Spectroscopy-based method for detecting infection markers in respiratory secretions

Abstract

The present application relates to the technical field of spectral detection, and particularly relates to a method for detecting infection markers in respiratory secretions based on spectrum, which comprises: collecting and preprocessing spectral data of respiratory secretions, and dividing a plurality of spectral sections; screening possible spectral sections containing infection marker signals based on the spectral sections, dividing the possible spectral sections to obtain a plurality of sections, and determining characteristic expression degrees of the sections as infection marker spectral peaks according to spectral peak expression of the infection marker signals; analyzing spectral data before and after preprocessing of the spectral sections, and determining interference degrees of the spectral sections; evaluating a cooperative change mode of infection marker signals corresponding to each section based on the spectral data of another section, and combining the interference degrees; determining a cross-section characteristic vector based on the characteristic expression degrees and the cooperative change mode of the spectral sections; inputting the cross-section characteristic vector into a classifier, and outputting a detection result; and realizing high-sensitivity and high-specificity detection of the infection markers.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of spectral detection technology, and in particular to a method for detecting infection markers in respiratory secretions based on spectrum. BACKGROUND

[0002] Respiratory infection is an important cause of morbidity and mortality, and early and rapid detection of pathogens is crucial. Respiratory secretions, such as sputum, tracheal aspirate, and throat swab solution, are biological samples derived from the respiratory tract, containing nucleic acids, proteins, and other pathogen components, host immune proteins, metabolites, and other biological information. They can be non-labeled and non-destructively detected by spectroscopy to identify characteristic molecules reflecting the infection status in respiratory secretions, including viral RNA (Ribonucleic Acid), bacterial proteins, and other pathogen-related molecules and cytokines, acute phase proteins, and host response molecules, to distinguish between any type of viral or bacterial infection and disease severity. Surface-enhanced Raman spectroscopy (SERS) can quickly and accurately detect biomarkers related to viral or bacterial infection in respiratory secretions due to its high sensitivity and molecular fingerprint specificity.

[0003] In actual situations, respiratory secretions are a highly complex biological matrix containing a large number of non-target components in addition to the target infection markers, such as mucus, host proteins, cell debris, blood components, and other metabolites. During spectral detection, these non-target components produce significant background signals and fluorescence interference in different wavelength segments. For example, in Raman spectroscopy, the characteristic vibration mode of hemoglobin produces a strong fluorescence background in the visible to near-infrared region, with a wide coverage range and high intensity, which easily overlaps with the Raman signal of the target infection marker. Due to the large difference in spectral response of the components in respiratory secretions, some segments may be covered by the strong signal of mucus or hemoglobin, causing the characteristic peaks of bacterial lipopolysaccharide, viral capsid protein, and inflammatory factors to be masked or signal distorted, causing matrix interference, which leads to false positives or false negatives in the detection of infection markers in respiratory secretions, affecting the accuracy of the detection of infection markers in respiratory secretions, and affecting the early diagnosis and treatment of patients. SUMMARY

[0004] To solve the technical problem of low accuracy in the detection of infection markers in respiratory secretions due to matrix interference caused by complex matrix and analysis of only a single segment in the existing spectral detection process, the present application aims to provide a method for detecting infection markers in respiratory secretions based on spectrum, and the technical solution is as follows:

[0005] Collecting and pre-processing spectral data of respiratory secretions, dividing multiple spectral segments;

[0006] Screening possible spectral segments containing infection marker signals based on spectral segments, dividing possible spectral segments to obtain several segments, and determining the characteristic expression degree of the segments as infection marker spectral peaks according to the spectral peak expression of the infection marker signals;

[0007] Analyzing the spectral data before and after pre-processing the spectral segments, determining the interference degree of the spectral segments; and evaluating the synergistic change mode of the infection marker signals corresponding to each segment based on the corresponding spectral data of another segment, combined with the interference degree;

[0008] Determining the cross-segment feature vector based on the comprehensive characteristic expression degree and the synergistic change mode of the spectral segments;

[0009] Obtaining a classifier, inputting the cross-segment feature vector into the classifier, and outputting a detection result.

[0010] Preferably, collecting and pre-processing spectral data of respiratory secretions, dividing multiple spectral segments, includes:

[0011] Collecting and processing respiratory secretion samples to form a signal-enhanced hot spot region;

[0012] Collecting spectral data based on the signal-enhanced hot spot region, and sequentially performing median filtering, smoothing processing, baseline correction, beam axis correction and intensity normalization processing on the spectral data;

[0013] Dividing the processed spectral data into multiple spectral segments, and removing abnormal spectra.

[0014] Preferably, collecting and processing respiratory secretion samples to form a signal-enhanced hot spot region includes:

[0015] Preserving the respiratory secretion sample at low temperature, and obtaining a clear supernatant by degradation and centrifugal processing;

[0016] Obtaining a base, mixing the clear supernatant with the base, combining salt induction to obtain a mixture, carrying the mixture by a carrier and drying to form a signal-enhanced hot spot region.

[0017] Preferably, the spectral peak expression of the infection marker signal includes any peak shape of sharp peak, medium peak and wide peak, and is single peak or multiple peak combination.

[0018] Preferably, screening possible spectral segments containing infection marker signals based on spectral segments, dividing possible spectral segments to obtain several segments, and determining the characteristic expression degree of the segments as infection marker spectral peaks according to the spectral peak expression of the infection marker signals includes:

[0019] obtaining a spectrum curve corresponding to the spectrum section, and defining a spectrum section with a clear peak shape in the spectrum curve as a possible spectrum section containing an infection marker signal;

[0020] determining a fitting baseline of the possible spectrum section by an algorithm, comparing the spectrum curve with the fitting baseline, and performing segmented processing on the possible spectrum section to obtain a plurality of segments;

[0021] analyzing the spectrum curve of each segment according to the spectral peak performance of the infection marker signal, and determining a characteristic performance degree of the infection marker spectral peak in each segment.

[0022] Preferably, the spectrum data before and after the spectrum section is preprocessed is analyzed to determine the interference degree of the spectrum section, specifically:

[0023] The spectrum data before and after preprocessing of each spectrum section is analyzed to obtain an interference processing difference corresponding to the spectrum section, an average value is obtained based on the average processing difference of all spectrum sections, and the interference degree of the corresponding spectrum section is determined according to the interference processing difference of each spectrum section and the average value.

[0024] Preferably, the spectrum data corresponding to each segment and another segment are combined based on the interference degree to evaluate the cooperative change mode of the infection marker signal corresponding to the segment, including:

[0025] The spectrum data corresponding to the two segments is analyzed to obtain the associated change degree of the infection marker signal corresponding to the two segments;

[0026] According to the current analysis of the characteristic performance degree of the two segments, the consistency of the cooperative change of the infection marker signal corresponding to the two segments is obtained;

[0027] The associated change degree and the consistency of the cooperative change are combined to determine the cooperative change relationship of the infection marker signal corresponding to the two segments;

[0028] The maximum value of the cooperative change relationship obtained by any segment and all remaining segments is screened, and the interference degree of the spectrum section corresponding to the current segment is combined to evaluate the cooperative change mode of the infection marker signal corresponding to the current segment.

