A quantitative detection method for chikungunya virus combining silicon shell enhanced probe and ai algorithm

Abstract

The application discloses a quantitative detection method for chikungunya virus by combining a silicon shell enhanced probe with an AI algorithm, the probe is prepared by taking graphene oxide as a skeleton, quantum dots are loaded on the graphene oxide by electrostatic adsorption and layer-by-layer self-assembly technology, and the fluorescence signal can be significantly amplified; and a dense silicon dioxide shell is coated on the surface of the probe to shield the fluorescence quenching effect caused by complex biological sample matrixes such as whole blood or serum. The application combines a smartphone imaging technology with an artificial intelligence algorithm, establishes a quantitative analysis model based on a physical information neural network and a deep monotone constraint (PINN-DMC), can accurately extract the RGB color features of a detection line in a test strip image photographed by a smartphone, and reports virus concentration data. The application has the advantages of low cost, high sensitivity, strong anti-interference capability and independence on expensive professional fluorescence reading equipment, and is particularly suitable for instant quantitative detection of viruses in resource-limited areas.

Description

Technical Field

[0001] This invention relates to the field of biomedical detection technology, specifically to a quantitative detection method for chikungunya virus that combines a silicon-shell-enhanced probe with an AI algorithm. Background Technology

[0002] Chikungunya virus (CHIKV) is an arbovirus primarily transmitted by Aedes mosquitoes. In recent years, its geographical distribution has rapidly expanded from traditional endemic areas in Asia and Africa to Europe and the Americas, becoming a global public health threat. The clinical symptoms caused by CHIKV infection (such as high fever, rash, and severe joint pain) are highly similar to those of dengue fever (DENV) and Zika virus (ZIKV), making clinical differential diagnosis extremely challenging. Failure to achieve rapid and accurate screening in the early stages of infection (viremia) can easily lead to misdiagnosis and missed diagnosis, thus delaying treatment and exacerbating the spread of the epidemic. Currently, the "gold standard" for laboratory diagnosis of CHIKV mainly relies on virus isolation and culture and reverse transcription polymerase chain reaction (RT-PCR). Although these methods have extremely high sensitivity and specificity, their procedures are cumbersome and time-consuming (usually taking several hours to several days), and highly dependent on expensive centralized laboratory equipment and professionally trained technicians. This makes it difficult to widely implement the "gold standard" method in areas with scarce medical resources, at border quarantine sites, and on the front lines of sudden outbreaks.

[0003] To address the needs of point-of-care testing (POCT), lateral flow immunochromatography (LFIA) has emerged. Traditional colloidal gold test strips, using colloidal gold nanoparticles (AuNPs) as markers, are widely used due to their low cost and ease of operation. However, the colloidal gold method has significant technical limitations: the signal reading method based on visual colorimetry has low sensitivity, often failing to detect low viral loads, leading to a high false-negative rate; furthermore, the colloidal gold method typically only provides qualitative (negative / positive) results, making precise quantitative analysis difficult. To overcome these sensitivity limitations, fluorescence immunochromatography (FLFIA) based on fluorescent nanomaterials (such as quantum dots and fluorescent microspheres) has become a research hotspot. Theoretically, fluorescent probes can improve the limit of detection (LOD) by 1-2 orders of magnitude. However, in practical applications, existing FLFIA technologies still face the following serious challenges: i) Existing quantum dot probes are highly susceptible to non-specific protein adsorption or ionic strength in complex biological matrices (such as whole blood and serum), resulting in fluorescence quenching or aggregation, leading to significant signal attenuation and poor detection reproducibility. ii) Traditional monolayer quantum dot coupling methods have a limited fluorescent load per unit probe, which restricts the overall signal brightness and makes it difficult to achieve ultra-trace detection. iii) Reading traditional fluorescence chromatography results requires dedicated benchtop or handheld fluorescence readers. These instruments are expensive, bulky, and require a power supply, which significantly diminishes the "portability and low cost" advantages of POCT technology and limits its widespread adoption in resource-constrained scenarios.

[0004] Therefore, there is an urgent need to develop a new type of immunochromatographic platform that can significantly improve fluorescence intensity and anti-interference ability through material structure innovation, and can also get rid of dependence on dedicated instruments, so as to achieve low-cost, intelligent on-site quantitative detection. Summary of the Invention

[0005] To address the shortcomings of existing technologies, one objective of this invention is to provide a method for preparing graphene oxide-quantum dot core-shell nanoprobes (GO@TQD@Si). Specifically, a three-layer self-assembly method using polyethyleneimine (PEI) is employed to adsorb three layers of quantum dots (QDs) onto the surface of graphene oxide (GO), forming a core with strong fluorescence intensity, which is then coated with a layer of silica (SiO2). The large specific surface area of ​​graphene oxide provides an efficient loading framework, the multilayer quantum dot structure provides ultra-strong fluorescence signal output, and the SiO2 shell effectively shields the fluorescence quenching effect of complex biological samples on the quantum dots and significantly improves the colloidal stability of the probe.

[0006] The second objective of this invention is to combine the GO@TQD@Si probe with fluorescence immunochromatography (FLFIA) technology, and to combine a silicon-enhanced probe with an AI algorithm for quantitative detection of chikungunya virus, along with a smartphone-based AI algorithm analysis system, to achieve highly sensitive and non-destructive quantitative detection of chikungunya virus (CHIKV).

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for quantitative detection of CHIKV using an AI algorithm-assisted graphene-quantum dot fluorescent probe is based on the GO@TQD@Si nanoprobe, which consists of four main parts: graphene oxide (GO) as the central framework, a three-layer quantum dot (TQD) layer to provide enhanced fluorescence signal, a silica (Si) shell to provide protection and hydrophilicity, and a surface-coupled specific antibody to provide the ability to capture chikungunya virus.

[0009] Specifically, the quantitative detection method for chipore Kenyan virus combining silicon-shell-enhanced probes and AI algorithms includes the following steps:

[0010] Step S1: Construct a GO@TQD@Si fluorescent nanoprobe with a core-shell structure. The probe uses graphene oxide (GO) nanosheets as a two-dimensional framework, loads multilayer quantum dots (QDs) through electrostatic self-assembly, and coats the outermost layer with a silica (SiO2) shell. Finally, a specific antibody against chikungunya virus (CHIKV) is coupled to the surface of the shell.

[0011] Step S2: The biological sample to be tested is dropped onto an immunochromatographic test strip containing the fluorescent nanoprobe prepared in step S1, and the sample is driven to flow by capillary force. If the sample contains Chikungunya virus (CHIKV) antigen, it forms a double antibody sandwich complex of "probe-antigen-antibody" with the fluorescent nanoprobe and the capture antibody immobilized on the nitrocellulose membrane, which is enriched at the detection line (T line) and emits a fluorescent signal.

[0012] Step S3: Under ultraviolet excitation light source, use the camera of the mobile terminal device to capture the original image of the test strip detection area;

[0013] Step S4: Input the captured image into the pre-built deep multi-constraint network model (PINN-DMC). This model extracts the RGB color features of the detection line T region, combines them with the physical constraint loss function for calculation, and outputs the concentration value of Chikungunya virus (CHIKV) in the sample.

