range that comes with a smartphone as the sole excitation source for the system, which greatly simplifies the optical structure of the system. Second, in contrast to broad-spectrum absorption, the emission spectrum of quantum dots exhibits a very narrow band and a symmetrical peak shape with a near-Gaussian distribution. The full width at half maximum (FWHM) of their emission spectrum is typically only 20-40 nanometers, thus enabling extremely high spectral resolution between the emission peaks of different colored quantum dots, with minimal signal crosstalk or overlap. To ensure effective differentiation of multiple signals, the central emission wavelengths of the two quantum dots used for encoding must meet the following conditions: ; in: and The first species and first The central emission wavelength of a quantum dot; and These are the full width at half maximum (FWHM) of the emission spectrum of the corresponding quantum dot; This is a preset dimensionless separation coefficient used to ensure spectral resolution, and its value is usually greater than 0.5.

[0059] Based on the aforementioned spectral characteristics, this invention constructs N unique identifiers, distinguishable by an optical system, for N different targets using wavelength encoding. Specifically, for the first... Type of target to be tested ( ), select an emission wavelength of quantum dots ( ), and the quantum dot is specifically recognized by energy. The probe molecules are coupled together. In this way, the detection of N targets is transformed into the detection of N quantum dot signals with unique emission wavelengths. In the final image analysis, the wavelength is... The identification of fluorescence signals represents the presence of [something] in the sample being tested. The existence of; and the strength of the signal, which is related to The concentration is directly related to the target, thus enabling simultaneous qualitative and quantitative analysis of multiple targets.

[0060] In order to practically apply the above-mentioned wavelength encoding principle, the quantum dots, which serve as signal reporting units, must be stably coupled with molecules that can specifically recognize the target to be tested, thereby constructing a functionalized quantum dot fluorescent probe. Therefore, the preparation process must be a key step in endowing the detection system with specificity.

[0061] Specifically, this step first requires surface modification of the quantum dots themselves. Since the core-shell structure of quantum dots is typically synthesized in organic phases, their surface ligands are hydrophobic. Therefore, it is necessary to transform them into water-soluble ligands through methods such as ligand exchange or surface coating, while simultaneously introducing active chemical functional groups onto their surface for subsequent reactions. In a preferred embodiment, the surface of the quantum dots is modified with a large number of hydrophilic ligands with end groups of carboxyl (-COOH) or amino (-NH2).

[0062] Target-specific recognition molecules are the foundation for selective detection. Their specific types depend on the nature of the target to be tested, including but not limited to: monoclonal antibodies, polyclonal antibodies or their antibody fragments (such as Fab, scFv) that can bind to antigen targets; nucleic acid aptamers that can form specific three-dimensional structures and bind to small molecules, ions or protein targets; or oligonucleotide probes that can hybridize with complementary nucleic acid sequences.

[0063] The surface-modified quantum dots are then coupled to the target-specific recognition molecule, typically using covalent bonding to ensure the highest probe stability. This invention provides a preferred coupling method utilizing carbodiimide chemistry. The specific steps of this method are as follows: Quantum dots (QD-COOH) with carboxyl groups on their surface are dispersed in a suitable buffer, such as MES buffer.

[0064] 1-Ethyl-3-(3-dimethylaminopropyl)carbodiimide (EDC) and N-hydroxysuccinimide (NHS) are added to the dispersion. In this step, EDC first reacts with the carboxyl groups on the quantum dot surface to form an unstable O-acylisourea intermediate.

[0065] The intermediate immediately reacts with NHS present in the system to generate a more stable NHS-ester active intermediate (QD-NHS) that is effectively resistant to hydrolysis.

[0066] A solution containing a target-specific recognition molecule (e.g., an antibody) is added to the activated quantum dot solution. The primary amino group (-NH2) of the lysine residues on the surface of the antibody protein molecule acts as a nucleophile, attacking the carbonyl carbon of the NHS- ester, ultimately forming a stable amide bond (-CO-NH-), thereby covalently linking the antibody molecule to the quantum dot surface.