[0029] Preferably, the spectrum data corresponding to the two segments is analyzed to obtain the associated change degree of the infection marker signal corresponding to the two segments, specifically:

[0030] Obtaining a case base corresponding to respiratory tract secretions, based on spectrum data, analyzing two segments corresponding to Raman shift segments, respectively determining the number of cases in the case base in which two segments simultaneously exist and any segment exists infection marker signal abnormality, and combining the correlation between the concentration sequences of infection marker signals in the case base in which two segments simultaneously exist infection marker signal abnormality, obtaining the correlation degree of change of the infection marker signals corresponding to the two segments.

[0031] Preferably, according to the current analysis of the characteristic performance degree of the two segments, the cooperative change consistency of the infection marker signals corresponding to the two segments is obtained, specifically:

[0032] Respectively determining the smaller value and the larger value of the characteristic performance degree of the current analysis of the two segments, screening all cases with the same characteristic performance degree as the larger value in the case base, obtaining the mean value of the characteristic performance degree based on the smaller value corresponding to the segment and all cases, and obtaining the cooperative change consistency of the infection marker signals corresponding to the two segments according to the smaller value and the mean value of the characteristic performance degree.

[0033] Preferably, the cross-segment feature vector is determined based on the comprehensive characteristic performance degree and the cooperative change mode of the spectrum segment, including:

[0034] Based on the mean value of the characteristic performance degree of all segments in each spectrum segment as the infection marker spectrum peak and the cooperative change mode of the corresponding infection marker signal, the characteristic performance degree and the cooperative change mode of each spectrum segment are respectively determined;

[0035] The characteristic performance degree and the cooperative change mode of each spectrum segment are integrated to obtain the splicing weight of the corresponding spectrum segment;

[0036] The corresponding latent space vector is obtained by analyzing each spectrum segment, and the cross-segment feature vector is obtained by weighted average of the latent space vector through the splicing weight.

[0037] The present application has the following beneficial effects:

[0038] The spectrum section is determined according to spectrum data of respiratory tract secretions, the possible spectrum section containing the infection marker signal is selected, the whole spectrum is avoided to be indiscriminately processed, redundant information and invalid noise are reduced, and the calculation complexity is reduced; the possible spectrum section is processed in sections, and the characteristic expression degree of each section as the infection marker spectrum peak is determined, that is, the sensitivity and specificity of the infection marker signal, which is the target signal, are improved, and excessive background interference caused by non-related spectrum sections is avoided; the interference degree of the spectrum section is determined according to the spectrum data before and after the spectrum section preprocessing, that is, the intensity of the background interference, such as mucus, hemoglobin or host protein signal, suffered by each spectrum section is evaluated, the noise contribution can be reduced or the spectrum section with serious interference can be removed during feature extraction, so that the target peak is prevented from being submerged or misjudged; the cooperative change mode of the infection marker signal corresponding to the section is evaluated, so that the dispersed signal information is integrated, the discrimination ability for the infection state is improved, and the robustness in the case of single spectrum section interference is enhanced; the cross-section feature vector is determined by comprehensively considering the characteristic expression degree and the cooperative change mode, so that the problem that the single spectrum section is easily interfered by the matrix to cause signal distortion is effectively overcome, information complementation and redundancy elimination are realized, the robustness and discrimination ability of feature extraction are significantly improved, the expression of the infection marker signal in the complex respiratory tract secretion matrix is strengthened, and then the overall recognition is improved, and the influence of background interference on a single peak position is reduced; finally, the cross-section feature vector is input into the classifier, and a detection result is output, so that high-sensitivity and high-specificity detection of the infection marker is realized, and the overall detection stability is improved. BRIEF DESCRIPTION OF DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, and the advantages thereof, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0040] Figure 1 A step flow chart of a spectrum-based infection marker detection method in respiratory tract secretions provided by an embodiment of the present application;

[0041] Figure 2 A spectrum curve schematic diagram of a spectrum-based infection marker detection method in respiratory tract secretions provided by an embodiment of the present application;

[0042] Figure 3 A spectrum section segmentation processing schematic diagram of a spectrum-based infection marker detection method in respiratory tract secretions provided by an embodiment of the present application;

[0043] Figure 4A spectral curve schematic diagram before spectral section pretreatment of a spectral-based infection marker detection method in respiratory tract secretions provided by one embodiment of the present application;

[0044] Figure 5 A spectral curve schematic diagram after spectral section pretreatment of a spectral-based infection marker detection method in respiratory tract secretions provided by one embodiment of the present application. DETAILED DESCRIPTION

[0045] In order to further illustrate the technical means and effects taken by the present application to achieve the predetermined purposes, the following describes in detail the specific implementation, structure, features and effects of a spectral-based infection marker detection method in respiratory tract secretions according to the present application, in combination with the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0046] 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.

[0047] The following specifically describes the specific scheme of a spectral-based infection marker detection method in respiratory tract secretions provided by the present application in combination with the accompanying drawings.

[0048] In the process of using surface-enhanced Raman spectroscopy to detect infection markers in respiratory tract secretions in the prior art, the biological matrix of respiratory tract secretions is complex, and metabolites may cause matrix interference, resulting in misjudgment or missed detection of the characteristic peaks of infection markers in the spectral data, reducing the detection accuracy of infection markers. Therefore, the spectral data of respiratory tract secretions is divided into multiple spectral sections, and possible spectral sections containing infection marker signals are screened and processed in sections to determine the degree of characteristic performance of the section as an infection marker spectral peak. Then, the degree of interference of the spectral section is determined, and the cooperative change mode of the infection marker signal corresponding to the section is evaluated, and the cross-section characteristic vector is determined by comprehensively considering the degree of characteristic performance and the cooperative change mode. The classifier outputs the detection result according to the cross-section characteristic vector, thereby improving the sensitivity, specificity and overall stability of the detection of infection markers in respiratory tract secretions.

[0049] Please refer to Figure 1 which shows a step flowchart of a spectral-based infection marker detection method in respiratory tract secretions provided by one embodiment of the present application, the method comprising:

[0050] Step S1: Collect and pretreat the spectral data of respiratory tract secretions, and divide the spectral data into multiple spectral sections;

[0051] Step S2: screening possible spectral segments containing infection marker signals based on spectral segments, dividing possible spectral segments into several segments, and determining the characteristic expression degree of the segments as infection marker spectral peaks according to the spectral peak expression of the infection marker signals;

[0052] Step S3: analyzing the spectral data before and after the preprocessing of the spectral segments to determine the interference degree of the spectral segments; and based on the corresponding spectral data of each segment and another segment, the interference degree is combined to evaluate the synergistic change mode of the infection marker signals corresponding to the segments;

[0053] Step S4: determining the cross-segment feature vector based on the comprehensive characteristic expression degree and the synergistic change mode of the spectral segments;

[0054] Step S5: obtaining a classifier, inputting the cross-segment feature vector into the classifier, and outputting a detection result.

