Multi-modal detection method for pathogenic bacteria based on nanocomposites and neural networks

Abstract

The application relates to the technical field of food safety detection, and discloses a pathogenic bacteria multi-modal detection method based on a nano-composite material and a neural network, which comprises the following steps: preparing a nano-composite material based on the metal chelating property and interface regulation technology of cuttlefish ink source nanoparticles; constructing a dual-mode immunosensing detection platform based on the nano-composite material and a lateral flow immunoassay detection platform; preliminarily detecting pathogenic bacteria based on the dual-mode immunosensing detection platform to obtain dual-mode detection signals; extracting multi-modal signal characteristics of the dual-mode detection signals based on a fusion architecture of a multilayer perceptron and a radial basis function; and analyzing the multi-modal signal characteristics based on a neural network algorithm and outputting detection results and prediction results of the pathogenic bacteria. The application combines nano-composite engineering, multi-modal signal transduction and machine learning organically, breaks through the bottleneck of traditional detection technology, and realizes dual-mode detection of rapid and accurate detection and risk assessment of foodborne pathogenic bacteria.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of food safety detection, in particular to a pathogenic bacteria multi-modal detection method based on nanocomposites and neural networks. BACKGROUND

[0002] Foodborne pathogenic bacteria (such as Salmonella) are the main pathogens that cause global food poisoning events. The existing detection methods (such as culture method, PCR technology) have problems such as long cycle (24-72 hours), high dependence on equipment, complex operation, etc., which are difficult to meet the needs of on-site rapid screening. At the same time, although the lateral flow immunochromatography technology (LFIA) has become the mainstream point-of-care testing (POCT) tool due to its portability, its inherent defects still limit its practical application: (1) insufficient sensitivity, relying on traditional markers such as colloidal gold, which is difficult to meet the detection needs of trace pathogens; (2) lack of quantitative ability, only visual interpretation through test strip color intensity, lacking precise quantitative analysis function; (3) single result interpretation, unable to correlate detection data with risk level, unable to provide decision support information.

[0003] Nanoparticles (such as Prussian blue nanoparticles, PBNPs) are widely used in LFIA signal amplification due to their peroxidase-like activity. However, there is an inherent contradiction between nanoparticle activity and stability: reducing the size of PBNPs can improve catalytic activity, but it leads to an increase in surface energy and exacerbates particle aggregation. Existing stabilization strategies (such as high molecular coating) often sacrifice accessibility to active sites, limiting the space for sensitivity improvement.

[0004] Existing technologies focus on optimizing detection sensitivity, while ignoring data depth analysis and risk prediction capabilities. Supervised learning in machine learning can achieve "end-to-end" prediction from detection data to result output through neural network algorithm pre-model training, which has the ability to analyze test results.

[0005] Therefore, it is necessary to design a pathogenic bacteria multi-modal detection method based on nanocomposites and neural networks to solve the above problems. SUMMARY

[0006] Therefore, the present application proposes a pathogenic bacteria multi-modal detection method based on nanocomposites and neural networks, which aims to break through the bottleneck of traditional detection technology by organically combining nanocomposite engineering, multi-modal signal transduction and machine learning, and to realize rapid and accurate detection of foodborne pathogenic bacteria and risk assessment, providing a new solution for food and public health safety risk prevention and control.

[0007] In one aspect, the present application proposes a pathogenic bacteria multi-modal detection method based on nanocomposites and neural networks, comprising:

[0008] Preparation of nanocomposites based on metal chelation and interface regulation technology of cuttlefish ink-derived nanoparticles;

[0009] Based on the performance detection data of the nanocomposites and the preset qualified performance threshold of the nanocomposites, it is judged whether the nanocomposites are usable, and a dual-mode immunosensing detection platform is constructed based on the judgment result and a lateral flow immunoassay detection platform;

[0010] Based on the dual-mode immunosensing detection platform, a preliminary detection of pathogenic bacteria is performed to obtain a dual-mode detection signal;

[0011] Based on the fusion architecture of multilayer perceptron and radial basis function, the dual-mode detection signal is analyzed, and multi-modal signal features are extracted;

[0012] Based on the end-to-end analysis model constructed by the neural network algorithm, the multi-modal signal features are analyzed, and the detection result and the prediction result of the pathogenic bacteria are output according to the analysis result.

[0013] Further, the preparation of nanocomposites based on the metal chelation and interface regulation technology of cuttlefish ink-derived nanoparticles includes:

[0014] Mechanically stirring and centrifugally purifying cuttlefish ink to obtain a cuttlefish ink-derived nanoparticle dispersion;

[0015] Adding cuttlefish ink-derived nanoparticles to a Fe3+ mixed solution to inhibit the growth of Prussian blue nanoparticle crystal nuclei through competitive metal chelation;

[0016] The dropwise addition rate of the cuttlefish ink-derived nanoparticle dispersion is 160 μL / 30 s;

[0017] Stirring at room temperature for 6 hours completes the interface self-assembly;

[0018] After centrifugal washing, the nanocomposites are obtained by redispersion in an aqueous solution.