[0067] As another preferred implementation, the coupling process can also be achieved through maleimide chemistry. This method requires that the quantum dot surface be modified with maleimide groups, and that the target-specific recognition molecule (e.g., a modified antibody fragment or a synthetic nucleic acid aptamer) possess a free thiol group (-SH). In a weakly alkaline buffer environment, the double bond of the maleimide group can undergo an efficient and specific Michael addition reaction with the thiol group to form a stable thioether bond, thus achieving covalent coupling between the two.

[0068] After the coupling reaction is complete, the reaction product needs to be a mixture containing components such as the target probe, unreacted quantum dots, unreacted recognition molecules, and coupling reagents. This mixture must be purified to obtain a high-purity functionalized probe. Purification methods can employ size exclusion chromatography, dialysis, or ultracentrifugation, utilizing the significant differences in molecular size between the product and impurities to effectively separate them. The purified product is the quantum dot fluorescent probe of this invention, where each quantum dot carries a recognition molecule with targeting function, ready for subsequent chip immobilization steps.

[0069] In one embodiment, the fluorescence signal generated after the functionalized fluorescent probe binds to the target needs to be decoded and quantified by an analysis platform to obtain the final detection result. Therefore, this invention provides a signal analysis subsystem for implementation. However, when a consumer-grade smartphone is used as the hardware foundation of this analysis platform, inherent differences and non-standardization exist between different smartphones in terms of hardware and built-in image processing algorithms. Specifically: First, the CMOS image sensors used in different smartphone models come from different manufacturers, and their key performance parameters, such as pixel size, quantum efficiency (QE) spectrum, and spectral transmission curve of Bayer color filter array, are different. Therefore, the original electrical signal strength generated and recorded by the CMOS sensors of two different smartphone models will be inconsistent even if the incident photons of the same wavelength emitted by quantum dots are received by the CMOS sensors of two different smartphone models.

[0070] Secondly, each smartphone has an opaque image signal processing flow embedded within it, namely the algorithm executed by the image signal processor (ISP). Its design goal is to optimize the subjective visual perception of the image, rather than to ensure the linear fidelity of the light signal. Therefore, after the CMOS sensor outputs raw data, the ISP will perform a series of complex, non-linear, and automatic processing on it, which varies from phone model to phone model.

[0071] These processing steps include, but are not limited to: performing proprietary demosaic algorithms to reconstruct full-color images from Bayer format; applying a color correction matrix (CCM) to map the sensor’s own color space to a standard color space (such as sRGB); performing non-linear tone mapping or gamma correction to adjust the dynamic range and contrast of the image, which severely disrupts the linear relationship between the original light intensity and the final pixel brightness value; and applying various automatic white balance, noise reduction, and sharpening algorithms designed to enhance the image.

[0072] Therefore, for an objectively existing and real fluorescence signal intensity distribution, when it is collected by two different smartphones (denoted as phone A and phone B), the actual final image data obtained will differ. and The following relationship exists: ; ; In the formula, and These represent the physical response functions of the CMOS sensors and their optical components in mobile phones A and B, respectively. and These represent a complex set of non-linear image processing functions executed by the image signal processors of mobile phone A and mobile phone B, respectively. and These represent the random noise introduced by each of the two systems; This is a real signal; Furthermore, since the hardware response function and ISP processing function are different for any two different mobile phones, this leads to discrepancies even when the actual signal is exactly the same. The input is the final output image data. and They are not the same either.

[0073] Therefore, in order to solve the aforementioned problems of signal distortion and incomparability caused by differences in different smartphone hardware and built-in algorithms, this invention introduces a specially designed and trained artificial intelligence (AI) decoding model to learn and reverse this complex, device-dependent signal transformation process, thereby achieving standardized restoration of the original fluorescence signal.