[0055] For better illustration, respiratory secretions refer to sputum, tracheal aspirate, and throat swab solution, containing nucleic acid, protein, and other pathogen components, host immune proteins, metabolites, and other biological samples; infection markers refer to virus RNA, bacterial proteins, and other pathogen-related molecules and cytokines, acute phase proteins, and other host response molecules used to reflect the characteristics of infection markers. The presence or absence of infection markers and their concentration can reflect the type, severity of infection, and immune response status of the body, providing a basis for rapid diagnosis of respiratory infections.

[0056] Respiratory infections are an important cause of morbidity and mortality. For example, acute respiratory infections caused by influenza viruses are highly contagious and can lead to seasonal epidemics or even pandemics. Bacterial pneumonia can cause severe lung inflammation and can rapidly progress to respiratory failure and other critical conditions if not treated promptly. Therefore, early and accurate pathogen detection is crucial, and the present application proposes a spectral-based method for detecting infection markers in respiratory secretions to achieve high sensitivity and specificity in detecting infection markers and improve diagnostic efficiency.

[0057] As an optional implementation, in the embodiment, the collected spectrum data of respiratory tract secretion is analyzed based on a Raman spectrometer using surface-enhanced Raman spectroscopy technology. The Raman spectrometer is a precision optical instrument for detecting information such as molecular vibration and rotation of a substance. The Raman spectrometer irradiates the processed respiratory tract secretion sample with monochromatic laser light to make the sample produce Raman scattering, collects scattered light, and performs spectral processing to generate a corresponding Raman spectrum, which is used to reflect the chemical structure and functional group information of molecules in the sample. The surface-enhanced Raman spectroscopy technology is a technology enhanced on the basis of the Raman spectrum. When molecules in the sample are adsorbed on the surface of a metal nanostructure, the Raman scattering cross-section is greatly increased due to the local enhancement of the electromagnetic field and the chemical interaction, so that extremely low concentration molecules or even single molecules can be detected, and the sensitivity and detection limit of the Raman spectrum are improved.

[0058] Further, in step S1, the following steps are included:

[0059] Step S11: Collect and process the respiratory tract secretion sample to form a signal-enhanced hot spot area.

[0060] Preferably, in the embodiment, the respiratory tract secretion sample refers to sputum, throat swab eluate, or tracheal aspirate. The sputum refers to the deep respiratory tract secretion of a patient discharged by coughing, which contains a large amount of mucus, exfoliated epithelial cells, and pathogenic microorganisms from the trachea, bronchus, and even the alveoli. The throat swab eluate refers to the liquid obtained by dissolving the secretion of the throat of a patient with a specific eluent. The tracheal aspirate refers to the secretion aspirated from the airway of a patient through medical operations such as tracheal intubation or bronchoscopy.

[0061] Further, in step S11, the following steps are included:

[0062] Step S111: Store the respiratory tract secretion sample at low temperature, and obtain a clear supernatant by degradation and centrifugal processing.

[0063] Specifically, the respiratory tract secretion sample is stored at 4°C for a short period of time, and a solution such as 0.1%-0.2% N-acetyl cysteine is used for mucus degradation treatment. That is, by destroying the disulfide bonds between mucin molecules, the viscosity of the respiratory tract secretion sample is reduced, and the mucus components in the respiratory tract secretion sample are effectively decomposed to improve the accuracy of subsequent detection. Cell debris and large particles are removed by low-speed centrifugation to obtain a clear supernatant, which provides a high-quality sample for subsequent data acquisition.

[0064] Step S112: Obtain a substrate, mix the clear supernatant with the substrate, combine salt induction to obtain a mixture, carry the mixture on a carrier, and dry to form a signal-enhanced hot spot area.

[0065] As an optional embodiment, the substrate refers to a SERS (Surface-Enhanced Raman Scattering) substrate, i.e., a gold nanoparticle colloid; and the carrier refers to a glass sheet.

[0066] Specifically, after the substrate is mixed with the clear supernatant, the surface plasmon resonance effect can significantly enhance the Raman scattering signal of the molecules adsorbed on or near the surface of the nanoparticles, realizing high-sensitivity detection; then, the Raman signal is improved by salt-induced aggregation, i.e., a salt solution is added to the mixture of the clear supernatant and the substrate after mixing, and the intermolecular interaction is promoted by stirring or ultrasonic treatment, so that the originally dispersed molecules or nanoparticles are aggregated to form larger aggregates, the Raman scattering cross section is enhanced, the Raman signal intensity is improved, and then the mixture is obtained; the mixture is carried by the carrier and dried, the solvent is gradually volatilized, the components are aggregated or orderly arranged on the surface of the carrier, a local area with a higher concentration is formed, i.e., a signal enhancement hot spot area is formed, and analysis is performed on the area, which can significantly enhance the intensity and sensitivity of the target signal required.

[0067] Step S12: Collecting spectral data based on the signal enhancement hot spot area, and sequentially performing median filtering, smoothing processing, baseline correction, beam axis correction and intensity normalization processing on the spectral data.

[0068] Specifically, the surface-enhanced Raman spectroscopy technology is used to collect spectral data of the signal enhancement hot spot area, wherein the excitation light source is selected as a 785 nm laser, which belongs to the near-infrared band, and the fluorescence quenching effect is more significant, which can reduce the fluorescence interference; the laser power is controlled to be 1-10 mW, which can ensure sufficient excitation energy to produce clear Raman scattering signals and avoid local overheating of the sample caused by excessive power, resulting in ablation or structural damage; the collection range includes the fingerprint region of 400-1800 cm -1 and the high band of 2600-3200 cm -1 . The fingerprint region covers the characteristic vibration modes of various chemical bonds in the molecule, and has high specificity and uniqueness; the high band corresponds to the stretching vibration of the molecule and the vibration mode of some heavy atoms, which can provide more information about the functional group composition and molecular structure of the sample, and is beneficial to the comprehensive characterization of the chemical composition of the sample; in the collection process, a multi-point repeated measurement strategy is adopted, i.e., multiple measurements are performed on different positions, and the obtained data are averaged to reduce the deviation introduced by the position difference, eliminate the systematic error and random deviation caused by the non-uniformity of the sample and the difference in the distribution of the small particles, and improve the accuracy of the measurement results.

[0069] Then, the spectral data is subjected to data quality control to eliminate saturated signals and abnormal spectra, and to confirm the accuracy of wave number calibration; cosmic rays and spike noise are removed by median filtering, i.e. abnormal values are replaced by the median of data within a sliding window, effectively retaining the overall spectral data while removing isolated noise points; random noise is reduced by Savitzky-Golay smoothing, i.e. based on polynomial fitting, the characteristic peak information and peak position accuracy of the spectrum are retained while smoothing the noise; baseline correction is performed by Asymmetric Least Squares (AsLS), i.e. by setting an asymmetric weight function, the baseline and effective signal are effectively distinguished, the fluorescence background and baseline drift are eliminated, and the spectral data is clearer; the spectral data is subjected to wave number axis correction and intensity normalization processing to ensure data comparability, to eliminate possible systematic bias, and to ensure the reliability of the spectral data.