[0019] Further, the judgment of whether the nanocomposites are usable based on the performance detection data of the nanocomposites and the preset qualified performance threshold of the nanocomposites includes:

[0020] Detecting the performance data of the colorimetric performance, antibody coupling efficiency, and catalytic characteristics of the nanocomposites;

[0021] If the performance data reaches the preset qualified performance threshold of the nanocomposites, it is judged that the nanocomposites are usable for detecting pathogenic bacteria;

[0022] If the performance data does not reach the preset qualified performance threshold of the nanocomposites, it is judged that the nanocomposites are not usable for detecting pathogenic bacteria.

[0023] Further, the double-mode immunosensing detection platform is constructed based on the judgment result and the lateral flow immunoassay detection platform, comprising:

[0024] In the judgment of the nanocomposite material for detecting pathogenic bacteria, the nanocomposite material is combined with the lateral flow immunoassay detection platform to construct a double-mode immunosensing detection platform.

[0025] Two parallel detection line components coated with antibody probes are arranged on the test strip of the lateral flow immunoassay detection platform, which are T1 detection line and T2 quality control line.

[0026] The nanocomposite material labeled pathogenic bacteria antibody is arranged on the conjugate pad of the lateral flow immunoassay detection platform.

[0027] The nanocomposite material is co-incubated with the sample to be tested to obtain a test complex.

[0028] The test complex is combined with the nanocomposite material labeled pathogenic bacteria antibody on the conjugate pad to form a test mixture.

[0029] The test mixture flows to the two detection lines on the lateral flow immunoassay platform to form a double-mode immunosensing detection platform.

[0030] Further, two parallel detection line components coated with antibody probes are arranged on the test strip of the lateral flow immunoassay detection platform, which are T1 detection line and T2 quality control line, comprising: the T1 detection line is coated with specific antibodies against pathogenic bacteria, and the T2 quality control line is coated with quality control antibodies against pathogenic bacteria; the T1 detection line is further provided with a colorimetric signal capture area and a catalytic reaction signal amplification area.

[0031] Further, the double-mode immunosensing detection platform is used for preliminary detection of pathogenic bacteria to obtain a double-mode detection signal, comprising:

[0032] The test mixture first flows to the T1 detection line, and the nanocomposite material is aggregated on the T1 detection line to form a black-brown strip visible to the naked eye, forming a colorimetric mode detection platform, based on which the pathogenic bacteria are visually qualitatively detected, and a colorimetric detection signal is generated.

[0033] 3,3',5,5'-tetramethylbenzidine substrate is added to the catalytic reaction signal amplification area of the T1 detection line for catalytic reaction, and the test mixture flows to the T2 quality control line to form a catalytic mode detection platform.

[0034] The catalytic signal recorded by the mobile device and the catalytic data of the catalytic mode detection platform are used for quantitative detection of the pathogenic bacteria, and a catalytic detection signal is generated.

[0035] Further, the fusion architecture based on the multi-layer perceptron and the radial basis function analyzes the dual-mode detection signal and extracts multi-modal signal features, including:

[0036] The multi-layer perceptron extracts signal intensity gradient change features of the catalytic detection signal to generate global structured signal features.

[0037] The radial basis function network extracts mutation points and boundary features of the colorimetric detection signal to generate local discriminative signal features.

[0038] Further, the end-to-end analysis model based on the neural network algorithm analyzes the multi-modal signal features, and outputs the detection result and the prediction result of the pathogenic bacteria according to the analysis result, including:

[0039] Based on the neural network algorithm, a data training set and a data validation set are constructed.

[0040] Based on the data training set and the data validation set, a classification prediction model and a quantitative prediction model are constructed.

[0041] Based on the classification prediction model, the local discriminative signal features are analyzed to obtain the detection result.

[0042] Based on the quantitative prediction model, the global structured signal features are analyzed to obtain the prediction result.

[0043] Further, the detection result is the concentration value of the pathogenic bacteria, and the prediction result is the risk assessment level of the pathogenic bacteria, which is divided into potential risk, medium risk and high risk.

[0044] Compared with the prior art, the beneficial effects of the present application are:

[0045] The pathogenic bacteria multi-modal detection method based on the nano-composite material and the neural network of the present application uses cuttlefish ink source nanoparticles as a carrier, utilizes the metal chelation characteristics of the cuttlefish ink source nanoparticles, generates a low-crystallinity Prussian blue nanoparticle composite system, i.e. a nano-composite material, through an interfacial self-assembly technology, breaks through the technical bottleneck that the catalytic activity and stability of traditional Prussian blue nanoparticles cannot be compatible. The nano-composite material obtained by the present application and the lateral flow immunochromatography detection platform significantly improve the detection sensitivity; the present application combines the lateral flow immunochromatography detection platform and the neural network analysis and prediction system to realize the dual-mode biosensing detection and risk warning of pathogenic bacteria.

[0046] The application combines nanocomposite engineering, multi-modal signal transduction and machine learning organically, realizes end-to-end analysis through integrated neural network algorithm, directly correlates original data and risk level, provides basis for measures to be taken after detection, enhances explainability, can improve the prediction accuracy of pathogenic bacteria, realizes rapid and accurate detection and prediction of foodborne pathogenic bacteria, and provides a new solution with advanced technology and reliable results for food and public health safety risk prevention and control.

[0047] In another aspect, the application also provides an application of the multi-modal detection method of pathogenic bacteria based on nanocomposite and neural network according to any one of claims 1-9. The application includes that the detection method is applied to on-site rapid detection and pollution tracing of foodborne salmonella typhimurium in food processing, cold chain transportation terminal or clinical sample.

