Antibiotic identification method based on absorption and fluorescence spectrum internal correlation

Abstract

The invention relates to an antibiotic identification method based on absorption and fluorescence spectrum internal correlation. The method comprises the following steps: firstly, carrying out scattering removal and internal filtering effect correction on original three-dimensional fluorescence excitation-emission matrix data; then, accurately measuring an ultraviolet-visible absorption spectrum of a water sample, constructing an absorption corrected fluorescence spectrum model, and carrying out deep fusion and enhancement on the corrected fluorescence spectrum by taking the absorption characteristic as a weight, so as to effectively highlight and enhance a real fluorescence fingerprint closely related to the absorption characteristic of the antibiotic; and finally, carrying out singular value decomposition on the enhanced fluorescence spectrum data, and carrying out similarity matching and threshold judgment on the enhanced fluorescence spectrum data and a known antibiotic standard characteristic database to realize accurate qualitative recognition and high-precision quantitative analysis on different antibiotics in the water sample. According to the method, the real spectrum fingerprints of the antibiotics can be recovered to the maximum extent, and the sensitivity, accuracy and reliability of antibiotic pollution detection are remarkably improved.

Description

Technical Field

[0001] This invention relates to an antibiotic identification method based on the intrinsic correlation between absorption and fluorescence spectra. Background Technology

[0002] Antibiotics, as a widely used class of drugs, enter the aquatic environment during their production, use, and discharge, becoming a global environmental problem. Antibiotic pollution poses a serious threat to ecosystems and human health, potentially leading to increased bacterial resistance and disrupting the balance of aquatic ecosystems. Therefore, rapid and accurate detection of antibiotics in water bodies is crucial.

[0003] Three-dimensional fluorescence excitation-emission matrix (3D-EEM) spectroscopy has demonstrated significant advantages in environmental monitoring and aquatic antibiotic detection due to its speed, non-destructive nature, and rich information content. By simultaneously scanning excitation and emission wavelengths, it can generate unique "fluorescent fingerprints" reflecting the type and concentration of organic matter and specific fluorescent antibiotics in the water. However, interference from Rayleigh and Raman scattering, as well as internal filtration effects, often occur during EEM spectroscopy acquisition, which can affect the accuracy of the fluorescence signal. Furthermore, single fluorescence spectral information may not be sufficient to fully reveal the characteristics of complex antibiotic pollutants in water.

[0004] On the other hand, ultraviolet-visible (UV-Vis) absorption spectroscopy can provide information on the absorption characteristics of dissolved substances in water. Although absorption spectroscopy itself is difficult to use for qualitative and quantitative analysis of complex mixtures, it has an inherent relationship with fluorescence spectroscopy, i.e., absorption is a prerequisite for fluorescence emission. Traditionally, absorption spectroscopy and fluorescence spectroscopy are often used independently in water quality analysis, failing to fully utilize the synergistic information between the two.

[0005] In existing technologies, although some studies have attempted to combine multiple spectroscopic techniques for water quality analysis, few methods can fully utilize the inherent physicochemical relationship between absorption and fluorescence spectroscopy to extract deeper characteristic information closely related to the internal properties of antibiotic molecules, thereby achieving a more comprehensive and accurate analysis of antibiotic pollutants in complex water bodies. Therefore, developing a novel analytical method for antibiotic pollutants in water bodies that can integrate absorption and fluorescence spectral information and extract richer and deeper characteristics is of great significance. Summary of the Invention

[0006] To overcome the shortcomings of insufficient single spectral information in existing technologies, this invention provides an antibiotic identification method based on the intrinsic correlation between absorption and fluorescence spectra. By integrating absorption and fluorescence spectral data, it reveals deeper molecular internal characteristics of antibiotics, thereby improving the accuracy and information richness of antibiotic pollutant identification and analysis in water bodies.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0008] An antibiotic identification method based on the intrinsic correlation between absorption and fluorescence spectra is proposed. This method first acquires and preprocesses the three-dimensional fluorescence excitation-emission matrix (3D-EEM) spectrum and ultraviolet-visible (UV-Vis) absorption spectrum of water samples to eliminate the effects of scattering and internal filtration. Then, by constructing an absorption-corrected fluorescence spectrum, an intrinsic correlation model between absorption and fluorescence is established, deeply fusing the two spectral information to generate joint spectral data that more comprehensively reflects the molecular characteristics of antibiotics. Finally, singular value decomposition (SVD) is performed on the enhanced fluorescence spectral data. Through similarity matching and threshold judgment with a known antibiotic standard feature database, accurate qualitative identification and high-precision quantitative analysis of different antibiotics in the water sample are achieved.