[0014] Preferably, the concentration of PEI in step S1 is 1 mg / mL to 5 mg / mL; the amount of TEOS added is 5-15 μL, preferably 10 μL; the thickness of the silica shell is controlled between 15 nm and 45 nm; the amount of APTES added is 0.1-0.8% (v / v); and the amount of CHIKV specific antibody added is 5-15 μg / mL.

[0015] In any of the above schemes, it is preferred that the concentration of the capture antibody immobilized on the nitrocellulose membrane in step S2 is 0.8-1.6 mg / mL, preferably 1.2 mg / mL.

[0016] In any of the above schemes, the preferred method is that the image acquisition and processing method in step S3 is as follows: using a smartphone camera to photograph the test strip in a dark environment, automatically locating the detection line area through an image processing algorithm, cropping out the region of interest, and calculating the average gray value of the pixels in the R, G, and B channels within the region as the input feature vector.

[0017] Preferably, in any of the above schemes, the artificial intelligence algorithm model construction method described in step S4 includes the following steps:

[0018] (1) Establishing a dataset: Prepare standard samples of chikungunya virus with different known concentrations, and obtain the corresponding image data according to steps S2 and S3 in claim 1;

[0019] (2) Feature extraction: Crop the region of interest of the T-line in the image and extract the pixel intensity values ​​of the R, G, and B channels;

[0020] (3) Model training: A neural network is used as the regression model, with RGB features as input and virus concentration as output;

[0021] (4) Model optimization: Physical constraints are introduced into the loss function to improve the model’s generalization ability and prediction accuracy under small sample data.

[0022] In any of the above schemes, it is preferred that the training of the model in step S4 uses a Physics-Informed Loss Function, the formula of which includes the following constraints:

[0023] L_total =L MSE +λ_monotonic·L_monotonic + λ_boundary·L_boundary + λ_smooth·L_smooth + λ_beer_lambert·L_beer_lambert

[0024] Where LMSE The mean squared error between the predicted and actual concentrations is defined by: L_monotonic, a monotonicity constraint term used to force the model output signal to increase monotonically with increasing concentration; L_beer_lambert, a Beer-Lambert law constraint term used to constrain the logarithmic linear relationship between fluorescence intensity and concentration; L_boundary, a boundary constraint term used to limit the output results to a physically feasible range; and L_smooth, a smoothness constraint loss. λ_monotonic, λ_boundary, λ_smooth, and λ_beer_lambert are the weighting coefficients of each constraint term.

[0025] This invention also provides a method for preparing the above-mentioned GO@TQD@Si fluorescent nanoprobes (based on graphene-quantum dot core-shell structure nanoprobes). The method involves adding a graphene oxide substrate to a PEI aqueous solution, using positively charged polyethyleneimine as a linker, and allowing PEI to self-assemble on the GO surface to form positively charged GO@PEI. GO@PEI is then added to a negatively charged carboxylated CdSe / ZnS quantum dot solution, where the negatively charged carboxylated CdSe / ZnS quantum dots are alternately adsorbed onto the graphene oxide surface. This process is repeated three times with a "PEI adsorption-washing-QD adsorption" cycle to form a GO@TQD composite with three layers of quantum dot loading. Finally, GO@TQD is silanized by dispersing GO@TQD nanosheets in an ethanol / water mixed solvent, with tetraethyl orthosilicate added as a silicon source (the amount of tetraethyl orthosilicate added is 5-15 μL, preferably 10 μL). Under alkaline conditions, a hydrolysis-condensation reaction is carried out to grow a uniform silica shell in situ on the core surface. The silica shell is then aminated using 3-aminopropyltriethoxysilane (APTES addition amount: 0.1-0.8% (v / v), preferably 0.4% (v / v)). Subsequently, a carboxyl group is introduced by reaction with succinic anhydride. Finally, the carboxyl group is activated by EDC / NHS (1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride / N-hydroxysuccinimide) to achieve covalent coupling with the antibody (CHIKV specific antibody addition amount: 5-15 μg / mL, preferably 10 μg / mL). The specific steps include the following:

[0026] (1) Preparation of graphene oxide (GO) substrate;

[0027] (2) The GO obtained in step (1) is added to the PEI aqueous solution. Under a fixed ultrasonic time, PEI self-assembles on the GO surface to form GO@PEI with a positive surface charge.

[0028] (3) Add the GO@PEI obtained in step (2) to the negatively charged carboxylated CdSe / ZnS quantum dot solution, and adsorb a layer of quantum dots through electrostatic interaction, and alternately adsorb the negatively charged carboxylated CdSe / ZnS quantum dots on the surface of graphene oxide.

[0029] (4) Repeat steps (2) and (3) to assemble the GO surface layer by layer using PEI as a mediator. By repeating the “PEI adsorption-washing-QD adsorption” cycle 3 times, a GO@TQD core with three layers of quantum dots is formed until a composite material (GO@TQD) loaded with three layers of quantum dots is obtained.

[0030] (5) The GO@TQD obtained in step (4) is subjected to silanization treatment. The GO@TQD nanosheets are dispersed in a mixed solvent of ethanol / water / ammonia, and tetraethyl orthosilicate is added as a silicon source. Under alkaline conditions, a hydrolysis-condensation reaction is carried out to grow a uniform and dense silica shell on its surface in situ, thereby obtaining a core-shell structured nanomaterial (GO@TQD@Si).

[0031] (6) The surface of GO@TQD@Si was modified by carboxylation using 3-aminopropyltriethoxysilane (APTES) and succinic anhydride, and anti-chikungunya virus-specific antibody was coupled on the shell surface. Anti-chikungunya virus monoclonal antibody was then coupled by EDC / NHS method.

[0032] The GO@TQD@Si probe provided by this invention comprises a multilayer cationic polymer strongly positively charged interlayer and a three-layer quantum dot fluorescent interlayer loaded on the GO surface; the quantum dot layer and GO are connected through the cationic polymer strongly positively charged interlayer, and the Si shell thickness ranges from 15 to 25 nm. The cationic polymer strongly positively charged interlayer is preferably a polyethyleneimine (PEI) layer.

[0033] Preferably, the graphene oxide in step (1) is a single-layer or few-layer sheet structure with a large specific surface area.

[0034] In any of the above schemes, the preferred embodiment is that, in steps (2), (3), and (4), the PEI solution is characterized by having a molecular weight of PEI of 10,000 to 80,000 Da, specifically any value within the range of 10,000 to 80,000 Da, such as 10,000 Da, 20,000 Da, 30,000 Da, 40,000 Da, 50,000 Da, 60,000 Da, 70,000 Da, or 80,000 Da; a concentration of PEI solution of any value within the range of 1 mg / mL to 5 mg / mL, such as 1 mg / mL, 2 mg / mL, 3 mg / mL, 4 mg / mL, or 5 mg / mL, preferably 2.5 mg / mL; and an ultrasonic time of any value within the range of 20 to 60 min, such as 20 min, 30 min, 40 min, 50 min, or 60 min, preferably 40 min. The quantum dot layer has quantum dot particle sizes of any value in the range of 5-20 nm, such as 5 nm, 10 nm, 15 nm, 20 nm, preferably 10 nm; and quantum dot concentrations of any value in the range of 1-10 mg / mL, such as 1 mg / mL, 3 mg / mL, 5 mg / mL, 8 mg / mL, 10 mg / mL, preferably 5 mg / mL.