[0048] It can be understood that the multi-modal detection method of pathogenic bacteria based on nanocomposite and neural network and the application have the same beneficial effects, which will not be repeated. BRIEF DESCRIPTION OF DRAWINGS

[0049] Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The accompanying drawings are included to provide a description of the preferred embodiments and are not intended to limit the scope of the application. Moreover, the same reference numerals are used throughout the same figures. In the drawings:

[0050] Figure 1 The flow chart of the multi-modal detection method of pathogenic bacteria based on nanocomposite and neural network provided by the embodiment of the application;

[0051] Figure 2 The nanocomposite synthesis mechanism and characterization diagram provided by the embodiment of the application;

[0052] Figure 3 The performance detection and evaluation comparison diagram of the nanocomposite and prussian blue nanoparticles provided by the embodiment of the application;

[0053] Figure 4 The schematic diagram of the test strip and the binding pad of the lateral flow immunoassay detection platform provided by the embodiment of the application. DETAILED DESCRIPTION

[0054] Exemplary embodiments of the present disclosure will be described in greater detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the drawings, it is understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that the present disclosure can be more thoroughly understood, and the scope of the present disclosure can be accurately conveyed to those skilled in the art. It should be noted that the embodiments in the present disclosure and the features in the embodiments can be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0055] Foodborne pathogens such as Salmonella are the main pathogens causing global food poisoning events. The existing detection methods (such as culture method, PCR technology) have problems such as long cycle (24-72 hours), high dependence on equipment, complex operation, etc., which are difficult to meet the needs of on-site rapid screening. At the same time, although the lateral flow immunochromatography technology (LFIA) has become the mainstream point-of-care testing (POCT) tool due to its portability, its inherent defects still limit its practical application: (1) insufficient sensitivity, relying on traditional markers such as colloidal gold, which is difficult to meet the detection needs of trace pathogens; (2) lack of quantitative ability, only visual interpretation by test strip color intensity, lacking precise quantitative analysis function; (3) single result interpretation, unable to correlate detection data with risk level, unable to provide decision support information.

[0056] Nanoparticles (such as Prussian blue nanoparticles, PBNPs) are widely used in LFIA signal amplification due to their peroxidase-like activity. However, there is an inherent contradiction between nanoparticle activity and stability: reducing the size of PBNPs can improve catalytic activity, but it will lead to an increase in surface energy and exacerbate particle aggregation. Existing stabilization strategies (such as polymer coating) often sacrifice accessibility to active sites, limiting the space for sensitivity improvement.

[0057] Existing technologies focus on optimizing detection sensitivity, while ignoring data depth analysis and risk prediction capabilities. Supervised learning in machine learning can achieve "end-to-end" prediction from detection data to result output through neural network algorithms, which has the ability to analyze test results.

[0058] Therefore, the present application proposes a pathogenic bacteria multi-modal detection method based on nanocomposites and neural networks to solve the above problems.

[0059] Referring to Figure 1 In some embodiments of the present application, the pathogenic bacteria multi-modal detection method based on nanocomposites and neural networks comprises:

[0060] S1, preparing a nanocomposite based on the metal chelating property of cuttlefish ink-derived nanoparticles and interface regulation technology.

[0061] S2, judging whether the nanocomposite is usable based on performance detection data of the nanocomposite and a preset qualified performance threshold value of the nanocomposite, and constructing a dual-mode immunosensing detection platform based on a judgment result and a lateral flow immunodetection platform.

[0062] S3, performing preliminary detection on pathogenic bacteria based on the dual-mode immunosensing detection platform to obtain a dual-mode detection signal.

[0063] S4, analyzing the dual-mode detection signal based on a fusion architecture of a multilayer perceptron and a radial basis function, and extracting a multi-modal signal feature.

[0064] S5, analyzing the multi-modal signal feature based on an end-to-end analysis model constructed based on a neural network algorithm, and outputting a detection result and a prediction result of the pathogenic bacteria according to an analysis result.

[0065] It can be seen that the application organically combines nanomaterial engineering, multi-modal biosensing and machine learning to form a complete technical chain of "material preparation-signal detection-intelligent analysis".

[0066] It can be seen that the dual-mode detection of the application takes into account the qualitative convenience and accuracy, and solves the single signal defect of the traditional lateral flow immunodetection platform.

[0067] The neural network of the application realizes intelligent conversion from the original signal to the risk level through end-to-end analysis, and improves the accuracy and interpretability of the detection result.

[0068] Referring to Figures 1-2 It can be seen that the application prepares the nanocomposite based on metal chelation of cuttlefish ink source nanoparticles and interface regulation technology, including:

[0069] S11, obtaining a cuttlefish ink source nanoparticle dispersion liquid by mechanically stirring and centrifugal purification of cuttlefish ink;

[0070] S12, adding the cuttlefish ink source nanoparticles in a Fe3+ mixed solution to inhibit the growth of Prussian blue nanoparticle crystal nuclei through competitive metal chelation;

[0071] S13, the dropwise addition rate of the cuttlefish ink source nanoparticle dispersion liquid is 160 μL / 30 s;

[0072] S14, completing interface self-assembly under stirring for 6 hours at room temperature;

[0073] S15, after centrifugal washing, redispersing in an aqueous solution to obtain the nanocomposite.

