Pathogenic bacteria distinguishing method and system based on surface enhanced Raman spectroscopy and AI assistance

Abstract

The invention provides a pathogenic bacterium distinguishing method and system based on surface enhanced Raman spectroscopy and AI assistance, and relates to the technical field of pathogenic bacterium analysis and detection.The pathogenic bacterium distinguishing method comprises the steps that original Raman spectrum data of a pathogenic bacterium sample capable of collecting Raman spectrum signals is obtained; preprocessing and dimensionality reduction processing are carried out on the original Raman spectrum data, characteristic peaks used for classification are screened out, and Raman shifts of the characteristic peaks comprise any one or more of 832 cm <-1 >, 1001 cm <-1 >, 1122 cm <-1 >, 1165 cm <-1 >, 1293 cm <-1 >, 1353 cm <-1 > and 1580 cm <-1 >; and inputting any one or more Raman shifts in the characteristic peaks into a pre-trained one-dimensional convolutional neural network model, analyzing the Raman spectrums of different pathogenic bacteria by the model, and outputting a classification result of the pathogenic bacteria. According to the method, the surface enhanced Raman spectroscopy technology is utilized, the AI auxiliary analysis method based on the convolutional neural network is introduced, different pathogenic bacteria are distinguished, and the method has the advantages of being easy and convenient to operate, short in consumed time, high in sensitivity and high in accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pathogenic bacteria analysis and detection, and particularly relates to a pathogenic bacteria distinguishing method and system based on surface-enhanced Raman spectroscopy and AI assistance. BACKGROUND

[0002] The statements in this section merely provide background technology related to the present disclosure and do not necessarily constitute prior art.

[0003] More than 300 million cases of serious or even fatal diseases are caused by bacterial infection or contamination every year (such as drug-resistant "super bacteria" and tuberculosis infection), resulting in more than 2 million deaths. The academic community has reached the following consensus: if early and effective diagnosis of bacterial infection can be achieved, the survival rate of patients will be greatly improved. Therefore, it is an urgent need to detect pathogenic bacteria quickly, sensitively and at low cost.

[0004] Raman spectroscopy is a label-free and non-destructive analysis technology based on molecular vibration spectroscopy, which can identify the composition of a substance through "molecular fingerprints", and is particularly suitable for rapid identification of microorganisms. Its advantage is that it can directly capture the characteristic signals of bacterial intrinsic molecules (such as lipopolysaccharide and peptidoglycan) without pre-culture or amplification, achieving minute-level single-cell recognition.

[0005] However, the traditional Raman spectroscopy technology still has the following problems in pathogenic bacteria detection: (1) The Raman signal is weak, which leads to low detection sensitivity and cannot be directly applied to complex clinical samples; (2) Complex background interference (such as blood proteins and other matrices) further reduces the signal-to-noise ratio; (3) The spectral difference is not obvious, especially between different types of gram-positive and gram-negative bacteria, lacking high-resolution classification ability; (4) Existing methods rely on large databases and complex preprocessing, which have problems such as complicated operation, long time consumption, poor reproducibility, etc. in actual application.

[0006] Therefore, there is an urgent need for a pathogenic bacteria rapid distinguishing method that can enhance Raman signals, improve classification accuracy, and have high reproducibility. SUMMARY

[0007] In order to solve the above problems, the present application provides a pathogenic bacteria distinguishing method and system based on surface-enhanced Raman spectroscopy and AI assistance, which utilizes surface-enhanced Raman spectroscopy (SERS) technology to effectively improve the Raman signal of pathogenic bacteria, and introduces an AI-assisted analysis method based on convolutional neural network to realize the distinguishing of different pathogenic bacteria, having the advantages of simple operation, short time consumption, high sensitivity and high accuracy.

[0008] The first aspect of the present application provides a pathogenic bacteria sample processing method based on surface-enhanced Raman spectroscopy, comprising: preparing a SERS substrate with a nano-bowl array structure; preparing Au@Cu 2-x Se core-shell structure nanoparticles, co-incubating them with a pathogenic bacteria sample to form a probe-pathogenic bacteria complex; attaching the probe-pathogenic bacteria complex to the SERS substrate, and after drying treatment, obtaining a pathogenic bacteria sample available for collecting Raman spectroscopy signals.

[0009] Further, the preparation of the SERS substrate comprises: magnetron sputtering of a gold layer on a silicon wafer; self-assembly of a polystyrene microsphere monolayer at a liquid-gas interface by Langmuir-Blodgett method; after electrochemical deposition of gold nanostructures, etching to remove the polystyrene microspheres to form the nano-bowl array structure.