[0035] In any of the above schemes, it is preferred that in step (2), the volume ratio of graphene oxide to PEI aqueous solution is 1:35-40, and more preferably, the volume ratio of graphene oxide to PEI aqueous solution is 1:40; the final concentration of GO graphene is 0.1 mg / mL, and the concentration of PEI aqueous solution is 1 mg / mL.

[0036] In any of the above schemes, it is preferred that in step (3), the volume ratio of GO@PEI to the negatively charged carboxylated CdSe / ZnS quantum dot solution is 35-45:1, specifically 35:1, 40:1, 45:1, preferably 40:1; and the concentration of carboxylated CdSe / ZnS quantum dots is 1 mg / mL.

[0037] In any of the above schemes, it is preferred that the thickness of the silica shell in step (5) is between 15 nm and 45 nm, specifically 15 nm, 20 nm, 25 nm, 30 nm, 35 nm, 40 nm, preferably 25 nm, so as to ensure the best fluorescence protection effect and not affect the flowability of chromatography.

[0038] In any of the above schemes, preferably, in step (6), the silica shell is aminated using 3-aminopropyltriethoxysilane, followed by reaction with succinic anhydride to introduce carboxyl groups, and finally the carboxyl groups are activated by EDC / NHS (1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride / N-hydroxysuccinimide) to achieve covalent coupling with the antibody. The amount of APTES added is any value in the range of 0.1-0.8% (v / v), such as 0.1% (v / v), 0.2% (v / v), 0.3% (v / v), 0.4% (v / v), 0.5% (v / v), 0.6% (v / v), 0.7% (v / v), 0.8% (v / v), preferably 0.4% (v / v); the amount of CHIKV specific antibody added is any value in the range of 5-15 μg / mL, such as 5 μg / mL. μg / mL, 10 μg / mL, 15 μg / mL, preferably 10 μg / mL.

[0039] This invention provides an immunochromatographic test strip for quantitative detection, using the GO@TQD@Si nanoprobe as a signal tag.

[0040] In this invention, the nitrocellulose membrane is loaded with one detection line (T line) and one control line (C line). The detection line modifier is preferably an anti-CHIKV antibody, and the control line modifier is preferably biotinylated IgG. The concentration of the anti-CHIKV antibody is 0.8-1.6 mg / mL, preferably 1.2 mg / mL, and the concentration of the biotinylated IgG is 0.5-1.5 mg / mL, preferably 1 mg / mL. The temperature of the constant temperature drying oven is set to 34-36℃, preferably 35℃.

[0041] This invention also provides a method for detection using AI algorithms, comprising the following steps:

[0042] The GO@TQD@Si probe is mixed with the sample to be tested and dropped onto the sample pad of the immunochromatographic test strip. After the chromatographic reaction, the test strip image is captured by a smartphone under ultraviolet light excitation. The RGB color feature values ​​of the detection line (T line) area are extracted by AI algorithm. The extracted RGB features are input into a pre-trained machine learning model, and the model calculates and outputs the quantitative concentration of the virus.

[0043] Another aspect of this invention discloses a detection system employing the aforementioned method for quantitative detection of chikungunya virus using a combination of a silicon-shell-enhanced probe and an AI algorithm. This system is used for quantitative detection of chikungunya virus and includes: a detection kit, vacuum-sealed GO@TQD@Si fluorescent immunochromatographic test strips and sample diluent, and intelligent analysis software. The intelligent analysis software is an application installed on a smartphone, integrating the PINN-DMC algorithm model for real-time display of image acquisition, feature extraction, and concentration calculation results.

[0044] Beneficial effects:

[0045] Compared with the prior art, the advantages of the present invention are:

[0046] This invention discloses a quantitative detection method for chipore Kenyan virus combining a silicon-shell-enhanced probe and an AI algorithm. The probe uses graphene oxide (GO) as a framework and employs electrostatic adsorption and layer-by-layer self-assembly technology to load quantum dots (QDs) at high density, significantly amplifying the fluorescence signal. A dense silica (SiO2) shell covers the surface to shield against fluorescence quenching effects caused by complex biological sample matrices such as whole blood or serum. This invention combines smartphone imaging technology with artificial intelligence algorithms to establish a quantitative analysis model based on physical information neural networks and deep monotonic constraints (PINN-DMC). This model can accurately extract the RGB color features of the detection lines in test strip images captured by a mobile phone and report virus concentration data. This invention has advantages such as low cost, high sensitivity, strong anti-interference ability, and no reliance on expensive professional fluorescence reading equipment. It is particularly suitable for point-of-care testing (POCT) of viruses in resource-limited areas. Specifically:

[0047] (1) Ultra-high sensitivity: The GO@TQD@Si nanoprobe proposed in this invention uses graphene oxide as a framework and utilizes the large specific surface area of ​​GO and multilayer assembly technology to achieve high-density loading of quantum dots. Experiments show that the detection limit of this platform for CHIKV is as low as 3.18 pg / mL, and the sensitivity is about 157 times higher than that of the traditional colloidal gold method and about 305 times higher than that of commercial ELISA kits.

[0048] (2) Excellent stability and anti-interference ability: The GO@TQD@Si nanoprobe proposed in this invention has a three-layer quantum dot core and a SiO2 shell structure. The three-layer quantum dots provide fluorescence intensity far exceeding that of traditional single-layer probes. The dense SiO2 shell effectively isolates the quantum dots from the external environment, prevents fluorescence quenching caused by biological matrix, and significantly improves the colloidal stability of the probe in complex matrices such as serum.

[0049] (3) The preparation method is simple and easy to industrialize: The present invention proposes a method for preparing multilayer probes by using PEI layer-by-layer self-assembly, which is simple to operate, has good repeatability, and is easy to achieve mass production and promotion.

[0050] (4) Intelligent Detection: This invention innovatively introduces a machine learning-based image analysis algorithm, combined with AI algorithms and smartphone imaging technology, to establish a precise mapping model between RGB color features and viral concentration. This technology only requires a smartphone to complete high-precision quantitative detection, without relying on expensive large-scale fluorescence reading equipment, greatly reducing detection costs and improving the accessibility of the technology. Validation data shows that the platform achieves a sensitivity of 94.24% and a specificity of 96.1% in clinical samples, respectively, showing high consistency with the gold standard method, and is particularly suitable for on-site real-time screening in resource-scarce areas. Attached Figure Description

[0051] Figure 1 The preparation process of the GO@TQD@Si nanoprobe described in this invention;

[0052] Figure 2 The characterization results are as follows: GO@TQD@Si nanoprobe described in this invention;

[0053] Figure 3 A comparison of the fluorescence properties of the GO@TQD@Si nanoprobes described in this invention;

[0054] Figure 4 To optimize the silica shell thickness of the GO@TQD@Si nanoprobe described in this invention;

[0055] Figure 5 For the stability evaluation of the GO@TQD@Si nanoprobe described in this invention;

[0056] Figure 6 This is a schematic diagram and verification of the bioconjugation of the GO@TQD@Si nanoprobe described in this invention;

[0057] Figure 7 This is a schematic diagram of the detection principle based on GO@TQD@Si-LFIA as described in this invention;

[0058] Figure 8 This invention provides an evaluation of the detection performance based on GO@TQD@Si-LFIA.