[0074] Specifically, the purification of the cuttlefish ink nanoparticles was performed by a reagent-free process, i.e. sequential water washing and centrifugation. The specific steps were as follows: first, food-grade cuttlefish ink was diluted with deionized water at a ratio of more than 15:1 by volume, and thoroughly stirred to ensure uniform mixing. Subsequently, the resulting suspension was centrifuged at a speed of 10,000 rpm for 15 minutes for five times in succession, and fresh deionized water was replaced after each centrifugation. The final purified cuttlefish ink nanoparticles were stored for later use after being treated under vacuum freeze-drying conditions for 12 hours.

[0075] The nanocomposite was prepared by depositing Prussian blue nanoparticles on the surface of the cuttlefish ink nanoparticles through an interfacial self-assembly technique in solution. The specific synthesis process was as follows: first, 5 mg of cuttlefish ink nanoparticles were dispersed in 10 ml of deionized water to form solution A; at the same time, solution B was prepared, which contained FeCl3(25 mg) and K3Fe(CN)6(28.8 mg) dissolved in 5 ml of deionized water. Subsequently, 0.8 ml of solution B was diluted with deionized water to a total volume of 10 ml, and was slowly injected into solution A at a controlled rate (160 μl / 30 s) by a syringe pump at ambient temperature (25 °C) while maintaining continuous stirring. After 6 hours of reaction, the system was centrifuged at a speed of 10,000 rpm for 10 minutes and washed five times, and the target product nanocomposite was finally obtained.

[0076] It can be understood that the nanocomposite is a naturally occurring nanoparticle mainly composed of melanin, which provides multiple biosorption sites for metal ions. Therefore, the synthesis of the nanocomposite can be attributed to the preferential chelation of Fe3+ on the surface of the cuttlefish ink nanoparticles, followed by interfacial self-assembly with [Fe(CN)6]4- to form the nanocomposite deposited on the surface of the cuttlefish ink nanoparticles. The degree of interfacial growth of the nanocomposite can be adjusted by adjusting the concentration of various precursor solutions. For example, Figure 2 The TEM (Transmission Electron Microscope) images of A and 2B show that the surface of individual CINPs is relatively smooth, while the surface of the assembled composite is enhanced in particle size, and small-sized PBNPs are clearly visible in a curved square morphology. CINPs are mainly distributed between 120-140 nm, with an average diameter of 131.75 nm. In contrast, PBNPs formed by interfacial self-assembly have a more uniform size distribution, with an average diameter of 26.25 nm, referred to as PBNPs (30) . This phenomenon can be related to rapid nucleation during the self-assembly process. In addition, PBNPs of larger size (designated PBNPs (200) ) were synthesized in the absence of CINPs. TEM-based electron diffraction analysis Figure 2 C) shows that PBNPs (200)showed single crystal structure with clear lattice fringes, indicating high crystallinity, while the self-assembled PBNPs (30) showed low crystallinity. Fourier transform infrared (FTIR) spectra Figure 2 D) showed a broad peak around 3427 cm-1and a distinct peak at 1600 cm-1, corresponding to the O-H stretching vibration and C-O bond vibration of the hydroxyl group in CINPs, respectively. The peak at 2090 cm-1was attributed to the C≡N stretching vibration of cyanide-bridged, which is characteristic of PBNPs. The comparative analysis of CPBNPs and PBNPs (30) confirmed the successful synthesis of PBNPs (30) on the surface of CINPs. CPBNPs showed enhanced C-O vibration signals, which might be attributed to the presence of CINPs.

[0077] It can be seen that the application inhibits the crystallization rate of Prussian blue nanoparticles by interface regulation, forms an amorphous structure, and obtains a nanocomposite with both activity and stability. Moreover, the application uses cuttlefish ink sac waste to prepare Prussian blue nanoparticles, conforms to the principle of circular economy, reduces material cost, and reduces environmental pollution.

[0078] Referring to Figure 1 and 3 , in some embodiments of the application, whether the nanocomposite is usable is determined based on performance detection data of the nanocomposite and a preset qualified performance threshold of the nanocomposite, including:

[0079] detecting performance data of colorimetric performance, antibody coupling efficiency, and catalytic characteristics of the nanocomposite;

[0080] if the performance data reaches the preset qualified performance threshold of the nanocomposite, it is determined that the nanocomposite is usable for detecting pathogenic bacteria;

[0081] if the performance data does not reach the preset qualified performance threshold of the nanocomposite, it is determined that the nanocomposite is not usable for detecting pathogenic bacteria.

[0082] Specifically, under the condition of continuous dilution, the colorimetric performance of the nanocomposite and Prussian blue nanoparticles at the same concentration is compared. As shown in Figure 3 A, compared with Prussian blue nanoparticles, the nanocomposite reaches the visual signal threshold that is not detectable at a lower concentration. In addition, the nanocomposite shows superior colorimetric signal intensity at each dilution gradient. These findings show that the composite structure of cuttlefish ink nanoparticles and Prussian blue nanoparticles in the nanocomposite enhances the colorimetric responsiveness and slows down the color development rate at the same dilution level.

[0083] It can be understood that this phenomenon may be related to the broad absorption characteristics of cuttlefish ink nanoparticles in the visible light region.