[0070] Step S13: dividing the processed spectral data into multiple spectral segments and eliminating abnormal spectra.

[0071] For better illustration, in the prior art, the case library corresponding to respiratory secretions refers to a database that has systematically collected, sorted and stored clinical case information related to respiratory secretions, which covers patient identity, sample characteristics, detection results and other multi-dimensional data, and which contains signals of segments where infection markers may appear, such as specific pathogen gene fragment sequence signals, signal intervals corresponding to changes in inflammatory factor levels, and specific antibody reaction signal intensity ranges in immunological detection, etc., providing important reference for respiratory secretions infection.

[0072] Preferably, in the present embodiment, the spectral segments are divided according to the segments where infection markers exist in the case library, and abnormal spectral data caused by instrument errors, sample contamination, operation errors, etc. is eliminated, i.e. spectral baseline drift is too large, signal-to-noise ratio is too low, or there is obvious noise interference, ensuring the quality of spectral data for subsequent analysis.

[0073] Please refer to Figure 2 , Figure 2A schematic diagram of a spectrum curve of a method for detecting an infection marker in respiratory tract secretions based on spectrum in the first embodiment of the present application is shown, wherein the abscissa is the Raman shift and the ordinate is the Raman scattering intensity. Understandably, the matrix of respiratory tract secretions is very complex, containing proteins, lipids, nucleic acids, cell fragments, and drug residues and various components, which will produce a large amount of background signal or fluorescence interference in different wave number ranges, resulting in high overall spectral information noise. For example, the proteins in respiratory tract secretions produce strong Raman scattering signals in specific wave numbers, and the lipid components produce characteristic peaks in their related wave bands, which mask the characteristic peaks of the infection pathogens. Therefore, if the full spectrum is analyzed, the detection of the infection marker will not only be disturbed by irrelevant information, but also increase the dimension, leading to the increase of the calculation complexity and the dilution of the characteristics. Therefore, the spectrum segment where the Raman characteristic peak of the infection marker signal is located is analyzed to reduce the influence of other irrelevant spectrum segments on the detection of the infection marker, while reducing the data dimension and improving the sensitivity of the infection marker detection.

[0074] Further, the spectrum peak of the infection marker signal includes any peak shape of sharp peak, medium peak and wide peak, and is single peak or multi-peak combination.

[0075] It can be explained that the sharp peak usually corresponds to the characteristic vibration mode of a specific functional group in the infection marker signal, has a higher peak height and a narrower half-peak width; the medium peak usually originates from the superimposed vibration of multiple functional groups in the infection marker signal or the contribution of partial disordered structure, has a moderate peak height and peak width; the wide peak is usually related to the disordered region of the infection marker molecule, the aggregate state of macromolecules or the solvation effect, has a relatively flat peak shape, a larger half-peak width and a relatively low peak height; wherein the uniformity of the peak is that the farther the distance from the peak point, the smaller the Raman scattering intensity, which can help to distinguish the background noise and other interference signals; if it is a multi-peak, it is approximately symmetrical on both sides of the segment midpoint after segmentation.

[0076] Further, in step S2, it includes:

[0077] Step S21: Obtain the spectrum curve corresponding to the spectrum segment, and define the spectrum segment with clear peak shape in the spectrum curve as the possible spectrum segment containing the infection marker signal.

[0078] It is explained that when analyzing multiple spectrum segments in the spectrum data, all spectrum segments where the infection marker signal may exist are identified, which means that there is a clear peak shape on the spectrum curve corresponding to the spectrum data, and the segment where the spectrum peak performance of the infection marker signal appears is the possible spectrum segment.

[0079] Please refer to Figure 3Each dashed box corresponds to a segment; Step S22: Determine the fitting baseline of possible spectral segments through the algorithm, compare the spectral curve with the fitting baseline, and perform segmentation processing on the possible spectral segments to obtain several segments.

[0080] Specifically, the airPLS algorithm is used to obtain the fitting baseline for possible spectral segments. A penalty term is introduced to balance the smoothness of the baseline fitting and the degree of fit to the original spectral data. During the iteration process, the baseline estimate is continuously updated until the preset convergence condition is met, resulting in a continuous, smooth fitting baseline that reflects the spectral background trend. Based on the deviation of the spectral curve from the fitting baseline, the deviation region exceeding the threshold is segmented. That is, while retaining weak peaks, the threshold is selected by a fixed multiple method based on statistics. In this embodiment, the threshold is the mean of the fitting baseline plus the standard deviation of the deviation of the spectral curve from the fitting baseline, where the deviation refers to the difference in Raman scattering intensity corresponding to the spectral curve and the fitting baseline. In each spectral segment, the spectral signal with a deviation exceeding the threshold is marked, and the continuous marked spectral signal is divided into a segment, thus obtaining several segments of possible spectral segments containing infection marker signals. In addition, the unmarked spectral signals, i.e., spectral signals less than or equal to the threshold, are directly excluded without processing. Similarly, all segments of all possible spectral segments are obtained.

[0081] Step S23: Based on the spectral peak characteristics of the infection marker signals, analyze the segmented spectral curves to determine the characteristic performance of each segment as the spectral peak of the infection marker.

[0082] Specifically, the corresponding calculation formula is:

[0083] ;

[0084] in, Indicates the first Each segment represents the characteristic expression level of the spectral peaks of infection markers; Indicates the number of spectral signals in the segment; Indicates the first Within the segment, the first Raman shift of a spectral signal; Indicates the first Raman shift of the spectral signal with the maximum Raman scattering intensity within each segment; Indicates the first Within the segment, the first Raman scattering intensity of each spectral signal; Indicates the first Symmetry of the spectral signal in each segment; Indicates An exponential function with base 0.

[0085] It can be explained that the number of spectral signals refers to the number of effective spectral data points in the segment; the spectral signal symmetry refers to the DTW (Dynamic Time Warping) matching distance of the two curves of the spectral curve corresponding to any segment from the midpoint to the two end points, that is, taking the midpoint as the reference, the spectral curve corresponding to the segment is divided into left and right two parts, and the matching distance of the two parts is calculated. The smaller the value, the more similar the two curves from the middle to the two sides, the greater the symmetry, and the more likely it is the spectral peak of the infection marker. represents the position feature of the i-th spectral signal in the j-th segment, when the i-th spectral signal is farther away from the peak point, the Raman scattering intensity is smaller, the segment forms the peak shape, and it is more likely that the segment is the spectral peak of the infection marker.