[0010] Further, the preparation of the Au@Cu 2-x Se nanoparticles comprises: reducing chloroauric acid with sodium citrate at 160℃ to form Au cores; under the reduction condition of citric acid, introducing Cu 2+ and SeO2 into the solution containing the Au cores to form a Cu 2-x Se shell layer covering the outside of the Au cores; stabilizing the obtained core-shell structure nanoparticles by using mercapto-modified polyethylene glycol.

[0011] The second aspect of the present application provides an AI-assisted pathogenic bacteria differentiation method based on the pathogenic bacteria sample processing method of the first aspect, comprising: obtaining original Raman spectroscopy data of a pathogenic bacteria sample available for collecting Raman spectroscopy signals; preprocessing and dimensionality reduction processing the original Raman spectroscopy data to screen out characteristic peaks for classification, the Raman shift of the characteristic peaks comprising any one or several of the following: 832 cm -1 , 1001 cm -1 , 1122 cm -1 , 1165 cm -1 , 1293 cm -1 , 1353 cm -1 , and 1580 cm -1 ; In the pathogenic bacteria differentiation, a one-dimensional convolutional neural network model analyzes the pre-processed different pathogenic bacteria spectrum images input, and different types of pathogenic bacteria are differentiated based on the characteristic Raman peaks screened out by dimension reduction analysis, and a classification result is output.

[0012] Further, the preprocessing includes baseline correction using an asymmetric least squares algorithm and smoothing processing using a Savitzky-Golay algorithm.

[0013] Further, the dimension reduction processing includes principal component analysis and t-distributed stochastic neighbor embedding algorithm.

[0014] Further, the one-dimensional convolutional neural network model comprises the following layers connected in sequence: A first convolutional layer uses 32 filters with a size of 5; A second convolutional layer uses 64 filters with a size of 3; A first max-pooling layer; A second max-pooling layer; A fully connected layer comprising 128 neurons; A classification output layer.

[0015] The third aspect of the present application provides a pathogenic bacteria differentiation system based on surface-enhanced Raman spectroscopy and AI assistance, comprising: A spectrum acquisition module for acquiring original Raman spectrum data of pathogenic bacteria samples available for collecting Raman spectrum signals; A data processing module for preprocessing and dimension reduction processing of the original Raman spectrum data, screening out characteristic peaks for classification, and the Raman shift of the characteristic peaks comprising any one or several of the following: 832 cm -1 , 1001 cm -1 , 1122 cm -1 , 1165 cm -1 , 1293 cm -1 , 1353 cm -1 and 1580 cm -1 ; A classification and identification module, the pre-processed different pathogenic bacteria spectrum images are input to a pre-trained one-dimensional convolutional neural network model for analysis, and different pathogenic bacteria Raman spectrum are analyzed based on the characteristic Raman peaks screened out by the model based on dimension reduction analysis, and the classification result of pathogenic bacteria is output.

[0016] The fourth aspect of the present application provides a pathogenic bacteria differentiation device based on surface-enhanced Raman spectroscopy and AI assistance, the device comprising a memory and a processor; the memory is used to store a computer program; the processor is used to realize the AI-assisted pathogenic bacteria differentiation method described above when the computer program is executed.

[0017] The fifth aspect of the present application provides a computer readable storage medium, wherein the storage medium stores a computer program, and the computer program, when executed by a processor, implements the AI-assisted pathogenic bacteria distinguishing method.

[0018] Compared with the prior art, the AI-assisted pathogenic bacteria distinguishing method and system based on surface-enhanced Raman spectrum provided by the present application has the following beneficial effects: (1) In view of the technical problems of weak Raman signal and low sensitivity, the present application uses the SERS substrate and Au@Cu 2-x Se core-shell structure nanoparticles with electromagnetic hot spot enhancement effect to significantly enhance the Raman signal of pathogenic bacteria, realize high-sensitivity detection of low-concentration pathogenic bacteria, and the minimum detection limit can reach 10 -7 mol / L, which provides a reliable signal basis for rapid identification.

[0019] (2) In view of the technical problems of complex matrix interference and insufficient spectral difference, the present application performs preprocessing and dimensionality reduction processing on the data, uses principal component analysis (PCA) and t-SNE algorithm to reduce the dimensionality of the spectral data, and combines asymmetric least squares baseline correction and Savitzky-Golay smoothing processing to effectively remove background interference, enhance the spectral difference between different bacteria, and improve the accuracy of subsequent classification.

[0020] (3) In view of the technical problems of low accuracy of pathogenic bacteria classification and dependence on manual feature extraction, the present application uses a 1D-CNN model with a specific structure to automatically extract local spectral features, realizes high-precision classification of six common pathogenic bacteria and gram-positive / negative bacteria, and the accuracy is more than 99%, which is significantly better than traditional machine learning methods.