[0084] Specifically, in order to ensure the specific recognition ability to the target, the coupling relationship between the nanocomposite and the antibody recognition molecule must be considered when constructing the sensing platform. Gold nanoparticles (AuNPs) were selected as a control to evaluate the antibody coupling efficiency of the nanocomposite. Three dose groups (1, 5, and 10 μg) were coupled with AuNPs and the nanocomposite, respectively. After centrifugation, the residual antibody in the supernatant was quantified by enzyme-linked immunosorbent assay (ELISA), and the coupling efficiency was calculated. As shown in FIGS. Figure 3 As shown in FIGS. B and 3C, the conjugation efficiencies of AuNPs under the three gradients were 99.82%, 97.18%, and 83.14%, respectively, while the conjugation efficiencies of the nanocomposite under the corresponding conditions were 99.94%, 94.90%, and 80.41%, respectively. Complete binding of the two signal tags to the antibody was observed at low antibody concentrations, while excessive antibody may induce binding saturation, resulting in an increase in the content of free antibody.

[0085] It can be understood that the nanocomposite exhibits a binding efficiency comparable to AuNPs for the antibody recognition molecule in the immunoassay, which may be attributed to the presence of cuttlefish ink nanoparticles. Studies have shown that in cuttlefish ink nanoparticles, melanin imparts various functional groups to the nanocomposite, enabling it to form covalent or non-covalent interactions with various molecules.

[0086] Specifically, the catalytic properties of the nanocomposite were detected. First, the source of catalytic activity was studied by comparing the color development ability of the supernatant after centrifugation with that of the nanocomposite precipitate. As shown in FIG. Figure 3 D, only the substrate catalyzed by the nanocomposite precipitate showed a distinct characteristic absorption peak in the ultraviolet-visible spectrum, confirming that the catalytic function originated from the nanocomposite composite nanomaterial rather than ion leaching. In addition, two color substrates, TMB and OPD, were selected to evaluate the catalytic performance of the nanocomposite. As shown in FIG. Figure 3 E, the signal intensity obtained with TMB as the catalytic substrate was significantly higher than that based on oxidase activity. Therefore, TMB was used as the optimal catalytic substrate for subsequent experiments. Subsequent studies on the enzymatic specificity of the nanocomposite ( Figure 3 F) showed that when the nanocomposite only interacted with the TMB substrate, the color development signal was weak, while the color intensity was significantly enhanced in the presence of H2O2. This observation indicates that the nanocomposite exhibits significantly better peroxidase (POD)-like activity compared to similar oxidase functions. Based on these findings, we systematically optimized the pH conditions for nanocomposite-mediated catalysis ( Figure 3(G). The catalytic system exhibits the highest activity under weakly alkaline conditions, with significant performance degradation under strongly acidic or alkaline environments. Therefore, pH 5 was chosen as the optimal reaction parameter for subsequent experimental verification. To quantitatively characterize the enzymatic performance, we measured the Michaelis-Menten kinetic parameters (Km and Vmax) using TMB and H2O2 as substrates, respectively. Comparative analysis of these parameters provides important insights into optimizing the catalytic affinity and turnover capacity of the nanocomposite material under reaction conditions. When using TMB as a substrate, the Km of the nanomaterial was 0.49 mM, and the Vmax was 31.89 × 10⁻⁶. -8 Ms -1 ( Figure 3 H). When H2O2 is used as the substrate, these parameters become Km = 21.75 mM and Vmax = 27.7 × 10⁻⁶, respectively. -8 Ms -1 ( Figure 3 I). The significantly lower Km value of TMB indicates a higher substrate binding affinity, while the Vmax value indicates an effective catalytic rate for both substrates.

[0087] It is understandable that nanocomposites possess superior catalytic activity and have significant potential as signaling tags in catalytic applications.

[0088] It can be seen that multi-dimensional performance testing ensures that nanocomposite materials meet testing requirements, avoids testing errors caused by material defects, and improves the reliability of the testing method. This application provides a quantitative standard for material screening by setting a pre-defined pass threshold, which facilitates quality control and batch-to-batch consistency verification in industrial production.

[0089] Reference Figure 1 and 4 As shown in some embodiments of this application, the construction of a dual-mode immunosensing detection platform based on the judgment result and the lateral flow immunoassay detection platform includes:

[0090] When it is determined that the nanocomposite material can be used to detect pathogenic bacteria, the nanocomposite material is combined with the lateral flow immunoassay platform to construct a dual-mode immunosensing detection platform;

[0091] Among them, the test strip of the lateral flow immunoassay platform is equipped with two parallel detection line components coated with antibody probes, which are set as T1 detection line and T2 control line.

[0092] The nanocomposite material-labeled pathogenic bacteria antibody is configured on the binding pad of the lateral flow immunoassay detection platform;

[0093] The nanocomposite material was co-incubated with the sample to be tested to obtain the complex to be tested;

[0094] The test complex is combined with the nano-composite material labeled pathogenic bacteria antibody on the binding pad to form a test mixture;

[0095] The test mixture flows to two detection lines on the lateral flow immunochromatographic platform to form a dual-mode immunosensor detection platform.

[0096] Referring to Figure 1 and 4 In some embodiments of the present application, the test strip of the lateral flow immunoassay detection platform is provided with two parallel detection line assemblies coated with antibody probes, which are T1 detection line and T2 quality control line, comprising: the T1 detection line is coated with specific antibodies against pathogenic bacteria, and the T2 quality control line is coated with quality control antibodies against pathogenic bacteria; the T1 detection line is further provided with a colorimetric signal capture area and a catalytic reaction signal amplification area.