[0086] It can be understood that, due to the different interference intensities of mucus, hemoglobin, host proteins and other background components in respiratory secretions in different spectral segments, only analyzing a single spectral segment may mask the spectral signal of the infection marker. For example, in the near-infrared spectral segment, water molecules and protein amide bond vibrations in mucus will produce strong scattering and absorption signals, which may mask the characteristic vibration of viral RNA; in the mid-infrared spectral segment, the characteristic absorption peak of hemoglobin may interfere with the amide I band of pathogen proteins. Due to the complex chemical composition of viral RNA, pathogen proteins, host inflammation-related proteins and other infection markers, the molecular vibration information is distributed in multiple spectral segments, and different chemical groups produce characteristic signals in different wave number ranges. For example, the purine ring, phosphate skeleton and sugar group of nucleic acid molecules are located in different wavebands of 700-800 cm -1 , 1050-1100 cm -1 and 1250-1500 cm -1 , respectively. The overall cross-segment information can still reflect the presence of infection markers. When the concentration of infection markers changes, the characteristic signals in different spectral segments will show a synergistic increasing and decreasing trend, that is, there is a chemical and statistical correlation between multiple characteristic peaks. Therefore, by using the synergistic mode across the segments, the detection stability and reliability can be enhanced, and the recognition accuracy of infection markers in respiratory secretion samples can be improved when a certain characteristic peak is interfered or lost.

[0087] Please refer to Figure 4 and Figure 5 , wherein, Figure 4 represents the preprocessed spectral data; Figure 5 ​​​This represents the spectral data before preprocessing, i.e., the raw spectral data of respiratory secretions; further, in step S3, the spectral data before and after preprocessing of the spectral segments are analyzed to determine the degree of interference in the spectral segments, specifically as follows:

[0088] Analyze the spectral data before and after preprocessing for each spectral segment to obtain the interference processing differences for the corresponding spectral segment. Obtain the mean value based on the average processing differences for all spectral segments. Determine the degree of interference for the corresponding spectral segment based on the interference processing differences and the mean value for each spectral segment.

[0089] It is explained that respiratory secretions are complex in composition, with mucus, cell debris, blood, etc., generating background interference. After preprocessing the spectral data, segments with weak characteristic features in each spectral region may be infection marker signals after being interfered with by the matrix. Therefore, the degree of interference in each spectral region is analyzed. If the spectral region is severely interfered with, it indicates that the segment with weak characteristic features is more likely to be an infection marker signal; conversely, if the interference is slight, it indicates that the superposition between the infection marker signal and the background signal is small. This method can more easily identify spectral peak signals containing infection markers, avoiding full-spectrum analysis and reducing the amount of data.

[0090] Specifically, if the spectral data of a spectral segment differs significantly before and after preprocessing, it indicates a greater impact from matrix interference from respiratory secretions. Therefore, based on the spectral data before and after preprocessing, the interference processing difference for the corresponding spectral segment can be obtained, and the corresponding calculation formula is as follows:

[0091] ;

[0092] in, Indicates the first Differences in interference processing across different spectral regions; Indicates the number of spectral signals within a spectral segment; Indicates the first Within the spectral segment, the first Raman scattering intensity of each spectral signal before data preprocessing; Indicates the first Within the spectral segment, the first Raman scattering intensity of each spectral signal after data preprocessing; This represents absolute value operations.

[0093] Next, since there are multiple interfering components in respiratory secretions, they have different interference effects in different spectral bands. The interference is significant in some spectral bands, while it may not be affected in others. Therefore, a comparative analysis is performed using other spectral bands to obtain the degree of interference in each spectral band. The corresponding calculation formula is as follows:

[0094] ;

[0095] wherein, represents the interference degree of the first spectral segment; represents the interference processing difference of the first spectral segment; represents the mean of the average processing difference of all spectral segments; represents the normalization function.

[0096] It can be explained that the greater the interference processing difference of the first spectral segment relative to other segments, the greater the influence of the interference on the spectral segment, and the greater the value of the interference degree .

[0097] It can be understood that in the case data of respiratory diseases, the spectral curves of different infection markers present a correlation change, wherein the concentration of the spectral peaks of the infection marker signals presents a synergistic change trend, and the concentration negatively correlates; and further, due to the existence of corresponding chemical correlations between different molecular signals of the infection markers, in order to prevent misjudgment and missed judgment of the infection marker detection, the correlation degree between different infection marker signals is analyzed through the spectral existence of the infection marker signals in the case of respiratory secretions, if the concentrations of two infection marker signals present a correlation change, and the abnormal signal concentrations always appear at the same time, then there is a greater correlation between the two signals; for example, in patients with bacterial pneumonia, the concentrations of C-reactive protein (CRP) and procalcitonin (PCT) increase synchronously, both of which reflect the intensity of the body's inflammatory response, and their signal change trends are highly consistent; that is, through the simultaneous detection of different infection markers, false positives or false negatives that occur when analyzing a single infection marker are avoided, and the reliability of the subsequent detection results is ensured.

[0098] Further, in step S3, based on the spectral data corresponding to each segment and another segment, the synergistic change mode of the infection marker signal corresponding to the segment is evaluated in combination with the interference degree, including:

[0099] Step S31: analyzing the spectral data corresponding to two segments to obtain the correlation change degree of the infection marker signals corresponding to the two segments.

[0100] It is explained that the correlation change degree refers to the correlation and synergistic change degree of the signal intensity, expression level or concentration change between different infection markers.

[0101] Further, in step S31, specifically:

[0102] Obtaining a case library corresponding to respiratory tract secretions, analyzing the Raman shift segments corresponding to the two segments based on spectral data, respectively determining the number of cases in the case library in which infection marker signal abnormalities exist in both segments and in any one segment, and combining the correlation between the concentration sequences of infection marker signals in which infection marker signal abnormalities exist in both segments in the case library, to obtain the correlation change degree of the infection marker signals corresponding to the two segments.

[0103] Preferably, the case library used at this site is the aforementioned case library data, which reflects the case data of multiple respiratory diseases and provides a data basis for the analysis of infection marker signals.

[0104] For better illustration, the presence of infection marker signal abnormalities means that the concentration value detected by the Raman shift segment corresponding to the segment exceeds the normal concentration range of healthy human body or non-infectious state, i.e. if the concentration data detected by the Raman shift segment corresponding to any segment exceeds 2 times or even more of the normal physiological level, then the segment is determined to have infection marker signal abnormalities; and based on the analysis of any two segments in all segments, in this embodiment, the first segment and the second segment are analyzed.