[0021] (4) The preparation method of the SERS substrate and nanoparticles provided by the present application ensures the uniformity of the substrate and the monodispersity of the nanoparticles, reduces the signal relative deviation between different batches, has good reproducibility and stability, and is suitable for clinical practical application. BRIEF DESCRIPTION OF DRAWINGS

[0022] The drawings accompanying the specification of the present disclosure serve to provide a further understanding of the present disclosure, and the illustrative embodiments of the present disclosure and their descriptions are used to explain the present disclosure, and do not constitute an improper limitation on the present disclosure.

[0023] Figure 1 is a step flow chart of the pathogenic bacteria sample processing method based on surface-enhanced Raman spectrum provided by the first embodiment of the present application.

[0024] Figure 2is a step flow chart of the AI-assisted pathogenic bacteria distinguishing method based on the pathogenic bacteria sample processing method provided in Embodiment Two of the present application.

[0025] Figure 3 is Au@Cu 2-x is a schematic diagram and structural characterization of the synthesis of Au@Cu

[0026] Figure 4 is a scanning electron microscope image of the PS microspheres and nanobowl cavities prepared in Embodiment One of the present application.

[0027] Figure 5 is Au@Cu 2-x is a spectrum of different concentrations of R6G as a SERS substrate of Au@Cu

[0028] Figure 6 is a processing process and final Raman spectrum of pathogenic bacteria in Embodiment Two of the present application.

[0029] Figure 7 is a principal component analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE) dimensionality reduction analysis result and correlation of each strain in Embodiment Two of the present application.

[0030] Figure 8 is a one-dimensional convolutional neural network learning process based on Raman spectrum in Embodiment Two of the present application and a confusion matrix and ROC curve of the accuracy rate of the test set verification result of different bacterial species.

[0031] Figure 9 is a one-dimensional convolutional neural network learning process based on Raman spectrum in Embodiment Two of the present application and a confusion matrix of the accuracy rate of the test set verification result of gram-positive bacteria and gram-negative bacteria.

[0032] Figure 10 is a schematic diagram of the pathogenic bacteria distinguishing system based on surface-enhanced Raman spectrum and AI assistance provided in Embodiment Three of the present application. DETAILED DESCRIPTION

[0033] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.

[0034] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit exemplary embodiments according to the present invention. As used herein, unless the context clearly indicates otherwise, the singular form is also intended to include the plural form. In addition, it should be understood that the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0035] In the absence of conflict, the embodiments of the present invention and the features thereof may be combined with each other.

[0036] All data in this embodiment is obtained in compliance with laws and regulations and based on the consent of the user, and is used legally.

[0037] Example 1 like Figure 1 As shown, the present invention provides a method for distinguishing pathogens based on surface-enhanced Raman spectroscopy and AI assistance, comprising: A method for processing pathogen samples based on surface-enhanced Raman spectroscopy, comprising: Prepare a SERS substrate with a nanobowl array structure; Preparation of Au@Cu 2-x Se core-shell structured nanoparticles are co-incubated with pathogen samples to form a probe-pathogen complex; The probe-pathogen complex is attached to the SERS substrate and dried to obtain a pathogen sample for collecting Raman spectrum signals.

[0038] Substrate / complex preparation and spectrum acquisition solve the technical problems of traditional Raman spectroscopy, namely "weak signal and difficult to capture", and provide a high-quality signal source for subsequent analysis.

[0039] Specifically, the preparation of the SERS substrate includes: magnetron sputtering of a gold layer on a silicon wafer; The Langmuir-Blodgett method was used to self-assemble a polystyrene microsphere monolayer at the liquid-air interface; After electrochemically depositing the gold nanostructure, the polystyrene microspheres are etched away to form the nanobowl array structure.

[0040] The "cavity structure" of the nanobowl array cooperates with the gold layer to form dense electromagnetic hot spots, thereby achieving effective enhancement of the Raman signal of the analyte; during the preparation of the substrate, the uniform self-assembly of the monolayer of PS microspheres on the surface of the substrate and the controllable electrochemical deposition of the gold nanoparticles in the gap between the microspheres (the thickness of the gold deposition is controlled at 50% of the radius of the microspheres) avoid the protrusions or depressions on the surface of the substrate, ensure the uniform attachment of the probe-pathogen complex, reduce the signal fluctuation caused by uneven attachment, solve the technical problems of large batch-to-batch difference and unstable detection results of the traditional substrate, and ensure the consistency of the signal enhancement effect of different batches of substrates (RSD < 10%).

[0041] Specifically, the Au@Cu 2-x The preparation of the Se nanoparticles includes: The Au core is formed by reducing chloroauric acid with sodium citrate at 160°C; Under the reduction condition of citric acid, Cu 2+ and SeO2 are introduced into the solution containing the Au core to form a Cu 2-x Se shell layer covering the Au core; The obtained core-shell structure nanoparticles are stabilized by using mercapto-modified polyethylene glycol.