[0097] Referring to Figure 1 and 4 In some embodiments of the present application, the dual-mode immunosensor detection platform is used for preliminary detection of pathogenic bacteria to obtain dual-mode detection signals, comprising:

[0098] The test mixture first flows to the T1 detection line, and the nano-composite material is aggregated on the T1 detection line to form a black-brown strip visible to the naked eye, constituting a colorimetric mode detection platform. Based on the colorimetric mode detection platform, the pathogenic bacteria are visually qualitatively detected, and a colorimetric detection signal is generated;

[0099] The 3,3',5,5'-tetramethylbenzidine substrate is added to the catalytic reaction signal amplification area of the T1 detection line for catalytic reaction, and the test mixture flows to the T2 quality control line to form a catalytic mode detection platform;

[0100] Based on the catalytic signal recorded by the mobile device and the catalytic data of the catalytic mode detection platform, the pathogenic bacteria are quantitatively detected, and a catalytic detection signal is generated.

[0101] Specifically, the T1 detection line of the test strip is coated with Salmonella monoclonal antibody (1 mg / mL) at a line drawing speed of 1 μL / cm; the T2 quality control line is coated with goat anti-mouse IgG (1 mg / mL) at a line drawing speed of 1 μL / cm; the catalytic liquid is 20 mM TMB 4 μL and 3% H2O2 28 μL.

[0102] Specifically, 100 μL of different concentrations (10-10 7The T1 detection line is divided into two areas, which realizes "qualitative-quantitative" signal separation, facilitates subsequent feature extraction and model analysis, and improves the reliability of the detection result.

[0103] It can be understood that the T1 detection line is divided into two areas, which realizes "qualitative-quantitative" signal separation, facilitates subsequent feature extraction and model analysis, and improves the reliability of the detection result.

[0104] It can be understood that the nanocomposite material is aggregated on the T1 detection line to form a black-brown strip, and a colorimetric signal is generated by visual observation or image acquisition, so that the presence or absence of pathogenic bacteria can be qualitatively determined.

[0105] It can be understood that the nanocomposite material catalytic reaction generates a blue product by adding 3,3',5,5'-tetramethylbenzidine substrate to the T1 detection line catalytic reaction area, and the catalytic signal (absorbance value) is recorded by a mobile device (such as a smart phone), and a quantitative detection signal is generated by combining a standard curve.

[0106] It can be seen that the colorimetric mode detection platform of the present application is suitable for on-site rapid screening, and the catalytic mode detection platform is suitable for accurate quantification, and the present application can meet the needs of different scenes.

[0107] Referring to Figure 1 In some embodiments of the present application, the fusion architecture based on multi-layer perceptron and radial basis function analyzes the dual-mode detection signal and extracts multi-modal signal features, including:

[0108] The multi-layer perceptron extracts the signal intensity gradient change feature of the catalytic detection signal to generate a global structured signal feature;

[0109] The radial basis function network extracts the mutation point and boundary feature of the colorimetric detection signal to generate a local discriminative signal feature.

[0110] It can be understood that the global analysis of the multi-layer perceptron is suitable for quantitative models such as concentration prediction, and the local analysis of the radial basis function network is suitable for classification models such as existence judgment.

[0111] It can be understood that, in view of the possible visual interpretation error of the colorimetric signal, the radial basis function network reduces the influence of noise through boundary feature extraction; the multi-layer perceptron filters random fluctuations in the catalytic signal through gradient analysis, thereby improving the reliability of the features.

[0112] Referring toFigure 1 As shown, in some embodiments of the present application, the end-to-end analysis model constructed based on the neural network algorithm analyzes the multi-modal signal features, and outputs the detection result and prediction result of the pathogenic bacteria according to the analysis result, including:

[0113] The data training set and the data verification set are constructed based on the neural network algorithm;

[0114] The classification prediction model and the quantitative prediction model are respectively constructed based on the data training set and the data verification set;

[0115] The detection result is obtained by analyzing the local discriminative signal features based on the classification prediction model;

[0116] The prediction result is obtained by analyzing the global structured signal features based on the quantitative prediction model.

[0117] Specifically, the historical detection data is divided into a training set and a verification set according to a ratio of 7:3, and is input into the model after standardization processing.

[0118] Specifically, the classification prediction model inputs the local discriminative signal features and outputs the concentration value of the pathogenic bacteria; and the quantitative prediction model inputs the global structured signal features and outputs the risk assessment level.

[0119] Specifically, the parameters are adjusted through cross-validation to ensure that the R 2 of the training set is greater than or equal to 0.942, the R 2 of the test set is greater than or equal to 0.962, and the mean absolute error (MAE) is less than or equal to 0.21.

[0120] Referring to Figure 1 As shown, in some embodiments of the present application, the detection result is the concentration value of the pathogenic bacteria, and the prediction result is the risk assessment level of the pathogenic bacteria, which is divided into potential risk, medium risk and high risk.