[0105] Specifically, the number of cases in which infection marker signal abnormalities exist in both the first segment and the second segment, or in any one segment, is counted from the case library, i.e. the case library is traversed, the case in which infection marker signal abnormalities exist in both the first segment and the second segment is screened, and the corresponding number of cases is counted and denoted as ; similarly, the number of cases in which infection marker signal abnormalities exist in the first segment or the second segment is obtained and denoted as ; and the correlation between the concentration sequences of infection marker signals in which infection marker signal abnormalities exist in both segments is determined, i.e. the Pearson correlation coefficient is denoted as , to quantify the strength and direction of linear correlation between the two concentration sequences. The value range of the Pearson correlation coefficient is , where 1 represents complete positive correlation, -1 represents complete negative correlation, and 0 represents no linear correlation. For example, when the Pearson correlation coefficient of CRP or PCT is high and positive, it indicates that the correlation between the two infection marker signals is strong; and the correlation change degree of the infection marker signals corresponding to the two segments is obtained, and the corresponding calculation formula is:

[0106] ; ​​​​​​​​

[0107] wherein, represents the correlation degree of the infection marker signal corresponding to the first segment and the first segment; represents the correlation degree of the infection marker signal corresponding to the first segment and the first segment; represents the number of cases in the case library in which the Raman shift segment corresponding to the first segment or the first segment has an abnormal infection marker signal, and the value is not 0; represents the correlation between the concentration sequence of the infection marker signal in which the Raman shift segment corresponding to the first segment and the first segment has an abnormal infection marker signal in the case library.

[0108] It can be explained that, represents the frequency of occurrence of the infection marker signal in which the Raman shift segment corresponding to the first segment or the first segment has an abnormal infection marker signal in the case library, and the correlation degree is greater, indicating that the infection marker signals corresponding to the two abnormal segments have a chemical correlation; similarly, the correlation degree of the infection marker signals corresponding to all pairs of segments is obtained.

[0109] Step S32: According to the characteristic performance degree of the two segments currently analyzed, the cooperative change consistency of the infection marker signals corresponding to the two segments is obtained.

[0110] It is explained that if the infection marker signals with obvious correlation meet the cooperative change trend in the spectrum data of the respiratory secretion sample in the embodiment, it is explained that the spectrum segment in which the segments are located is a key spectrum segment for detecting the infection marker; wherein, the cooperative change consistency refers to the consistency degree of the change trend, amplitude and direction of different infection markers in the spectrum segment.

[0111] Further, in step S32, specifically,

[0112] The smaller value and the larger value of the characteristic performance degree of the two segments currently analyzed are determined respectively, all cases with the same characteristic performance degree as the larger value are screened in the case library, the mean value of the characteristic performance degree is obtained based on the segment corresponding to the smaller value and all cases, and the cooperative change consistency of the infection marker signals corresponding to the two segments is obtained according to the smaller value and the mean value of the characteristic performance degree.

[0113] Specifically, the calculation formula for the consistency of coordinated changes is as follows:

[0114] ;

[0115] in, Indicates the first The segment and the first Consistent cooperative changes in infection biomarker signals corresponding to each segment; Indicates An exponential function with base 0; Indicates the first The segment and the first The smaller value of the feature representation degree corresponding to each segment; Indicates the first The segment and the first The larger value of the feature manifestation degree corresponding to each segment is the mean of the feature manifestation degree of all cases with the same feature manifestation degree in the case database and the segment corresponding to the smaller value. This represents absolute value operations.

[0116] To better illustrate, regarding the first The segment and the first To illustrate the degree of feature representation corresponding to each segment, let's assume the segment number... Each segment corresponds to a feature representation level of 2, that is... ;No. Each segment corresponds to a feature representation level of 4, that is... Therefore, the smaller value of the feature representation degree corresponding to the two segments is... The larger value is Based on the first The degree of feature representation corresponding to each segment Select all cases with the same degree of manifestation from the case database, and then calculate the first... The mean degree of characteristic manifestations of each segment and these identical cases, i.e., through the first The segmented auxiliary judgment of the first The relevant factors of each segment.

[0117] It can be explained that, with the first The segment and the first Two segments within the three segments were used as reference standards, representing the segments with a greater degree of characteristic expression of the spectral peaks of the infection markers. Cases with the same degree of characteristic expression as the larger segment were obtained, and the average degree of characteristic expression of the corresponding cases and the smaller segment were calculated. In this sample, if the degree of characteristic expression of the smaller segment is close to the average degree of characteristic expression of the segment in the same cases, it indicates that the co-change consistency of the infection marker signals in the two segments is high.

[0118] Step S33: Determine the synergistic change relationship of the infection marker signals corresponding to the two segments based on the correlation change degree and the synergistic change consistency.

[0119] It should be noted that the synergistic change relationship refers to the mutual correlation, mutual action or mutual influence change pattern between different infection marker signals in different segments during the infection process of respiratory tract secretions.

[0120] Specifically, the synergistic change relationship of the infection marker signals corresponding to the two segments is determined, and the corresponding calculation formula is:

[0121]

[0122] Wherein, represents the synergistic change relationship of the infection marker signals corresponding to the i-th segment and the j-th segment; represents the correlation change degree of the infection marker signals corresponding to the i-th segment and the j-th segment; represents the exponential function with base e; represents the synergistic change consistency of the infection marker signals corresponding to the i-th segment and the j-th segment; represents the absolute value operation. It should be noted that in the i-th segment and the j-th segment currently analyzed, when the Raman shift corresponding to any segment corresponds to the infection marker signal of respiratory disease, there is another segment with strong correlation degree based on the statistical data of the case library, and the spectral characteristics of the infection marker signals corresponding to the two segments are close to the characteristic performance degree in the statistical data of the case library, which meets the correlation change degree of the infection marker signals corresponding to the two segments. It is proved that the synergistic change relationship of the infection marker signals corresponding to the two segments is strong; similarly, the synergistic change relationship of the infection marker signals corresponding to any segment and each of the other segments is obtained. Step S34: Screen the maximum value of the synergistic change relationship of any segment and the remaining all segments, and combine the interference degree of the spectrum segment corresponding to the current segment to evaluate the synergistic change pattern of the infection marker signal corresponding to the current segment.

[0123]

[0124] Step S34: Screen the maximum value of the synergistic change relationship of any segment and the remaining all segments, and combine the interference degree of the spectrum segment corresponding to the current segment to evaluate the synergistic change pattern of the infection marker signal corresponding to the current segment.

[0125] ​​​​​​​​​It is explained that when the spectral peaks of infection biomarkers are potentially obscured by background interference from respiratory secretions, they appear as weak peaks in the Raman spectral curve. For example, when the Raman characteristic peaks of infection biomarkers such as specific inflammatory factors or pathogen metabolites are reduced in intensity and signal-to-noise ratio due to scattering and absorption by complex matrices such as respiratory mucus and proteins, the infection biomarkers become difficult to accurately identify and quantify. Therefore, using another spectral signal that is correlated with the infection biomarker signal and exhibits a similar trend of coordinated change is used to assist in identification, effectively distinguishing background interference from the true infection biomarker signal.

[0126] Specifically, based on the current analysis, the first The segment is calculated according to the aforementioned step S31. The segment and the first The coordinated changes in infection biomarker signals corresponding to each segment are similarly obtained. The co-variation relationship between the infection biomarker signals of each segment and all remaining segments was analyzed, and the maximum value was selected and denoted as . The co-variation pattern of infection biomarker signals corresponding to the current segment is assessed, and the corresponding calculation formula is as follows:

[0127] ;

[0128] in, Indicates the first Coordinated change patterns of infection biomarker signals corresponding to each segment; Indicates the first The maximum value of the co-variation relationship between the infection biomarker signals corresponding to each segment and all remaining segments; Indicates the first The degree of interference in the spectral regions corresponding to each segment.