[0042] After the mercapto PEG modification, the particles do not aggregate in complex clinical matrices (such as blood and urine), thereby avoiding the signal weakening or distortion caused by aggregation. The composite lattice of the core-shell structure (Au@Cu 2-x Se) can produce a synergistic electromagnetic enhancement effect, and the Raman signal enhancement effect is increased by 3-5 times compared with single Au nanoparticles, thereby further reducing the detection limit.

[0043] In one specific embodiment, the Au@Cu 2-x Se is prepared by the method as shown in Figure 3 a, which specifically includes the following steps: 1) Take 4 mL of 1% (w / v) HAuCl4 and add it to 96 mL of boiling ultrapure water, then add 10 mL of 11.4 mg / ml sodium citrate solution, keep boiling and reflux for 15 min, and finally cool it to room temperature naturally, centrifuge and wash, obtain AuNPs, and store the solution in a 4°C refrigerator for standby.

[0044] 2) 8 mL of the AuNPs aqueous solution prepared above was mixed with 2.39 mL of 0.4031 mg / mL hyaluronic acid (HA) and gently stirred at 30°C. 0.9 mL of 0.2 mol / L ascorbic acid and 0.3 mL of 0.45 mol / L SeO2 were then added in sequence and stirring continued for 10 min. A mixture of 0.3 mL of 0.65 mol / L CuSO4·5H2O and 1.2 mL of 0.2 mol / L ascorbic acid solution was then added under vigorous stirring and stirred for 24 h until the solution turned light green. The product was washed with deionized water, centrifuged three times, and then freeze-dried to obtain dry Au@Cu ... 2-x Se.

[0045] The core-shell structure obtained in this embodiment: Figure 3 b, as confirmed by TEM images. Further characterization using high-angle annular dark-field scanning transmission electron microscopy (HAADF-STEM) and corresponding energy-dispersive X-ray spectroscopy (EDS) revealed a uniform distribution of Se and Cu elements within the shell ( Figure 3 c). Statistical analysis of Figure 1e shows that the gold core diameter is 14.3nm and the overall core-shell structure size is 41.7nm, indicating an estimated shell thickness of approximately 13.7nm. High-resolution TEM image ( Figure 3 f, 3g, 3h) show different lattice spacings: the core exhibits an interplanar spacing corresponding to the (111) planes of Au, while the shell 2-x Se (311) plane alignment. XRD analysis ( Figure 3 i) The composite crystal structure was further confirmed, showing the 2-x Se consistent diffraction pattern. X-ray photoelectron spectroscopy (XPS, Figure 3 j, 3k) elucidate the elemental composition and oxidation state: crucially, Au is mainly in the metallic state (A 0 ) exists, which is essential for plasmon resonance, enabling chemical catalysis and strong light-induced electromagnetic fields. The signals observed at 932.2 eV and 951.8 eV correspond to the Cu2p3 / 2 and Cu2p1 / 2 orbitals of Cu, respectively, while the binding energies at 933.4 eV and 953.4 eV are assigned to Cu 2+ .Cu 2+ and Cu 1+ The coexistence of substances indicates that Cu 2-x The presence of copper vacancies in Se is beneficial to its catalytic activity.

[0046] In order to achieve stable Raman signal acquisition, use Figure 4 The process shown in a prepared a large-area homogeneous SERS substrate.

[0047] Polystyrene (PS) microspheres were self-assembled into a monolayer on an ion-sputtered silicon wafer by liquid-gas interfacial assembly to form a flat hexagonal lattice stabilized by van der Waals interactions. Figure 4 b). Gold nanoparticles were then electrochemically deposited within the gaps to generate a hexagonally ordered array of nanobowls, with the PS microspheres acting as sacrificial templates. After removal of the PS microspheres with dichloromethane (c), the inventors' team obtained a substrate with uniform nanoscale cavities (diameter ~ 200 nm). The optimal gold deposition thickness was kept at ~ 50% of the microsphere radius to ensure the planarity of the substrate. Figure 4

[0048] The SERS enhancement capability and stability of the flexible SERS sensor prepared by the method provided in the embodiment were tested: Au@Cu 2-x Se and the cavity nanobowl double-layer structure as a sensor SERS substrate, rhodamine 6G (R6G) as a probe molecule, different concentrations (10 -3 mol / L, 10 -4 mol / L, 10 -5 mol / L, 10 -6 mol / L, 10 -7 mol / L) of R6G aqueous solution were added dropwise on the sensor, and a Raman spectrometer was used for testing at a wavelength of 785 nm. As Figure 5 a is a test SERS spectrum, showing that the sensor has excellent SERS enhancement capability, and the lowest detection limit of R6G can be detected to 10 -7 mol / L, indicating that the SERS sensor has high sensitivity, and the SERS hot spots formed by uniform gold nanoparticles significantly enhance the Raman signal of the adsorbed analyte molecules in these regions. Figure 5 c is the SERS signal intensity of 10 -7 mol / L R6G aqueous solution using different batches of sensor samples as SERS substrates, showing high consistency, and the relative deviation of the Raman characteristic peak intensity is kept at a low level (< 10%), indicating that the SERS sensor has good reproducibility and stability.