[0121] Specifically, the specific concentration value of the pathogenic bacteria is output through the classification prediction model, and the concentration value is mapped to the three-level risk level through the quantitative prediction model:

[0122] Specifically, the potential risk means that the specific concentration value of the pathogenic bacteria is less than or equal to 10 3 CFU / mL, the risk of normal population is low, but the susceptible population needs attention;

[0123] The medium risk means that the specific concentration value of the pathogenic bacteria is 10 3 -10 5 CFU / mL, the risk of gastroenteritis is 60%, and preliminary intervention is needed;

[0124] The high risk means that the specific concentration value of the pathogenic bacteria is greater than 10 5CFU / mL, 80% risk of severe diarrhea, and emergency treatment is required.

[0125] It can be seen that the present application can provide accurate concentration values and risk levels at the same time, the former is used for scientific research and standard compliance judgment, and the latter is used for regulatory decision and public health warning, meeting the needs of different users.

[0126] It can be seen that the complex concentration data is converted into intuitive risk levels, which is convenient for non-professionals (such as food processing practitioners) to quickly understand and take action, and improves the practical application value of the detection results.

[0127] The pathogenic bacteria multi-modal detection method based on the nanocomposite and the neural network in the above embodiment breaks through the technical bottleneck that the catalytic activity and stability of the traditional Prussian blue nanoparticle cannot be compatible. The nanocomposite obtained by the present application and the lateral flow immunochromatography technology detection platform significantly improve the detection sensitivity; the present application combines the lateral flow immunochromatography technology detection platform and the neural network analysis and prediction system to realize the dual-modal biosensing detection and risk warning of pathogenic bacteria.

[0128] The present application combines nanocomposite engineering, multi-modal signal transduction and machine learning, realizes end-to-end analysis through integrated neural network algorithm, directly correlates the original data and the risk level, provides a basis for the measures that should be taken after detection, enhances the explainability, improves the prediction accuracy of pathogenic bacteria, realizes the rapid and accurate detection and prediction of foodborne pathogenic bacteria, and provides a new type of solution with advanced technology and reliable results for food and public health safety risk prevention and control.

[0129] In another preferred mode based on the above embodiment, referring to Figure 1 Figure 1 The present embodiment provides an application of the pathogenic bacteria multi-modal detection method based on the nanocomposite and the neural network according to any one of claims 1-9, including: the detection method is applied to the on-site rapid detection and pollution tracing of foodborne salmonella typhimurium in food processing links, cold chain transportation terminals or clinical samples.

[0130] It can be understood that the pathogenic bacteria multi-modal detection method based on the nanocomposite and the neural network and the application in the above embodiments of the present application have the same beneficial effects, which will not be repeated.

[0131] It should be noted that:

[0132] In the description provided herein, numerous specific details are set forth. However, it is understood that embodiments of the application can be practiced without these specific details. In some instances, well-known structures and techniques have not been shown in detail in order not to obscure the understanding of this description.

[0133] Similarly, it is to be understood that the embodiments of the application can be used in other ways, and that the application can have other embodiments, without departing from the scope of the present application. For example, the description above of exemplary embodiments of the application has, at times, been presented in the context of individual features being grouped together into single embodiments, figures, or descriptions of embodiments. However, this does not imply that the disclosed systems reflect a suggestion that the application requires more features than are explicitly recited in each claim.

[0134] More specifically, as will now be apparent to those of ordinary skill in the art, embodiments of the application are neither limited to the details of the foregoing exemplary embodiments nor do they represent the complete particulars of the application. Rather the claims, as recited below, define the scope of the application. The application is therefore to be interpreted in light of the claims and the full scope of equivalents, while employing only the principles of the application.

[0135] Furthermore, those of ordinary skill in the art will recognize that, although some embodiments described herein include certain features that are not included in other embodiments, descriptions of the different embodiments do not mean that a system that is not otherwise claimed is reflected in the drawings. Rather, the different embodiments are presented as examples of the application.

[0136] For example, in the claims below, any of the embodiments of the application can be used in any combination.

[0137] The above description is the preferred specific embodiments of the present application. However, any variations and replacements of the technical features disclosed in the present application within the scope of the present application are also included in the scope of the present application. Therefore, the scope of the present application should be defined by the scope of the claims.

Claims

1. A multi-modal detection method of pathogenic bacteria based on nanocomposites and neural networks, characterized in that, The method comprises the following steps: Preparation of nanocomposites based on the metal chelating property and interface regulation technology of cuttlefish ink-derived nanoparticles; Based on the performance detection data of the nanocomposites and the preset qualified performance threshold of the nanocomposites, it is judged whether the nanocomposites are usable, and a dual-mode immunosensing detection platform is constructed based on the judgment result and a lateral flow immunoassay detection platform; Based on the dual-mode immunosensing detection platform, a preliminary detection of pathogenic bacteria is performed, and a dual-mode detection signal is obtained; Based on the fusion architecture of multilayer perceptron and radial basis function, the dual-mode detection signal is analyzed, and multi-modal signal features are extracted; Based on an end-to-end analysis model constructed based on a neural network algorithm, the multi-modal signal features are analyzed, and a detection result and a prediction result of the pathogenic bacteria are output according to the analysis result; The preparation of nanocomposites based on the metal chelating property and interface regulation technology of cuttlefish ink-derived nanoparticles comprises the following steps: Mechanical stirring and centrifugal purification of cuttlefish ink are performed to obtain a cuttlefish ink-derived nanoparticle dispersion liquid; Cuttlefish ink-derived nanoparticles are added into a mixed Fe3+ solution, and competitive metal chelation is used to inhibit the growth of Prussian blue nanoparticle crystal nuclei; The dropwise addition rate of the cuttlefish ink-derived nanoparticle dispersion liquid is 160 μL / 30 s; Interface self-assembly is completed after stirring for 6 hours at room temperature; After centrifugal washing, the cuttlefish ink-derived nanoparticle dispersion liquid is redispersed in an aqueous solution to obtain nanocomposites. The detection result is a concentration value of the pathogenic bacteria, and the prediction result is a risk assessment level of the pathogenic bacteria, which is divided into potential risk, medium risk and high risk.