[0129] It can be explained that when the first The cooperative change pattern of the infection biomarker signals corresponding to the first segment is not obvious, indicating a strong correlation. The Raman characteristic peak formed by the other segment is a characteristic of the infection biomarker, and the degree of consistency is generally low, indicating possible matrix interference. The degree of interference in the spectral region where each segment is located is evaluated. If the degree of interference is large, the co-variation relationship of the segment is corrected to improve the co-variation relationship of the infection marker signal corresponding to the segment. Similarly, the co-variation pattern of the infection marker signal corresponding to all segments in the respiratory secretion sample in this embodiment is determined.

[0130] It can be understood that the infection marker signals in respiratory secretions are distributed in multiple spectral segments, and each spectral segment can be interfered by different background components such as mucus, protein, blood, etc. The information of infection marker signals cannot be fully reflected by using a single spectral segment, and signal loss or distortion may occur, thereby reducing the sensitivity and specificity of detection. For example, when only relying on the near-infrared light segment, the strong scattering of the protein background can reduce the signal-to-noise ratio of the infection marker signal, resulting in deviation in the judgment of low-concentration infection. Therefore, by fusing and splicing multiple spectral segments, the useful information of different spectral segments is integrated together, the signal complementation is realized, the background interference of a single spectral segment is eliminated, the overall performance of the infection marker signal is enhanced, and the sensitivity and specificity of detection are improved.

[0131] Further, in step S4, the following steps are included:

[0132] Step S41: Based on the mean value of the characteristic performance degree of each spectral segment and the corresponding cooperative change mode of the infection marker signal, the characteristic performance degree and the cooperative change mode of each spectral segment are determined respectively.

[0133] It is explained that based on the analysis of the first spectral segment, the characteristic performance degree of the infection marker spectrum peak in all segments in the spectral segment and the cooperative change mode of the corresponding infection marker signal are sequentially counted, and the number of segments in the spectral segment is determined. Both are subjected to mean value processing, and the mean values of the characteristic performance degree and the cooperative change mode of the first spectral segment are obtained respectively, which are denoted as and respectively. .

[0134] Step S42: The characteristic performance degree and the cooperative change mode of each spectral segment are integrated to obtain the splicing weight of the corresponding spectral segment.

[0135] It is explained that the composition of respiratory secretions is complex, resulting in significant differences in interference of different spectral segments. The adjustment by the splicing weight obtained by calculation can weaken the high-interference spectral segment and retain the high-value spectral segment. Specifically, the splicing weight of the spectral segment is obtained, and the corresponding calculation formula is:

[0136] ;

[0137] Wherein, represents the splicing weight of the first spectral segment; represents the mean value of the characteristic performance degree of the infection marker spectrum peak in all segments in the first spectral segment; represents the mean value of the characteristic performance degree of the infection marker spectrum peak in all segments in the first spectral segment; represents the mean value of the characteristic performance degree of the infection marker spectrum peak in all segments in the first spectral segment; represents the mean value of the characteristic performance degree of the infection marker spectrum peak in all segments in the first spectral segment; ​​​the mean of the synergistic change pattern of the infection marker signals corresponding to all the subsegments in the spectrum segment.

[0138] It can be explained that when detecting the infection markers of respiratory secretions, the goal is to analyze the spectrum data containing the characteristic peaks of infection markers, and the spectrum segment with the characteristic peaks of infection markers has a high splicing weight. At the same time, in order to prevent the interference of secretions from covering the infection signal, the spectrum segment with a synergistic change pattern of infection marker signals is given a higher splicing weight. Then, the information of multi-dimensional infection marker signals is integrated, and the influence of respiratory secretions on single infection marker signal is reduced.

[0139] Step S43: analyze each spectrum segment to obtain the corresponding latent space vector, and weight average the latent space vectors by the splicing weight to obtain the cross-segment feature vector.

[0140] It can be explained that in order to prevent the background interference of complex components of respiratory secretions and strengthen the cross-segment synergistic signal, multi-spectrum segment feature fusion is performed. Autoencoder (Autoencoder) fusion algorithm is used to map each spectrum segment to a low-dimensional latent space to obtain a latent space vector. This vector can effectively retain the core information of each spectrum segment and remove redundant noise. Preferably, in the present embodiment, the spectrum data of each spectrum segment is subjected to feature extraction by PCA (Principal Component Analysis, Principal Component Analysis) to obtain a feature vector of each spectrum segment. That is, the spectrum data of the spectrum segment is input into the PCA model, the eigenvalues and eigenvectors of the data covariance matrix are calculated to determine the number of principal components, each principal component corresponds to a feature vector, which is used to retain the information of the original spectrum data to the greatest extent, and the principal components are independent of each other, reducing the data dimension and reducing the redundant information. Autoencoder fusion algorithm is a feature fusion method based on the structure of autoencoder. By compressing the feature vector of the spectrum data corresponding to each spectrum segment into a low-dimensional latent feature, and then reconstructing the latent feature into an output similar to the feature vector of the spectrum data corresponding to the spectrum segment through the decoder, the latent space vector is obtained. The cross-segment feature vector is obtained by weighting and averaging the latent space vectors corresponding to each spectrum segment by the splicing weight of each spectrum segment. That is, the cross-segment feature vector comprehensively and accurately reflects the true state of each respiratory secretion sample.

[0141] As an optional implementation, in the embodiment, the classifier is SVM and SVR, wherein SVM (Support Vector Machine) is to find an optimal hyperplane to maximize the margin between different classes of samples on both sides of the hyperplane to achieve accurate classification of new samples; SVR (Support Vector Regression) is to find a hyperplane that can best fit the training data, while allowing a certain range of errors, which is measured by The insensitive loss function is implemented.

[0142] It is explained that in step S5, the cross-section feature vector in step S4 is taken as the input of the classifier, and the detection result, i.e., the infection state corresponding to the respiratory secretion sample, including virus, bacteria or negative, is output; at the same time, the concentration data of each infection marker signal is output to intuitively reflect the concentration data of different infection marker signals, identify and detect the infection markers in the respiratory secretion.