[0049] Embodiment Two As Figure 2 , the application provides an AI-assisted pathogenic bacteria differentiation method based on the pathogenic bacteria sample processing method of embodiment one, comprising: Obtaining original Raman spectrum data of a pathogenic bacteria sample that can be used to collect Raman spectrum signals; Pretreating and reducing the dimensionality of the original Raman spectrum data, and screening characteristic peaks for classification, wherein the Raman shift of the characteristic peaks includes any one or several of the following: 832 cm -1 , 1001 cm -1 ​, 1122 cm -1 , 1165 cm -1 , 1293 cm -1 , 1353 cm -1 , and 1580 cm -1 ; inputting any one or more of the characteristic peaks into a pre-trained one-dimensional convolutional neural network model, analyzing the Raman spectra of different pathogenic bacteria by the model, and outputting a classification result of the pathogenic bacteria.

[0050] The original spectrum data is pre-processed, background interference is effectively removed, the spectral difference between different bacteria is enhanced, and dimension reduction analysis is performed to screen out spectrum characteristic peak data for subsequent classification and identification of pathogenic bacteria. The pre-processed spectrum of different pathogenic bacteria is input into a pre-trained one-dimensional convolutional neural network model for analysis, and the model analyzes the Raman spectra of different pathogenic bacteria based on the characteristic Raman peaks screened out by the dimension reduction analysis, and outputs a classification result of the pathogenic bacteria.

[0051] Dimension reduction processing extracts and screens key features from high-dimensional and high-redundancy original spectrum data, and 1DCNN classification realizes automatic and high-precision identification of pathogenic bacteria, which can not only distinguish single bacteria, but also can realize gram-positive and gram-negative bacteria typing.

[0052] Specifically, the dimension reduction processing includes principal component analysis and t-distributed stochastic neighbor embedding algorithm. Principal component analysis extracts the main variance of the spectrum data, removes redundant information (such as background noise and irrelevant molecular vibration signals), and compresses high-dimensional data to low-dimensional data; t-distributed stochastic neighbor embedding further optimizes the data clustering effect based on PCA, clusters similar spectra into one class, expands the inter-cluster distance of different bacterial spectra, and improves the feature difference. Through dimension reduction analysis, spectrum characteristic peaks for subsequent pathogenic bacteria differentiation are screened out.

[0053] Specifically, the one-dimensional convolutional neural network model comprises the following layers connected in sequence: a first convolutional layer using 32 filters with a size of 5; a second convolutional layer using 64 filters with a size of 3; a first max-pooling layer; a second max-pooling layer; a fully connected layer comprising 128 neurons; a classification output layer.

[0054] The 32 filters with a size of 5 in the convolution layer capture the wide-range features of the spectrum (such as the molecular vibration modes of a wide range of wave numbers), and the 64 filters with a size of 3 capture the local fine features (such as the characteristic peaks of specific wave numbers), so that the macro and micro features of the spectrum can be covered by the two convolution layers; the maximum pooling layer performs down-sampling on the feature map after convolution, reduces the number of parameters (such as reducing the number of parameters by 60%), avoids overfitting of the model, and at the same time preserves the position information of the key features; the fully connected layer integrates the features after pooling into a 128-dimensional vector, highlights the contribution of key features to classification through neuron weight distribution, and finally outputs the pathogenic bacteria class (such as 6 species of bacteria + 2 types of gram).

[0055] Specifically, the method further comprises: preprocessing the original spectrum data to obtain preprocessed spectrum data; The preprocessed spectrum data is subjected to dimensionality reduction processing to obtain feature data after dimensionality reduction.

[0056] The data preprocessing is performed on the original spectrum before dimensionality reduction to solve the technical problems of baseline drift and noise interference of the original spectrum: during the acquisition of the original spectrum, the baseline drift and random noise may occur due to the influence of laser intensity fluctuation, substrate background, environmental interference and the like, and the preprocessing can correct these problems; the signal-to-noise ratio (SNR) of the preprocessed spectrum data is significantly improved, avoiding the noise being retained and the key features being covered due to the direct dimensionality reduction of the original data.

[0057] Specifically, the preprocessing comprises: performing baseline correction using an asymmetric least squares algorithm and performing smoothing processing using a Savitzky-Golay algorithm.