2. The nanocomposite and neural network based multi-modal detection method of pathogenic bacteria as claimed in claim 1 wherein, The judgment of whether the nanocomposites are usable based on the performance detection data of the nanocomposites and the preset qualified performance threshold of the nanocomposites comprises the following steps: Performance data of the colorimetric performance, antibody coupling efficiency and catalytic characteristics of the nanocomposites are detected; If the performance data reaches the preset qualified performance threshold of the nanocomposites, it is judged that the nanocomposites are usable for detecting pathogenic bacteria; If the performance data does not reach the preset qualified performance threshold of the nanocomposites, it is judged that the nanocomposites are not usable for detecting pathogenic bacteria.

3. The nanocomposite and neural network based multi-modal detection method of pathogenic bacteria as claimed in claim 2, wherein, The construction of a dual-mode immunosensing detection platform based on the judgment result and a lateral flow immunoassay detection platform comprises the following steps: When it is judged that the nanocomposites are usable for detecting pathogenic bacteria, the nanocomposites are combined with a lateral flow immunoassay detection platform to construct a dual-mode immunosensing detection platform; Two parallel detection line assemblies coated with antibody probes are arranged on a test strip of the lateral flow immunoassay detection platform, and are set as T1 detection line and T2 quality control line; The nanocomposite-labeled pathogenic bacteria antibodies are arranged on a conjugate pad of the lateral flow immunoassay detection platform; The nanocomposites are co-incubated with a to-be-detected sample to obtain a to-be-detected complex; The to-be-detected complex is combined with the nanocomposite-labeled pathogenic bacteria antibodies on the conjugate pad to form a to-be-detected mixture; The to-be-detected mixture flows to the two detection lines on the lateral flow immunoassay detection platform to form a dual-mode immunosensing detection platform.

4. The nanocomposite and neural network based multi-modal detection method of pathogenic bacteria as claimed in claim 3, wherein, The test strip of the lateral flow immunoassay detection platform is provided with two parallel detection line assemblies coated with antibody probes, which are T1 detection line and T2 quality control line, comprising: the T1 detection line is coated with specific antibodies for pathogenic bacteria, and the T2 quality control line is coated with quality control antibodies for pathogenic bacteria; the T1 detection line is further provided with a colorimetric signal capture area and a catalytic reaction signal amplification area.

5. The nanocomposite and neural network based multi-modal detection method of pathogenic bacteria as claimed in claim 4, wherein, The double-mode immunosensor detection platform is used for preliminary detection of pathogenic bacteria, and double-mode detection signals are obtained, comprising: The test mixture first flows to the T1 detection line, and the nano-composite material aggregates on the T1 detection line to form a black-brown strip visible to the naked eye, constituting a colorimetric mode detection platform; based on the colorimetric mode detection platform, visual qualitative detection of pathogenic bacteria is carried out, and a colorimetric detection signal is generated; 3,3',5,5'-tetramethylbenzidine substrate is added to the catalytic reaction signal amplification area of the T1 detection line for catalytic reaction, and the test mixture flows to the T2 quality control line, constituting a catalytic mode detection platform; Based on the catalytic signal recorded by the mobile device and the catalytic data of the catalytic mode detection platform, quantitative detection of pathogenic bacteria is carried out, and a catalytic detection signal is generated.

6. The nanocomposite and neural network based multi-modal detection method of pathogenic bacteria as claimed in claim 5 wherein, The fusion architecture based on multilayer perceptron and radial basis function analyzes the double-mode detection signal and extracts multi-modal signal features, comprising: The multilayer perceptron extracts the signal intensity gradient change feature of the catalytic detection signal to generate a global structured signal feature; The radial basis function extracts the mutation point and boundary feature of the colorimetric detection signal to generate a local discriminative signal feature.

7. The nanocomposite and neural network based multi-modal detection method of pathogenic bacteria as claimed in claim 6 wherein, The end-to-end analysis model constructed based on the neural network algorithm analyzes the multi-modal signal features, and outputs the detection result and the prediction result of the pathogenic bacteria according to the analysis result, comprising: A data training set and a data validation set are constructed based on the neural network algorithm; A classification prediction model and a quantitative prediction model are constructed based on the data training set and the data validation set, respectively; The local discriminative signal feature is analyzed based on the classification prediction model to obtain the detection result; The global structured signal feature is analyzed based on the quantitative prediction model to obtain the prediction result.

8. Use of a nanocomposite and neural network based multi-modal detection method of pathogenic bacteria according to any one of claims 1 to 7, characterized in that, The detection method is applied to the on-site rapid detection and pollution tracing of foodborne Salmonella typhimurium in food processing links, cold chain transportation terminals or clinical samples.

Images