[0074] Specifically, a convolutional neural network (CNN) is preferably used as the basic architecture of this decoding model. This network structure can automatically extract spatially hierarchical features from the raw image data layer by layer. Therefore, its input is the raw image region containing one or more reaction units captured from a smartphone. The front end of the network consists of multiple convolutional and pooling layers stacked alternately, used to identify low-level features (such as the edges and textures of fluorescent dots) and high-level features (such as the complete shape and relative position of fluorescent dots) in the image. The middle and back ends of the network integrate and map the extracted high-level feature maps through fully connected layers or global average pooling layers, ultimately outputting one or a set of values, which is the corrected normalized signal intensity.

[0075] Furthermore, the key to building this AI decoding model lies in the training process. Therefore, it is trained by constructing a training dataset containing device noise, and the steps for constructing this dataset are as follows: First, a series of standard solutions with known concentration echelons are prepared. These standards contain the target analyte, and their concentrations are precisely calibrated, covering the entire dynamic range from the limit of detection to the saturation concentration.

[0076] The aforementioned standards were then reacted on functionalized chips to obtain a series of fluorescent chips corresponding to different target concentrations. These chips were then imaged using a high-precision, calibrated research-grade fluorescence imaging device (e.g., a cooled CCD camera on a research-grade microscope) to obtain the true signal intensity, denoted as . And this It is considered an ideal, standardized signal that is independent of the device.

[0077] Then, a large number of representative consumer smartphones (e.g., dozens of different models) from the market are selected, covering different brands, release years, and operating system versions. Using each smartphone, images of the fluorescent chips at all the aforementioned concentrations are acquired under conditions identical to actual detection (i.e., via adapter coupling), and saved in raw data format (RAW format). Through these steps, a training set containing tens of thousands or millions of image samples can be constructed. Each training data point in this dataset is a pair of data, in the form of a raw image acquired by the phone. The reference true signal strength corresponding to this image .

[0078] During the training phase, this dataset is used to optimize the internal parameters of the convolutional neural network (i.e., weights and biases, denoted as ). The goal of optimization is to minimize the error between the network's predicted output and the baseline true value. This process can be quantified by a loss function, which in a preferred implementation is the mean squared error. Therefore, the entire training process is essentially solving the following optimization problem: ; In the formula, For parameters The convolutional neural network model for input images The given predicted signal strength; This is the device-independent reference true signal strength corresponding to the image; These are the optimized network parameters obtained after training. This optimization process typically employs gradient descent-based algorithms, such as the Adam optimizer, to progressively adjust the network parameters through multiple iterations on the dataset. This continues until the loss function converges to a minimum value; This represents the entire dataset used to train the model; This represents the square of the L2 norm of the difference.

[0079] Thus, the neural network model trained using the above method is complete. Essentially, it has learned a highly complex, non-linear mapping function. Therefore, this function can accept non-standard fluorescence images taken by any smartphone (within the device distribution covered by the training set). As input, it automatically identifies signal distortion patterns introduced by the CMOS and ISP of a specific mobile phone, and then performs targeted reverse correction. The output corrected signal strength numerically eliminates the bias caused by device differences, thereby restoring the inherent, device-independent linear or nonlinear relationship between the signal and the target concentration in the sample.

[0080] Finally, the fully trained AI decoding model is integrated into a smartphone application (APP) for a complete analysis process from signal calibration to result output.

[0081] Furthermore, during installation, the application software will include the previously trained data containing optimal parameters. The convolutional neural network model file is packaged within its application package. To ensure efficient operation on mobile devices, the model is pre-converted to a lightweight format suitable for mobile inference, such as TensorFlow-Lite or ONNX-Runtime. When a user performs a detection, the application loads this model file through its built-in AI inference engine, along with the raw image data (including the detection region array) captured by the camera. The signal strength is input to the loaded model. The model then performs a forward propagation calculation and outputs a device-independent, normalized, corrected signal strength. This is an intermediate step in achieving accurate quantification, not the final result. Therefore, it is necessary to convert this standardized optical signal value into a target concentration value with practical physical meaning, and the application software (APP) can pre-load at least one standard curve. This standard curve describes the target concentration value. With the corrected signal strength The functional relationship between them. This functional relationship can be established during the system development phase by using a series of standards with known concentrations for testing and recording the corresponding baseline true value signal intensity. Obtained by fitting.