[0058] The asymmetric least squares baseline correction sets a weight parameter (such as a smoothing factor λ = 10 6 ), and only corrects the baseline drift (such as removing the baseline rise caused by instrument light source fluctuation and sample scattering), without affecting the intensity and position of the characteristic peaks (such as a characteristic peak intensity retention rate > 98%); The Savitzky-Golay smoothing adopts a polynomial fitting (such as a 3rd order polynomial) and a sliding window (such as 11 data points), filters random noise (such as reducing the noise intensity by 60%), while preserving the sharpness of the characteristic peaks (such as a change in the half-height width of the characteristic peaks < 5%), avoiding the characteristic blurring caused by smoothing.

[0059] The embodiment is applied to detection of pathogenic bacteria More than 300 million cases of serious or even fatal diseases are caused by bacterial infection or contamination every year, resulting in more than 2 million deaths. Therefore, it is an urgent need to detect pathogenic bacteria quickly, sensitively and at low cost. The sensor prepared by the method provided in the embodiment is used for detecting pathogenic bacteria. The detection method is as follows: The Au@Cu prepared by the method provided in this example 2-x Se was co-cultured with each strain to an OD of ~0.6, and then dripped onto the surface of the nanobowl cavity prepared by the method provided in this example. After drying, the SERS raw spectra were detected using a Raman spectrometer at a wavelength of 785 nm. The SERS raw spectra were scanned and baseline corrected using an asymmetric least squares algorithm and smoothed using Savitzky-Golay ( Figure 6 ).

[0060] The inventors implemented dimensionality reduction techniques for efficient spectral analysis. Principal component analysis (PCA) and t-distributed stochastic neighbor embedding (t-SNE) were used ( Figure 7 a) Seven characteristic Raman bands were screened for subsequent pathogen differentiation. The specific peak positions are shown in the table below.

[0061] The PCA loading plot identified the main sources of variance corresponding to the characteristic Raman bands. The two-dimensional PCA scores showed species-specific clustering with clear separation boundaries, where larger inter-cluster distances indicated stronger spectral differences. t-SNE projections confirmed these species discrimination patterns. In contrast, multivariate analysis of unprocessed spectra resulted in less compact clusters with blurred boundaries, indicating that rigorous preprocessing (baseline correction and smoothing) is essential before performing multivariate spectral analysis. To assess the spectral correlations between different bacterial species, the inventors' team used linear discriminant analysis based on principal component analysis (PCA-LDA). No significant pairwise correlations were observed between any bacterial groups ( Figure 7 b).

[0062] Then, based on the dimensionality reduction process, the integrated convolutional neural network (1DCNN) extracts local features ( Figure 8 a), specifically: the first layer uses 32 filters of size 5, and the second layer uses 64 filters of size 3. This is followed by two max pooling layers and a fully connected layer with 128 neurons, ultimately generating the classification output. In terms of dataset partitioning, the total spectral data volume consists of 71×11 samples, including augmented data generated to expand the dataset. The data is randomly partitioned, with 60% used for training, 20% for validation, and the remaining 20% ​​for testing. To ensure robustness, the full training-validation-test cycle was repeated 20 times to ensure that every sample appears in the training, validation, and test sets during the selection process.

[0063] like Figure 8 b shows the confusion matrix derived from the model, where the values ​​represent the percentage distribution of classification results. The diagonal elements represent the prediction accuracy, which is all above 90%. It is worth noting that Bacillus subtilis ( B.subtilisMDR- E.coli MRSA and Salmonella heidelberg (S. heidelberg) S.heidelberg The accuracy of the proposed 1D-CNN based bacterial classification model reached 100% for all the six combinations. Figure 8 The corresponding receiver operating characteristic (ROC) curves are shown in Fig. 6c, and the area under the curve (AUC) values of all the curves are more than 0.995. This proves the excellent performance of the proposed 1D-CNN based bacterial classification model (AUC > 0.9), which confirms its high classification accuracy. Specifically, in the classification between P. aeruginosa and MDR-E. coli, P. aeruginosa reached a recognition accuracy of 97.8%; in the comparison between P. aeruginosa and MRSA, P. aeruginosa reached an accuracy of 96.6%; for the classification of B. subtilis and S. aureus, S. aureus reached an accuracy of 99.2%; in the pair of P. aeruginosa and S. aureus, S. aureus obtained an accuracy of 99.6%, and the remaining 11 combinations showed a classification accuracy of 100%.

[0064] The present embodiment is applied to the classification of Gram bacteria Among the six bacteria, S. aureus, B. subtilis (corrected from repeated S. aureus) and MRSA represent Gram-positive bacteria, while MDR-E. coli, P. aeruginosa and S. heidelberg constitute Gram-negative bacteria. Figure 9 The binary confusion matrix for the classification of Gram-positive bacteria and Gram-negative bacteria is shown in Fig. 6d, showing that the recognition accuracy of the two types of bacteria is more than 99%. In addition, the AUC value of the corresponding receiver operating characteristic curve reaches 0.999.