[0059] Figure 9 This invention provides a comparison of GO@TQD@Si-LFIA with other methods.

[0060] Figure 10 The training-validation accuracy curve and overall accuracy on the independent test set of the PINN-DMC model described in this invention are shown.

[0061] Figure 11The results show the comparison between the PINN-DMC model described in this invention and the traditional ImageJ software analysis.

[0062] Figure 12 This invention provides a concordance rate analysis of GO@TQD@Si-LFIA and PCR in the qualitative detection of clinical samples. Detailed Implementation

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. However, the present invention is not limited to the various configurations described below. Various modifications can be made within the scope of the claims. Embodiments and examples obtained by appropriately combining different implementation methods and the technical means disclosed in each embodiment are also included in the technical scope of the present invention. The following examples further illustrate the present invention, but do not constitute a limitation. In the present invention, unless otherwise specified, all raw materials / components are commercially available products well known to those skilled in the art.

[0064] This invention provides an AI-assisted graphene-quantum dot fluorescent probe (GO@TQD@Si-mAb). Utilizing the layer-by-layer self-assembly function of PEI, three layers of carboxylated quantum dots are loaded onto a GO backbone with a large specific surface area and coated with a dense SiO2 shell to form a composite nanotag that generates a strong fluorescent signal and is resistant to quenching. Furthermore, carboxyl groups are modified on the surface of GO@TQD@Si, thereby enabling the chemical modification to couple with an anti-CHIKV monoclonal antibody, achieving highly sensitive capture of viral antigens.

[0065] This invention discloses a method for quantitative detection of CHIKV using an AI algorithm-assisted graphene-quantum dot fluorescent probe. The core of this method lies in the GO@TQD@Si nanoprobe used, which is mainly composed of four parts: graphene oxide (GO) as the central framework, three layers of quantum dot (TQD) layers to provide enhanced fluorescence signal, a silica (Si) shell to provide protection and hydrophilicity, and surface-coupled specific antibodies to provide the ability to capture chikungunya virus.

[0066] Specifically, the quantitative detection method for chipore Kenyan virus combining silicon-shell-enhanced probes and AI algorithms includes the following steps:

[0067] Step S1: Construct a GO@TQD@Si fluorescent nanoprobe with a core-shell structure. The probe uses graphene oxide (GO) nanosheets as a two-dimensional framework, loads multilayer quantum dots (QDs) through electrostatic self-assembly, and coats the outermost layer with a silica (SiO2) shell. Finally, a specific antibody against chikungunya virus (CHIKV) is coupled to the surface of the shell.

[0068] Step S2: The biological sample to be tested is dropped onto an immunochromatographic test strip containing the fluorescent nanoprobe prepared in step S1, and the sample is driven to flow by capillary force. If the sample contains Chikungunya virus (CHIKV) antigen, it forms a double antibody sandwich complex of "probe-antigen-antibody" with the fluorescent nanoprobe and the capture antibody immobilized on the nitrocellulose membrane, which is enriched at the detection line (T line) and emits a fluorescent signal.

[0069] Step S3: Under the ultraviolet excitation light source wavelength (365 nm), use the camera of the mobile terminal device to capture the original image of the test strip detection area;

[0070] Step S4: Input the captured image into the pre-built deep multi-constraint network model (PINN-DMC). This model extracts the RGB color features of the detection line T region, combines them with the physical constraint loss function for calculation, and outputs the concentration value of Chikungunya virus (CHIKV) in the sample.

[0071] The further optimized technical solution of this embodiment is that the concentration of PEI in step S1 is 1 mg / mL to 5 mg / mL; the amount of TEOS added is 5-15 μL, preferably 10 μL; the thickness of the silica shell is controlled between 15 nm and 45 nm; the amount of APTES added is 0.1-0.8% (v / v); and the amount of CHIKV specific antibody added is 5-15 μg / mL.

[0072] A further optimized technical solution in this embodiment is that the concentration of the capture antibody immobilized on the nitrocellulose membrane in step S2 is 0.8-1.6 mg / mL, preferably 1.2 mg / mL.

[0073] The further optimized technical solution in this embodiment is that the image acquisition and processing method in step S3 is as follows: the test strip is photographed in a dark environment using a smartphone camera, the detection line area is automatically located through an image processing algorithm, the region of interest is cropped out, and the average gray value of the pixels in the R, G, and B channels in the region is calculated as the input feature vector.

[0074] A further optimized technical solution in this embodiment is that the artificial intelligence algorithm model construction method described in step S4 includes the following steps:

[0075] (1) Establishing a dataset: Prepare standard samples of chikungunya virus with different known concentrations, and obtain the corresponding image data according to steps S2 and S3 in claim 1;

[0076] (2) Feature extraction: Crop the region of interest of the T-line in the image and extract the pixel intensity values ​​of the R, G, and B channels;

[0077] (3) Model training: A neural network is used as the regression model, with RGB features as input and virus concentration as output;

[0078] (4) Model optimization: Physical constraints are introduced into the loss function to improve the model’s generalization ability and prediction accuracy under small sample data.

[0079] A further optimized technical solution in this embodiment is that the training of the model in step S4 adopts the Physics-Informed Loss Function, the formula of which includes the following constraints:

[0080] L_total =L MSE +λ_monotonic·L_monotonic + λ_boundary·L_boundary + λ_smooth·L_smooth + λ_beer_lambert·L_beer_lambert

[0081] Where L MSE The mean squared error between the predicted and actual concentrations is defined by: L_monotonic, a monotonicity constraint term used to force the model output signal to increase monotonically with increasing concentration; L_beer_lambert, a Beer-Lambert law constraint term used to constrain the logarithmic linear relationship between fluorescence intensity and concentration; L_boundary, a boundary constraint term used to limit the output results to a physically feasible range; and L_smooth, a smoothness constraint loss. λ_monotonic, λ_boundary, λ_smooth, and λ_beer_lambert are the weighting coefficients of each constraint term.

[0082] Example 1: Construction, System Characterization and Optimization of GO@TQD@Si Core-Shell Structured Nanoprobes

[0083] 1. Preparation method of GO@TQD@Si nanoprobes, such as Figure 1 As shown, a multilayer fluorescent core was constructed: GO's large specific surface area was used as a two-dimensional carrier, and the positive charge properties of PEI were utilized to modify the GO surface; negatively charged carboxylated CdSe / ZnS quantum dots were adsorbed via electrostatic attraction. By repeating the cyclical steps of "PEI adsorption-quantum dot adsorption," the layer-by-layer controlled assembly of quantum dots on the GO surface was achieved, ultimately forming a GO@TQD core loaded with three layers of quantum dots. In-situ growth of a SiO2 protective shell: Using a modified Stöber method with TEOS as the silicon source, a dense SiO2 shell was grown in-situ on the GO@TQD surface, forming a GO@TQD@Si structure.