[0082] In a preferred embodiment, the standard curve can be fitted using a four-parameter logistic regression model (4PL), the functional expression of which is: ; The application actually stores the inverse function of that function. and the four parameters required ( , , , ),in: The asymptote of the response value of the curve is the maximum signal strength; This is the lower asymptote of the curve's response value, i.e., the minimum signal strength; The concentration value at which a 50% response is achieved, i.e., the half-maximal effect concentration; The Hill coefficient is related to the slope at the inflection point of the curve. By solving the above equation, the inverse function used to calculate the concentration can be obtained: ; When the application software obtains the corrected signal strength from the AI ​​model Then, this built-in function is automatically called. For multi-target detection chips, the software will call the curve parameters of the specific target corresponding to each reaction unit based on the spatial position of each unit, and set the curve parameters of each unit accordingly. By substituting the values ​​into the inverse function formula above, the final concentration of each target substance in the sample can be accurately calculated. .

[0083] The user interface of this application software is responsible for presenting the calculated concentration values ​​to the user in a clear and intuitive way, such as directly displaying the numerical value and its unit (e.g., ng / mL, g / kg, or ppm). In some application scenarios, the software can also perform qualitative judgments on the results based on built-in threshold standards, such as displaying information like qualified, exceeded, negative, or positive.

[0084] In addition, the application software also has data management and traceability functions. The results of each test, including but not limited to the final concentration value, test time, sample information (which can be entered by the user), and the original fluorescence image used for analysis, are all stored as an independent record in the local database of the mobile phone. Users can query historical test records at any time through the software interface, view result trends, or export single or batch test results in report form (such as PDF or CSV files) for archiving or sharing.

[0085] In one application embodiment, if the present invention is required to detect nitrofuran drug residues in food (e.g., livestock and poultry products), antigens of nitrofuran metabolites or their conjugates can be pre-immobilized on the surface of the reaction detection area of ​​the microfluidic detection chip. Simultaneously, a monoclonal antibody capable of specifically recognizing this type of antigen is conjugated to quantum dots with a specific emission wavelength to prepare a quantum dot fluorescent probe. During detection, the extract of the food sample to be tested is premixed with the aforementioned quantum dot fluorescent probe and then added together to the microfluidic detection chip. Subsequently, the nitrofurans present in the sample bind to the antigen and quantum dot fluorescent probe immobilized on the chip surface. The intensity of the fluorescence signal is then collected and analyzed, and calculated against a built-in standard curve based on competitive inhibition, thereby achieving highly sensitive quantitative detection of nitrofurans.

[0086] In another application embodiment, if the present invention is required to simultaneously detect multiple targets in food (e.g., pesticide residues, heavy metal ions, pathogens, and viruses), different microreaction units can be functionalized in the reaction detection area array of the microfluidic detection chip. For example, a monoclonal antibody for capturing specific pesticide molecules (e.g., chlorpyrifos) can be immobilized in the first region; a nucleic acid aptamer that specifically binds to specific heavy metal ions (e.g., lead or mercury ions) can be immobilized in the second region; a capture antibody for recognizing surface antigens of specific pathogens (e.g., Salmonella) can be immobilized in the third region; and an oligonucleotide probe for complementary hybridization with the nucleic acid sequence of a specific virus (e.g., norovirus) can be immobilized in the fourth region. Simultaneously, each of the above-mentioned recognition molecules is coupled to quantum dots with different emission wavelengths, thereby constructing a probe array with both spatial and wavelength encoding characteristics. When the sample flows through the entire reaction detection area, multiple targets that may be present in the sample will be specifically captured in their corresponding regions and generate fluorescence signals of specific colors. The system can simultaneously acquire fluorescence signals of different locations and colors in a single imaging process. The application software then uses an artificial intelligence decoding model and multiple independent standard curves for synchronous analysis, thereby enabling high-throughput simultaneous detection of various target objects of different categories.