[0009] Furthermore, the antibiotic analysis method includes the following steps:

[0010] Step 1, data acquisition and preprocessing, as follows:

[0011] Acquire three-dimensional fluorescence excitation-emission matrix (3D-EEM) spectral data of the water sample to be tested. raw (W ex W em The fluorescence intensity F is then subjected to descattering processing, including Rayleigh and Raman scattering removal, to obtain the preliminary fluorescence intensity F. pre_scatter (W ex W em );

[0012] Subsequently, the ultraviolet-visible (UV-Vis) absorption spectrum data A(W) of the water sample to be tested were acquired. abs ), combined with absorption spectrum A(W abs ), for F pre_scatter (W ex W em Internal filtration effect correction was performed to obtain the corrected fluorescence intensity F. corr (W ex W em The calculation formula is as follows:

[0013]

[0014] Among them, A(W) ex ) and A(W em The following are the fluorescence excitation wavelengths of the water sample: W and W, respectively. ex and fluorescence emission wavelength W em Absorbance at that location;

[0015] Step 2, establish an intrinsic correlation model between absorption and fluorescence, as follows:

[0016] Constructing absorption-corrected fluorescence spectra F final (W ex W em This model uses the characteristics of the absorption spectrum as weights to weight the corrected fluorescence spectrum, highlighting antibiotic components that are more likely to fluoresce at specific absorption wavelengths. The calculation formula is as follows:

[0017] F final (W ex W em ) = F corr (W ex W em )×A(W ex (2)

[0018] Among them, A(W) ex ) is related to the fluorescence excitation wavelength W ex The corresponding absorption value. By directly incorporating the absorption intensity at the excitation wavelength into the fluorescence intensity, this model strengthens the intrinsic correlation between absorption and fluorescence, resulting in a final F... final (W ex W em Spectral information can more directly and comprehensively reflect the absorption and fluorescence emission processes of antibiotic molecules;

[0019] Step 3, extract the intrinsic correlation characteristics, as follows:

[0020] F final (W ex W em The matrix is ​​converted into a two-dimensional matrix, and features are extracted using Singular Value Decomposition (SVD). The SVD decomposition is expressed as:

[0021] F final =UΣV T (3)

[0022] Where U is an m×m orthogonal matrix, and V is an n×n orthogonal matrix, V T It is the transpose of V. Σ is an m×n diagonal matrix whose diagonal elements are singular values ​​σ1≥σ2≥…≥0;

[0023] Step 4, antibiotic analysis, as follows:

[0024] The singular values ​​and singular vectors extracted in step 3 are combined to form a high-dimensional feature vector X. sampleThis serves as the "absorption-fluorescence combined fingerprint" of antibiotics in the water sample to be tested. Subsequently, this combined fingerprint is compared with a pre-established "known antibiotic characteristic database," which should contain the corresponding SVD feature vectors X obtained by processing various common antibiotics and their standard solutions under the same conditions through steps 1 to 3. ref,i ;

[0025] By calculating the feature vector X of the water sample to be tested sample Compared with the feature vectors X of known antibiotics in the database ref,i Cosine similarity between Sim(X) sample ,X ref,i This method identifies the types of antibiotics present in water samples. The formula for calculating cosine similarity is:

[0026]

[0027] Among them, X sample X is the feature vector of the water sample to be tested; ref,i It is the feature vector of the i-th known antibiotic in the antibiotic feature database. The closer the cosine similarity value is to 1, the smaller the angle between the two feature vectors and the higher the similarity.

[0028] For each known antibiotic i in the database, if its cosine similarity to the water sample to be tested is Sim(X) sample ,X ref,i The similarity exceeds the preset threshold T. sim If the condition is met, it is determined that the antibiotic is present in the water sample being tested. Multiple antibiotics may simultaneously meet this condition, indicating that there may be mixed contamination of multiple antibiotics in the water sample.