[0084] Specifically, a graphene oxide substrate is added to a PEI aqueous solution, and positively charged polyethyleneimine is used as a linker. PEI self-assembles on the GO surface to form GO@PEI with a positively charged surface. GO@PEI is then added to a negatively charged carboxylated CdSe / ZnS quantum dot solution, and the negatively charged carboxylated CdSe / ZnS quantum dots are alternately adsorbed onto the graphene oxide surface. By repeating the "PEI adsorption-washing-QD adsorption" cycle three times, a GO@TQD composite with three layers of quantum dots is formed. GO@TQD is then silanized. The process involves dispersing GO@TQD nanosheets in an ethanol / water mixed solvent, adding tetraethyl orthosilicate as a silicon source, and performing a hydrolysis-condensation reaction under alkaline conditions to grow a uniform silica shell in situ on the core surface. The silica shell is then aminated using 3-aminopropyltriethoxysilane, followed by a reaction with succinic anhydride to introduce carboxyl groups. Finally, the carboxyl groups are activated via EDC / NHS (1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride / N-hydroxysuccinimide) to achieve covalent coupling with the antibody. The specific preparation method includes the following steps:

[0085] (1) Preparation of graphene oxide (GO) substrate: Take an appropriate amount of GO aqueous solution and place it in an ultrasonic cleaner for moderate ultrasonic treatment (15 min). Transfer the ultrasonically treated GO suspension to a centrifuge tube and centrifuge at low speed (2000~3000 rpm, centrifuge for 15 minutes). After centrifugation, discard the bottom precipitate (unpeeled graphite and ultra-large GO flakes) and collect the supernatant. Centrifuge the collected supernatant at medium / high speed (8000~10000 rpm, centrifuge for 15 minutes). After centrifugation, collect the bottom precipitate and discard the supernatant. Wash the GO precipitate collected in the previous step with deionized water 1-2 times to remove residual impurities.

[0086] (2) The GO obtained in step (1) is added to the PEI aqueous solution. Under the action of a fixed ultrasonic time, PEI self-assembles on the GO surface to form GO@PEI with a positive charge on the surface. The volume ratio of graphene oxide to PEI aqueous solution is 1:40, the final concentration of GO graphene is 0.1 mg / mL, and the concentration of PEI aqueous solution is 1 mg / mL.

[0087] (3) The GO@PEI obtained in step (2) is added to the negatively charged carboxylated CdSe / ZnS quantum dot solution to adsorb a layer of quantum dots through electrostatic interaction, and the negatively charged carboxylated CdSe / ZnS quantum dots are alternately adsorbed on the surface of graphene oxide; the volume ratio of GO@PEI to the negatively charged carboxylated CdSe / ZnS quantum dot solution is 40:1, and the concentration of carboxylated CdSe / ZnS quantum dots is 1 mg / mL.

[0088] (4) Repeat steps (2) and (3) to assemble the GO surface layer by layer using PEI as a mediator. By repeating the “PEI adsorption-washing-QD adsorption” cycle 3 times, a GO@TQD core with three layers of quantum dots is formed until a composite material (GO@TQD) loaded with three layers of quantum dots is obtained.

[0089] (5) The GO@TQD obtained in step (4) was subjected to silanization treatment. The GO@TQD nanosheets were dispersed in a mixed solvent of ethanol / water / ammonia (volume ratio of ethanol / water / ammonia was 18:1:1) to achieve a final concentration of 0.5 mg / mL. Tetraethyl orthosilicate (5~15 μL) was added as a silicon source, and a hydrolysis-condensation reaction was carried out under alkaline conditions (pH range of 9-11) to grow a uniform and dense silica shell on its surface in situ, thereby obtaining a core-shell structured nanomaterial (GO@TQD@Si).

[0090] (6) The surface of GO@TQD@Si was modified by carboxylation using 3-aminopropyltriethoxysilane (APTES) and succinic anhydride, and then coupled with anti-Chikungunya virus monoclonal antibody by EDC / NHS method. The amount of APTES added was 0.1-0.8% (v / v), and the concentration of succinic anhydride was 0.1 mol / L.

[0091] The GO@TQD@Si probe provided by this invention comprises a multilayer cationic polymer strongly positively charged interlayer and a three-layer quantum dot fluorescent interlayer loaded on the GO surface; the quantum dot layer and GO are both connected through the cationic polymer strongly positively charged interlayer. The cationic polymer strongly positively charged interlayer is preferably a polyethyleneimine (PEI) layer.

[0092] The further optimized technical solution of this embodiment is: the graphene oxide in step (1) is preferably a single-layer or few-layer sheet structure with a large specific surface area.

[0093] The further optimized technical solution of this embodiment is as follows: In steps (2), (3) and (4), the PEI solution is characterized in that the molecular weight of PEI is 10,000 to 80,000 Da, specifically any value in the range of 10,000 to 80,000 Da, such as 10,000 Da, 20,000 Da, 30,000 Da, 40,000 Da, 50,000 Da, 60,000 Da, 70,000 Da, 80,000 Da; the concentration of the PEI solution is any value in the range of 1 mg / mL to 5 mg / mL, such as 1 mg / mL, 2 mg / mL, 3 mg / mL, 4 mg / mL, 5 mg / mL, preferably 2.5 mg / mL; the ultrasonic time is any value in the range of 20 to 60 min, such as 20 min, 30 min, 40 min, 50 min, 60 min, preferably 40 min. The quantum dot layer has quantum dot particle sizes of any value in the range of 5-20 nm, such as 5 nm, 10 nm, 15 nm, 20 nm, preferably 10 nm; and quantum dot concentrations of any value in the range of 1-10 mg / mL, such as 1 mg / mL, 3 mg / mL, 5 mg / mL, 8 mg / mL, 10 mg / mL, preferably 5 mg / mL.

[0094] The further optimized technical solution of this embodiment is: in step (5), the thickness of the silica shell is between 15 nm and 45 nm, specifically 15 nm, 20 nm, 25 nm, 30 nm, 35 nm, 40 nm, preferably 25 nm, so as to ensure the best fluorescence protection effect and not affect the flowability of chromatography.

[0095] The further optimized technical solution of this embodiment is as follows: In step (6), 3-aminopropyltriethoxysilane (APTES) is used to aminate the silica shell, followed by reaction with succinic anhydride to introduce carboxyl groups, and finally the carboxyl groups are activated by EDC / NHS (1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride / N-hydroxysuccinimide) to achieve covalent coupling with the antibody. The purity of APTES is above 98%, and the amount of APTES added is any value in the range of 0.1-0.8% (v / v), such as 0.1% (v / v), 0.2% (v / v), 0.3% (v / v), 0.4% (v / v), 0.5% (v / v), 0.6% (v / v), 0.7% (v / v), 0.8% (v / v), preferably 0.4% (v / v); the amount of CHIKV specific antibody added is any value in the range of 5-15 μg / mL, such as 5 μg / mL, 10 μg / mL, 15 μg / mL, preferably 10 μg / mL.