[0087] This invention also provides a quantum dot fluorescence-encoded multi-target detection method, comprising the following steps: Step a, System Assembly and Alignment: A microfluidic detection chip is placed in the positioning mechanism of a portable signal acquisition module. The chip has a reaction detection area array composed of multiple spatially isolated micro-reaction units, and each micro-reaction unit has a specific quantum dot fluorescent probe fixed inside. Then, the module is coupled to a smart terminal so that the LED flash of the smart terminal is optically aligned with the reaction detection area of ​​the chip through an excitation filter, and the rear camera is optically aligned with the reaction detection area of ​​the chip through an emission filter. Step b, Excitation and Image Acquisition: Using the application software on the smart terminal, control the LED flash to excite the reaction detection area; at the same time, after locking the sensitivity, exposure time, focus mode and white balance parameters of the rear camera, control it to acquire the raw (RAW) format fluorescence image of the reaction detection area. Step c, Signal Decoding and Standardization: The pre-built artificial intelligence decoding model in the application software is invoked to analyze the original format fluorescence image, thereby converting the image data containing inherent distortions of the device into a standardized fluorescence signal intensity that is independent of the device hardware and built-in image processing algorithms. Step d, Quantitative calculation and result output: Substitute the standardized fluorescence signal intensity into the inverse function of the standard curve pre-stored in the application software to calculate the concentration value of the target to be tested, and display the concentration value on the smart terminal interface.

Claims

1. A quantum dot fluorescently encoded multi-target detection chip system, characterized in that, include: A microfluidic detection chip, wherein the chip has an array of reaction detection regions consisting of multiple spatially isolated micro-reaction units, and each micro-reaction unit has a quantum dot fluorescent probe fixed by chemical covalent coupling. The quantum dot fluorescent probe is made by chemically coupling quantum dots with a target-specific recognition molecule through carbodiimide. A portable signal acquisition module is used to couple an adapter between a smart terminal and the microfluidic detection chip. The adapter is provided with a chip positioning mechanism for positioning the microfluidic detection chip. The intelligent terminal, serving as both an excitation light source and an imaging unit, achieves position calibration through the adapter and the microfluidic detection chip. This allows the LED flash to illuminate the reaction detection area after passing through the excitation filter, while the rear camera acquires fluorescence images of the reaction detection area through the emission filter. The application software, installed in the smart terminal, executes the following processes: controlling the operation of the LED flash; under the premise of locking the camera's sensitivity, exposure time, focus, and white balance parameters, controlling the rear camera to acquire RAW format fluorescence images of the reaction detection area; calling the built-in, pre-trained artificial intelligence decoding model to analyze the RAW images, outputting device-independent standardized signal strength, and calculating and displaying the target concentration results based on the pre-stored standard curve.

2. The quantum dot fluorescently encoded multi-target detection chip system according to claim 1, characterized in that, The application software specifically includes an artificial intelligence decoding model, which takes a RAW format fluorescence image as input, maps it, and outputs a standardized fluorescence signal intensity that is independent of the device hardware and built-in image processing algorithms.

3. The quantum dot fluorescently encoded multi-target detection chip system according to claim 2, characterized in that, The artificial intelligence decoding model is a convolutional neural network, trained using image datasets collected from various types and models of consumer-grade smart terminals. The training set is based on the same set of standard fluorescent chips and also includes the baseline real value signal intensity obtained by a fluorescent imaging device.