[0065] The present application adopts the strategy of "bacterial culture → connection of SERS hot spot nano probes → attachment to nano bowl array → AI assisted spectral processing", adds nano particles that can capture pathogenic bacteria and amplify their Raman spectrum signals, obtains pathogenic bacteria in good growth state by inoculating liquid culture medium, and the operation is simple; then SERS probes that can bind to the surface of bacteria are added for co-culture, and then attached to the nano bowl array SERS substrate, through the enhancement of the hot spot area of the substrate to the signal of pathogenic bacteria, the time of pathogenic bacteria spectrum acquisition is shortened, and the sensitivity and reproducibility of spectral analysis are improved, at the same time, with the help of AI assisted strategy, the Raman spectrum of the six pathogenic bacteria and Gram bacteria is analyzed and processed, realizing the accurate distinction of the six pathogenic bacteria. The method has the advantages of strong specificity, high sensitivity and good reproducibility, and solves the bottleneck of the actual application of the existing Raman spectrum detection method in distinguishing different types of pathogenic bacteria.

[0066] Firstly, the Au@Cu 2-xThe electromagnetic hot spot of the Se and cavity nanobowl double-layer structure SERS substrate enhances the collected Raman spectrum signal, provides excellent signal uniformity and sensitivity, and lays a foundation for improving the accuracy of spectral analysis. The pre-processing algorithm reduces the background interference caused by complex media, the dimension reduction technology enhances the difference of spectral data, and the 1D-CNN deep learning model has a faster data extraction speed than the traditional machine learning model, and can better extract the local important features of the spectrum, which shortens the data acquisition time and effectively improves the recognition accuracy of different pathogenic bacteria. This solves the problems of slow turnaround and limited reliability of traditional methods. In the analysis of gram-positive bacteria, the accuracy of distinguishing different types of positive and negative bacteria is more than 99%, and the recognition accuracy of the same type of bacteria is also more than 97%. The method provided by the present application has the advantages of simple steps, short time consumption, fast differentiation of different types of pathogenic bacteria, high accuracy and good reproducibility, and has important significance for realizing early real-time and accurate treatment guidance of bacterial infection.

[0067] Embodiment three As shown in Figure 10 , the present embodiment provides a pathogenic bacteria differentiation system based on surface enhanced Raman spectrum and AI assistance, comprising: A spectrum acquisition module is configured to acquire original Raman spectrum data of a pathogenic bacteria sample available for collecting Raman spectrum signals; A data processing module is configured to pre-process and dimensionally reduce the original Raman spectrum data, and screen out characteristic peaks for classification, wherein the Raman shift of the characteristic peaks includes any one or more of 832 cm -1 , 1001 cm -1 , 1122 cm -1 , 1165 cm -1 , 1293 cm -1 , 1353 cm -1 and 1580 cm -1 ; A classification and recognition module is configured to input the pre-processed spectrum of different pathogenic bacteria into a pre-trained one-dimensional convolutional neural network model for analysis, analyze the Raman spectrum of different pathogenic bacteria based on the characteristic Raman peaks screened out by the model based on dimension reduction analysis, and output the classification result of the pathogenic bacteria.

[0068] Embodiment four The present embodiment provides a pathogenic bacteria differentiation device based on surface enhanced Raman spectrum and AI assistance, which comprises a memory and a processor; the memory is used for storing a computer program; the processor is used for realizing the pathogenic bacteria differentiation method based on surface enhanced Raman spectrum and AI assistance as described above when the computer program is executed.

[0069] The processor is connected with the memory, and the one or more computer programs are stored in the memory. When the electronic device is running, the processor executes the one or more computer programs stored in the memory, so that the electronic device executes the method in the embodiment one.

[0070] It should be understood that, in the embodiment, the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), ready programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0071] The memory can include read-only memory and random access memory, and provide instructions and data for the processor. A part of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0072] In the implementation process, each step of the above method can be completed by integrated logic circuits of hardware in the processor or instructions in the form of software.

[0073] The method in the embodiment one can be directly embodied as execution completed by a hardware processor, or executed by a combination of hardware and software modules in the processor. The software module can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, register, etc. The storage medium is located in the memory, and the processor reads information in the memory and combines hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0074] Those skilled in the art can realize that the units and algorithm steps of each example described in combination with the embodiment can be realized in electronic hardware or a combination of computer software and electronic hardware. Whether the functions are executed in hardware or software mode depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0075] Embodiment five Another embodiment of the present application provides a computer readable storage medium storing a computer program, when the computer program is executed by a processor, a pathogenic bacteria distinguishing method based on surface enhanced Raman spectroscopy and AI assistance is realized.