[0029] The technical concept of this invention is to introduce an antibiotic identification method based on the intrinsic correlation between absorption and fluorescence spectra. This method accurately measures the ultraviolet-visible (UV-Vis) absorption spectrum and the three-dimensional fluorescence excitation-emission matrix (3D-EEM) spectrum of the water sample to be tested, and utilizes the inherent energy absorption and emission relationship between the two to construct an absorption-corrected fluorescence spectral model. This model performs deep weighting and fusion of the original fluorescence data, thereby effectively highlighting and strengthening the fluorescence signal closely related to the absorption characteristics of antibiotics, compensating for the influence of internal filtration effects, and purifying spectral features. Based on this, a high-dimensional joint feature vector is extracted from the absorption-corrected fluorescence spectrum using the singular value decomposition (SVD) method. This vector more comprehensively characterizes the molecular internal properties of antibiotics in the water. By calculating the similarity between the joint feature vector and the standard feature vector in a known antibiotic feature database, and setting corresponding thresholds for judgment, accurate qualitative identification of different antibiotics is achieved. Simultaneously, by establishing a standard curve of features versus concentration, precise quantitative analysis of antibiotics is achieved, thereby significantly improving the accuracy and reliability of antibiotic pollution detection in water bodies.

[0030] The beneficial effects of this invention are:

[0031] 1. Enhanced information richness: It integrates both absorption and fluorescence information, providing richer and more comprehensive data on the internal properties of antibiotic molecules than a single spectrum, which helps to more accurately identify and analyze complex antibiotic pollutants in water.

[0032] 2. Enhanced feature correlation: By constructing an absorption-corrected fluorescence spectral model, the extracted features more directly reflect the energy absorption, transfer, and emission processes of antibiotic molecules, and these features are more closely related to the structure and composition of antibiotic molecules.

[0033] 3. Improve detection accuracy: Analysis based on features with more intrinsic physical meaning can effectively reduce the impact of interference such as scattering and internal filtering effects on the results in traditional methods, thereby improving the accuracy of qualitative and quantitative analysis of antibiotics.

[0034] 4. Expanding the scope of application: This method is not only applicable to the detection of antibiotics with strong fluorescence, but also to antibiotics with weak fluorescence but specific absorption characteristics. It can be used to indirectly analyze the influence of antibiotics on the overall absorption-fluorescence correlation characteristics of water samples, thus broadening the scope of antibiotic detection. Attached Figure Description

[0035] Figure 1 This is a flowchart of the method of the present invention.

[0036] Figure 2 This is a matching diagram of absorption and fluorescence spectra.

[0037] Figure 3These are the spectral characteristics of two compounds whose absorption peaks are separated but whose fluorescence peaks are close in position. Compound 1 represents compound 1 (ciprofloxacin), and Compound 2 represents compound 2 (norfloxacin). Detailed Implementation

[0038] The technical solution of the present invention will be further described below with reference to the accompanying drawings.

[0039] refer to Figures 1-3 This paper presents an antibiotic identification method based on the intrinsic correlation between absorption and fluorescence spectra. The method first acquires and preprocesses the three-dimensional fluorescence excitation-emission matrix (3D-EEM) spectrum and ultraviolet-visible (UV-Vis) absorption spectrum of water samples to eliminate the effects of scattering and internal filtration. Then, by constructing an absorption-corrected fluorescence spectrum, an intrinsic correlation model between absorption and fluorescence is established, deeply fusing the two spectral information to generate joint spectral data that more comprehensively reflects the molecular characteristics of antibiotics. Finally, singular value decomposition (SVD) is performed on the enhanced fluorescence spectral data. Through similarity matching and threshold judgment with a known antibiotic standard feature database, accurate qualitative identification and high-precision quantitative analysis of different antibiotics in water samples are achieved.

[0040] The method includes the following steps:

[0041] Step 1, data acquisition and preprocessing, as follows;

[0042] Acquire three-dimensional fluorescence excitation-emission matrix (3D-EEM) spectral data of the water sample to be tested. raw (W ex W em The fluorescence intensity F is then subjected to descattering processing, including Rayleigh and Raman scattering removal, to obtain the preliminary fluorescence intensity F. pre_scatter (W ex W em );

[0043] Subsequently, the ultraviolet-visible (UV-Vis) absorption spectrum data A(W) of the water sample to be tested were acquired. abs Combined with absorption spectrum A(W) abs ), for F pre_scatter (W ex W em Internal filtration effect correction was performed to obtain the corrected fluorescence intensity F. corr (W ex W em The calculation formula is as follows:

[0044]

[0045] Among them, A(W) ex ) and A(Wem The following are the fluorescence excitation wavelengths of the water sample: W and W, respectively. ex and fluorescence emission wavelength W em Absorbance at that location;

[0046] Step 2, establish an intrinsic correlation model between absorption and fluorescence, as follows:

[0047] Constructing absorption-corrected fluorescence spectra F final (W ex W em This model uses the characteristics of the absorption spectrum as weights to weight the corrected fluorescence spectrum, highlighting antibiotic components that are more likely to fluoresce at specific absorption wavelengths. The calculation formula is as follows:

[0048] F final (W ex W em ) = F corr (W ex W em )×A(W ex (6)

[0049] Among them, A(W) ex ) is related to the fluorescence excitation wavelength W ex The corresponding absorption value. By directly incorporating the absorption intensity at the excitation wavelength into the fluorescence intensity, this model strengthens the intrinsic correlation between absorption and fluorescence, resulting in a final F... final (W ex W em Spectral information can more directly and comprehensively reflect the energy absorption and fluorescence emission processes of antibiotic molecules;

[0050] Step 3, extract the intrinsic correlation features, as follows:

[0051] F final (W ex W em The matrix is ​​converted into a two-dimensional matrix, and features are extracted using Singular Value Decomposition (SVD). The SVD decomposition is expressed as:

[0052] F final =UΣV T (7)

[0053] Where U is an m×m orthogonal matrix, and V is an n×n orthogonal matrix, V T It is the transpose of V. Σ is an m×n diagonal matrix whose diagonal elements are singular values ​​σ1≥σ2≥…≥0;

[0054] Step 4, antibiotic analysis, as follows:

[0055] The singular values ​​and singular vectors extracted in step 3 are combined to form a high-dimensional feature vector X. sample This serves as the "absorption-fluorescence combined fingerprint" of antibiotics in the water sample to be tested. Subsequently, this combined fingerprint is compared with a pre-established "known antibiotic characteristic database," which should contain the corresponding SVD feature vectors X obtained by processing various common antibiotics and their standard solutions under the same conditions through steps 1 to 3. ref,i ;

[0056] By calculating the feature vector X of the water sample to be tested sample Compared with the feature vectors X of known antibiotics in the database ref,i Cosine similarity between Sim(X) sample ,X ref,i To identify the types of antibiotics present in a water sample, the cosine similarity is calculated using the following formula:

[0057]

[0058] Among them, X sample X is the feature vector of the water sample to be tested; ref,i It is the feature vector of the i-th known antibiotic in the antibiotic feature database. The closer the cosine similarity value is to 1, the smaller the angle between the two feature vectors and the higher the similarity.

[0059] For each known antibiotic i in the database, if its cosine similarity to the water sample to be tested is Sim(X) sample ,X ref,i The similarity exceeds the preset threshold T. sim If the condition is met, it is determined that the antibiotic is present in the water sample being tested; if multiple antibiotics may meet this condition simultaneously, it indicates that there may be mixed contamination of multiple antibiotics in the water sample.

[0060] Based on this, to further improve the accuracy of antibiotic identification and the robustness of quantitative analysis in water bodies, this invention introduces a feature synergistic enhancement mechanism based on the intrinsic correlation between absorption and fluorescence spectra. This method accurately measures the ultraviolet-visible (UV-Vis) absorption spectrum of the water sample and combines it with a three-dimensional fluorescence excitation-emission matrix (3D-EEM) spectrum. Utilizing the inherent relationship between energy absorption and fluorescence emission, an absorption-corrected fluorescence spectral model is constructed. This model performs deep weighting and fusion of the original fluorescence data (i.e., multiplicatively correcting the corrected fluorescence intensity using absorption values), effectively highlighting and strengthening the true fluorescence signal closely related to antibiotic absorption characteristics, compensating for signal loss caused by internal filtration effects, and thus purifying and enhancing the fluorescent fingerprint of antibiotics. After obtaining the enhanced fluorescence features, singular value decomposition (SVD) is used to analyze the independent fluorescence component features in the water sample. By calculating the similarity between the analyzed features and a database of known antibiotic standard features, combined with a preset threshold, accurate qualitative identification of different antibiotics is achieved. Simultaneously, by establishing a standard curve of features versus concentration, precise quantitative analysis of specific antibiotics in water bodies is realized. This method can not only more accurately determine whether a specific antibiotic exists in the water, but also effectively quantify its concentration, providing more reliable decision support for the precise monitoring of antibiotic pollution in the water environment.