[0096] 2. Structural Characterization

[0097] To verify the microstructure of the prepared probe, it was characterized using transmission electron microscopy (TEM) and energy-dispersive X-ray spectroscopy (EDS). Figure 2 As shown in Figure A, the initial GO exhibits a typical two-dimensional wrinkled sheet structure with a smooth surface. After multi-layer assembly (GO@TQD), a large number of high-electron-density black dots (i.e., quantum dots) can be clearly observed to uniformly and densely cover the GO sheet surface without obvious agglomeration. After coating with a Si shell (GO@TQD@Si), a continuous, uniform, and semi-transparent thin layer structure is formed on the nanosheet surface, clearly showing a "graphene-quantum dot-silica" sandwich core-shell structure. Elemental surface scan results (e.g.) Figure 2 (As shown in B) This further confirms that element C constitutes the sheet-like framework. Elements Cd, Se, Zn, and S (characteristic elements of quantum dots) and elements Si and O (characteristic elements of silicon dioxide shells) are highly overlapping and uniformly distributed in space, proving the successful construction of the core-shell structure.

[0098] 3. Performance optimization

[0099] (1) Optimization of fluorescence signal: This example compares the effect of different quantum dot loading layers (1 layer, 2 layers, 3 layers) on probe fluorescence intensity. For example Figure 3The results showed that the fluorescence emission peak intensity of the probe significantly increased in a stepwise manner with the increase of the number of assembly layers. Compared with the monolayer (GO@QD) and bilayer (GO@DQD) structures, the three-layer loaded GO@TQD exhibited the highest fluorescence quantum yield and signal intensity. This indicates that the large specific surface area of ​​GO can support high-density quantum dots, and the spacing effect of the PEI layers effectively suppresses the self-quenching effect caused by fluorescence resonance energy transfer (FRET) between quantum dot layers. Therefore, the three-layer assembly structure is preferred as the core of this invention.

[0100] (2) Optimization of Si shell thickness: In this embodiment, probes with different shell thicknesses were prepared by adjusting the amount of TEOS added (e.g., 10 μL, 20 μL, 40 μL, etc.). TEM observation showed that as the amount of TEOS increased, such as... Figure 4 As shown in Figure A, the Si shell thickness gradually increases from approximately 15 nm to 45 nm. Fluorescence spectroscopy measurements indicate (e.g.) Figure 4 As shown in Figure B), the fluorescence intensity of the probe remained at its highest level when the shell thickness was in the range of 15-25 nm; however, when the thickness increased to approximately 45 nm, the fluorescence intensity showed a significant decrease. This is because an excessively thick Si layer increases light scattering and absorption losses. Meanwhile, polydispersity index (PDI) measurements showed (as shown in Figure B)... Figure 4 As shown in Figure C, all silica-coated probes exhibited significantly improved stability (manifested as lower PDI values) compared to the uncoated precursors. Probes with shell thicknesses of approximately 15 nm and 45 nm showed lower PDI values, indicating optimal colloidal stability, while probes with a shell thickness of approximately 25 nm exhibited improved but relatively higher PDI values. Therefore, this invention optimizes the Si shell thickness to 25 nm to balance protection and light transmittance.

[0101] 4. Stability Comparison Experiment

[0102] To verify the crucial role of the SiO2 shell in enhancing probe performance, GO@TQD without Si shell protection and GO@TQD@Si of this invention were compared in a simulated physiological environment. Experimental results (e.g.) Figure 5 A and such Figure 5 As shown in Figure B, under 365 nm UV excitation, uncoated GO@TQD nanosheets exhibited significant aggregation in low pH buffer (< 4.0) accompanied by significant fluorescence quenching. In contrast, GO@DQD@Si maintained uniform dispersion and bright fluorescence throughout the tested pH range.

[0103] Example 2: Construction and performance evaluation of GO@TQD@Si nanoprobe surface antibody conjugation, immunochromatographic test strip.

[0104] 1. Antibody conjugation process on the surface of GO@TQD@Si nanoprobes

[0105] This invention conjugates anti-CHIKV monoclonal antibodies to the probe surface through a series of chemical modifications, endowing inorganic nanoprobes with the ability to specifically recognize biomolecules. For example... Figure 6 As shown in Figure A, firstly, the hydroxyl groups on the GO@TQD@Si surface are modified using APTES to introduce amino groups (-NH2); subsequently, the surface is converted to carboxyl groups (-COOH) through the reaction of succinic anhydride with the amino group; finally, the carboxyl groups are activated by EDC / NHS to form an active ester intermediate, which forms a stable amide bond with the amino group in the antibody molecule, completing the bioconjugation. Changes in surface potential (e.g.) Figure 6 As shown in Figure B, the success of the modification process is visually confirmed. The original GO@TQD@Si exhibits a negative charge (approximately -25 mV) due to the presence of silanol groups on its surface; after amination, the potential flips to a positive charge (approximately +30 mV); after carboxylation, it reverts to a negative charge (approximately -35 mV); after antibody conjugation, due to the introduction of the protein, the potential shifts slightly to the positive direction but remains negative (approximately -20 mV). This is evident in the infrared spectrum (as shown in Figure B). Figure 6 As shown in C), the coupled probe exhibited distinct characteristic absorption peaks at 1640 cm⁻¹ and 1540 cm⁻¹, corresponding to the vibrations of the amide I and amide II bands, respectively, confirming that the antibody had been successfully covalently bound to the surface of the nanoprobe.

[0106] 2. Immunochromatographic test strip detection

[0107] like Figure 7 As shown, this invention employs a "double antibody sandwich method" for detection. The test sample (80 μL) is premixed with a GO@TQD@Si-Ab fluorescent probe (3-4 μL). If the CHIKV antigen is present in the sample, the antibody on the probe surface will specifically bind to it, forming a "probe-antigen" complex. After the mixture is added to the sample pad, it flows towards the absorbent pad under capillary action. When flowing through the detection line (T line), the complex is intercepted by the capture antibody immobilized on the T line, forming a "probe-antigen-capture antibody" sandwich structure, which displays a bright fluorescent band under UV excitation. Unbound probes continue to flow to the control line (C line), where they are captured by the secondary antibody, forming a fluorescent control band used to verify the effectiveness of the chromatography process.

[0108] 3. Quantitative detection performance evaluation

[0109] This embodiment utilizes smartphone imaging and AI algorithms to quantitatively analyze the detection results. A series of CHIKV protein standard solutions of different concentrations (0 pg / mL - 100 ng / mL) were prepared for detection, and the results are as follows. Figure 8As shown in Figure A, the fluorescence intensity of the T-line gradually weakens as the antigen concentration decreases. RGB feature values ​​of the T-line were extracted using an AI algorithm, and a log-linear standard curve (R² > 0.98) was constructed to show the relationship between fluorescence intensity and virus concentration. The calculated limit of detection (LOD) of the method of this invention is as low as 3.18 pg / mL. Cross-reactivity tests were performed on structurally similar arboviruses such as dengue virus, Zika virus (ZIKV), yellow fever virus (YFV), and Japanese encephalitis virus (JEV). The results showed (e.g.) Figure 8 As shown in Figure B), only the CHIKV sample produced a significant fluorescent signal at the T line; the signal intensity of other virus samples was no different from the blank control, indicating that the probe has extremely high specificity. Intra-assay and inter-assay repeatability tests were performed on samples of the same concentration. The results showed (as shown in Figure B). Figure 8 As shown in C), the intra-batch coefficient of variation (CV) is less than 6.5%, and the inter-batch coefficient of variation is less than 9.2%, indicating that the test strip has good manufacturing process stability and detection precision.