[0076] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing relevant hardware, and the program can be stored in a computer readable storage medium. When the program is executed, the program can include the processes of the above-mentioned embodiment methods. The storage medium can be a magnetic disc, an optical disc, a read-only memory (ROM), a random access memory (RAM), or the like. In the present application, the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or they can be distributed on multiple network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiment of the present application. In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can exist physically independently, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0077] Although the present application is disclosed as above, the protection scope of the present application is not limited to this. Those skilled in the art can make various changes and modifications without departing from the spirit and scope of the present application, and these changes and modifications will fall within the protection scope of the present application.

Claims

1. A pathogen sample processing method based on surface enhanced Raman spectroscopy, characterized in that: include: Prepare a SERS substrate with a nanobowl array structure; Preparation of Au@Cu 2-x Se core-shell structured nanoparticles are co-incubated with pathogen samples to form a probe-pathogen complex; The probe-pathogen complex is attached to the SERS substrate and dried to obtain a pathogen sample for collecting Raman spectrum signals.

2. The method for processing pathogen samples according to claim 1, wherein: The preparation of the SERS substrate includes: magnetron sputtering of a gold layer on a silicon wafer; The Langmuir-Blodgett method was used to self-assemble a polystyrene microsphere monolayer at the liquid-air interface; After electrochemically depositing the gold nanostructure, the polystyrene microspheres are etched away to form the nanobowl array structure.

3. The method for processing pathogen samples according to claim 1, wherein: The Au@Cu 2-x The preparation of Se nanoparticles includes: Au nuclei were formed by reducing chloroauric acid with sodium citrate at 160°C; Under the condition of citric acid reduction, Cu was introduced into the solution containing the Au core. 2+ and SeO2, forming a Cu layer wrapped around the Au core. 2-x Se shell; The obtained core-shell structured nanoparticles were stabilized using thiol-modified polyethylene glycol.

4. An AI-assisted pathogen differentiation method based on the pathogen sample processing method according to any one of claims 1 to 3, characterized in that: include: Acquiring original Raman spectral data of a pathogen sample for collecting Raman spectral signals; The raw Raman spectral data is preprocessed and dimensionally reduced to screen out characteristic peaks for classification. The Raman shift of the characteristic peaks includes any one or more of the following: 832 cm -1 , 1001 cm -1 、1122 cm -1 、1165 cm -1 、1293cm -1 、1353 cm -1 and 1580 cm -1 ; When distinguishing pathogens, the one-dimensional convolutional neural network model analyzes the input preprocessed spectra of different pathogens, and distinguishes different types of pathogens based on the characteristic Raman peaks screened out by dimensionality reduction analysis, and outputs the classification results.

5. The AI-assisted pathogen differentiation method according to claim 4, wherein: The preprocessing includes: performing baseline correction using an asymmetric least squares algorithm and performing smoothing using a Savitzky-Golay algorithm.

6. The AI-assisted pathogen differentiation method according to claim 4, wherein: The dimensionality reduction process includes principal component analysis and t-distributed random neighbor embedding algorithm.

7. The AI-assisted pathogen differentiation method according to claim 4, wherein: The one-dimensional convolutional neural network model includes the following layers connected in sequence: The first convolutional layer uses 32 filters of size 5; The second convolutional layer uses 64 filters of size 3; First max pooling layer; Second maximum pooling layer; Fully connected layer, containing 128 neurons; Classification output layer.

8. A pathogen differentiation system based on surface-enhanced Raman spectroscopy and AI assistance, characterized in that: include: A spectrum acquisition module, used to obtain original Raman spectrum data of pathogen samples for collecting Raman spectrum signals; The data processing module is used to pre-process and reduce the dimension of the original Raman spectrum data, and screen out the characteristic peaks for classification. The Raman shift of the characteristic peaks includes any one or more of the following: 832 cm -1 , 1001 cm -1 、1122cm -1 、1165 cm -1 、1293 cm -1 、1353 cm -1 and 1580 cm -1 ; The classification and recognition module inputs the preprocessed spectra of different pathogens into a pre-trained one-dimensional convolutional neural network model for analysis. The model analyzes the Raman spectra of different pathogens based on the characteristic Raman peaks screened out by dimensionality reduction analysis and outputs the classification results of the pathogens.

9. A pathogen differentiation device based on surface-enhanced Raman spectroscopy and AI assistance, characterized in that: The device includes a memory and a processor; the memory is used to store a computer program; and the processor is used to implement the AI-assisted pathogen differentiation method as described in any one of claims 4 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The storage medium stores a computer program, and when the computer program is executed by the processor, the AI-assisted pathogen differentiation method according to any one of claims 4 to 7 is implemented.

Images