[0061] This embodiment uses the detection of two common antibiotics, norfloxacin and ciprofloxacin, in a river sample near a pharmaceutical factory as an example for verification. The specific steps are as follows:

[0062] Step 1: Collect water samples from the river surrounding the pharmaceutical factory. Obtain the raw 3D-EEM spectral data of the water samples using a three-dimensional fluorescence spectrometer. raw (W ex W em Preliminary observation of the raw EEM data shows that in W... ex ≈270-285nm, W em A broad but weak overlapping peak exists in the ≈430-460 nm range, making it difficult to clearly distinguish the independent fluorescence signals of norfloxacin and ciprofloxacin. (Regarding F...) raw (W ex W em Descattering is performed: first-order Rayleigh scattering (W) is removed using a nonlinear interpolation method. em =W ex ) and second-order Rayleigh scattering (W em =2W ex The influence of ) was also considered. Simultaneously, Raman scattering was removed by subtracting the EEM data of blank deionized water under the same conditions. The fluorescence intensity F of the preliminary treatment was obtained. pre_scatter (W ex W emThe ultraviolet-visible absorption spectrum (A(W)) of the water sample in the 200-600 nm range was measured. abs ). Found in W abs The absorbance at 270 nm is A(270nm) = 0.37, and at W abs The absorbance at 280 nm is A(280nm) = 0.21. For F... pre_scatter (W ex W em Internal filtration effect correction is performed. For example, for W... ex =280nm,W em The fluorescence signal at 450 nm is known to be F. pre_scatter (280nm, 450nm) = 1800, and A(280nm) = 0.21 and A(450nm) = 0.07 were measured. Using the above formula (1), the corrected fluorescence intensity F can be obtained. pre_scatter (280nm, 450nm)≈2485.

[0063] Step 2: Using the absorption-corrected fluorescence spectral model, perform a calibration of the entire corrected F-wavelength spectrum. corr (W ex W em The matrix W is used for processing. ex =280nm,W em Taking the point at 450 nm as an example, its corrected fluorescence intensity F corr (280nm, 450nm) = 2485, corresponding to an absorbance value A(280nm) = 0.21. Using the above formula (2), the absorption-corrected fluorescence intensity F can be obtained. final (280nm, 450nm) = 521.85. After absorption correction processing of the entire EEM matrix, the previously blurred overlapping peaks were significantly enhanced and distinguished. Visually, this can be seen in W... ex =280nm,W em A clear peak appears at 450 nm (corresponding to norfloxacin), and in W ex =275nm,W em Another clear peak appeared at 440 nm (corresponding to ciprofloxacin). Compared with the original EEM data, the fluorescence intensity increased by an average of about 2-3 times, the peak shape became sharper and the separation was greatly improved, significantly enhancing the ability to resolve weak signals and overlapping peaks.

[0064] Step 3, the absorption-corrected fluorescence spectrum F obtained in Step 2 final (W ex W emThe sample is treated as a two-dimensional matrix and input into a singular value decomposition (SVD) model for feature extraction. After SVD decomposition, the singular value matrix Σ and the left and right singular vector matrices U and V are obtained. By analyzing the contribution rate of the singular values, the top two largest singular values ​​and their corresponding singular vectors are selected as the main features, since norfloxacin and ciprofloxacin are the two main antibiotics found in the water sample. These two largest singular values ​​and their corresponding left and right singular vectors are flattened and concatenated to form a high-dimensional feature vector X. sample .

[0065] Step 4, convert the feature vector X of the water sample to be tested. sample The similarity is compared with a pre-established "database of known antibiotic characteristics". Empirically, a similarity threshold T is set. sim =0.62. Calculate X respectively. sample With X ref,3 (i=3 corresponds to ciprofloxacin) and X ref,5 The cosine similarity between (i=5 corresponding to norfloxacin) and ciprofloxacin was calculated to obtain Sim(X) of ciprofloxacin. sample ,X ref,3 )=0.721>T sim Norfloxacin's Sim(X) sample ,X ref,5 ) = 0.212 <T sim Therefore, it can be determined that the river water contains the antibiotic ciprofloxacin, but not norfloxacin.

[0066] The embodiments described in this specification are merely examples of implementations of the inventive concept and are for illustrative purposes only. The scope of protection of this invention should not be considered limited to the specific forms described in these embodiments; rather, it extends to equivalent technical means conceived by those skilled in the art based on the inventive concept.