[0110] Under the same experimental conditions, the method of this invention was compared with traditional colloidal gold test strips and commercial ELISA kits (data attached). Figure 9 A and Figure 9 (As shown in B). The detection limit of the homemade colloidal gold test strip is approximately 0.5 ng / mL. The sensitivity of the method of this invention is approximately 157 times higher than that of the colloidal gold method. The detection limit of a commercially available ELISA kit is 0.97 ng / mL. The sensitivity of the method of this invention is approximately 305 times higher than that of the ELISA method.

[0111] Example 3: Establishment of a machine learning algorithm model based on smartphone image processing and evaluation of its clinical diagnostic efficacy.

[0112] After constructing a core-shell structured GO@TQD@Si fluorescent nanoprobe according to Example 1, the biological sample to be tested was dropped onto an immunochromatographic test strip containing the fluorescent nanoprobe, and the sample flow was driven by capillary force. If the sample contained chikungunya virus antigen, it formed a "probe-antigen-antibody" double antibody sandwich complex with the fluorescent nanoprobe and the capture antibody immobilized on the nitrocellulose membrane, which was enriched at the detection line and emitted a fluorescent signal. Under ultraviolet excitation, the original image of the detection area of ​​the test strip was captured by the camera of a mobile terminal device. Then, the captured image was input into a pre-constructed deep multi-constraint network model, which extracted the RGB color features of the detection line area, combined with the physical constraint loss function, and calculated to output the concentration value of chikungunya virus in the sample.

[0113] The above image acquisition and processing method is as follows: use a smartphone camera to take a picture of the test strip in a dark environment, use an image processing algorithm to automatically locate the detection line area, crop out the region of interest, and calculate the average gray value of the R, G, and B channels in the region as the input feature vector.

[0114] The method for constructing artificial intelligence algorithm models includes the following steps:

[0115] (1) Establishing a dataset: Prepare standard samples of Chikungunya virus at different known concentrations. According to steps S2 and S3 of the quantitative detection method of Chikungunya virus combining silicon-shell enhanced probe and AI algorithm (Step S2: The biological sample to be tested is dropped onto the immunochromatographic test strip containing the fluorescent nanoprobe prepared in step S1, and the sample is driven to flow by capillary force; if the sample contains Chikungunya virus CHIKV antigen, it forms a double antibody sandwich complex of "probe-antigen-antibody" with the fluorescent nanoprobe and the capture antibody fixed on the nitrocellulose membrane, which is enriched at the detection line (T line) and emits a fluorescent signal);

[0116] Step S3: Under an ultraviolet excitation light source wavelength of 365 nm, capture the original image of the test strip detection area using the camera of the mobile terminal device; obtain the corresponding image data;

[0117] (2) Feature extraction: Crop the region of interest of the T-line in the image and extract the pixel intensity values ​​of the R, G, and B channels;

[0118] (3) Model training: A neural network is used as the regression model, with RGB features as input and virus concentration as output;

[0119] (4) Model optimization: Physical constraints are introduced into the loss function to improve the model’s generalization ability and prediction accuracy under small sample data.

[0120] The above model is trained using a Physics-Informed Loss Function, the formula of which includes the following constraints:

[0121] L_total =L MSE +λ_monotonic·L_monotonic + λ_boundary·L_boundary + λ_smooth·L_smooth + λ_beer_lambert·L_beer_lambert

[0122] Where L MSEThe mean squared error between the predicted and actual concentrations is defined by: L_monotonic, a monotonicity constraint term used to force the model output signal to increase monotonically with increasing concentration; L_beer_lambert, a Beer-Lambert law constraint term used to constrain the logarithmic linear relationship between fluorescence intensity and concentration; L_boundary, a boundary constraint term used to limit the output results to a physically feasible range; and L_smooth, a smoothness constraint loss. λ_monotonic, λ_boundary, λ_smooth, and λ_beer_lambert are the weighting coefficients of each constraint term.

[0123] Establishment of machine learning algorithm models based on smartphone image processing.

[0124] In practice, a series of CHIKV positive and negative samples with known concentrations are first prepared. Images of the immunochromatographic test strips are then captured using a smartphone under controlled ultraviolet light. Image processing algorithms are used to automatically locate and crop the detection line (T-line) region, extracting the average pixel intensity of its red (R), green (G), and blue (B) channels as feature input vectors. An artificial neural network (ANN) is employed as the core algorithm framework. To ensure the physical interpretability of the model predictions and prevent overfitting, this invention introduces Dropout regularization and L2 weight decay mechanisms for synergistic adjustment. Figure 10 As shown in Figure A, the model's validation accuracy gradually increases with the number of training iterations (Epochs). After approximately 50 Epochs, the accuracy stabilizes, indicating that the model has converged and overfitting has not occurred. The optimal combination of hyperparameters determined through grid search and cross-validation enables the model to achieve a classification accuracy of 99% for samples of different concentrations on the independent test set (e.g., ...). Figure 10 As shown in B), its application potential in rapid biosensing quantitative detection has been fully verified.

[0125] To verify the model's practical superiority, PINN-DMC was benchmarked against standard ImageJ analysis under three concentration gradients (low, medium, and high) and high background scenarios. The signal readout process for analysis and comparison is as follows: Figure 11As shown in Figure 11A. Under all conditions, PINN-DMC consistently achieves a higher signal-to-noise ratio than ImageJ (as shown in Figure 11B), with the most significant advantage in high-background samples. By employing adaptive computation rather than fixed ROI selection, this algorithm effectively suppresses noise interference, demonstrating superior sensitivity to weak signals and robustness to complex matrices. Furthermore, PINN-DMC significantly reduces measurement variability, consistently maintaining a lower CV value compared to ImageJ's higher CV value due to the subjectivity of manual operation (e.g., ...). Figure 11 (As shown in C). Therefore, the framework combining immunochromatographic test strips with the PINN-DMC algorithm provides an innovative paradigm for quantitative immunochromatographic detection.

[0126] 1. Clinical diagnostic efficacy assessment

[0127] This study collected 293 clinical serum samples, including 139 CHIKV-positive samples confirmed by qRT-PCR and 154 negative samples, to verify the effectiveness of the platform in practical clinical applications. This invention compared the diagnostic performance of two modes: "manual visual interpretation" and "AI intelligent analysis." Qualitative judgment relies on the presence or absence of a fluorescent band on the T-line of the test strip observed by the testing personnel. Results showed (e.g.) Figure 12 As shown in Figure A), a total of 121 positive samples and 147 negative samples were identified. Compared with the PCR gold standard, there were 18 false negatives (FN) and 7 false positives (FP). The calculated clinical sensitivity was 87.05% (121 / 139), and the specificity was 95.45% (147 / 154). This indicates that it is difficult to detect weak fluorescence signals in low-concentration samples with the naked eye, which can easily lead to missed diagnoses. The AI ​​algorithm developed in this invention was used to extract and quantify signals from the same batch of test strips, and classify them based on preset thresholds. The results show (as shown in Figure A) Figure 12 As shown in B), the AI ​​successfully identified 131 positive samples and 148 negative samples. False negatives were significantly reduced to 8, and false positives to 6. Clinical sensitivity increased to 94.24% (131 / 139), and specificity increased to 96.10% (148 / 154). Compared to traditional manual interpretation, AI-assisted analysis significantly reduced the rates of missed diagnoses and misdiagnoses, especially demonstrating superior signal capture capabilities in the detection of samples with low viral loads.