[0143] It can be understood that the spectrum section is determined according to the spectrum data of the respiratory secretion, the possible spectrum section containing the infection marker signal is selected, the full spectrum is avoided to be processed without distinction, the redundant information and invalid noise are reduced, and the calculation complexity is reduced; each section is determined as the characteristic expression degree of the infection marker spectrum peak after the section processing of the possible spectrum section, the sensitivity and specificity of the infection marker signal, which is the target signal, are improved, and too much background interference caused by non-related spectrum sections is avoided; the interference degree of the spectrum section is determined according to the spectrum data before and after the spectrum section preprocessing, that is, the intensity of the background interference such as mucus, hemoglobin or host protein signal suffered by each spectrum section is evaluated, the noise contribution can be reduced or the spectrum section with serious interference can be removed during feature extraction, so as to prevent the target peak from being submerged or misjudged; the cooperative change mode of the infection marker signal corresponding to the section is evaluated to integrate the dispersed signal information, improve the discrimination ability of the infection state, and enhance the robustness under the interference of a single spectrum section; the cross-section feature vector is determined by comprehensively considering the characteristic expression degree and the cooperative change mode to effectively overcome the problem that the signal is distorted due to the interference of the single spectrum section, realize information complementation and redundancy elimination, significantly improve the robustness and discrimination ability of feature extraction, and strengthen the expression of the infection marker signal in the complex respiratory secretion matrix, thereby improving the overall recognition of the infection marker signal and reducing the influence of background interference on a single peak position; finally, the cross-section feature vector is input into the classifier to output the detection result, realize high sensitivity and high specificity detection of the infection marker, and improve the overall detection stability.

[0144] It is to be noted that the sequential order of the above-described embodiments of the present application only for the purpose of description, but not the advantages and disadvantages of the embodiments. The processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0145] Each of the embodiments in the specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other. Each embodiment focuses on the difference from other embodiments.

Claims

1. A method for detecting an infection marker in respiratory secretions based on spectroscopy, characterized in that, The method comprises: Collecting and preprocessing spectrum data of respiratory secretions, dividing multiple spectrum segments; Screening possible spectrum segments containing infection marker signals based on spectrum segments, dividing possible spectrum segments to obtain several segments, and determining the characteristic expression degree of the segments as infection marker spectrum peaks according to the spectrum peak expression of the infection marker signals; Analyzing the spectrum data before and after preprocessing of each spectrum segment to obtain the interference processing difference of the corresponding spectrum segment, obtaining the average value based on the average processing difference of all spectrum segments, and determining the interference degree of the corresponding spectrum segment according to the interference processing difference and the average value of each spectrum segment; Analyzing the spectrum data corresponding to the two segments to obtain the correlation change degree of the infection marker signals corresponding to the two segments; obtaining the consistency of the cooperative change of the infection marker signals corresponding to the two segments according to the current analysis of the characteristic expression degree of the two segments; comprehensively determining the cooperative change relationship of the infection marker signals corresponding to the two segments according to the correlation change degree and the consistency of the cooperative change; screening the maximum value of the cooperative change relationship obtained by any segment and all the remaining segments, and evaluating the cooperative change mode of the infection marker signals corresponding to the current segment in combination with the interference degree of the spectrum segment corresponding to the current segment. Determine the cross-segment feature vector based on the comprehensive characteristic expression degree and the cooperative change mode of the spectrum segment; Obtain a classifier, input the cross-segment feature vector into the classifier, and output a detection result.

2. The method according to claim 1, wherein Collecting and preprocessing spectrum data of respiratory secretions, dividing multiple spectrum segments, comprising: Collecting and processing respiratory secretion samples to form a signal enhancement hot spot region; Collecting spectrum data based on the signal enhancement hot spot region, and sequentially performing median filtering, smoothing processing, baseline correction, beam axis correction and intensity normalization processing on the spectrum data; Divide the processed spectrum data into multiple spectrum segments, and remove abnormal spectrum.

3. The method according to claim 2, wherein the method is a method for detecting an infection marker in respiratory tract secretions based on spectroscopy, characterized by, Collecting and processing respiratory secretion samples to form a signal enhancement hot spot region, comprising: Low-temperature preservation of respiratory secretion samples, degradation and centrifugal treatment to obtain clear supernatant; Obtain a base, mix with the clear supernatant, combine with salt induction to obtain a mixture, load the mixture on a carrier and dry to form a signal enhancement hot spot region.

4. The method of claim 1, wherein the method is a method for detecting an infection marker in respiratory tract secretions based on spectroscopy. The spectrum peak expression of the infection marker signal includes any peak shape of sharp peak, medium peak and wide peak, and is single peak or multiple peak combination.

5. The method of claim 4, wherein the method is based on the detection of an infection marker in respiratory secretions. Screening possible spectrum segments containing infection marker signals based on spectrum segments, dividing possible spectrum segments to obtain several segments, and determining the characteristic expression degree of the segments as infection marker spectrum peaks according to the spectrum peak expression of the infection marker signals, comprising: Obtaining the spectrum curve corresponding to the spectrum segment, and defining the spectrum segment with clear peak shape in the spectrum curve as the possible spectrum segment containing the infection marker signal; Determine the fitting baseline of the possible spectrum segment by algorithm, compare the spectrum curve with the fitting baseline, and segment the possible spectrum segment to obtain several segments; According to the spectrum peak expression of the infection marker signal, analyze the spectrum curve of the segment, and determine the characteristic expression degree of each segment as the infection marker spectrum peak.

6. The method of claim 1, wherein the method is a method of detecting an infection marker in respiratory secretions based on spectroscopy. The two-section corresponding spectral data are analyzed to obtain the correlation degree of the infection marker signal corresponding to the two sections, specifically as follows: A case library corresponding to respiratory secretions is obtained, the two-section corresponding Raman shift sections are analyzed based on spectral data, the number of cases in which the infection marker signal is abnormal in the case library is determined when the two sections exist simultaneously and when any section exists, and the correlation between the concentration sequences of the infection marker signal in which the infection marker signal is abnormal in the case library when the two sections exist simultaneously is combined to obtain the correlation degree of the infection marker signal corresponding to the two sections.

7. The method according to claim 6, wherein the method is a method for detecting an infection marker in respiratory tract secretions based on spectroscopy. According to the current analysis of the characteristic expression degree corresponding to the two sections, the consistency of the cooperative change of the infection marker signal corresponding to the two sections is obtained, specifically as follows: The smaller value and the larger value of the characteristic expression degree corresponding to the current analysis of the two sections are determined, all cases in which the characteristic expression degree corresponding to the larger value is the same are screened in the case library, the mean value of the characteristic expression degree is obtained based on the smaller value corresponding to the section and all cases, and the consistency of the cooperative change of the infection marker signal corresponding to the two sections is obtained according to the smaller value and the mean value of the characteristic expression degree.

8. The method of claim 1, wherein the method is a method of detecting an infection marker in respiratory secretions based on spectroscopy. The cross-section characteristic vector is determined based on the comprehensive characteristic expression degree and the cooperative change mode of the spectral section, including: The mean value of the characteristic expression degree and the cooperative change mode of the infection marker signal corresponding to each spectral section in which all sections are the infection marker spectral peak is determined to determine the characteristic expression degree and the cooperative change mode of each spectral section, respectively; The splicing weight of the corresponding spectral section is obtained by comprehensively analyzing the characteristic expression degree and the cooperative change mode of each spectral section; The corresponding latent space vector is obtained by analyzing each spectral section, and the cross-section characteristic vector is obtained by weighted average of the latent space vector through the splicing weight.

Images