4. The quantum dot fluorescently encoded multi-target detection chip system according to claim 3, characterized in that, The application software also has at least one pre-stored standard curve, which is used to describe the quantitative relationship between the target concentration and the device-independent standardized signal intensity output by the artificial intelligence decoding model; the software automatically calculates and outputs the quantitative concentration value of the target to be tested by substituting the standardized signal intensity into the inverse function of the standard curve.

5. The quantum dot fluorescence-encoded multi-target detection chip system according to claim 4, characterized in that, The standard curve is fitted using a four-parameter logistic regression model, which is determined by nonlinear regression of standardized signal intensity data obtained after testing known concentration standards.

6. The quantum dot fluorescently encoded multi-target detection chip system according to claim 1, characterized in that, Before controlling the rear camera to acquire fluorescence images, the application software locks the sensitivity, exposure time, focus mode, and white balance parameters of the rear camera; and the fluorescence images are saved as raw sensor data.

7. The quantum dot fluorescence-encoded multi-target detection chip system according to claim 1, characterized in that, The adapter of the portable signal acquisition module is made of an opaque polymer material to form a dark chamber. The adapter also has a physical slot that matches the shape of the microfluidic detection chip as a chip positioning mechanism, and integrates an excitation filter and an emission filter. The excitation filter and the emission filter are respectively located in the optical path of the smart terminal's LED flash and in the imaging optical path of the rear camera.

8. The quantum dot fluorescently encoded multi-target detection chip system according to claim 1, characterized in that, The substrate of the microfluidic detection chip is manufactured by integrating a three-dimensional computer-aided design model containing a preset microstructure through an additive manufacturing process; the additive manufacturing process is digital light processing three-dimensional printing or stereolithography three-dimensional printing.

9. The quantum dot fluorescently encoded multi-target detection chip system according to claim 1, characterized in that, The quantum dot fluorescent probe is fixed to the surface of the reaction detection region by chemical covalent coupling. The chemical covalent coupling is achieved by modifying the surface of the microfluidic detection chip with a first active group, introducing a second active group onto the quantum dot fluorescent probe molecule, and causing the first active group and the second active group to react and form a covalent bond. When the reaction detection region is fixed with two or more different types of quantum dot fluorescent probes, the distance between the center emission wavelengths of any two quantum dot fluorescent probes is greater than the product of the sum of the full width at half maximum (FWHM) of their respective emission spectra and the dimensionless separation coefficient.

10. A quantum dot fluorescence-encoded multi-target detection method, based on the quantum dot fluorescence-encoded multi-target detection chip system according to any one of claims 1-9, characterized in that, Includes the following steps: Step a, System Assembly and Alignment: A microfluidic detection chip is placed in the positioning mechanism of a portable signal acquisition module. The chip has a reaction detection area array composed of multiple spatially isolated micro-reaction units, and each micro-reaction unit has a specific quantum dot fluorescent probe fixed inside. Then, the module is coupled to a smart terminal so that the LED flash of the smart terminal is optically aligned with the reaction detection area of ​​the chip through an excitation filter, and the rear camera is optically aligned with the reaction detection area of ​​the chip through an emission filter. Step b, Excitation and Image Acquisition: Using the application software on the smart terminal, control the LED flash to excite the reaction detection area; at the same time, after locking the sensitivity, exposure time, focus mode and white balance parameters of the rear camera, control it to acquire the raw (RAW) format fluorescence image of the reaction detection area. Step c, Signal Decoding and Standardization: The pre-built artificial intelligence decoding model in the application software is invoked to analyze the original format fluorescence image, thereby converting the image data containing inherent distortions of the device into a standardized fluorescence signal intensity that is independent of the device hardware and built-in image processing algorithms. Step d, Quantitative calculation and result output: Substitute the standardized fluorescence signal intensity into the inverse function of the standard curve pre-stored in the application software to calculate the concentration value of the target to be tested, and display the concentration value on the smart terminal interface.

Images