[0128] The above description is only for illustrating the technical concept and features of the present invention, and its purpose is to enable those skilled in the art to understand the content of the present invention and implement it accordingly. However, the scope of protection of the present invention is not limited thereto. Equivalent substitutions or changes made to the technical solution and inventive concept of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for quantitative detection of chipore Kenyan virus combining a silicon-shell-enhanced probe and an AI algorithm, characterized in that, Includes the following steps: Step S1: Construct a GO@TQD@Si fluorescent nanoprobe with a core-shell structure. The probe uses graphene oxide nanosheets as a two-dimensional framework, loads multiple quantum dots through electrostatic self-assembly, and coats the outermost layer with a silica shell. Finally, a specific antibody against chikungunya virus is coupled to the surface of the shell. Step S2: The biological sample to be tested is dropped onto the immunochromatographic test strip containing the fluorescent nanoprobe prepared in step S1; if the sample contains chikungunya virus antigen, it forms a "probe-antigen-antibody" double antibody sandwich complex with the fluorescent nanoprobe and the capture antibody immobilized on the nitrocellulose membrane, which is enriched at the detection line and emits a fluorescent signal. Step S3: Under ultraviolet excitation light source, use the camera of the mobile terminal device to capture the original image of the test strip detection area; Step S4: Input the captured image into a pre-built deep multi-constraint network model of artificial intelligence algorithm. This model extracts the RGB color features of the test strip detection line area, combines them with the physical constraint loss function to calculate, and outputs the concentration value of chikungunya virus in the sample.

2. The method for quantitative detection of chipore Kenyan virus combining a silicon-shell enhanced probe and an AI algorithm according to claim 1, characterized in that, The preparation method of the GO@TQD@Si fluorescent nanoprobe described in step S1 specifically includes the following steps: (1) Add graphene oxide substrate to polyethyleneimine aqueous solution and use positively charged polyethyleneimine as a linker to form GO@PEI with positively charged surface on graphene oxide surface; add GO@PEI to negatively charged carboxylated CdSe / ZnS quantum dot solution and alternately adsorb negatively charged carboxylated CdSe / ZnS quantum dots on graphene oxide surface. Repeat the "PEI adsorption-washing-QD adsorption" cycle 3 times to form GO@TQD complex with three layers of quantum dot load. (2) GO@TQD was silanized by dispersing GO@TQD nanosheets in a mixed solvent of ethanol / water / ammonia, adding tetraethyl orthosilicate as a silicon source, and carrying out a hydrolysis-condensation reaction under alkaline conditions to grow a uniform silica shell on the core surface in situ. (3) The silica shell was modified by amino-modification with 3-aminopropyltriethoxysilane, followed by reaction with succinic anhydride to introduce carboxyl groups. Finally, the carboxyl groups were activated by EDC / NHS (1-(3-dimethylaminopropyl)-3-ethylcarbodiimide hydrochloride / N-hydroxysuccinimide) to couple anti-Chikungunya virus specific antibody on the silica shell surface, thus achieving covalent coupling with the antibody.

3. The method for quantitative detection of chipore Kenyan virus combining a silicon-shell enhanced probe and an AI algorithm according to claim 2, characterized in that, The concentration of polyethyleneimine in step (1) is 1 mg / mL to 5 mg / mL; the amount of tetraethyl orthosilicate added in step (2) is 5-15 μL, and the thickness of the silica shell is between 15 nm and 45 nm; the amount of 3-aminopropyltriethoxysilane added in step (3) is 0.1-0.8% (v / v), and the amount of specific antibody added is 5-15 μg / mL.

4. The method for quantitative detection of chipore Kenyan virus combining a silicon-shell enhanced probe and an AI algorithm according to claim 3, characterized in that, The concentration of the capture antibody immobilized on the nitrocellulose membrane in step S2 is 0.8-1.6 mg / mL.

5. The method for quantitative detection of chipore Kenyan virus combining a silicon-shell enhanced probe and an AI algorithm according to claim 4, characterized in that, The original image acquisition and processing method described in step S3 is as follows: the test strip is photographed in a dark environment using a smartphone camera, the detection line area is automatically located through an image processing algorithm, the region of interest is cropped out, and the average gray value of the pixels in the R, G, and B channels within the region is calculated as the input feature vector.

6. The method for quantitative detection of chipore Kenyan virus combining a silicon-shell enhanced probe and an AI algorithm according to claim 5, characterized in that, In step S4, the method for constructing a deep multi-constraint network model for artificial intelligence algorithms includes the following steps: (1) Establishing a dataset: Prepare standard samples of chikungunya virus with different known concentrations, and obtain the corresponding image data according to steps S2 and S3 in claim 1; (2) Feature extraction: Crop the region of interest of the T-line in the image and extract the pixel intensity values ​​of the R, G, and B channels; (3) Model training: A neural network is used as the regression model, with RGB features as input and virus concentration as output; (4) Model optimization: Physical constraints are introduced into the loss function to improve the model’s generalization ability and prediction accuracy under small sample data.

7. The method for quantitative detection of chipore Kenyan virus combining a silicon-shell enhanced probe and an AI algorithm according to claim 6, characterized in that, The model training uses a physical information loss function, the formula of which includes the following constraints: L_total =L MSE +λ_monotonic·L_monotonic + λ_boundary·L_boundary + λ_smooth·L_smooth + λ_beer_lambert·L_beer_lambert Where L MSE The mean squared error between the predicted and actual concentrations is defined by: L_monotonic, a monotonicity constraint term used to force the model output signal to increase monotonically with increasing concentration; L_beer_lambert, a Beer-Lambert law constraint term used to constrain the logarithmic linear relationship between fluorescence intensity and concentration; L_boundary, a boundary constraint term used to limit the output results to a physically feasible range; and L_smooth, a smoothness constraint loss. λ_monotonic, λ_boundary, λ_smooth, and λ_beer_lambert are the weighting coefficients of each constraint term.

8. An immunochromatographic test strip for quantitative detection, using the GO@TQD@Si fluorescent nanoprobe prepared according to claim 2 as a signal tag.

9. A detection system employing the method for quantitative detection of chikungunya virus combining a silicon-shell enhanced probe and an AI algorithm as described in claim 1, for quantitative detection of chikungunya virus, comprising: Test kits, vacuum-sealed GO@TQD@Si fluorescent immunochromatographic test strips and sample diluents, and intelligent analysis software; The intelligent analysis software is an application installed on a smartphone that integrates the PINN-DMC algorithm model for real-time display of image acquisition, feature extraction, and concentration calculation results.

Images