Claims

1. A method for antibiotic identification based on the intrinsic correlation between absorption and fluorescence spectra, characterized in that, First, data acquisition and necessary preprocessing were performed on the three-dimensional fluorescence excitation-emission matrix spectrum and ultraviolet-visible absorption spectrum of the water sample; Subsequently, by constructing absorption-corrected fluorescence spectra, an intrinsic correlation model between absorption and fluorescence is established, and the two spectral information are deeply fused to generate joint spectral data that more comprehensively reflects the molecular characteristics of antibiotics. Finally, singular value decomposition (SVD) is performed on the enhanced fluorescence spectral data, and by similarity matching and threshold judgment with a known antibiotic standard feature database, accurate qualitative identification and high-precision quantitative analysis of different antibiotics in water samples are achieved.

2. The antibiotic identification method based on the intrinsic correlation between absorption and fluorescence spectra as described in claim 1, characterized in that, The method includes the following steps: Step 1, data acquisition and preprocessing, as follows: Acquire the three-dimensional fluorescence excitation-emission matrix (3D-EEM) spectral data of the water sample to be tested. raw (W ex W em The fluorescence intensity F is then subjected to descattering processing, including Rayleigh and Raman scattering removal, to obtain the preliminary fluorescence intensity F. pre_scatter (W ex W em ); Subsequently, the ultraviolet-visible (UV-Vis) absorption spectrum data A(W) of the water sample to be tested were acquired. abs ), combined with absorption spectrum A(W abs ), for F pre_scatter (W ex W em Internal filtration effect correction was performed to obtain the corrected fluorescence intensity F. corr (W ex W em The calculation formula is as follows: F corr (W ex ,W em )=F pre_scatter (W ex ,W em )·10 (A( W ex ) +A(W em )) / 2 (1) Among them, A(W) ex ) and A(W em The following are the fluorescence excitation wavelengths of the water sample: W and W, respectively. ex and fluorescence emission wavelength W em Absorbance at that location; Step 2, establish an intrinsic correlation model between absorption and fluorescence, as follows: Constructing absorption-corrected fluorescence spectra F final (W ex W em This model uses the characteristics of the absorption spectrum as weights to weight the corrected fluorescence spectrum, highlighting antibiotic components that are more likely to fluoresce at specific absorption wavelengths. The calculation formula is as follows: F final (W ex ,W em )=F corr (W ex ,W em )×A(W ex ) (2) Among them, A(W) ex ) is related to the fluorescence excitation wavelength W ex The corresponding absorption value, by directly incorporating the absorption intensity at the excitation wavelength into the fluorescence intensity, strengthens the intrinsic correlation between absorption and fluorescence, resulting in a final F... final (W ex W em Spectral information can more directly and comprehensively reflect the absorption and fluorescence emission processes of antibiotic molecules; Step 3, extract the intrinsic correlation characteristics, as follows: F final (W ex W em The matrix is ​​converted into a two-dimensional matrix, and features are extracted using the Singular Value Decomposition (SVD) method. The SVD decomposition is expressed as: F final =UΣV T (3) Where U is an m×m orthogonal matrix, and V is an n×n orthogonal matrix, V T Σ is the transpose of V, and Σ is an m×n diagonal matrix whose diagonal elements are singular values ​​σ1≥σ2≥…≥0; Step 4, antibiotic analysis, as follows: The singular values ​​and singular vectors extracted in step 3 are combined to form a high-dimensional feature vector X. sample This serves as the "absorption-fluorescence combined fingerprint" of antibiotics in the water sample to be tested. Subsequently, this combined fingerprint is compared with a pre-established "known antibiotic characteristic database," which should contain the corresponding SVD feature vectors X obtained by processing various common antibiotics and their standard solutions under the same conditions through steps 1 to 3. ref,i ; By calculating the feature vector X of the water sample to be tested sample Compared with the feature vectors X of known antibiotics in the database ref,i Cosine similarity between Sim(X) sample ,X ref,i To identify the types of antibiotics present in a water sample, the cosine similarity is calculated using the following formula: Among them, X sample X is the feature vector of the water sample to be tested; ref,i It is the feature vector of the i-th known antibiotic in the antibiotic feature database. The closer the cosine similarity value is to 1, the smaller the angle between the two feature vectors and the higher the similarity. For each known antibiotic i in the database, if its cosine similarity to the water sample to be tested is Sim(X) sample ,X ref,i The similarity exceeds the preset threshold T. sim If the condition is met, it is determined that the antibiotic is present in the water sample being tested. Multiple antibiotics may simultaneously meet this condition, indicating that there may be mixed contamination of multiple antibiotics in the water sample.

Images